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 128 XPUs • 5 selected

Alibaba

Hanguang 800

AMD

MI100

23.1 TFLOPs

AMD

MI210

181 TFLOPs

AMD

MI250X

383 TFLOPs

AMD

MI300X

1,307 TFLOPs

AMD

MI325X

1,400 TFLOPs

AMD

MI350X

2,100 TFLOPs

AMD

MI355X

5,300 TFLOPs

AMD

Radeon Pro V520

23.04 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

FuriosaAI

RNGD (Renegade)

256 TFLOPs

FuriosaAI

Warboy

Google

TPU v4

275 TFLOPs

Google

TPU v5e

197 TFLOPs

Google

TPU v5p

459 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	Intel Habana Gaudi 2	Groq LPU Inference Engine	NVIDIA GeForce RTX 5060 Ti	NVIDIA GeForce GTX 1060	NVIDIA GeForce RTX 3090
Architecture	Gaudi Gen2	TSP (Tensor Streaming Processor)	Blackwell	Pascal	Ampere
Form Factor	OAM	—	PCIe	PCIe	PCIe
VRAM	96 GB	230 GB	16 GB	6 GB	24 GB
Memory Bandwidth	2,450 GB/s	—	544 GB/s	192 GB/s	936 GB/s
TFLOPs (FP32)	—	—	25	4.375	35.6
TFLOPs (FP16)	—	—	—	—	—
TFLOPs	432	—	88	4.375	71
TFLOPs (FP8)	—	—	—	—	—
TDP	600 W	300 W	220 W	120 W	350 W
Launch Date	May 2022	Feb 2024	Mar 2025	Jul 2016	Sep 2020

Efficiency Metrics

Metric	Gaudi 2	LPU Inference Engine	GeForce RTX 5060 Ti	GeForce GTX 1060	GeForce RTX 3090
TFLOPs per Watt (FP32-eq)	0.36	—	0.20	0.04	0.10
Memory Bandwidth per GB	25.5 GB/s	—	34.0 GB/s	32.0 GB/s	39.0 GB/s

Performance Equivalence

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

To match 1x Intel Habana Gaudi 2

Groq LPU Inference Engine

VRAM

0.42x

LPU Inference Engine has 2.40x more

NVIDIA GeForce RTX 5060 Ti

Compute (FP32-eq)

4.91x

Need 4.91x GeForce RTX 5060 Ti

VRAM

6.00x

Need 6.00x GeForce RTX 5060 Ti

Memory Bandwidth

4.50x

Need 4.50x GeForce RTX 5060 Ti

NVIDIA GeForce GTX 1060

Compute (FP32-eq)

49.37x

Need 49.37x GeForce GTX 1060

VRAM

16.00x

Need 16.00x GeForce GTX 1060

Memory Bandwidth

12.76x

Need 12.76x GeForce GTX 1060

NVIDIA GeForce RTX 3090

Compute (FP32-eq)

6.08x

Need 6.08x GeForce RTX 3090

VRAM

4.00x

Need 4.00x GeForce RTX 3090

Memory Bandwidth

2.62x

Need 2.62x GeForce RTX 3090

To match 1x Groq LPU Inference Engine

Intel Habana Gaudi 2

VRAM

2.40x

Need 2.40x Gaudi 2

NVIDIA GeForce RTX 5060 Ti

VRAM

14.38x

Need 14.38x GeForce RTX 5060 Ti

NVIDIA GeForce GTX 1060

VRAM

38.33x

Need 38.33x GeForce GTX 1060

NVIDIA GeForce RTX 3090

VRAM

9.58x

Need 9.58x GeForce RTX 3090

To match 1x NVIDIA GeForce RTX 5060 Ti

Intel Habana Gaudi 2

Compute (FP32-eq)

0.20x

Gaudi 2 is 4.91x faster

VRAM

0.17x

Gaudi 2 has 6.00x more

Memory Bandwidth

0.22x

Gaudi 2 has 4.50x more

Groq LPU Inference Engine

VRAM

0.07x

LPU Inference Engine has 14.38x more

NVIDIA GeForce GTX 1060

Compute (FP32-eq)

10.06x

Need 10.06x GeForce GTX 1060

FP32 Compute

5.71x

Need 5.71x GeForce GTX 1060

VRAM

2.67x

Need 2.67x GeForce GTX 1060

Memory Bandwidth

2.83x

Need 2.83x GeForce GTX 1060

NVIDIA GeForce RTX 3090

Compute (FP32-eq)

1.24x

Need 1.24x GeForce RTX 3090

FP32 Compute

0.70x

GeForce RTX 3090 is 1.42x faster

VRAM

0.67x

GeForce RTX 3090 has 1.50x more

Memory Bandwidth

0.58x

GeForce RTX 3090 has 1.72x more

To match 1x NVIDIA GeForce GTX 1060

Intel Habana Gaudi 2

Compute (FP32-eq)

0.02x

Gaudi 2 is 49.37x faster

VRAM

0.06x

Gaudi 2 has 16.00x more

Memory Bandwidth

0.08x

Gaudi 2 has 12.76x more

Groq LPU Inference Engine

VRAM

0.03x

LPU Inference Engine has 38.33x more

NVIDIA GeForce RTX 5060 Ti

Compute (FP32-eq)

0.10x

GeForce RTX 5060 Ti is 10.06x faster

FP32 Compute

0.17x

GeForce RTX 5060 Ti is 5.71x faster

VRAM

0.38x

GeForce RTX 5060 Ti has 2.67x more

Memory Bandwidth

0.35x

GeForce RTX 5060 Ti has 2.83x more

NVIDIA GeForce RTX 3090

Compute (FP32-eq)

0.12x

GeForce RTX 3090 is 8.11x faster

FP32 Compute

0.12x

GeForce RTX 3090 is 8.14x faster

VRAM

0.25x

GeForce RTX 3090 has 4.00x more

Memory Bandwidth

0.21x

GeForce RTX 3090 has 4.88x more

To match 1x NVIDIA GeForce RTX 3090

Intel Habana Gaudi 2

Compute (FP32-eq)

0.16x

Gaudi 2 is 6.08x faster

VRAM

0.25x

Gaudi 2 has 4.00x more

Memory Bandwidth

0.38x

Gaudi 2 has 2.62x more

Groq LPU Inference Engine

VRAM

0.10x

LPU Inference Engine has 9.58x more

NVIDIA GeForce RTX 5060 Ti

Compute (FP32-eq)

0.81x

GeForce RTX 5060 Ti is 1.24x faster

FP32 Compute

1.42x

Need 1.42x GeForce RTX 5060 Ti

VRAM

1.50x

Need 1.50x GeForce RTX 5060 Ti

Memory Bandwidth

1.72x

Need 1.72x GeForce RTX 5060 Ti

NVIDIA GeForce GTX 1060

Compute (FP32-eq)

8.11x

Need 8.11x GeForce GTX 1060

FP32 Compute

8.14x

Need 8.14x GeForce GTX 1060

VRAM

4.00x

Need 4.00x GeForce GTX 1060

Memory Bandwidth

4.88x

Need 4.88x GeForce GTX 1060

Pricing

Price Type	Gaudi 2	LPU Inference Engine	GeForce RTX 5060 Ti	GeForce GTX 1060	GeForce RTX 3090
CAPEX (Street Price)	—	—	—	—	—
OPEX (per hour)	—	—	$0.09/hr	$0.04/hr	$0.11/hr
Price per TFLOPs (FP32-eq)	—	—	—	—	—