Compute.Exchange

GPUS

Inventory

Referral Program

ABOUT

BLOG

Compute.Exchange

HARDWARE CATALOG

GPU

SPECIFICATIONS

Detailed technical specifications and performance metrics for the world's most powerful AI accelerators. Compare architectures and find the right compute for your workload.

DATABASE UPDATED

JAN 2026

NVIDIA H100 GPU

Access NVIDIA H100 GPUs across global providers. Compare pricing, locations, configurations, and deployment options — without vendor lock-in

Availability

In Stock

Architecture

NVIDIA Hopper

MEMORY

80 GB HBM3

MEMORY BANDWIDTH

up to 3.35 TB/s (SXM)

PRECISION SUPPORT

FP8 / FP16 / BF16 / TF32

INTERCONNECT

NVLink (up to 900 GB/s)

FORM FACTORS

SXM5 / PCIe Gen 5

PRIMARY USE CASES

LLM Training, Inference at Scale

TDP

700W (SXM) / 350W (PCIe)

* Performance metrics based on sparsity enabled

VIEW FULL SPECIFICATIONS

NVIDIA H100 GPU

Availability

In Stock

NVIDIA H100 GPU

Access NVIDIA H100 GPUs across global providers. Compare pricing, locations, configurations, and deployment options — without vendor lock-in

Architecture

NVIDIA Hopper

MEMORY

80 GB HBM3

MEMORY BANDWIDTH

up to 3.35 TB/s (SXM)

PRECISION SUPPORT

FP8 / FP16 / BF16 / TF32

INTERCONNECT

NVLink (up to 900 GB/s)

FORM FACTORS

SXM5 / PCIe Gen 5

PRIMARY USE CASES

LLM Training, Inference at Scale

TDP

700W (SXM) / 350W (PCIe)

* Performance metrics based on sparsity enabled

VIEW FULL SPECIFICATIONS

NVIDIA B200 GPU

Access reserved NVIDIA B200 capacity across global providers. Compare availability, regions, configurations, and contract terms — without vendor lock-in.

Availability

In Stock

Architecture

NVIDIA Blackwell

MEMORY

192 GB HBM3e

MEMORY BANDWIDTH

Up to ~8 TB/s

PRECISION SUPPORT

FP8 / FP16 / BF16 / TF32

INTERCONNECT

~1.8 TB/s NVLink

FORM FACTORS

SXM

PRIMARY USE CASES

LLM Training, Inference at Scale

TDP

700W (SXM) / 350W (PCIe)

* Performance metrics based on sparsity enabled

VIEW FULL SPECIFICATIONS

NVIDIA B200 GPU

Availability

In Stock

NVIDIA B200 GPU

Access reserved NVIDIA B200 capacity across global providers. Compare availability, regions, configurations, and contract terms — without vendor lock-in.

Architecture

NVIDIA Blackwell

MEMORY

192 GB HBM3e

MEMORY BANDWIDTH

Up to ~8 TB/s

PRECISION SUPPORT

FP8 / FP16 / BF16 / TF32

INTERCONNECT

~1.8 TB/s NVLink

FORM FACTORS

SXM

PRIMARY USE CASES

LLM Training, Inference at Scale

TDP

700W (SXM) / 350W (PCIe)

* Performance metrics based on sparsity enabled

VIEW FULL SPECIFICATIONS

NVIDIA H200 GPU

Access NVIDIA H200 GPUs across global providers. Compare pricing, locations, configurations, and deployment options — without vendor lock-in

Availability

In Stock

Architecture

NVIDIA Hopper

MEMORY

141 GB HBM3e

MEMORY BANDWIDTH

Up to ~4.8 TB/s

PRECISION SUPPORT

FP8 / FP16 / BF16 / TF32

INTERCONNECT

Up to 900 GB/s NVLink

FORM FACTORS

SXM5

PRIMARY USE CASES

LLM Training, Inference at Scale

TDP

700W (SXM) / 350W (PCIe)

* Performance metrics based on sparsity enabled

VIEW FULL SPECIFICATIONS

NVIDIA H200 GPU

Availability

In Stock

NVIDIA H200 GPU

Access NVIDIA H200 GPUs across global providers. Compare pricing, locations, configurations, and deployment options — without vendor lock-in

Architecture

NVIDIA Hopper

MEMORY

141 GB HBM3e

MEMORY BANDWIDTH

Up to ~4.8 TB/s

PRECISION SUPPORT

FP8 / FP16 / BF16 / TF32

INTERCONNECT

Up to 900 GB/s NVLink

FORM FACTORS

SXM5

PRIMARY USE CASES

LLM Training, Inference at Scale

TDP

700W (SXM) / 350W (PCIe)

* Performance metrics based on sparsity enabled

VIEW FULL SPECIFICATIONS

AMD MI300X GPU

Access AMD MI300X GPUs across global providers. Compare pricing, locations, configurations, and deployment options — without vendor lock-in.

Availability

In Stock

Architecture

AMD CDNA 3

MEMORY

192 GB HBM3

MEMORY BANDWIDTH

Up to ~5.3 TB/s

PRECISION SUPPORT

FP16 / BF16 / FP8 / INT8

INTERCONNECT

Infinity Fabric

FORM FACTORS

OAM

PRIMARY USE CASES

LLM Training, Inference at Scale

TDP

700W (SXM) / 350W (PCIe)

* Performance metrics based on sparsity enabled

VIEW FULL SPECIFICATIONS

AMD MI300X GPU

Availability

In Stock

AMD MI300X GPU

Access AMD MI300X GPUs across global providers. Compare pricing, locations, configurations, and deployment options — without vendor lock-in.

Architecture

AMD CDNA 3

MEMORY

192 GB HBM3

MEMORY BANDWIDTH

Up to ~5.3 TB/s

PRECISION SUPPORT

FP16 / BF16 / FP8 / INT8

INTERCONNECT

Infinity Fabric

FORM FACTORS

OAM

PRIMARY USE CASES

LLM Training, Inference at Scale

TDP

700W (SXM) / 350W (PCIe)

* Performance metrics based on sparsity enabled

VIEW FULL SPECIFICATIONS

NVIDIA H100 GPU

Access NVIDIA H100 GPUs across global providers. Compare pricing, locations, configurations, and deployment options — without vendor lock-in

Availability

In Stock

Architecture

NVIDIA Hopper

MEMORY

80 GB HBM3

MEMORY BANDWIDTH

up to 3.35 TB/s (SXM)

PRECISION SUPPORT

FP8 / FP16 / BF16 / TF32

INTERCONNECT

NVLink (up to 900 GB/s)

FORM FACTORS

SXM5 / PCIe Gen 5

PRIMARY USE CASES

LLM Training, Inference at Scale

TDP

700W (SXM) / 350W (PCIe)

* Performance metrics based on sparsity enabled

VIEW FULL SPECIFICATIONS

NVIDIA B200 GPU

Access reserved NVIDIA B200 capacity across global providers. Compare availability, regions, configurations, and contract terms — without vendor lock-in.

Availability

In Stock

Architecture

NVIDIA Blackwell

MEMORY

192 GB HBM3e

MEMORY BANDWIDTH

Up to ~8 TB/s

PRECISION SUPPORT

FP8 / FP16 / BF16 / TF32

INTERCONNECT

~1.8 TB/s NVLink

FORM FACTORS

SXM

PRIMARY USE CASES

LLM Training, Inference at Scale

TDP

700W (SXM) / 350W (PCIe)

* Performance metrics based on sparsity enabled

VIEW FULL SPECIFICATIONS

NVIDIA H200 GPU

Access NVIDIA H200 GPUs across global providers. Compare pricing, locations, configurations, and deployment options — without vendor lock-in

Availability

In Stock

Architecture

NVIDIA Hopper

MEMORY

141 GB HBM3e

MEMORY BANDWIDTH

Up to ~4.8 TB/s

PRECISION SUPPORT

FP8 / FP16 / BF16 / TF32

INTERCONNECT

Up to 900 GB/s NVLink

FORM FACTORS

SXM5

PRIMARY USE CASES

LLM Training, Inference at Scale

TDP

700W (SXM) / 350W (PCIe)

* Performance metrics based on sparsity enabled

VIEW FULL SPECIFICATIONS

WHY COMPARE

GPU SPECIFICATIONS?

Choosing the right GPU for your AI workload is critical for optimizing cost and performance. While the NVIDIA H100 offers the highest throughput for large-scale training, the A100 remains a cost-effective workhorse for inference and smaller models. Understanding memory bandwidth (HBM3 vs HBM2e) and interconnect speeds (NVLink vs PCIe) helps in architecting efficient clusters.

Training

Requires massive memory bandwidth and high interconnect speeds. H100 and H200 are preferred for their NVLink capabilities.

Training

Requires massive memory bandwidth and high interconnect speeds. H100 and H200 are preferred for their NVLink capabilities.

Training

Requires massive memory bandwidth and high interconnect speeds. H100 and H200 are preferred for their NVLink capabilities.

Inference

Latency and VRAM capacity are key. L40S and A100s offer excellent price/performance for serving models.

Inference

Latency and VRAM capacity are key. L40S and A100s offer excellent price/performance for serving models.

Inference

Latency and VRAM capacity are key. L40S and A100s offer excellent price/performance for serving models.

Fine-tuning

A balance of compute and memory. Single or multi-node A100 setups are often the sweet spot for LoRA/QLoRA.

Fine-tuning

A balance of compute and memory. Single or multi-node A100 setups are often the sweet spot for LoRA/QLoRA.

Fine-tuning

A balance of compute and memory. Single or multi-node A100 setups are often the sweet spot for LoRA/QLoRA.

COMPUTE

EXCHANGE

The transparent GPU marketplace for AI infrastructure. Built for builders.

ALL SYSTEMS OPERATIONAL

INFORMATION

GPUS

Inventory

Referral Program

ABOUT

BLOG

LEGAL

Marketplace Terms

Compute Service Terms

Fees

E-sign Disclosure

Ask AI for a summary of Compute Exchange

TWITTER

GITHUB

BUILT FOR THE AI ERA

COMPUTE

EXCHANGE

The transparent GPU marketplace for AI infrastructure. Built for builders.

ALL SYSTEMS OPERATIONAL

INFORMATION

GPUS

Inventory

Referral Program

ABOUT

BLOG

LEGAL

Marketplace Terms

Compute Service Terms

Fees

E-sign Disclosure

Ask AI for a summary of Compute Exchange

TWITTER

GITHUB

BUILT FOR THE AI ERA

COMPUTE

EXCHANGE

The transparent GPU marketplace for AI infrastructure. Built for builders.

ALL SYSTEMS OPERATIONAL

INFORMATION

GPUS

Inventory

Referral Program

ABOUT

BLOG

LEGAL

Marketplace Terms

Compute Service Terms

Fees

E-sign Disclosure

Ask AI for a summary of Compute Exchange

TWITTER

GITHUB

BUILT FOR THE AI ERA