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 L40S	AMD MI250X	Google TPU v4	NVIDIA GeForce RTX 4080 Super
Architecture	Ada Lovelace	CDNA 2	TPU v4	Ada Lovelace
Form Factor	PCIe	OAM	Mezzanine	PCIe
VRAM	48 GB	128 GB	—	16 GB
Memory Bandwidth	864 GB/s	3,277 GB/s	—	736 GB/s
TFLOPs (FP32)	91.6	47.9	—	52.2
TFLOPs (FP16)	733	383	—	104.4
TFLOPs	733	383	275	104.4
TFLOPs (FP8)	1,466	—	—	—
TDP	350 W	560 W	300 W	320 W
Launch Date	Oct 2023	Nov 2021	May 2021	Jan 2024

Efficiency Metrics

Metric	L40S	MI250X	TPU v4	GeForce RTX 4080 Super
TFLOPs per Watt (FP32-eq)	1.05	0.34	0.46	0.16
Memory Bandwidth per GB	18.0 GB/s	25.6 GB/s	—	46.0 GB/s

Performance Equivalence

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

To match 1x NVIDIA L40S

AMD MI250X

Compute (FP32-eq)

1.91x

Need 1.91x MI250X

FP32 Compute

1.91x

Need 1.91x MI250X

VRAM

0.38x

MI250X has 2.67x more

Memory Bandwidth

0.26x

MI250X has 3.79x more

Google TPU v4

Compute (FP32-eq)

2.67x

Need 2.67x TPU v4

NVIDIA GeForce RTX 4080 Super

Compute (FP32-eq)

7.02x

Need 7.02x GeForce RTX 4080 Super

FP32 Compute

1.75x

Need 1.75x GeForce RTX 4080 Super

VRAM

3.00x

Need 3.00x GeForce RTX 4080 Super

Memory Bandwidth

1.17x

Need 1.17x GeForce RTX 4080 Super

To match 1x AMD MI250X

NVIDIA L40S

Compute (FP32-eq)

0.52x

L40S is 1.91x faster

FP32 Compute

0.52x

L40S is 1.91x faster

VRAM

2.67x

Need 2.67x L40S

Memory Bandwidth

3.79x

Need 3.79x L40S

Google TPU v4

Compute (FP32-eq)

1.39x

Need 1.39x TPU v4

NVIDIA GeForce RTX 4080 Super

Compute (FP32-eq)

3.67x

Need 3.67x GeForce RTX 4080 Super

FP32 Compute

0.92x

GeForce RTX 4080 Super is 1.09x faster

VRAM

8.00x

Need 8.00x GeForce RTX 4080 Super

Memory Bandwidth

4.45x

Need 4.45x GeForce RTX 4080 Super

To match 1x Google TPU v4

NVIDIA L40S

Compute (FP32-eq)

0.38x

L40S is 2.67x faster

AMD MI250X

Compute (FP32-eq)

0.72x

MI250X is 1.39x faster

NVIDIA GeForce RTX 4080 Super

Compute (FP32-eq)

2.63x

Need 2.63x GeForce RTX 4080 Super

To match 1x NVIDIA GeForce RTX 4080 Super

NVIDIA L40S

Compute (FP32-eq)

0.14x

L40S is 7.02x faster

FP32 Compute

0.57x

L40S is 1.75x faster

VRAM

0.33x

L40S has 3.00x more

Memory Bandwidth

0.85x

L40S has 1.17x more

AMD MI250X

Compute (FP32-eq)

0.27x

MI250X is 3.67x faster

FP32 Compute

1.09x

Need 1.09x MI250X

VRAM

0.13x

MI250X has 8.00x more

Memory Bandwidth

0.22x

MI250X has 4.45x more

Google TPU v4

Compute (FP32-eq)

0.38x

TPU v4 is 2.63x faster

Pricing

Price Type	L40S	MI250X	TPU v4	GeForce RTX 4080 Super
CAPEX (Street Price)	$10,000	$12,000	—	—
OPEX (per hour)	$1.50/hr	$2.00/hr	$3.00/hr	—
Price per TFLOPs (FP32-eq)	$27	$63	—	—