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	AMD MI100	Google TPU v5p	Cambricon MLU370	Etched Sohu	Baidu Kunlun II
Architecture	CDNA	TPU v5	MLU Core	Transformer ASIC	Kunlun Core
Form Factor	PCIe	Mezzanine	—	PCIe	—
VRAM	32 GB	—	48 GB	144 GB	32 GB
Memory Bandwidth	1,228 GB/s	—	—	4,800 GB/s	—
TFLOPs (FP32)	23.1	—	—	10,000	—
TFLOPs (FP16)	184.6	—	—	—	256
TFLOPs	23.1	459	256	10,000	—
TFLOPs (FP8)	—	—	—	—	—
TDP	300 W	400 W	300 W	700 W	200 W
Launch Date	Nov 2020	Aug 2023	Nov 2022	Jun 2024	Aug 2021

Efficiency Metrics

Metric	MI100	TPU v5p	MLU370	Sohu	Kunlun II
TFLOPs per Watt (FP32-eq)	0.08	0.57	0.43	14.29	—
Memory Bandwidth per GB	38.4 GB/s	—	—	33.3 GB/s	—

Performance Equivalence

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

To match 1x AMD MI100

Google TPU v5p

Compute (FP32-eq)

0.10x

TPU v5p is 9.94x faster

Cambricon MLU370

Compute (FP32-eq)

0.18x

MLU370 is 5.54x faster

VRAM

0.67x

MLU370 has 1.50x more

Etched Sohu

Compute (FP32-eq)

0.00x

Sohu is 432.90x faster

FP32 Compute

0.00x

Sohu is 432.90x faster

VRAM

0.22x

Sohu has 4.50x more

Memory Bandwidth

0.26x

Sohu has 3.91x more

Baidu Kunlun II

VRAM

1.00x

Kunlun II has 1.00x more

To match 1x Google TPU v5p

AMD MI100

Compute (FP32-eq)

9.94x

Need 9.94x MI100

Cambricon MLU370

Compute (FP32-eq)

1.79x

Need 1.79x MLU370

Etched Sohu

Compute (FP32-eq)

0.02x

Sohu is 43.57x faster

To match 1x Cambricon MLU370

AMD MI100

Compute (FP32-eq)

5.54x

Need 5.54x MI100

VRAM

1.50x

Need 1.50x MI100

Google TPU v5p

Compute (FP32-eq)

0.56x

TPU v5p is 1.79x faster

Etched Sohu

Compute (FP32-eq)

0.01x

Sohu is 78.13x faster

VRAM

0.33x

Sohu has 3.00x more

Baidu Kunlun II

VRAM

1.50x

Need 1.50x Kunlun II

To match 1x Etched Sohu

AMD MI100

Compute (FP32-eq)

432.90x

Need 432.90x MI100

FP32 Compute

432.90x

Need 432.90x MI100

VRAM

4.50x

Need 4.50x MI100

Memory Bandwidth

3.91x

Need 3.91x MI100

Google TPU v5p

Compute (FP32-eq)

43.57x

Need 43.57x TPU v5p

Cambricon MLU370

Compute (FP32-eq)

78.13x

Need 78.13x MLU370

VRAM

3.00x

Need 3.00x MLU370

Baidu Kunlun II

VRAM

4.50x

Need 4.50x Kunlun II

To match 1x Baidu Kunlun II

AMD MI100

VRAM

1.00x

MI100 has 1.00x more

Cambricon MLU370

VRAM

0.67x

MLU370 has 1.50x more

Etched Sohu

VRAM

0.22x

Sohu has 4.50x more

Pricing

Price Type	MI100	TPU v5p	MLU370	Sohu	Kunlun II
CAPEX (Street Price)	—	—	—	—	—
OPEX (per hour)	—	$5.00/hr	—	—	—
Price per TFLOPs (FP32-eq)	—	—	—	—	—