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 A10	AMD MI250X	Cerebras WSE-3	Meta MTIA v1
Architecture	Ampere	CDNA 2	Wafer Scale Engine	Meta Training & Inference Accelerator
Form Factor	PCIe	OAM	—	—
VRAM	24 GB	128 GB	44 GB	128 GB
Memory Bandwidth	600 GB/s	3,277 GB/s	—	—
TFLOPs (FP32)	31.2	47.9	—	—
TFLOPs (FP16)	125	383	—	—
TFLOPs	125	383	—	—
TFLOPs (FP8)	—	—	—	—
TDP	150 W	560 W	23000 W	400 W
Launch Date	Apr 2021	Nov 2021	Mar 2024	May 2023

Efficiency Metrics

Metric	A10	MI250X	WSE-3	MTIA v1
TFLOPs per Watt (FP32-eq)	0.42	0.34	—	—
Memory Bandwidth per GB	25.0 GB/s	25.6 GB/s	—	—

Performance Equivalence

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

To match 1x NVIDIA A10

AMD MI250X

Compute (FP32-eq)

0.33x

MI250X is 3.06x faster

FP32 Compute

0.65x

MI250X is 1.54x faster

VRAM

0.19x

MI250X has 5.33x more

Memory Bandwidth

0.18x

MI250X has 5.46x more

Cerebras WSE-3

VRAM

0.55x

WSE-3 has 1.83x more

Meta MTIA v1

VRAM

0.19x

MTIA v1 has 5.33x more

To match 1x AMD MI250X

NVIDIA A10

Compute (FP32-eq)

3.06x

Need 3.06x A10

FP32 Compute

1.54x

Need 1.54x A10

VRAM

5.33x

Need 5.33x A10

Memory Bandwidth

5.46x

Need 5.46x A10

Cerebras WSE-3

VRAM

2.91x

Need 2.91x WSE-3

Meta MTIA v1

VRAM

1.00x

MTIA v1 has 1.00x more

To match 1x Cerebras WSE-3

NVIDIA A10

VRAM

1.83x

Need 1.83x A10

AMD MI250X

VRAM

0.34x

MI250X has 2.91x more

Meta MTIA v1

VRAM

0.34x

MTIA v1 has 2.91x more

To match 1x Meta MTIA v1

NVIDIA A10

VRAM

5.33x

Need 5.33x A10

AMD MI250X

VRAM

1.00x

MI250X has 1.00x more

Cerebras WSE-3

VRAM

2.91x

Need 2.91x WSE-3

Pricing

Price Type	A10	MI250X	WSE-3	MTIA v1
CAPEX (Street Price)	—	$12,000	—	—
OPEX (per hour)	—	$2.00/hr	—	—
Price per TFLOPs (FP32-eq)	—	$63	—	—