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 A100 SXM	NVIDIA L40S	NVIDIA H200 SXM	Alibaba Hanguang 800
Architecture	Ampere	Ada Lovelace	Hopper	XuanTie
Form Factor	SXM	PCIe	SXM	—
VRAM	80 GB	48 GB	141 GB	—
Memory Bandwidth	2,039 GB/s	864 GB/s	4,800 GB/s	—
TFLOPs (FP32)	19.5	91.6	67	—
TFLOPs (FP16)	312	733	1,979	—
TFLOPs	312	733	1,979	—
TFLOPs (FP8)	—	1,466	3,958	—
TDP	400 W	350 W	700 W	150 W
Launch Date	May 2020	Oct 2023	Nov 2023	Sep 2019

Efficiency Metrics

Metric	A100 SXM	L40S	H200 SXM	Hanguang 800
TFLOPs per Watt (FP32-eq)	0.39	1.05	1.41	—
Memory Bandwidth per GB	25.5 GB/s	18.0 GB/s	34.0 GB/s	—

Performance Equivalence

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

To match 1x NVIDIA A100 SXM

NVIDIA L40S

Compute (FP32-eq)

0.43x

L40S is 2.35x faster

FP32 Compute

0.21x

L40S is 4.70x faster

VRAM

1.67x

Need 1.67x L40S

Memory Bandwidth

2.36x

Need 2.36x L40S

NVIDIA H200 SXM

Compute (FP32-eq)

0.16x

H200 SXM is 6.34x faster

FP32 Compute

0.29x

H200 SXM is 3.44x faster

VRAM

0.57x

H200 SXM has 1.76x more

Memory Bandwidth

0.42x

H200 SXM has 2.35x more

To match 1x NVIDIA L40S

NVIDIA A100 SXM

Compute (FP32-eq)

2.35x

Need 2.35x A100 SXM

FP32 Compute

4.70x

Need 4.70x A100 SXM

VRAM

0.60x

A100 SXM has 1.67x more

Memory Bandwidth

0.42x

A100 SXM has 2.36x more

NVIDIA H200 SXM

Compute (FP32-eq)

0.37x

H200 SXM is 2.70x faster

FP32 Compute

1.37x

Need 1.37x H200 SXM

VRAM

0.34x

H200 SXM has 2.94x more

Memory Bandwidth

0.18x

H200 SXM has 5.56x more

To match 1x NVIDIA H200 SXM

NVIDIA A100 SXM

Compute (FP32-eq)

6.34x

Need 6.34x A100 SXM

FP32 Compute

3.44x

Need 3.44x A100 SXM

VRAM

1.76x

Need 1.76x A100 SXM

Memory Bandwidth

2.35x

Need 2.35x A100 SXM

NVIDIA L40S

Compute (FP32-eq)

2.70x

Need 2.70x L40S

FP32 Compute

0.73x

L40S is 1.37x faster

VRAM

2.94x

Need 2.94x L40S

Memory Bandwidth

5.56x

Need 5.56x L40S

Pricing

Price Type	A100 SXM	L40S	H200 SXM	Hanguang 800
CAPEX (Street Price)	$15,000	$10,000	$40,000	—
OPEX (per hour)	$4.05/hr	$1.50/hr	—	—
Price per TFLOPs (FP32-eq)	$96	$27	$40	—