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	AMD MI100	Google TPU v4	AMD MI300X	NVIDIA H200 PCIe
Architecture	CDNA	TPU v4	CDNA 3	Hopper
Form Factor	PCIe	Mezzanine	OAM	PCIe
VRAM	32 GB	—	192 GB	141 GB
Memory Bandwidth	1,228 GB/s	—	5,300 GB/s	4,800 GB/s
TFLOPs (FP32)	23.1	—	163	51
TFLOPs (FP16)	184.6	—	1,307	1,513
TFLOPs	23.1	275	1,307	1,513
TFLOPs (FP8)	—	—	2,614	—
TDP	300 W	300 W	750 W	350 W
Launch Date	Nov 2020	May 2021	Dec 2023	Mar 2024

Efficiency Metrics

Metric	MI100	TPU v4	MI300X	H200 PCIe
TFLOPs per Watt (FP32-eq)	0.08	0.46	0.87	2.16
Memory Bandwidth per GB	38.4 GB/s	—	27.6 GB/s	34.0 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 v4

Compute (FP32-eq)

0.17x

TPU v4 is 5.95x faster

AMD MI300X

Compute (FP32-eq)

0.04x

MI300X is 28.29x faster

FP32 Compute

0.14x

MI300X is 7.06x faster

VRAM

0.17x

MI300X has 6.00x more

Memory Bandwidth

0.23x

MI300X has 4.32x more

NVIDIA H200 PCIe

Compute (FP32-eq)

0.03x

H200 PCIe is 32.75x faster

FP32 Compute

0.45x

H200 PCIe is 2.21x faster

VRAM

0.23x

H200 PCIe has 4.41x more

Memory Bandwidth

0.26x

H200 PCIe has 3.91x more

To match 1x Google TPU v4

AMD MI100

Compute (FP32-eq)

5.95x

Need 5.95x MI100

AMD MI300X

Compute (FP32-eq)

0.21x

MI300X is 4.75x faster

NVIDIA H200 PCIe

Compute (FP32-eq)

0.18x

H200 PCIe is 5.50x faster

To match 1x AMD MI300X

AMD MI100

Compute (FP32-eq)

28.29x

Need 28.29x MI100

FP32 Compute

7.06x

Need 7.06x MI100

VRAM

6.00x

Need 6.00x MI100

Memory Bandwidth

4.32x

Need 4.32x MI100

Google TPU v4

Compute (FP32-eq)

4.75x

Need 4.75x TPU v4

NVIDIA H200 PCIe

Compute (FP32-eq)

0.86x

H200 PCIe is 1.16x faster

FP32 Compute

3.20x

Need 3.20x H200 PCIe

VRAM

1.36x

Need 1.36x H200 PCIe

Memory Bandwidth

1.10x

Need 1.10x H200 PCIe

To match 1x NVIDIA H200 PCIe

AMD MI100

Compute (FP32-eq)

32.75x

Need 32.75x MI100

FP32 Compute

2.21x

Need 2.21x MI100

VRAM

4.41x

Need 4.41x MI100

Memory Bandwidth

3.91x

Need 3.91x MI100

Google TPU v4

Compute (FP32-eq)

5.50x

Need 5.50x TPU v4

AMD MI300X

Compute (FP32-eq)

1.16x

Need 1.16x MI300X

FP32 Compute

0.31x

MI300X is 3.20x faster

VRAM

0.73x

MI300X has 1.36x more

Memory Bandwidth

0.91x

MI300X has 1.10x more

Pricing

Price Type	MI100	TPU v4	MI300X	H200 PCIe
CAPEX (Street Price)	—	—	$35,000	—
OPEX (per hour)	—	$3.00/hr	$10.40/hr	—
Price per TFLOPs (FP32-eq)	—	—	$54	—