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	NVIDIA L4	NVIDIA A40	Google TPU v4	SambaNova SN40L	Biren Technology BR100
Architecture	Ada Lovelace	Ampere	TPU v4	Reconfigurable Dataflow Unit (RDU)	Biren GPU
Form Factor	PCIe	PCIe	Mezzanine	—	—
VRAM	24 GB	48 GB	—	640 GB	64 GB
Memory Bandwidth	300 GB/s	696 GB/s	—	—	—
TFLOPs (FP32)	30.3	37.4	—	—	—
TFLOPs (FP16)	242	150	—	—	1,000
TFLOPs	121	150	275	—	—
TFLOPs (FP8)	485	—	—	—	—
TDP	72 W	300 W	300 W	700 W	550 W
Launch Date	Mar 2023	Oct 2020	May 2021	Jan 2024	Aug 2022

Efficiency Metrics

Metric	L4	A40	TPU v4	SN40L	BR100
TFLOPs per Watt (FP32-eq)	0.84	0.25	0.46	—	—
Memory Bandwidth per GB	12.5 GB/s	14.5 GB/s	—	—	—

Performance Equivalence

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

To match 1x NVIDIA L4

NVIDIA A40

Compute (FP32-eq)

0.81x

A40 is 1.24x faster

FP32 Compute

0.81x

A40 is 1.23x faster

VRAM

0.50x

A40 has 2.00x more

Memory Bandwidth

0.43x

A40 has 2.32x more

Google TPU v4

Compute (FP32-eq)

0.44x

TPU v4 is 2.27x faster

SambaNova SN40L

VRAM

0.04x

SN40L has 26.67x more

Biren Technology BR100

VRAM

0.38x

BR100 has 2.67x more

To match 1x NVIDIA A40

NVIDIA L4

Compute (FP32-eq)

1.24x

Need 1.24x L4

FP32 Compute

1.23x

Need 1.23x L4

VRAM

2.00x

Need 2.00x L4

Memory Bandwidth

2.32x

Need 2.32x L4

Google TPU v4

Compute (FP32-eq)

0.55x

TPU v4 is 1.83x faster

SambaNova SN40L

VRAM

0.07x

SN40L has 13.33x more

Biren Technology BR100

VRAM

0.75x

BR100 has 1.33x more

To match 1x Google TPU v4

NVIDIA L4

Compute (FP32-eq)

2.27x

Need 2.27x L4

NVIDIA A40

Compute (FP32-eq)

1.83x

Need 1.83x A40

To match 1x SambaNova SN40L

NVIDIA L4

VRAM

26.67x

Need 26.67x L4

NVIDIA A40

VRAM

13.33x

Need 13.33x A40

Biren Technology BR100

VRAM

10.00x

Need 10.00x BR100

To match 1x Biren Technology BR100

NVIDIA L4

VRAM

2.67x

Need 2.67x L4

NVIDIA A40

VRAM

1.33x

Need 1.33x A40

SambaNova SN40L

VRAM

0.10x

SN40L has 10.00x more

Pricing

Price Type	L4	A40	TPU v4	SN40L	BR100
CAPEX (Street Price)	$4,000	—	—	—	—
OPEX (per hour)	$0.80/hr	—	$3.00/hr	—	—
Price per TFLOPs (FP32-eq)	$66	—	—	—	—