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	Google TPU v4	AMD MI325X	Qualcomm Cloud AI 100
Architecture	Ampere	TPU v4	CDNA 3.5	Qualcomm AI
Form Factor	SXM	Mezzanine	OAM	PCIe
VRAM	80 GB	—	256 GB	16 GB
Memory Bandwidth	2,039 GB/s	—	6,000 GB/s	134 GB/s
TFLOPs (FP32)	19.5	—	175	50
TFLOPs (FP16)	312	—	1,400	400
TFLOPs	312	275	1,400	50
TFLOPs (FP8)	—	—	2,800	—
TDP	400 W	300 W	800 W	75 W
Launch Date	May 2020	May 2021	Oct 2024	Sep 2020

Efficiency Metrics

Metric	A100 SXM	TPU v4	MI325X	Cloud AI 100
TFLOPs per Watt (FP32-eq)	0.39	0.46	0.88	0.67
Memory Bandwidth per GB	25.5 GB/s	—	23.4 GB/s	8.4 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

Google TPU v4

Compute (FP32-eq)

1.13x

Need 1.13x TPU v4

AMD MI325X

Compute (FP32-eq)

0.22x

MI325X is 4.49x faster

FP32 Compute

0.11x

MI325X is 8.97x faster

VRAM

0.31x

MI325X has 3.20x more

Memory Bandwidth

0.34x

MI325X has 2.94x more

Qualcomm Cloud AI 100

Compute (FP32-eq)

3.12x

Need 3.12x Cloud AI 100

FP32 Compute

0.39x

Cloud AI 100 is 2.56x faster

VRAM

5.00x

Need 5.00x Cloud AI 100

Memory Bandwidth

15.22x

Need 15.22x Cloud AI 100

To match 1x Google TPU v4

NVIDIA A100 SXM

Compute (FP32-eq)

0.88x

A100 SXM is 1.13x faster

AMD MI325X

Compute (FP32-eq)

0.20x

MI325X is 5.09x faster

Qualcomm Cloud AI 100

Compute (FP32-eq)

2.75x

Need 2.75x Cloud AI 100

To match 1x AMD MI325X

NVIDIA A100 SXM

Compute (FP32-eq)

4.49x

Need 4.49x A100 SXM

FP32 Compute

8.97x

Need 8.97x A100 SXM

VRAM

3.20x

Need 3.20x A100 SXM

Memory Bandwidth

2.94x

Need 2.94x A100 SXM

Google TPU v4

Compute (FP32-eq)

5.09x

Need 5.09x TPU v4

Qualcomm Cloud AI 100

Compute (FP32-eq)

14.00x

Need 14.00x Cloud AI 100

FP32 Compute

3.50x

Need 3.50x Cloud AI 100

VRAM

16.00x

Need 16.00x Cloud AI 100

Memory Bandwidth

44.78x

Need 44.78x Cloud AI 100

To match 1x Qualcomm Cloud AI 100

NVIDIA A100 SXM

Compute (FP32-eq)

0.32x

A100 SXM is 3.12x faster

FP32 Compute

2.56x

Need 2.56x A100 SXM

VRAM

0.20x

A100 SXM has 5.00x more

Memory Bandwidth

0.07x

A100 SXM has 15.22x more

Google TPU v4

Compute (FP32-eq)

0.36x

TPU v4 is 2.75x faster

AMD MI325X

Compute (FP32-eq)

0.07x

MI325X is 14.00x faster

FP32 Compute

0.29x

MI325X is 3.50x faster

VRAM

0.06x

MI325X has 16.00x more

Memory Bandwidth

0.02x

MI325X has 44.78x more

Pricing

Price Type	A100 SXM	TPU v4	MI325X	Cloud AI 100
CAPEX (Street Price)	$15,000	—	$45,000	—
OPEX (per hour)	$4.05/hr	$3.00/hr	—	—
Price per TFLOPs (FP32-eq)	$96	—	$64	—