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	NVIDIA L4	NVIDIA H200 SXM	AMD MI350X	Etched Sohu
Architecture	Ada Lovelace	Hopper	CDNA 4	Transformer ASIC
Form Factor	PCIe	SXM	OAM	PCIe
VRAM	24 GB	141 GB	288 GB	144 GB
Memory Bandwidth	300 GB/s	4,800 GB/s	—	4,800 GB/s
TFLOPs (FP32)	30.3	67	—	10,000
TFLOPs (FP16)	242	1,979	—	—
TFLOPs	121	1,979	2,100	10,000
TFLOPs (FP8)	485	3,958	—	—
TDP	72 W	700 W	850 W	700 W
Launch Date	Mar 2023	Nov 2023	Jun 2025	Jun 2024

Efficiency Metrics

Metric	L4	H200 SXM	MI350X	Sohu
TFLOPs per Watt (FP32-eq)	0.84	1.41	1.24	14.29
Memory Bandwidth per GB	12.5 GB/s	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 NVIDIA L4

NVIDIA H200 SXM

Compute (FP32-eq)

0.06x

H200 SXM is 16.36x faster

FP32 Compute

0.45x

H200 SXM is 2.21x faster

VRAM

0.17x

H200 SXM has 5.88x more

Memory Bandwidth

0.06x

H200 SXM has 16.00x more

AMD MI350X

Compute (FP32-eq)

0.06x

MI350X is 17.36x faster

VRAM

0.08x

MI350X has 12.00x more

Etched Sohu

Compute (FP32-eq)

0.01x

Sohu is 165.29x faster

FP32 Compute

0.00x

Sohu is 330.03x faster

VRAM

0.17x

Sohu has 6.00x more

Memory Bandwidth

0.06x

Sohu has 16.00x more

To match 1x NVIDIA H200 SXM

NVIDIA L4

Compute (FP32-eq)

16.36x

Need 16.36x L4

FP32 Compute

2.21x

Need 2.21x L4

VRAM

5.88x

Need 5.88x L4

Memory Bandwidth

16.00x

Need 16.00x L4

AMD MI350X

Compute (FP32-eq)

0.94x

MI350X is 1.06x faster

VRAM

0.49x

MI350X has 2.04x more

Etched Sohu

Compute (FP32-eq)

0.10x

Sohu is 10.11x faster

FP32 Compute

0.01x

Sohu is 149.25x faster

VRAM

0.98x

Sohu has 1.02x more

Memory Bandwidth

1.00x

Sohu has 1.00x more

To match 1x AMD MI350X

NVIDIA L4

Compute (FP32-eq)

17.36x

Need 17.36x L4

VRAM

12.00x

Need 12.00x L4

NVIDIA H200 SXM

Compute (FP32-eq)

1.06x

Need 1.06x H200 SXM

VRAM

2.04x

Need 2.04x H200 SXM

Etched Sohu

Compute (FP32-eq)

0.10x

Sohu is 9.52x faster

VRAM

2.00x

Need 2.00x Sohu

To match 1x Etched Sohu

NVIDIA L4

Compute (FP32-eq)

165.29x

Need 165.29x L4

FP32 Compute

330.03x

Need 330.03x L4

VRAM

6.00x

Need 6.00x L4

Memory Bandwidth

16.00x

Need 16.00x L4

NVIDIA H200 SXM

Compute (FP32-eq)

10.11x

Need 10.11x H200 SXM

FP32 Compute

149.25x

Need 149.25x H200 SXM

VRAM

1.02x

Need 1.02x H200 SXM

Memory Bandwidth

1.00x

H200 SXM has 1.00x more

AMD MI350X

Compute (FP32-eq)

9.52x

Need 9.52x MI350X

VRAM

0.50x

MI350X has 2.00x more

Pricing

Price Type	L4	H200 SXM	MI350X	Sohu
CAPEX (Street Price)	$4,000	$40,000	—	—
OPEX (per hour)	$0.80/hr	—	—	—
Price per TFLOPs (FP32-eq)	$66	$40	—	—