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 v4	AMD MI325X	AMD MI350X	NVIDIA RTX 5000 Ada Generation
Architecture	TPU v4	CDNA 3.5	CDNA 4	Ada Lovelace
Form Factor	Mezzanine	OAM	OAM	PCIe
VRAM	—	256 GB	288 GB	32 GB
Memory Bandwidth	—	6,000 GB/s	—	576 GB/s
TFLOPs (FP32)	—	175	—	65.3
TFLOPs (FP16)	—	1,400	—	130.6
TFLOPs	275	1,400	2,100	130.6
TFLOPs (FP8)	—	2,800	—	—
TDP	300 W	800 W	850 W	250 W
Launch Date	May 2021	Oct 2024	Jun 2025	Aug 2023

Efficiency Metrics

Metric	TPU v4	MI325X	MI350X	RTX 5000 Ada Generation
TFLOPs per Watt (FP32-eq)	0.46	0.88	1.24	0.26
Memory Bandwidth per GB	—	23.4 GB/s	—	18.0 GB/s

Performance Equivalence

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

To match 1x Google TPU v4

AMD MI325X

Compute (FP32-eq)

0.20x

MI325X is 5.09x faster

AMD MI350X

Compute (FP32-eq)

0.13x

MI350X is 7.64x faster

NVIDIA RTX 5000 Ada Generation

Compute (FP32-eq)

2.11x

Need 2.11x RTX 5000 Ada Generation

To match 1x AMD MI325X

Google TPU v4

Compute (FP32-eq)

5.09x

Need 5.09x TPU v4

AMD MI350X

Compute (FP32-eq)

0.67x

MI350X is 1.50x faster

VRAM

0.89x

MI350X has 1.13x more

NVIDIA RTX 5000 Ada Generation

Compute (FP32-eq)

10.72x

Need 10.72x RTX 5000 Ada Generation

FP32 Compute

2.68x

Need 2.68x RTX 5000 Ada Generation

VRAM

8.00x

Need 8.00x RTX 5000 Ada Generation

Memory Bandwidth

10.42x

Need 10.42x RTX 5000 Ada Generation

To match 1x AMD MI350X

Google TPU v4

Compute (FP32-eq)

7.64x

Need 7.64x TPU v4

AMD MI325X

Compute (FP32-eq)

1.50x

Need 1.50x MI325X

VRAM

1.13x

Need 1.13x MI325X

NVIDIA RTX 5000 Ada Generation

Compute (FP32-eq)

16.08x

Need 16.08x RTX 5000 Ada Generation

VRAM

9.00x

Need 9.00x RTX 5000 Ada Generation

To match 1x NVIDIA RTX 5000 Ada Generation

Google TPU v4

Compute (FP32-eq)

0.47x

TPU v4 is 2.11x faster

AMD MI325X

Compute (FP32-eq)

0.09x

MI325X is 10.72x faster

FP32 Compute

0.37x

MI325X is 2.68x faster

VRAM

0.13x

MI325X has 8.00x more

Memory Bandwidth

0.10x

MI325X has 10.42x more

AMD MI350X

Compute (FP32-eq)

0.06x

MI350X is 16.08x faster

VRAM

0.11x

MI350X has 9.00x more

Pricing

Price Type	TPU v4	MI325X	MI350X	RTX 5000 Ada Generation
CAPEX (Street Price)	—	$45,000	—	—
OPEX (per hour)	$3.00/hr	—	—	—
Price per TFLOPs (FP32-eq)	—	$64	—	—