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	Google TPU v5p	NVIDIA B200	AWS Trainium	NVIDIA GeForce RTX 4090
Architecture	TPU v5	Blackwell	Inferentia/Trainium	Ada Lovelace
Form Factor	Mezzanine	SXM	—	PCIe
VRAM	—	192 GB	32 GB	24 GB
Memory Bandwidth	—	8,000 GB/s	—	1,008 GB/s
TFLOPs (FP32)	—	90	—	82.6
TFLOPs (FP16)	—	2,250	—	165.2
TFLOPs	459	2,250	190	165.2
TFLOPs (FP8)	—	4,500	—	—
TDP	400 W	1000 W	200 W	450 W
Launch Date	Aug 2023	Mar 2024	Nov 2021	Oct 2022

Efficiency Metrics

Metric	TPU v5p	B200	Trainium	GeForce RTX 4090
TFLOPs per Watt (FP32-eq)	0.57	1.13	0.47	0.18
Memory Bandwidth per GB	—	41.7 GB/s	—	42.0 GB/s

Performance Equivalence

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

To match 1x Google TPU v5p

NVIDIA B200

Compute (FP32-eq)

0.20x

B200 is 4.90x faster

AWS Trainium

Compute (FP32-eq)

2.42x

Need 2.42x Trainium

NVIDIA GeForce RTX 4090

Compute (FP32-eq)

2.78x

Need 2.78x GeForce RTX 4090

To match 1x NVIDIA B200

Google TPU v5p

Compute (FP32-eq)

4.90x

Need 4.90x TPU v5p

AWS Trainium

Compute (FP32-eq)

11.84x

Need 11.84x Trainium

VRAM

6.00x

Need 6.00x Trainium

NVIDIA GeForce RTX 4090

Compute (FP32-eq)

13.62x

Need 13.62x GeForce RTX 4090

FP32 Compute

1.09x

Need 1.09x GeForce RTX 4090

VRAM

8.00x

Need 8.00x GeForce RTX 4090

Memory Bandwidth

7.94x

Need 7.94x GeForce RTX 4090

To match 1x AWS Trainium

Google TPU v5p

Compute (FP32-eq)

0.41x

TPU v5p is 2.42x faster

NVIDIA B200

Compute (FP32-eq)

0.08x

B200 is 11.84x faster

VRAM

0.17x

B200 has 6.00x more

NVIDIA GeForce RTX 4090

Compute (FP32-eq)

1.15x

Need 1.15x GeForce RTX 4090

VRAM

1.33x

Need 1.33x GeForce RTX 4090

To match 1x NVIDIA GeForce RTX 4090

Google TPU v5p

Compute (FP32-eq)

0.36x

TPU v5p is 2.78x faster

NVIDIA B200

Compute (FP32-eq)

0.07x

B200 is 13.62x faster

FP32 Compute

0.92x

B200 is 1.09x faster

VRAM

0.13x

B200 has 8.00x more

Memory Bandwidth

0.13x

B200 has 7.94x more

AWS Trainium

Compute (FP32-eq)

0.87x

Trainium is 1.15x faster

VRAM

0.75x

Trainium has 1.33x more

Pricing

Price Type	TPU v5p	B200	Trainium	GeForce RTX 4090
CAPEX (Street Price)	—	$70,000	—	—
OPEX (per hour)	$5.00/hr	—	—	—
Price per TFLOPs (FP32-eq)	—	$62	—	—