benchmark.name
large_stringclasses 1
value | benchmark.version
large_stringclasses 2
values | datetime
large_stringclasses 1
value | metrics.accuracy
large_stringlengths 65
120
| metrics.result
float64 64
129k
| metrics.result_per_accelerator
float64 23.6
16.1k
| metrics.units
large_stringclasses 1
value | model.architecture
large_stringclasses 1
value | model.mlperf_name
large_stringclasses 8
values | model.name
large_stringclasses 19
values | model.number_of_parameters
float64 6
671
| model.weight_data_types
large_stringclasses 7
values | software.framework
large_stringclasses 28
values | software.framework.ai
large_stringclasses 0
values | software.framework.cuda
large_stringclasses 6
values | software.framework.cudnn
large_stringclasses 0
values | software.framework.jetpack
large_stringclasses 1
value | software.framework.pytorch
large_stringclasses 2
values | software.framework.rocm
large_stringclasses 5
values | software.framework.sima
large_stringclasses 0
values | software.framework.tensorrt
large_stringclasses 9
values | software.framework.vllm
large_stringclasses 4
values | software.operating_system
large_stringclasses 31
values | submission.availability
large_stringclasses 2
values | submission.debug_uid
large_stringlengths 16
16
| submission.division
large_stringclasses 2
values | submission.organization
large_stringclasses 24
values | submission.scenario
large_stringclasses 2
values | system.accelerator.count_per_node
int64 1
8
| system.accelerator.memory_capacity
float64 23.6
2.05k
⌀ | system.accelerator.memory_config
large_stringclasses 9
values | system.accelerator.name
large_stringclasses 20
values | system.accelerator.total_count
int64 1
72
| system.accelerator.vendor
large_stringclasses 2
values | system.cpu.caches
large_stringclasses 15
values | system.cpu.core_count
int64 4
384
⌀ | system.cpu.count_per_node
int64 1
4
| system.cpu.frequency
float64 2
4.8k
⌀ | system.cpu.model
large_stringclasses 56
values | system.cpu.vcpu_count
large_stringclasses 2
values | system.cpu.vendor
large_stringclasses 4
values | system.interconnect.accelerator
large_stringclasses 7
values | system.interconnect.accelerator_host
large_stringclasses 4
values | system.memory.capacity
float64 64
8.19k
⌀ | system.memory.configuration
large_stringclasses 66
values | system.name
large_stringclasses 109
values | system.number_of_nodes
int64 1
18
| system.type
large_stringclasses 2
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
mlperf-inference
|
4.1
| null |
ROUGE1: 45.3528 ROUGE2: 23.2465 ROUGEL: 30.3132 TOKENS_PER_SAMPLE: 146.0 gsm8k_accuracy: 73.08 mbxp_accuracy: 59.9
| 50,797.2 | 6,349.65 |
Tokens/s
|
LLM
|
mixtral-8x7b
|
mixtral-8x7b
| 47 |
fp16
|
TensorRT 10.2.0, CUDA 12.4
| null |
12.4
| null | null | null | null | null |
10.2.0
| null |
Ubuntu 22.04.4
|
available
|
283cbbcf5adf4c20
|
closed
|
Sustainable_Metal_Cloud
|
Server
| 8 | null |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
| null | 32 | 2 | null |
Intel(R) Xeon(R) Platinum 8462Y+
| null |
Intel
|
NVLink
|
InfiniBand
| 2,048 |
32x 64GB DDR5
|
SMC H100 (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 43.0343 ROUGE2: 20.0912 ROUGEL: 29.9254 GEN_LEN: 4102192
| 19,889.7 | 2,486.2125 |
Tokens/s
|
LLM
|
gptj-99.9
|
gptj-99.9
| 6 |
fp16
|
TensorRT 10.2.0, CUDA 12.4
| null |
12.4
| null | null | null | null | null |
10.2.0
| null |
22.04.4 LTS
|
available
|
a0e7dd4d87e242a7
|
open
|
HPE
|
Offline
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 4.5 MiB (96 instances), L1i: 3 MiB (96 instances), L2: 192 MiB (96 instances), L3: 210 MiB (2 instances)
| 48 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8468
| null |
Intel
|
NVLink-4
| null | 2,048 |
32x 64GB MTC40F2046S1RC48BA1
|
HPE Cray XD670 (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 43.0346 ROUGE2: 20.0913 ROUGEL: 29.9259 GEN_LEN: 4102422
| 19,627.6 | 2,453.45 |
Tokens/s
|
LLM
|
gptj-99.9
|
gptj-99.9
| 6 |
fp16
|
TensorRT 10.2.0, CUDA 12.4
| null |
12.4
| null | null | null | null | null |
10.2.0
| null |
22.04.4 LTS
|
available
|
a7d6804cca3d4fca
|
open
|
HPE
|
Server
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 4.5 MiB (96 instances), L1i: 3 MiB (96 instances), L2: 192 MiB (96 instances), L3: 210 MiB (2 instances)
| 48 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8468
| null |
Intel
|
NVLink-4
| null | 2,048 |
32x 64GB MTC40F2046S1RC48BA1
|
HPE Cray XD670 (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 45.4026 ROUGE2: 23.228 ROUGEL: 30.3021 TOKENS_PER_SAMPLE: 145.3 gsm8k_accuracy: 73.64 mbxp_accuracy: 59.98
| 18,839 | 4,709.75 |
Tokens/s
|
LLM
|
mixtral-8x7b
|
mixtral-8x7b
| 47 |
fp16
|
TensorRT 10.2.0, CUDA 12.4
| null |
12.4
| null | null | null | null | null |
10.2.0
| null |
Ubuntu 22.04.4 LTS
|
available
|
2c6f2462927a433f
|
open
|
HPE
|
Offline
| 4 | 94 |
HBM3
|
NVIDIA H100-NVL-94GB
| 4 |
NVIDIA
|
L1d: 3 MiB, L1i: 2 MiB, L2: 128 MiB (64 instances), L3: 320 MiB (2 instances)
| 32 | 2 | 4 |
INTEL(R) XEON(R) GOLD 6530
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB
|
HPE ProLiant DL380a Gen11 (4x H100-NVL-94GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 45.4098 ROUGE2: 23.2338 ROUGEL: 30.3109 TOKENS_PER_SAMPLE: 145.3 gsm8k_accuracy: 73.66 mbxp_accuracy: 60.0
| 17,453.4 | 4,363.35 |
Tokens/s
|
LLM
|
mixtral-8x7b
|
mixtral-8x7b
| 47 |
fp16
|
TensorRT 10.2.0, CUDA 12.4
| null |
12.4
| null | null | null | null | null |
10.2.0
| null |
Ubuntu 22.04.4 LTS
|
available
|
f71a64b524a14526
|
open
|
HPE
|
Server
| 4 | 94 |
HBM3
|
NVIDIA H100-NVL-94GB
| 4 |
NVIDIA
|
L1d: 3 MiB, L1i: 2 MiB, L2: 128 MiB (64 instances), L3: 320 MiB (2 instances)
| 32 | 2 | 4 |
INTEL(R) XEON(R) GOLD 6530
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB
|
HPE ProLiant DL380a Gen11 (4x H100-NVL-94GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 44.5491 ROUGE2: 22.1761 ROUGEL: 28.7877 TOKENS_PER_SAMPLE: 293.9
| 1,468.24 | 367.06 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp8
|
vLLM 0.5.2
| null | null | null | null | null | null | null | null |
0.5.2
|
Ubuntu 22.04 (linux-6.5.0-35-generic-glibc2.35)
|
available
|
5199544dc41a427b
|
open
|
NeuralMagic
|
Server
| 4 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 4 |
NVIDIA
|
L1d cache: 3 MiB (64 instances), L1i cache: 2 MiB (64 instances), L2 cache: 128 MiB (64 instances), L3 cache: 120 MiB (2 instances)
| 32 | 2 | 4,100 |
Intel(R) Xeon(R) Platinum 8462Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| null |
undefined
|
SYS-821GE-TNHR H100 'beaker' (4x H100-SXM-80GB, vLLM, FP8)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 44.2003 ROUGE2: 21.8589 ROUGEL: 28.4858 TOKENS_PER_SAMPLE: 586.9
| 1,577.11 | 394.2775 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
int4
|
vLLM 0.5.2
| null | null | null | null | null | null | null | null |
0.5.2
|
Ubuntu 22.04 (linux-6.5.0-35-generic-glibc2.35)
|
available
|
581fb25395f645e5
|
open
|
NeuralMagic
|
Offline
| 4 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 4 |
NVIDIA
|
L1d cache: 3 MiB (64 instances), L1i cache: 2 MiB (64 instances), L2 cache: 128 MiB (64 instances), L3 cache: 120 MiB (2 instances)
| 32 | 2 | 4,100 |
Intel(R) Xeon(R) Platinum 8462Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| null |
undefined
|
SYS-821GE-TNHR H100 'beaker' (4x H100-SXM-80GB, vLLM, GPTQ)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 44.1657 ROUGE2: 21.8046 ROUGEL: 28.4386 TOKENS_PER_SAMPLE: 291.2
| 1,164.04 | 291.01 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
int4
|
vLLM 0.5.2
| null | null | null | null | null | null | null | null |
0.5.2
|
Ubuntu 22.04 (linux-6.5.0-35-generic-glibc2.35)
|
available
|
aa148b2a6f3547a4
|
open
|
NeuralMagic
|
Server
| 4 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 4 |
NVIDIA
|
L1d cache: 3 MiB (64 instances), L1i cache: 2 MiB (64 instances), L2 cache: 128 MiB (64 instances), L3 cache: 120 MiB (2 instances)
| 32 | 2 | 4,100 |
Intel(R) Xeon(R) Platinum 8462Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| null |
undefined
|
SYS-821GE-TNHR H100 'beaker' (4x H100-SXM-80GB, vLLM, GPTQ)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 44.0047 ROUGE2: 21.7276 ROUGEL: 28.4008 TOKENS_PER_SAMPLE: 291.9
| 424.895 | 212.4475 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp32
|
vLLM 0.5.2
| null | null | null | null | null | null | null | null |
0.5.2
|
Ubuntu 23.04 (linux-6.2.0-39-generic-glibc2.37)
|
available
|
6bf77e8b453b4c5e
|
open
|
NeuralMagic
|
Offline
| 2 | 23.647034 | null |
NVIDIA GeForce RTX 4090
| 2 |
NVIDIA
|
L1d cache: 1.1 MiB (24 instances), L1i cache: 768 KiB (24 instances), L2 cache: 48 MiB (24 instances), L3 cache: 45 MiB (1 instance)
| 24 | 1 | 4,800 |
Intel(R) Xeon(R) w7-2495X
| null |
Intel
| null | null | null |
undefined
|
GATE Overflow Intel Sapphire Rapids RTX 4090 (2x RTX 4090, vLLM, FP8)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 44.0047 ROUGE2: 21.7276 ROUGEL: 28.4008 TOKENS_PER_SAMPLE: 291.9
| 424.895 | 212.4475 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp32
|
vLLM 0.5.2
| null | null | null | null | null | null | null | null |
0.5.2
|
Ubuntu 23.04 (linux-6.2.0-39-generic-glibc2.37)
|
available
|
6f5de3b9e1454bf1
|
open
|
NeuralMagic
|
Offline
| 2 | 23.647034 | null |
NVIDIA GeForce RTX 4090
| 2 |
NVIDIA
|
L1d cache: 1.1 MiB (24 instances), L1i cache: 768 KiB (24 instances), L2 cache: 48 MiB (24 instances), L3 cache: 45 MiB (1 instance)
| 24 | 1 | 4,800 |
Intel(R) Xeon(R) w7-2495X
| null |
Intel
| null | null | null |
undefined
|
GATE Overflow Intel Sapphire Rapids RTX 4090 (2x RTX 4090, vLLM, FP8)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 44.1411 ROUGE2: 21.8139 ROUGEL: 28.4395 TOKENS_PER_SAMPLE: 292.9
| 923.333 | 115.416625 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp32
|
vLLM 0.5.2
| null | null | null | null | null | null | null | null |
0.5.2
|
Ubuntu 22.04 (linux-5.15.0-94-generic-glibc2.35)
|
available
|
d807e0b4adb9407c
|
open
|
NeuralMagic
|
Offline
| 8 | 48 |
GDDR6
|
NVIDIA L40S
| 8 |
NVIDIA
|
L1d cache: 1.3 MiB (40 instances), L1i cache: 1.3 MiB (40 instances), L2 cache: 40 MiB (40 instances), L3 cache: 160 MiB (5 instances)
| 4 | 2 | null |
AMD EPYC 9254 24-Core Processor
| null |
AMD
| null |
PCIe
| null |
undefined
|
Crusoe Cloud L40S (8x L40S PCIe, vLLM)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 44.1411 ROUGE2: 21.8139 ROUGEL: 28.4395 TOKENS_PER_SAMPLE: 292.9
| 923.333 | 115.416625 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp32
|
vLLM 0.5.2
| null | null | null | null | null | null | null | null |
0.5.2
|
Ubuntu 22.04 (linux-5.15.0-94-generic-glibc2.35)
|
available
|
e859a3c3abaf4a65
|
open
|
NeuralMagic
|
Offline
| 8 | 48 |
GDDR6
|
NVIDIA L40S
| 8 |
NVIDIA
|
L1d cache: 1.3 MiB (40 instances), L1i cache: 1.3 MiB (40 instances), L2 cache: 40 MiB (40 instances), L3 cache: 160 MiB (5 instances)
| 4 | 2 | null |
AMD EPYC 9254 24-Core Processor
| null |
AMD
| null |
PCIe
| null |
undefined
|
Crusoe Cloud L40S (8x L40S PCIe, vLLM)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 42.9783 ROUGE2: 24.3368 ROUGEL: 31.4041 TOKENS_PER_SAMPLE: 164.6
| 11,189.1 | 11,189.1 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp16
|
CUDA 12.4
| null |
12.4
| null | null | null | null | null | null | null |
Ubuntu 22.04.4
|
available
|
c2b7612d94aa4864
|
open
|
NVIDIA
|
Offline
| 1 | 141 |
HBM3
|
NVIDIA H200-SXM-141GB
| 1 |
NVIDIA
| null | 56 | 2 | null |
Intel(R) Xeon(R) Platinum 8480C
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x 64GB MTC40F2046S1RC48BA1
|
NVIDIA H200 (1x H200-SXM-141GB)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 44.46 ROUGE2: 22.4526 ROUGEL: 29.9493 TOKENS_PER_SAMPLE: 234.4
| 4,575.06 | 4,575.06 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp16
|
CUDA 12.4
| null |
12.4
| null | null | null | null | null | null | null |
Ubuntu 22.04.4
|
available
|
9fed7da170ed4cbc
|
open
|
NVIDIA
|
Offline
| 1 | 141 |
HBM3
|
NVIDIA H200-SXM-141GB
| 1 |
NVIDIA
| null | 56 | 2 | null |
Intel(R) Xeon(R) Platinum 8480C
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x 64GB MTC40F2046S1RC48BA1
|
NVIDIA H200 (1x H200-SXM-141GB)
| 1 |
datacenter
|
mlperf-inference
|
4.1
| null |
ROUGE1: 43.2263 ROUGE2: 24.3279 ROUGEL: 31.3816 TOKENS_PER_SAMPLE: 166.3
| 184.893 | 184.893 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp16
|
Jetpack 6.0, TensorRT 10.1, CUDA 12.2
| null |
12.2
| null |
6.0
| null | null | null |
10.1
| null |
Jetson r36.3.1 L4T
|
available
|
b0a00ef1d4c44da0
|
open
|
NVIDIA
|
Offline
| 1 | null |
LPDDR5
|
NVIDIA Jetson AGX Orin 64G
| 1 |
NVIDIA
| null | 12 | 1 | null |
12-core ARM Cortex-A78AE CPU
| null |
ARM
| null | null | 64 |
64GB 256-bit LPDDR5
|
NVIDIA Jetson AGX Orin Developer Kit 64G (TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGEL: 21.658590925735947 exact_match: 90.0034579748511 TOKENS_PER_SAMPLE: 648.3
| 428.895 | 53.611875 |
Tokens/s
|
LLM
|
llama3.1-405b
|
llama3_1-405b
| 405 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
44931392474a4c8a
|
open
|
HPE
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-NVL-141GB
| 8 |
NVIDIA
| null | 192 | 2 | null |
Intel(R) Xeon(R) 6740E
| null |
Intel
|
NVLink-4
|
PCIe-5
| 3,072 |
16x 128GB MTC40F2047S1RC64BB1 MWFF; 16x 64GB HMCG94AHBRA480N
|
HPE ProLiant DL380a Compute Gen12 (8x H200-NVL-141GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 30.6202 ROUGE2: 13.9221 ROUGEL: 18.9101 TOKENS_PER_SAMPLE: 581.8
| 2,631.93 | 2,631.93 |
Tokens/s
|
LLM
|
llama2-70b-99
|
DeepSeek-R1-Distill-Llama-8B
| 8 |
bfloat16
|
vLLM v0.7.3
| null | null | null | null | null | null | null | null | null |
Ubuntu 22.04.5 LTS (5.15.0-131-generic)
|
available
|
22cb991a91654451
|
open
|
FlexAI
|
Server
| 1 | null |
HBM3
|
NVIDIA H100 80GB HBM3
| 1 |
NVIDIA
|
L1d cache: 6.3 MiB (200 instances), L1i cache: 6.3 MiB (200 instances), L2 cache: 800 MiB (200 instances), L3 cache: 3.1 GiB (200 instances)
| 52 | 2 | null |
Intel Xeon Processor (SapphireRapids)
| null |
Intel
|
NVLink
|
PCIe
| null |
undefined
|
flexbench test node 0ef307db09d34a91 with 8xH100
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.246 ROUGE2: 22.4512 ROUGEL: 27.9398 TOKENS_PER_SAMPLE: 335.5
| 134.714 | 134.714 |
Tokens/s
|
LLM
|
llama2-70b-99
|
Llama-3.3-70B-Instruct-FP8-dynamic
| 70 |
bfloat16
|
vLLM v0.7.3
| null | null | null | null | null | null | null | null | null |
Ubuntu 22.04.5 LTS (5.15.0-131-generic)
|
available
|
383f5a943a6b472d
|
open
|
FlexAI
|
Server
| 1 | null |
HBM3
|
NVIDIA H100 80GB HBM3
| 1 |
NVIDIA
|
L1d cache: 6.3 MiB (200 instances), L1i cache: 6.3 MiB (200 instances), L2 cache: 800 MiB (200 instances), L3 cache: 3.1 GiB (200 instances)
| 52 | 2 | null |
Intel Xeon Processor (SapphireRapids)
| null |
Intel
|
NVLink
|
PCIe
| null |
undefined
|
flexbench test node 0ef307db09d34a91 with 8xH100
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.4619 ROUGE2: 22.0574 ROUGEL: 28.6452 TOKENS_PER_SAMPLE: 295.9
| 1,482.31 | 370.5775 |
Tokens/s
|
LLM
|
llama2-70b-99
|
Llama-2-70b-chat-hf-FP8
| 70 |
fp8
|
vLLM v0.7.3
| null | null | null | null | null | null | null | null | null |
Ubuntu 22.04.5 LTS (5.15.0-131-generic)
|
available
|
1d6bd9150ec04c03
|
open
|
FlexAI
|
Server
| 4 | null |
HBM3
|
NVIDIA H100 80GB HBM3
| 4 |
NVIDIA
|
L1d cache: 6.3 MiB (200 instances), L1i cache: 6.3 MiB (200 instances), L2 cache: 800 MiB (200 instances), L3 cache: 3.1 GiB (200 instances)
| 52 | 2 | null |
Intel Xeon Processor (SapphireRapids)
| null |
Intel
|
NVLink
|
PCIe
| null |
undefined
|
flexbench test node 0ef307db09d34a91 with 8xH100
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.8294 ROUGE2: 22.3814 ROUGEL: 29.1878 TOKENS_PER_SAMPLE: 271.8
| 573.765 | 573.765 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp16
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Jetson L4T
|
preview
|
bf96378f74a14cdc
|
open
|
NVIDIA
|
Offline
| 1 | null |
LPDDR5X
|
NVIDIA Jetson AGX Thor 128G
| 1 |
NVIDIA
| null | 14 | 1 | null |
14-core ARM Poseidon-AE CPU
| null |
ARM
| null | null | 128 |
128GB 256-bit LPDDR5X
|
NVIDIA Jetson AGX Thor Developer Kit 128G (TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.8294 ROUGE2: 22.3814 ROUGEL: 29.1878 TOKENS_PER_SAMPLE: 271.8
| 573.765 | 573.765 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp16
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Jetson L4T
|
preview
|
ff6f1f2e60974a79
|
open
|
NVIDIA
|
Offline
| 1 | null |
LPDDR5X
|
NVIDIA Jetson AGX Thor 128G
| 1 |
NVIDIA
| null | 14 | 1 | null |
14-core ARM Poseidon-AE CPU
| null |
ARM
| null | null | 128 |
128GB 256-bit LPDDR5X
|
NVIDIA Jetson AGX Thor Developer Kit 128G (TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 47.6186 ROUGE2: 25.2204 ROUGEL: 31.4595 TOKENS_PER_SAMPLE: 296.9
| 16,162.8 | 2,020.35 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama3_1-70b-fp8_dyn
| 70 |
fp16
|
SGLang v0.4.3
| null | null | null | null | null | null | null | null | null |
Ubuntu 22.04.4 LTS (Jammy Jellyfish)
|
available
|
da8f192701f04af7
|
open
|
Krai
|
Offline
| 8 | 192 |
HBM3
|
AMD Instinct MI300X 192GB HBM3
| 8 |
AMD
| null | 40 | 2 | null |
Intel(R) Xeon(R) Platinum 8460Y+
| null |
Intel
|
XGMI
|
PCIe-5
| 2,048 |
32x 64GB DDR5
|
Dell PowerEdge XE9680 (8x MI300X-192GB, SGLang v0.4.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGEL: 18.9480793556203 exact_match: 90.14579086697552 TOKENS_PER_SAMPLE: 641.1
| 243.449 | 30.431125 |
Tokens/s
|
LLM
|
llama3.1-405b
|
deepseek-v3
| 671 |
UNSET
|
SGLang v0.4.3
| null | null | null | null | null | null | null | null | null |
Ubuntu 22.04.4 LTS (Jammy Jellyfish)
|
available
|
62aa1de1cf414696
|
open
|
Krai
|
Offline
| 8 | 192 |
HBM3
|
AMD Instinct MI300X 192GB HBM3
| 8 |
AMD
| null | 40 | 2 | null |
Intel(R) Xeon(R) Platinum 8460Y+
| null |
Intel
|
XGMI
|
PCIe-5
| 2,048 |
32x 64GB DDR5
|
Dell PowerEdge XE9680 (8x MI300X-192GB, SGLang v0.4.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5375 ROUGE2: 22.202 ROUGEL: 28.7796 TOKENS_PER_SAMPLE: 299.2
| 16,229 | 2,028.625 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-fp8_dyn
| 70 |
UNSET
|
vLLM v0.7.3
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
af60de163f644337
|
open
|
Krai
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, vLLM v0.7.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.6113 ROUGE2: 22.2456 ROUGEL: 28.8393 TOKENS_PER_SAMPLE: 298.1
| 14,426.4 | 1,803.3 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-fp8_dyn
| 70 |
UNSET
|
vLLM v0.7.3
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
8a98ad7e17364bce
|
open
|
Krai
|
Server
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, vLLM v0.7.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGEL: 21.59129621595158 exact_match: 90.09433818663138 TOKENS_PER_SAMPLE: 641.0
| 188.604 | 23.5755 |
Tokens/s
|
LLM
|
llama3.1-405b
|
llama3_1-405b
| 405 |
fp16
|
vLLM v0.7.3
| null | null | null | null | null | null | null | null | null |
Ubuntu 22.04.4 LTS (Jammy Jellyfish)
|
available
|
be8c3d3156354651
|
open
|
Krai
|
Offline
| 8 | 192 |
HBM3
|
AMD Instinct MI300X 192GB HBM3
| 8 |
AMD
| null | 40 | 2 | null |
Intel(R) Xeon(R) Platinum 8460Y+
| null |
Intel
|
XGMI
|
PCIe-5
| 2,048 |
32x 64GB DDR5
|
Dell PowerEdge XE9680 (8x MI300X-192GB, vLLM v0.7.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGEL: 21.66444823059425 exact_match: 90.10726671078757 TOKENS_PER_SAMPLE: 632.1
| 278.649 | 34.831125 |
Tokens/s
|
LLM
|
llama3.1-405b
|
llama3_1-405b-fp8_dyn
| 405 |
fp16
|
vLLM v0.7.3
| null | null | null | null | null | null | null | null | null |
Ubuntu 22.04.4 LTS (Jammy Jellyfish)
|
available
|
39862e1b9c0a43a2
|
open
|
Krai
|
Offline
| 8 | 192 |
HBM3
|
AMD Instinct MI300X 192GB HBM3
| 8 |
AMD
| null | 40 | 2 | null |
Intel(R) Xeon(R) Platinum 8460Y+
| null |
Intel
|
XGMI
|
PCIe-5
| 2,048 |
32x 64GB DDR5
|
Dell PowerEdge XE9680 (8x MI300X-192GB, vLLM v0.7.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.4974 ROUGE2: 22.1155 ROUGEL: 28.7195 TOKENS_PER_SAMPLE: 298.1
| 17,128.9 | 2,141.1125 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-fp8_dyn
| 70 |
fp16
|
vLLM v0.7.3
| null | null | null | null | null | null | null | null | null |
Ubuntu 22.04.4 LTS (Jammy Jellyfish)
|
available
|
3f250f4875b34e64
|
open
|
Krai
|
Offline
| 8 | 192 |
HBM3
|
AMD Instinct MI300X 192GB HBM3
| 8 |
AMD
| null | 40 | 2 | null |
Intel(R) Xeon(R) Platinum 8460Y+
| null |
Intel
|
XGMI
|
PCIe-5
| 2,048 |
32x 64GB DDR5
|
Dell PowerEdge XE9680 (8x MI300X-192GB, vLLM v0.7.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5066 ROUGE2: 22.1409 ROUGEL: 28.7316 TOKENS_PER_SAMPLE: 297.1
| 12,989 | 1,623.625 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-fp8_dyn
| 70 |
fp16
|
vLLM v0.7.3
| null | null | null | null | null | null | null | null | null |
Ubuntu 22.04.4 LTS (Jammy Jellyfish)
|
available
|
f5ebd4e25f274ee5
|
open
|
Krai
|
Server
| 8 | 192 |
HBM3
|
AMD Instinct MI300X 192GB HBM3
| 8 |
AMD
| null | 40 | 2 | null |
Intel(R) Xeon(R) Platinum 8460Y+
| null |
Intel
|
XGMI
|
PCIe-5
| 2,048 |
32x 64GB DDR5
|
Dell PowerEdge XE9680 (8x MI300X-192GB, vLLM v0.7.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5183 ROUGE2: 22.1357 ROUGEL: 28.7267 TOKENS_PER_SAMPLE: 297.0
| 21,372.3 | 2,671.5375 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-fp8_dyn
| 70 |
fp16
|
vLLM v0.6.4 FlashInfer
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
bdc143a1a35a474d
|
open
|
Krai
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, vLLM v0.6.4)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5421 ROUGE2: 22.1749 ROUGEL: 28.8061 TOKENS_PER_SAMPLE: 295.1
| 19,618.1 | 2,452.2625 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-fp8_dyn
| 70 |
fp16
|
vLLM v0.6.4 FlashInfer
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
b33b1f283a184dca
|
open
|
Krai
|
Server
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, vLLM v0.6.4)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5823 ROUGE2: 22.2287 ROUGEL: 28.8178 TOKENS_PER_SAMPLE: 296.5
| 15,906.1 | 1,988.2625 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-fp8_dyn
| 70 |
fp16
|
vLLM v0.6.4
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
9701226c7ddb4327
|
open
|
Krai
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, vLLM v0.6.4)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5362 ROUGE2: 22.176 ROUGEL: 28.773 TOKENS_PER_SAMPLE: 295.7
| 13,716.7 | 1,714.5875 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-fp8_dyn
| 70 |
fp16
|
vLLM v0.6.4
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
504172c763c6410a
|
open
|
Krai
|
Server
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, vLLM v0.6.4)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.4973 ROUGE2: 22.1394 ROUGEL: 28.7255 TOKENS_PER_SAMPLE: 301.3
| 18,561.4 | 2,320.175 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-fp8_dyn
| 70 |
fp16
|
vLLM v0.6.4
| null | null | null | null | null | null | null | null | null |
Ubuntu 22.04.4 LTS (Jammy Jellyfish)
|
available
|
4ca3b67de0e74489
|
open
|
Krai
|
Offline
| 8 | 192 |
HBM3
|
AMD Instinct MI300X 192GB HBM3
| 8 |
AMD
| null | 40 | 2 | null |
Intel(R) Xeon(R) Platinum 8460Y+
| null |
Intel
|
XGMI
|
PCIe-5
| 2,048 |
32x 64GB DDR5
|
Dell PowerEdge XE9680 (8x MI300X-192GB, vLLM v0.6.4)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5398 ROUGE2: 22.1484 ROUGEL: 28.725 TOKENS_PER_SAMPLE: 299.5
| 13,916.2 | 1,739.525 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-fp8_dyn
| 70 |
fp16
|
vLLM v0.6.4
| null | null | null | null | null | null | null | null | null |
Ubuntu 22.04.4 LTS (Jammy Jellyfish)
|
available
|
004c206126434a93
|
open
|
Krai
|
Server
| 8 | 192 |
HBM3
|
AMD Instinct MI300X 192GB HBM3
| 8 |
AMD
| null | 40 | 2 | null |
Intel(R) Xeon(R) Platinum 8460Y+
| null |
Intel
|
XGMI
|
PCIe-5
| 2,048 |
32x 64GB DDR5
|
Dell PowerEdge XE9680 (8x MI300X-192GB, vLLM v0.6.4)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 47.5723 ROUGE2: 25.2003 ROUGEL: 31.4338 TOKENS_PER_SAMPLE: 299.6
| 24,624.1 | 3,078.0125 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama3_1-70b-fp8_dyn
| 70 |
fp16
|
SGLang v0.4.3
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
a3a76e816a97450b
|
open
|
Krai
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, SGLang v0.4.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 47.589 ROUGE2: 25.2681 ROUGEL: 31.545 TOKENS_PER_SAMPLE: 292.1
| 27,687.9 | 3,460.9875 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama3_1-70b-fp8_pre
| 70 |
fp8
|
SGLang v0.4.3
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
bcfc5b7027c2466c
|
open
|
Krai
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, SGLang v0.4.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 47.6531 ROUGE2: 25.2981 ROUGEL: 31.5596 TOKENS_PER_SAMPLE: 291.3
| 26,074.4 | 3,259.3 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama3_1-70b-fp8_pre
| 70 |
fp8
|
SGLang v0.4.3
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
b3549c26f4cb40a4
|
open
|
Krai
|
Server
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, SGLang v0.4.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 47.5684 ROUGE2: 25.2712 ROUGEL: 31.533 TOKENS_PER_SAMPLE: 292.8
| 27,949.8 | 3,493.725 |
Tokens/s
|
LLM
|
llama2-70b-interactive-99.9
|
llama3_1-70b-interactive-fp8_pre
| 70 |
fp8
|
SGLang v0.4.3
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
d55226ae5c8946ba
|
open
|
Krai
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, SGLang v0.4.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 47.6193 ROUGE2: 25.2818 ROUGEL: 31.5351 TOKENS_PER_SAMPLE: 291.8
| 12,192.4 | 1,524.05 |
Tokens/s
|
LLM
|
llama2-70b-interactive-99.9
|
llama3_1-70b-interactive-fp8_pre
| 70 |
fp8
|
SGLang v0.4.3
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
9b59b70d9c5046a4
|
open
|
Krai
|
Server
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, SGLang v0.4.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGEL: 21.634266581485935 exact_match: 90.12628722700197 TOKENS_PER_SAMPLE: 656.3
| 326.537 | 40.817125 |
Tokens/s
|
LLM
|
llama3.1-405b
|
llama3_1-405b-fp8_dyn
| 405 |
fp16
|
SGLang v0.4.3
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
d28a776809e24d0e
|
open
|
Krai
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, SGLang v0.4.3)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 46.5248 ROUGE2: 24.2311 ROUGEL: 30.2319 TOKENS_PER_SAMPLE: 323.9
| 30,530.1 | 3,816.2625 |
Tokens/s
|
LLM
|
llama2-70b-interactive-99.9
|
llama3_1-70b-interactive-fp8_nim
| 70 |
fp16
|
NIM v1.5.0
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
5b0bfe6676894a2f
|
open
|
Krai
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, NIM v1.5.0)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 46.5348 ROUGE2: 24.2605 ROUGEL: 30.246 TOKENS_PER_SAMPLE: 324.7
| 15,960.1 | 1,995.0125 |
Tokens/s
|
LLM
|
llama2-70b-interactive-99.9
|
llama3_1-70b-interactive-fp8_nim
| 70 |
fp16
|
NIM v1.5.0
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
60fca8b370f64cbd
|
open
|
Krai
|
Server
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, NIM v1.5.0)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 46.5273 ROUGE2: 24.2307 ROUGEL: 30.2329 TOKENS_PER_SAMPLE: 323.8
| 30,042.8 | 3,755.35 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama3_1-70b-fp8_nim
| 70 |
fp16
|
NIM v1.5.0
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
307a52a7c3274e35
|
open
|
Krai
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, NIM v1.5.0)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 46.5201 ROUGE2: 24.2273 ROUGEL: 30.2319 TOKENS_PER_SAMPLE: 323.9
| 28,421 | 3,552.625 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama3_1-70b-fp8_nim
| 70 |
fp16
|
NIM v1.5.0
| null | null | null | null | null | null | null | null | null |
RHEL 9.4
|
available
|
d9404fee957443a0
|
open
|
Krai
|
Server
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
|
L1d: 3 MiB (64 instances), L1i: 2 MiB (64 instances), L2: 128 MiB (64 instances), L3: 120 MiB (2 instances)
| 32 | 2 | 2.8 |
Intel(R) Xeon(R) Platinum 8562Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x HPE Cray Supercomputing XD 64GB (1x64GB) Dual Rank x4 DDR5-5600 Registered Standard Memory
|
HPE Cray XD670 (8x H200-SXM-141GB, NIM v1.5.0)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.7869 ROUGE2: 22.383 ROUGEL: 29.1434 TOKENS_PER_SAMPLE: 272.9
| 97,403.4 | 12,175.425 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp4
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04
|
available
|
0ff96f26d3c94725
|
open
|
Lambda
|
Offline
| 8 | 180 |
HBM3e
|
NVIDIA B200-SXM-180GB
| 8 |
NVIDIA
| null | 56 | 2 | null |
Intel(R) Xeon(R) Platinum 8570
| null |
Intel
|
NVLink-5
|
PCIe-5
| 2,048 |
32x 64GB HMCG94AGBRA179N
|
NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 45.6315 ROUGE2: 23.4063 ROUGEL: 30.5125 TOKENS_PER_SAMPLE: 144.1 gsm8k_accuracy: 73.06 mbxp_accuracy: 60.38
| 121,886 | 15,235.75 |
Tokens/s
|
LLM
|
mixtral-8x7b
|
mixtral-8x7b
| 47 |
fp4
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04
|
available
|
2c2d879784df4044
|
open
|
Lambda
|
Offline
| 8 | 180 |
HBM3e
|
NVIDIA B200-SXM-180GB
| 8 |
NVIDIA
| null | 56 | 2 | null |
Intel(R) Xeon(R) Platinum 8570
| null |
Intel
|
NVLink-5
|
PCIe-5
| 2,048 |
32x 64GB HMCG94AGBRA179N
|
NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.7869 ROUGE2: 22.383 ROUGEL: 29.1434 TOKENS_PER_SAMPLE: 272.9
| 97,403.4 | 12,175.425 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp4
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04
|
available
|
ca3b0c472de844ac
|
open
|
Lambda
|
Offline
| 8 | 180 |
HBM3e
|
NVIDIA B200-SXM-180GB
| 8 |
NVIDIA
| null | 56 | 2 | null |
Intel(R) Xeon(R) Platinum 8570
| null |
Intel
|
NVLink-5
|
PCIe-5
| 2,048 |
32x 64GB HMCG94AGBRA179N
|
NVIDIA DGX B200 (8x B200-SXM-180GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGEL: 21.738071137617286 exact_match: 90.02873593646592 TOKENS_PER_SAMPLE: 660.4
| 564.696 | 70.587 |
Tokens/s
|
LLM
|
llama3.1-405b
|
llama3_1-405b
| 405 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04
|
available
|
90ee928cec444285
|
open
|
Supermicro
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-SXM-141GB
| 8 |
NVIDIA
| null | 48 | 2 | null |
Intel(R) Xeon(R) Platinum 8568Y+
| null |
Intel
|
NVLink-4
|
PCIe-5
| 2,048 |
32x 64GB DDR5 4800MT/s @ 4400MT/s
|
SYS-821GE-TNHR (8x H200-SXM-141GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1186 ROUGE2: 20.1722 ROUGEL: 30.0185 GEN_LEN: 4156655
| 2,791.38 | 2,791.38 |
Tokens/s
|
LLM
|
gptj-99.9
|
gptj-99.9
| 6 |
fp8
|
CUDA 12.8
| null |
12.8
| null | null | null | null | null | null | null |
Ubuntu 24.04
|
available
|
399e936cb1804999
|
open
|
Supermicro
|
Offline
| 1 | 96 |
HBM3
|
NVIDIA GH200 Grace Hopper Superchip 96GB
| 1 |
NVIDIA
| null | 72 | 1 | null |
NVIDIA Grace CPU
| null |
NVIDIA
|
InfiniBand
|
NVLink
| 480 |
LPDDR5x
|
ARS-111GL-NHR (1x GH200-96GB_aarch64)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1186 ROUGE2: 20.1722 ROUGEL: 30.0185 GEN_LEN: 4156655
| 2,791.38 | 2,791.38 |
Tokens/s
|
LLM
|
gptj-99
|
gptj-99
| 6 |
fp8
|
CUDA 12.8
| null |
12.8
| null | null | null | null | null | null | null |
Ubuntu 24.04
|
available
|
7474d9ec0881489a
|
open
|
Supermicro
|
Offline
| 1 | 96 |
HBM3
|
NVIDIA GH200 Grace Hopper Superchip 96GB
| 1 |
NVIDIA
| null | 72 | 1 | null |
NVIDIA Grace CPU
| null |
NVIDIA
|
InfiniBand
|
NVLink
| 480 |
LPDDR5x
|
ARS-111GL-NHR (1x GH200-96GB_aarch64)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5498 ROUGE2: 22.134 ROUGEL: 28.7842 TOKENS_PER_SAMPLE: 291.4
| 8,231.9 | 4,115.95 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Red Hat 9.4 + OpenShift 4.15
|
available
|
480c7d04b2be4452
|
open
|
Supermicro
|
Offline
| 2 | 144 |
HBM3e
|
NVIDIA GH200 Grace Hopper Superchip 144GB
| 2 |
NVIDIA
| null | 144 | 2 | null |
NVIDIA Grace CPU
| null |
NVIDIA
|
InfiniBand
|
NVLink
| 960 |
15x 16DP (32GB) LPDDR5x
|
ARS-221GL-NHIR (2x GH200-144GB_aarch64, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5491 ROUGE2: 22.1346 ROUGEL: 28.7871 TOKENS_PER_SAMPLE: 292.3
| 7,538.8 | 3,769.4 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Red Hat 9.4 + OpenShift 4.15
|
available
|
7bbdd41d6e5e46ff
|
open
|
Supermicro
|
Server
| 2 | 144 |
HBM3e
|
NVIDIA GH200 Grace Hopper Superchip 144GB
| 2 |
NVIDIA
| null | 144 | 2 | null |
NVIDIA Grace CPU
| null |
NVIDIA
|
InfiniBand
|
NVLink
| 960 |
15x 16DP (32GB) LPDDR5x
|
ARS-221GL-NHIR (2x GH200-144GB_aarch64, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5498 ROUGE2: 22.134 ROUGEL: 28.7842 TOKENS_PER_SAMPLE: 291.4
| 8,231.9 | 4,115.95 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Red Hat 9.4 + OpenShift 4.15
|
available
|
21d7651f55ce4046
|
open
|
Supermicro
|
Offline
| 2 | 144 |
HBM3e
|
NVIDIA GH200 Grace Hopper Superchip 144GB
| 2 |
NVIDIA
| null | 144 | 2 | null |
NVIDIA Grace CPU
| null |
NVIDIA
|
InfiniBand
|
NVLink
| 960 |
15x 16DP (32GB) LPDDR5x
|
ARS-221GL-NHIR (2x GH200-144GB_aarch64, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5491 ROUGE2: 22.1346 ROUGEL: 28.7871 TOKENS_PER_SAMPLE: 292.3
| 7,538.8 | 3,769.4 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Red Hat 9.4 + OpenShift 4.15
|
available
|
d024bd99975b49c4
|
open
|
Supermicro
|
Server
| 2 | 144 |
HBM3e
|
NVIDIA GH200 Grace Hopper Superchip 144GB
| 2 |
NVIDIA
| null | 144 | 2 | null |
NVIDIA Grace CPU
| null |
NVIDIA
|
InfiniBand
|
NVLink
| 960 |
15x 16DP (32GB) LPDDR5x
|
ARS-221GL-NHIR (2x GH200-144GB_aarch64, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.4322 ROUGE2: 22.0403 ROUGEL: 28.6357 TOKENS_PER_SAMPLE: 294.3
| 31,238 | 3,904.75 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
a1d43c28dae54e7c
|
closed
|
HPE
|
Offline
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with HPE GreenLake for File Storage (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.4297 ROUGE2: 22.0409 ROUGEL: 28.6367 TOKENS_PER_SAMPLE: 295.4
| 30,457.7 | 3,807.2125 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
c8db1c2d11454d6d
|
closed
|
HPE
|
Server
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with HPE GreenLake for File Storage (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1118 ROUGE2: 20.1071 ROUGEL: 29.9676 GEN_LEN: 4153194
| 20,796 | 2,599.5 |
Tokens/s
|
LLM
|
gptj-99.9
|
gptj-99.9
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
85c932c5501144f6
|
closed
|
HPE
|
Offline
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with HPE GreenLake for File Storage (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1118 ROUGE2: 20.1071 ROUGEL: 29.9676 GEN_LEN: 4153194
| 19,363.3 | 2,420.4125 |
Tokens/s
|
LLM
|
gptj-99.9
|
gptj-99.9
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
cdd931035a0d480b
|
closed
|
HPE
|
Server
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with HPE GreenLake for File Storage (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1118 ROUGE2: 20.1071 ROUGEL: 29.9676 GEN_LEN: 4153194
| 20,993.9 | 2,624.2375 |
Tokens/s
|
LLM
|
gptj-99
|
gptj-99
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
92ec13d59e3a46cd
|
closed
|
HPE
|
Offline
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with HPE GreenLake for File Storage (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1118 ROUGE2: 20.1071 ROUGEL: 29.9676 GEN_LEN: 4153194
| 19,363.2 | 2,420.4 |
Tokens/s
|
LLM
|
gptj-99
|
gptj-99
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
85ce8172fb3f43c8
|
closed
|
HPE
|
Server
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with HPE GreenLake for File Storage (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.4546 ROUGE2: 22.0578 ROUGEL: 28.6551 TOKENS_PER_SAMPLE: 294.3
| 23,389.3 | 2,923.6625 |
Tokens/s
|
LLM
|
llama2-70b-interactive-99
|
llama2-70b-interactive-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
28d865cdd1d24504
|
closed
|
HPE
|
Offline
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with HPE GreenLake for File Storage (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5336 ROUGE2: 22.0916 ROUGEL: 28.7043 TOKENS_PER_SAMPLE: 294.3
| 14,008.5 | 1,751.0625 |
Tokens/s
|
LLM
|
llama2-70b-interactive-99
|
llama2-70b-interactive-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
ebd5cb0a11f74c1f
|
closed
|
HPE
|
Server
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with HPE GreenLake for File Storage (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.4546 ROUGE2: 22.0578 ROUGEL: 28.6551 TOKENS_PER_SAMPLE: 294.3
| 23,385 | 2,923.125 |
Tokens/s
|
LLM
|
llama2-70b-interactive-99.9
|
llama2-70b-interactive-99.9
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
45ea64e8e274434b
|
closed
|
HPE
|
Offline
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with HPE GreenLake for File Storage (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5144 ROUGE2: 22.0732 ROUGEL: 28.686 TOKENS_PER_SAMPLE: 294.4
| 13,987.6 | 1,748.45 |
Tokens/s
|
LLM
|
llama2-70b-interactive-99.9
|
llama2-70b-interactive-99.9
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
8c6ddd0256ed4ea5
|
closed
|
HPE
|
Server
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with HPE GreenLake for File Storage (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.4347 ROUGE2: 22.0435 ROUGEL: 28.6357 TOKENS_PER_SAMPLE: 294.3
| 31,252.1 | 3,906.5125 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
2ba60812bba3469a
|
closed
|
HPE
|
Offline
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with HPE GreenLake for File Storage (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.4184 ROUGE2: 22.033 ROUGEL: 28.6265 TOKENS_PER_SAMPLE: 295.6
| 30,500.8 | 3,812.6 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
b6894f1874004b68
|
closed
|
HPE
|
Server
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with HPE GreenLake for File Storage (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5458 ROUGE2: 22.1332 ROUGEL: 28.7834 TOKENS_PER_SAMPLE: 291.5
| 30,297.5 | 3,787.1875 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
f9f85e257d9a424e
|
closed
|
HPE
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-NVL-141GB
| 8 |
NVIDIA
| null | 192 | 2 | null |
Intel(R) Xeon(R) 6740E
| null |
Intel
|
NVLink-4
|
PCIe-5
| 3,072 |
16x 128GB MTC40F2047S1RC64BB1 MWFF; 16x 64GB HMCG94AHBRA480N
|
HPE ProLiant DL380a Compute Gen12 (8x H200-NVL-141GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5621 ROUGE2: 22.1437 ROUGEL: 28.7928 TOKENS_PER_SAMPLE: 292.3
| 27,638.9 | 3,454.8625 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
017f3264aba44c22
|
closed
|
HPE
|
Server
| 8 | 141 |
HBM3e
|
NVIDIA H200-NVL-141GB
| 8 |
NVIDIA
| null | 192 | 2 | null |
Intel(R) Xeon(R) 6740E
| null |
Intel
|
NVLink-4
|
PCIe-5
| 3,072 |
16x 128GB MTC40F2047S1RC64BB1 MWFF; 16x 64GB HMCG94AHBRA480N
|
HPE ProLiant DL380a Compute Gen12 (8x H200-NVL-141GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1118 ROUGE2: 20.1071 ROUGEL: 29.9676 GEN_LEN: 4153194
| 17,882.9 | 2,235.3625 |
Tokens/s
|
LLM
|
gptj-99.9
|
gptj-99.9
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
6f3a95ad227d4768
|
closed
|
HPE
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-NVL-141GB
| 8 |
NVIDIA
| null | 192 | 2 | null |
Intel(R) Xeon(R) 6740E
| null |
Intel
|
NVLink-4
|
PCIe-5
| 3,072 |
16x 128GB MTC40F2047S1RC64BB1 MWFF; 16x 64GB HMCG94AHBRA480N
|
HPE ProLiant DL380a Compute Gen12 (8x H200-NVL-141GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1118 ROUGE2: 20.1071 ROUGEL: 29.9676 GEN_LEN: 4153194
| 18,065.1 | 2,258.1375 |
Tokens/s
|
LLM
|
gptj-99.9
|
gptj-99.9
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
3cbc904ba1f94641
|
closed
|
HPE
|
Server
| 8 | 141 |
HBM3e
|
NVIDIA H200-NVL-141GB
| 8 |
NVIDIA
| null | 192 | 2 | null |
Intel(R) Xeon(R) 6740E
| null |
Intel
|
NVLink-4
|
PCIe-5
| 3,072 |
16x 128GB MTC40F2047S1RC64BB1 MWFF; 16x 64GB HMCG94AHBRA480N
|
HPE ProLiant DL380a Compute Gen12 (8x H200-NVL-141GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1118 ROUGE2: 20.1071 ROUGEL: 29.9676 GEN_LEN: 4153194
| 17,882.9 | 2,235.3625 |
Tokens/s
|
LLM
|
gptj-99
|
gptj-99
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
2427ec1dd4e14f5b
|
closed
|
HPE
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-NVL-141GB
| 8 |
NVIDIA
| null | 192 | 2 | null |
Intel(R) Xeon(R) 6740E
| null |
Intel
|
NVLink-4
|
PCIe-5
| 3,072 |
16x 128GB MTC40F2047S1RC64BB1 MWFF; 16x 64GB HMCG94AHBRA480N
|
HPE ProLiant DL380a Compute Gen12 (8x H200-NVL-141GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1118 ROUGE2: 20.1071 ROUGEL: 29.9676 GEN_LEN: 4153194
| 18,065.1 | 2,258.1375 |
Tokens/s
|
LLM
|
gptj-99
|
gptj-99
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
ffab353cf3cb4b1e
|
closed
|
HPE
|
Server
| 8 | 141 |
HBM3e
|
NVIDIA H200-NVL-141GB
| 8 |
NVIDIA
| null | 192 | 2 | null |
Intel(R) Xeon(R) 6740E
| null |
Intel
|
NVLink-4
|
PCIe-5
| 3,072 |
16x 128GB MTC40F2047S1RC64BB1 MWFF; 16x 64GB HMCG94AHBRA480N
|
HPE ProLiant DL380a Compute Gen12 (8x H200-NVL-141GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 45.3505 ROUGE2: 23.2113 ROUGEL: 30.312 TOKENS_PER_SAMPLE: 144.9 gsm8k_accuracy: 73.02 mbxp_accuracy: 59.82
| 52,205.9 | 6,525.7375 |
Tokens/s
|
LLM
|
mixtral-8x7b
|
mixtral-8x7b
| 47 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
0a410c987b734b99
|
closed
|
HPE
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-NVL-141GB
| 8 |
NVIDIA
| null | 192 | 2 | null |
Intel(R) Xeon(R) 6740E
| null |
Intel
|
NVLink-4
|
PCIe-5
| 3,072 |
16x 128GB MTC40F2047S1RC64BB1 MWFF; 16x 64GB HMCG94AHBRA480N
|
HPE ProLiant DL380a Compute Gen12 (8x H200-NVL-141GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 45.4369 ROUGE2: 23.2825 ROUGEL: 30.4229 TOKENS_PER_SAMPLE: 145.9 gsm8k_accuracy: 73.48 mbxp_accuracy: 59.94
| 50,123.9 | 6,265.4875 |
Tokens/s
|
LLM
|
mixtral-8x7b
|
mixtral-8x7b
| 47 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
678eab9f2cd54d18
|
closed
|
HPE
|
Server
| 8 | 141 |
HBM3e
|
NVIDIA H200-NVL-141GB
| 8 |
NVIDIA
| null | 192 | 2 | null |
Intel(R) Xeon(R) 6740E
| null |
Intel
|
NVLink-4
|
PCIe-5
| 3,072 |
16x 128GB MTC40F2047S1RC64BB1 MWFF; 16x 64GB HMCG94AHBRA480N
|
HPE ProLiant DL380a Compute Gen12 (8x H200-NVL-141GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5458 ROUGE2: 22.1332 ROUGEL: 28.7834 TOKENS_PER_SAMPLE: 291.5
| 30,297.5 | 3,787.1875 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
5b9651aad28641ef
|
closed
|
HPE
|
Offline
| 8 | 141 |
HBM3e
|
NVIDIA H200-NVL-141GB
| 8 |
NVIDIA
| null | 192 | 2 | null |
Intel(R) Xeon(R) 6740E
| null |
Intel
|
NVLink-4
|
PCIe-5
| 3,072 |
16x 128GB MTC40F2047S1RC64BB1 MWFF; 16x 64GB HMCG94AHBRA480N
|
HPE ProLiant DL380a Compute Gen12 (8x H200-NVL-141GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5621 ROUGE2: 22.1437 ROUGEL: 28.7928 TOKENS_PER_SAMPLE: 292.3
| 27,638.9 | 3,454.8625 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
21d55faba2d74f75
|
closed
|
HPE
|
Server
| 8 | 141 |
HBM3e
|
NVIDIA H200-NVL-141GB
| 8 |
NVIDIA
| null | 192 | 2 | null |
Intel(R) Xeon(R) 6740E
| null |
Intel
|
NVLink-4
|
PCIe-5
| 3,072 |
16x 128GB MTC40F2047S1RC64BB1 MWFF; 16x 64GB HMCG94AHBRA480N
|
HPE ProLiant DL380a Compute Gen12 (8x H200-NVL-141GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.4331 ROUGE2: 22.0455 ROUGEL: 28.6365 TOKENS_PER_SAMPLE: 294.2
| 31,174.7 | 3,896.8375 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
e172993599054f66
|
closed
|
HPE
|
Offline
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with Cray ClusterStor (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.4303 ROUGE2: 22.039 ROUGEL: 28.6356 TOKENS_PER_SAMPLE: 295.3
| 30,498.9 | 3,812.3625 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
fa04f5c827cb4036
|
closed
|
HPE
|
Server
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with Cray ClusterStor (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1118 ROUGE2: 20.1071 ROUGEL: 29.9676 GEN_LEN: 4153194
| 20,716.5 | 2,589.5625 |
Tokens/s
|
LLM
|
gptj-99.9
|
gptj-99.9
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
d941ae1d491744aa
|
closed
|
HPE
|
Offline
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with Cray ClusterStor (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1118 ROUGE2: 20.1071 ROUGEL: 29.9676 GEN_LEN: 4153194
| 20,795.3 | 2,599.4125 |
Tokens/s
|
LLM
|
gptj-99.9
|
gptj-99.9
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
da0d8e299caf4c7c
|
closed
|
HPE
|
Server
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with Cray ClusterStor (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1118 ROUGE2: 20.1071 ROUGEL: 29.9676 GEN_LEN: 4153194
| 20,925.3 | 2,615.6625 |
Tokens/s
|
LLM
|
gptj-99
|
gptj-99
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
26e687a564dc46ff
|
closed
|
HPE
|
Offline
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with Cray ClusterStor (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1118 ROUGE2: 20.1071 ROUGEL: 29.9676 GEN_LEN: 4153194
| 20,991.3 | 2,623.9125 |
Tokens/s
|
LLM
|
gptj-99
|
gptj-99
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
aabda9670981454e
|
closed
|
HPE
|
Server
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with Cray ClusterStor (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5218 ROUGE2: 22.1082 ROUGEL: 28.731 TOKENS_PER_SAMPLE: 293.1
| 23,435.4 | 2,929.425 |
Tokens/s
|
LLM
|
llama2-70b-interactive-99
|
llama2-70b-interactive-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
c2ae77d781804382
|
closed
|
HPE
|
Offline
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with Cray ClusterStor (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5203 ROUGE2: 22.0752 ROUGEL: 28.6917 TOKENS_PER_SAMPLE: 294.2
| 14,004.9 | 1,750.6125 |
Tokens/s
|
LLM
|
llama2-70b-interactive-99
|
llama2-70b-interactive-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
8e4d1633ac534080
|
closed
|
HPE
|
Server
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with Cray ClusterStor (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5218 ROUGE2: 22.1082 ROUGEL: 28.731 TOKENS_PER_SAMPLE: 293.1
| 23,653.7 | 2,956.7125 |
Tokens/s
|
LLM
|
llama2-70b-interactive-99.9
|
llama2-70b-interactive-99.9
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
3f6029efe017449f
|
closed
|
HPE
|
Offline
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with Cray ClusterStor (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5263 ROUGE2: 22.0802 ROUGEL: 28.6949 TOKENS_PER_SAMPLE: 294.6
| 14,014.1 | 1,751.7625 |
Tokens/s
|
LLM
|
llama2-70b-interactive-99.9
|
llama2-70b-interactive-99.9
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
ece71ccb3aa74467
|
closed
|
HPE
|
Server
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with Cray ClusterStor (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.4346 ROUGE2: 22.0435 ROUGEL: 28.6357 TOKENS_PER_SAMPLE: 294.3
| 31,147.4 | 3,893.425 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
4a4ae0df176243c4
|
closed
|
HPE
|
Offline
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with Cray ClusterStor (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.4347 ROUGE2: 22.0423 ROUGEL: 28.6376 TOKENS_PER_SAMPLE: 295.4
| 30,480.6 | 3,810.075 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
8291ffcc09e94e08
|
closed
|
HPE
|
Server
| 8 | 80 |
HBM3
|
NVIDIA H100-SXM-80GB
| 8 |
NVIDIA
|
L1d: 5.3 MiB (112 instances), L1i: 3.5 MiB (112 instances), L2: 224 MiB (112 instances), L3: 210 MiB (2 instances)
| 56 | 2 | 3.8 |
Intel(R) Xeon(R) Platinum 8480+
| null |
Intel
|
NVLink-4
| null | 1,024 |
16x 64GB DDR5-4800 Registered Standard Memory
|
HPE Cray XD670 with Cray ClusterStor (8x H100-SXM-80GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5024 ROUGE2: 22.0807 ROUGEL: 28.7026 TOKENS_PER_SAMPLE: 293.0
| 3,655.89 | 456.98625 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp16
|
TensorRT 9.0, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
9.0
| null |
Ubuntu 24.04.1 LTS
|
available
|
f0f863a12ad84765
|
closed
|
HPE
|
Offline
| 8 | 48 |
HBM2e
|
NVIDIA L40S
| 8 |
NVIDIA
|
L1d: 6 MiB, L1i: 12 MiB, L2: 192 MiB, L3: 192 MiB
| 96 | 2 | 2.4 |
Intel(R) Xeon(R) 6740E
| null |
Intel
| null |
PCIe-5
| 1,024 |
16x 64GB 36ASF8G72PZ-3G2E1
|
HPE ProLiant Compute DL380a Gen12 (8x L40S-PCIe-48GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5208 ROUGE2: 22.0821 ROUGEL: 28.7224 TOKENS_PER_SAMPLE: 291.9
| 3,181.65 | 397.70625 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp16
|
TensorRT 9.0, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
9.0
| null |
Ubuntu 24.04.1 LTS
|
available
|
07b6067f7f984c17
|
closed
|
HPE
|
Server
| 8 | 48 |
HBM2e
|
NVIDIA L40S
| 8 |
NVIDIA
|
L1d: 6 MiB, L1i: 12 MiB, L2: 192 MiB, L3: 192 MiB
| 96 | 2 | 2.4 |
Intel(R) Xeon(R) 6740E
| null |
Intel
| null |
PCIe-5
| 1,024 |
16x 64GB 36ASF8G72PZ-3G2E1
|
HPE ProLiant Compute DL380a Gen12 (8x L40S-PCIe-48GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5024 ROUGE2: 22.0807 ROUGEL: 28.7026 TOKENS_PER_SAMPLE: 293.0
| 3,655.89 | 456.98625 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp16
|
TensorRT 9.0, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
9.0
| null |
Ubuntu 24.04.1 LTS
|
available
|
746b553ca1464d33
|
closed
|
HPE
|
Offline
| 8 | 48 |
HBM2e
|
NVIDIA L40S
| 8 |
NVIDIA
|
L1d: 6 MiB, L1i: 12 MiB, L2: 192 MiB, L3: 192 MiB
| 96 | 2 | 2.4 |
Intel(R) Xeon(R) 6740E
| null |
Intel
| null |
PCIe-5
| 1,024 |
16x 64GB 36ASF8G72PZ-3G2E1
|
HPE ProLiant Compute DL380a Gen12 (8x L40S-PCIe-48GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5208 ROUGE2: 22.0821 ROUGEL: 28.7224 TOKENS_PER_SAMPLE: 291.9
| 3,181.65 | 397.70625 |
Tokens/s
|
LLM
|
llama2-70b-99.9
|
llama2-70b-99.9
| 70 |
fp16
|
TensorRT 9.0, CUDA 12.8
| null |
12.8
| null | null | null | null | null |
9.0
| null |
Ubuntu 24.04.1 LTS
|
available
|
fb068c46be304c9b
|
closed
|
HPE
|
Server
| 8 | 48 |
HBM2e
|
NVIDIA L40S
| 8 |
NVIDIA
|
L1d: 6 MiB, L1i: 12 MiB, L2: 192 MiB, L3: 192 MiB
| 96 | 2 | 2.4 |
Intel(R) Xeon(R) 6740E
| null |
Intel
| null |
PCIe-5
| 1,024 |
16x 64GB 36ASF8G72PZ-3G2E1
|
HPE ProLiant Compute DL380a Gen12 (8x L40S-PCIe-48GB, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5446 ROUGE2: 22.1336 ROUGEL: 28.7828 TOKENS_PER_SAMPLE: 291.5
| 4,609.05 | 4,609.05 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.7
| null |
12.7
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
2fa475fb252a43d4
|
closed
|
HPE
|
Offline
| 1 | 144 |
HBM3e
|
NVIDIA GH200 Grace Hopper Superchip 144GB
| 1 |
NVIDIA
| null | 72 | 1 | null |
NVIDIA Grace CPU
| null |
NVIDIA
|
InfiniBand
|
NVLink
| 480 |
15x 16DP (32GB) LPDDR5x
|
HPE ProLiant Compute DL384 Gen12 (1x GH200-144GB_aarch64, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 44.5531 ROUGE2: 22.1392 ROUGEL: 28.7903 TOKENS_PER_SAMPLE: 291.5
| 4,327.68 | 4,327.68 |
Tokens/s
|
LLM
|
llama2-70b-99
|
llama2-70b-99
| 70 |
fp8
|
TensorRT 10.8, CUDA 12.7
| null |
12.7
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
55f04b610c4049b9
|
closed
|
HPE
|
Server
| 1 | 144 |
HBM3e
|
NVIDIA GH200 Grace Hopper Superchip 144GB
| 1 |
NVIDIA
| null | 72 | 1 | null |
NVIDIA Grace CPU
| null |
NVIDIA
|
InfiniBand
|
NVLink
| 480 |
15x 16DP (32GB) LPDDR5x
|
HPE ProLiant Compute DL384 Gen12 (1x GH200-144GB_aarch64, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1186 ROUGE2: 20.1722 ROUGEL: 30.0185 GEN_LEN: 4156655
| 2,845.43 | 2,845.43 |
Tokens/s
|
LLM
|
gptj-99.9
|
gptj-99.9
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.7
| null |
12.7
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
92b6dcd9a183450f
|
closed
|
HPE
|
Offline
| 1 | 144 |
HBM3e
|
NVIDIA GH200 Grace Hopper Superchip 144GB
| 1 |
NVIDIA
| null | 72 | 1 | null |
NVIDIA Grace CPU
| null |
NVIDIA
|
InfiniBand
|
NVLink
| 480 |
15x 16DP (32GB) LPDDR5x
|
HPE ProLiant Compute DL384 Gen12 (1x GH200-144GB_aarch64, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1186 ROUGE2: 20.1722 ROUGEL: 30.0185 GEN_LEN: 4156655
| 2,526.03 | 2,526.03 |
Tokens/s
|
LLM
|
gptj-99.9
|
gptj-99.9
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.7
| null |
12.7
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
a426c61bc533447b
|
closed
|
HPE
|
Server
| 1 | 144 |
HBM3e
|
NVIDIA GH200 Grace Hopper Superchip 144GB
| 1 |
NVIDIA
| null | 72 | 1 | null |
NVIDIA Grace CPU
| null |
NVIDIA
|
InfiniBand
|
NVLink
| 480 |
15x 16DP (32GB) LPDDR5x
|
HPE ProLiant Compute DL384 Gen12 (1x GH200-144GB_aarch64, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1186 ROUGE2: 20.1722 ROUGEL: 30.0185 GEN_LEN: 4156655
| 2,845.43 | 2,845.43 |
Tokens/s
|
LLM
|
gptj-99
|
gptj-99
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.7
| null |
12.7
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
a6c1ca0a77a84560
|
closed
|
HPE
|
Offline
| 1 | 144 |
HBM3e
|
NVIDIA GH200 Grace Hopper Superchip 144GB
| 1 |
NVIDIA
| null | 72 | 1 | null |
NVIDIA Grace CPU
| null |
NVIDIA
|
InfiniBand
|
NVLink
| 480 |
15x 16DP (32GB) LPDDR5x
|
HPE ProLiant Compute DL384 Gen12 (1x GH200-144GB_aarch64, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 43.1186 ROUGE2: 20.1722 ROUGEL: 30.0185 GEN_LEN: 4156655
| 2,526.03 | 2,526.03 |
Tokens/s
|
LLM
|
gptj-99
|
gptj-99
| 6 |
fp8
|
TensorRT 10.8, CUDA 12.7
| null |
12.7
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
a27f5ade487a4e8c
|
closed
|
HPE
|
Server
| 1 | 144 |
HBM3e
|
NVIDIA GH200 Grace Hopper Superchip 144GB
| 1 |
NVIDIA
| null | 72 | 1 | null |
NVIDIA Grace CPU
| null |
NVIDIA
|
InfiniBand
|
NVLink
| 480 |
15x 16DP (32GB) LPDDR5x
|
HPE ProLiant Compute DL384 Gen12 (1x GH200-144GB_aarch64, TensorRT)
| 1 |
datacenter
|
mlperf-inference
|
5.0
|
2025-04-03T22:56:53
|
ROUGE1: 45.2956 ROUGE2: 23.2043 ROUGEL: 30.3791 TOKENS_PER_SAMPLE: 144.9 gsm8k_accuracy: 73.0 mbxp_accuracy: 59.96
| 8,324.43 | 8,324.43 |
Tokens/s
|
LLM
|
mixtral-8x7b
|
mixtral-8x7b
| 47 |
fp8
|
TensorRT 10.8, CUDA 12.7
| null |
12.7
| null | null | null | null | null |
10.8
| null |
Ubuntu 24.04.1 LTS
|
available
|
de05f7de95394a5d
|
closed
|
HPE
|
Offline
| 1 | 144 |
HBM3e
|
NVIDIA GH200 Grace Hopper Superchip 144GB
| 1 |
NVIDIA
| null | 72 | 1 | null |
NVIDIA Grace CPU
| null |
NVIDIA
|
InfiniBand
|
NVLink
| 480 |
15x 16DP (32GB) LPDDR5x
|
HPE ProLiant Compute DL384 Gen12 (1x GH200-144GB_aarch64, TensorRT)
| 1 |
datacenter
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.