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 A40	Google TPU v5e	NVIDIA B200	Intel Habana Gaudi 2
Architecture	Ampere	TPU v5	Blackwell	Gaudi Gen2
Form Factor	PCIe	Mezzanine	SXM	OAM
VRAM	48 GB	—	192 GB	96 GB
Memory Bandwidth	696 GB/s	—	8,000 GB/s	2,450 GB/s
TFLOPs (FP32)	37.4	—	90	—
TFLOPs (FP16)	150	—	2,250	—
TFLOPs	150	197	2,250	432
TFLOPs (FP8)	—	—	4,500	—
TDP	300 W	250 W	1000 W	600 W
Launch Date	Oct 2020	Aug 2023	Mar 2024	May 2022

Efficiency Metrics

Metric	A40	TPU v5e	B200	Gaudi 2
TFLOPs per Watt (FP32-eq)	0.25	0.39	1.13	0.36
Memory Bandwidth per GB	14.5 GB/s	—	41.7 GB/s	25.5 GB/s

Performance Equivalence

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

To match 1x NVIDIA A40

Google TPU v5e

Compute (FP32-eq)

0.76x

TPU v5e is 1.31x faster

NVIDIA B200

Compute (FP32-eq)

0.07x

B200 is 15.00x faster

FP32 Compute

0.42x

B200 is 2.41x faster

VRAM

0.25x

B200 has 4.00x more

Memory Bandwidth

0.09x

B200 has 11.49x more

Intel Habana Gaudi 2

Compute (FP32-eq)

0.35x

Gaudi 2 is 2.88x faster

VRAM

0.50x

Gaudi 2 has 2.00x more

Memory Bandwidth

0.28x

Gaudi 2 has 3.52x more

To match 1x Google TPU v5e

NVIDIA A40

Compute (FP32-eq)

1.31x

Need 1.31x A40

NVIDIA B200

Compute (FP32-eq)

0.09x

B200 is 11.42x faster

Intel Habana Gaudi 2

Compute (FP32-eq)

0.46x

Gaudi 2 is 2.19x faster

To match 1x NVIDIA B200

NVIDIA A40

Compute (FP32-eq)

15.00x

Need 15.00x A40

FP32 Compute

2.41x

Need 2.41x A40

VRAM

4.00x

Need 4.00x A40

Memory Bandwidth

11.49x

Need 11.49x A40

Google TPU v5e

Compute (FP32-eq)

11.42x

Need 11.42x TPU v5e

Intel Habana Gaudi 2

Compute (FP32-eq)

5.21x

Need 5.21x Gaudi 2

VRAM

2.00x

Need 2.00x Gaudi 2

Memory Bandwidth

3.27x

Need 3.27x Gaudi 2

To match 1x Intel Habana Gaudi 2

NVIDIA A40

Compute (FP32-eq)

2.88x

Need 2.88x A40

VRAM

2.00x

Need 2.00x A40

Memory Bandwidth

3.52x

Need 3.52x A40

Google TPU v5e

Compute (FP32-eq)

2.19x

Need 2.19x TPU v5e

NVIDIA B200

Compute (FP32-eq)

0.19x

B200 is 5.21x faster

VRAM

0.50x

B200 has 2.00x more

Memory Bandwidth

0.31x

B200 has 3.27x more

Pricing

Price Type	A40	TPU v5e	B200	Gaudi 2
CAPEX (Street Price)	—	—	$70,000	—
OPEX (per hour)	—	$1.80/hr	—	—
Price per TFLOPs (FP32-eq)	—	—	$62	—