Compare XPUs

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

Select XPUs to Compare

Clear all (5)Jump to results

Filter by Vendor

Showing 72 XPUs • 5 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	Google TPU v4	NVIDIA B200	SambaNova SN40L	Tenstorrent Grayskull	AWS Trainium2
Architecture	TPU v4	Blackwell	Reconfigurable Dataflow Unit (RDU)	Tensix Core	Trainium Gen2
Form Factor	Mezzanine	SXM	—	—	—
VRAM	—	192 GB	640 GB	8 GB	—
Memory Bandwidth	—	8,000 GB/s	—	—	—
TFLOPs (FP32)	—	90	—	—	—
TFLOPs (FP16)	—	2,250	—	—	—
TFLOPs	275	2,250	—	200	680
TFLOPs (FP8)	—	4,500	—	—	—
TDP	300 W	1000 W	700 W	65 W	300 W
Launch Date	May 2021	Mar 2024	Jan 2024	Mar 2021	Nov 2023

Efficiency Metrics

Metric	TPU v4	B200	SN40L	Grayskull	Trainium2
TFLOPs per Watt (FP32-eq)	0.46	1.13	—	1.54	1.13
Memory Bandwidth per GB	—	41.7 GB/s	—	—	—

Performance Equivalence

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

To match 1x Google TPU v4

NVIDIA B200

Compute (FP32-eq)

0.12x

B200 is 8.18x faster

Tenstorrent Grayskull

Compute (FP32-eq)

1.38x

Need 1.38x Grayskull

AWS Trainium2

Compute (FP32-eq)

0.40x

Trainium2 is 2.47x faster

To match 1x NVIDIA B200

Google TPU v4

Compute (FP32-eq)

8.18x

Need 8.18x TPU v4

SambaNova SN40L

VRAM

0.30x

SN40L has 3.33x more

Tenstorrent Grayskull

Compute (FP32-eq)

11.25x

Need 11.25x Grayskull

VRAM

24.00x

Need 24.00x Grayskull

AWS Trainium2

Compute (FP32-eq)

3.31x

Need 3.31x Trainium2

To match 1x SambaNova SN40L

NVIDIA B200

VRAM

3.33x

Need 3.33x B200

Tenstorrent Grayskull

VRAM

80.00x

Need 80.00x Grayskull

To match 1x Tenstorrent Grayskull

Google TPU v4

Compute (FP32-eq)

0.73x

TPU v4 is 1.38x faster

NVIDIA B200

Compute (FP32-eq)

0.09x

B200 is 11.25x faster

VRAM

0.04x

B200 has 24.00x more

SambaNova SN40L

VRAM

0.01x

SN40L has 80.00x more

AWS Trainium2

Compute (FP32-eq)

0.29x

Trainium2 is 3.40x faster

To match 1x AWS Trainium2

Google TPU v4

Compute (FP32-eq)

2.47x

Need 2.47x TPU v4

NVIDIA B200

Compute (FP32-eq)

0.30x

B200 is 3.31x faster

Tenstorrent Grayskull

Compute (FP32-eq)

3.40x

Need 3.40x Grayskull

Pricing

Price Type	TPU v4	B200	SN40L	Grayskull	Trainium2
CAPEX (Street Price)	—	$70,000	—	—	—
OPEX (per hour)	$3.00/hr	—	—	—	$1.34/hr
Price per TFLOPs (FP32-eq)	—	$62	—	—	—