Compare XPUs

Select up to 5 XPUs to compare side-by-side

Select XPUs to Compare

Clear all (4)Jump to results

Filter by Vendor

Showing 72 XPUs • 4 selected

Alibaba

Hanguang 800

AMD

MI100

23.1 TFLOPs

AMD

MI210

181 TFLOPs

AMD

MI250X

383 TFLOPs

AMD

MI300A

980.6 TFLOPs

AMD

MI300X

1,307 TFLOPs

AMD

MI325X

1,400 TFLOPs

AMD

MI350X

2,100 TFLOPs

AMD

Radeon PRO W7900

122 TFLOPs

AMD

Radeon RX 7900 XT

104 TFLOPs

AMD

Radeon RX 7900 XTX

122 TFLOPs

AWS

Inferentia2

190 TFLOPs

AWS

Trainium

190 TFLOPs

AWS

Trainium2

680 TFLOPs

Baidu

Kunlun II

Biren Technology

BR100

Cambricon

MLU370

256 TFLOPs

Cerebras

WSE-3

Enflame Technology

CloudBlazer T20

Etched

Sohu

10,000 TFLOPs

FuriosaAI

Warboy

Google

TPU v4

275 TFLOPs

Google

TPU v5e

197 TFLOPs

Google

TPU v5p

459 TFLOPs

Google

TPU v6e (Trillium)

918 TFLOPs

Graphcore

Bow IPU

Graphcore

IPU-M2000

Groq

LPU Inference Engine

Huawei

Ascend 910B

Iluvatar CoreX

BI-V150

300 TFLOPs

Intel

Data Center GPU Max 1100

177 TFLOPs

Intel

Data Center GPU Max 1550

419 TFLOPs

Intel Habana

Gaudi 2

432 TFLOPs

Intel Habana

Gaudi 3

1,835 TFLOPs

Multi-Metric Comparison

Relative performance across 5 key metrics (normalized to 100 = best in comparison)

Compute Performance (BF16)

Memory Capacity

Power Consumption

Power Efficiency

Specifications

Specification	SambaNova SN40L	NVIDIA H200 PCIe	Etched Sohu	Huawei Ascend 910B
Architecture	Reconfigurable Dataflow Unit (RDU)	Hopper	Transformer ASIC	Da Vinci
Form Factor	—	PCIe	PCIe	—
VRAM	640 GB	141 GB	144 GB	64 GB
Memory Bandwidth	—	4,800 GB/s	4,800 GB/s	—
TFLOPs (FP32)	—	51	10,000	—
TFLOPs (FP16)	—	1,513	—	640
TFLOPs	—	1,513	10,000	—
TFLOPs (FP8)	—	—	—	—
TDP	700 W	350 W	700 W	400 W
Launch Date	Jan 2024	Mar 2024	Jun 2024	Aug 2023

Efficiency Metrics

Metric	SN40L	H200 PCIe	Sohu	Ascend 910B
TFLOPs per Watt (FP32-eq)	—	2.16	14.29	—
Memory Bandwidth per GB	—	34.0 GB/s	33.3 GB/s	—

Performance Equivalence

How many units of each GPU are needed to match the performance of the others?

To match 1x SambaNova SN40L

NVIDIA H200 PCIe

VRAM

4.54x

Need 4.54x H200 PCIe

Etched Sohu

VRAM

4.44x

Need 4.44x Sohu

Huawei Ascend 910B

VRAM

10.00x

Need 10.00x Ascend 910B

To match 1x NVIDIA H200 PCIe

SambaNova SN40L

VRAM

0.22x

SN40L has 4.54x more

Etched Sohu

Compute (FP32-eq)

0.08x

Sohu is 13.22x faster

FP32 Compute

0.01x

Sohu is 196.08x faster

VRAM

0.98x

Sohu has 1.02x more

Memory Bandwidth

1.00x

Sohu has 1.00x more

Huawei Ascend 910B

VRAM

2.20x

Need 2.20x Ascend 910B

To match 1x Etched Sohu

SambaNova SN40L

VRAM

0.23x

SN40L has 4.44x more

NVIDIA H200 PCIe

Compute (FP32-eq)

13.22x

Need 13.22x H200 PCIe

FP32 Compute

196.08x

Need 196.08x H200 PCIe

VRAM

1.02x

Need 1.02x H200 PCIe

Memory Bandwidth

1.00x

H200 PCIe has 1.00x more

Huawei Ascend 910B

VRAM

2.25x

Need 2.25x Ascend 910B

To match 1x Huawei Ascend 910B

SambaNova SN40L

VRAM

0.10x

SN40L has 10.00x more

NVIDIA H200 PCIe

VRAM

0.45x

H200 PCIe has 2.20x more

Etched Sohu

VRAM

0.44x

Sohu has 2.25x more

Pricing

Price Type	SN40L	H200 PCIe	Sohu	Ascend 910B
CAPEX (Street Price)	—	—	—	—
OPEX (per hour)	—	—	—	—
Price per TFLOPs (FP32-eq)	—	—	—	—