46 GPUs · 7 Families

GPU Cloud Comparison

Compare every NVIDIA GPU for AI training, inference, and HPC.

From Volta (2017) to Rubin Ultra (2026+). Find the right GPU by specs, use case, and cloud provider pricing across 0 global regions.

0
GPU plans
0
Regions
$0.00
From /hour
Global GPU Network
0 regions·0 plans
GPU regionSelected
Drag to rotate · Scroll to zoom · Click a dot for details

Next-generation (2026+). Superchip (Vera CPU + Rubin GPU), HBM4 memory, NVLink 6, designed for trillion-parameter models and exascale HPC.

Blackwell8 GPUs

Current flagship (2024-2025). HBM3e memory, NVLink 5, 2.5× faster than H100 for LLM workloads. Up to 288GB VRAM.

Available2025

B300

Blackwell Ultra standalone GPU. 288GB HBM3e — 1.5× more VRAM than B200. Flagship for extreme-scale AI training.

VRAM288 GB HBM3e
Bandwidth8.0 TB/s
InterconnectNVLink 5 (1.8 TB/s)
Available2025

GB300 NVL72

NVIDIA's flagship rack-scale AI system with 72 Blackwell Ultra GPUs. Designed for trillion-parameter model training.

VRAM13,824 GB HBM3e (192 GB per GPU × 72)
Bandwidth8.0 TB/s per GPU
InterconnectNVLink 5 (1.8 TB/s per GPU)
Available2025

RTX PRO 6000 Blackwell

Professional workstation/datacenter GPU based on Blackwell architecture. 96GB GDDR7 for AI, rendering, and graphics workloads.

VRAM96 GB GDDR7
Bandwidth1,792 GB/s
InterconnectPCIe Gen5 (128 GB/s) / NVLink 5
Available2025

RTX PRO 6000 SE

Server Edition of RTX PRO 6000 Blackwell. 96GB GDDR7. Optimized for datacenter deployment with lower TDP (300W PCIe).

VRAM96 GB GDDR7
Bandwidth1,792 GB/s
InterconnectPCIe Gen5 (128 GB/s) — no NVLink
Available2024

B100

Entry-level Blackwell GPU. Lower power variant of B200 for mainstream AI workloads.

VRAM192 GB HBM3e
Bandwidth8.0 TB/s
InterconnectNVLink 5 (1.8 TB/s)
Available2024

B200

NVIDIA B200 Blackwell flagship accelerator. Available in two form factors: SXM for highest power density (1000W, used in HGX B200 baseboards with NVLink 5), and PCIe as B200 NVL for traditional air-cooled server chassis (800W). Features 192 GB HBM3e memory with 8.0 TB/s bandwidth. 2.5× faster AI training than H100.

VRAM192 GB HBM3e
Bandwidth8.0 TB/s
InterconnectNVLink 5 (1.8 TB/s)
Available2024

GB200

Grace + Blackwell superchip with unified memory. The foundation of the GB200 NVL72 rack system.

VRAM192 GB HBM3e
Bandwidth8.0 TB/s
InterconnectNVLink 5 (1.8 TB/s)
Available2024

GB200 NVL72

NVIDIA's flagship rack-scale AI system with 72 Blackwell GPUs and 36 Grace CPUs. Designed for trillion-parameter model training.

VRAM13,824 GB HBM3e (192 GB per GPU × 72)
Bandwidth8.0 TB/s per GPU
InterconnectNVLink 5 (1.8 TB/s per GPU)
Hopper9 GPUs

Industry standard (2022-2024). HBM3 memory, Transformer Engine, NVLink 4. The most widely-deployed AI GPU architecture.

Available2024

H200 NVL

NVL Gen5 form factor of H200. 141GB HBM3e but no NVLink — uses NVL Gen5 (128 GB/s) instead. Lower power (350W) for environments without SXM.

VRAM141 GB HBM3e
Bandwidth4.8 TB/s
InterconnectPCIe Gen5 (128 GB/s) — no NVLink
Available2024

H200 SXM

SXM5 form factor of H200. 141GB HBM3e with full NVLink 4 (900 GB/s). Highest bandwidth Hopper variant for large model training.

VRAM141 GB HBM3e
Bandwidth4.8 TB/s
InterconnectNVLink 4 (900 GB/s)
Available2023

GH200

Grace + Hopper superchip with unified memory. Designed for giant-scale AI and HPC workloads.

VRAM144 GB HBM3e (or 96 GB HBM3)
Bandwidth4.8 TB/s (HBM3e)
InterconnectNVLink 4 (900 GB/s)
Available2023

H100 NVL

PCIe variant of H100 with 94GB HBM3 (more VRAM than SXM version). Uses NVLink bridge for multi-GPU on PCIe.

VRAM94 GB HBM3
Bandwidth3.9 TB/s
InterconnectNVLink 4 (900 GB/s) — PCIe card with NVLink bridge
Available2023

H20

China compliance variant of the H100. Meets US export controls with reduced compute performance but retains full 96GB HBM3 memory and NVLink interconnect. Designed for large-model inference in the China market.

VRAM96 GB HBM3
Bandwidth4.0 TB/s
InterconnectNVLink 4 (900 GB/s)
Available2023

H800 SXM

China export variant of the H100 SXM. Meets US export controls with reduced NVLink bandwidth (400 GB/s vs 900 GB/s) while retaining 80GB HBM3 and full compute density. Built for AI training in compliant regions.

VRAM80 GB HBM3
Bandwidth3.35 TB/s
InterconnectNVLink 4 (400 GB/s)
Available2022

GH100

The GH100 is NVIDIA's Hopper architecture GPU silicon — the foundational chip that powers all H100 variants (SXM5, PCIe, NVL). Built on TSMC's custom 4N process with 80 billion transistors across an 814 mm² die, it introduced the Transformer Engine for accelerated LLM training and inference. The GH100 die is the largest and most complex GPU NVIDIA had shipped at its 2022 launch.

VRAM80 GB HBM3
Bandwidth3.35 TB/s
InterconnectNVLink 4 (900 GB/s)
Available2022

H100 PCIe

Standard PCIe Gen5 form factor of H100 (no NVLink). 350W TDP — lower performance than SXM but easier to deploy. No multi-GPU NVLink interconnect.

VRAM80 GB HBM3
Bandwidth3.35 TB/s
InterconnectPCIe Gen5 (128 GB/s) — no NVLink
Available2022

H100 SXM

SXM5 form factor of H100. 80GB HBM3 with full NVLink 4 (900 GB/s). 700W TDP — the highest performance H100 variant for multi-GPU training.

VRAM80 GB HBM3
Bandwidth3.35 TB/s
InterconnectNVLink 4 (900 GB/s)

Inference & rendering (2022-2024). GDDR6 memory, optimized for AI inference, graphics, and video processing.

Available2024

RTX 2000 Ada

Entry-level Ada Lovelace workstation GPU. 16GB GDDR6 with low 70W TDP for single-slot deployments. Optimized for AI inference, CAD, and content creation workloads.

VRAM16 GB GDDR6
Bandwidth256 GB/s
InterconnectPCIe Gen4
Available2023

L2

China market variant of the L4. 24GB GDDR6 in a low-profile 150W form factor for AI inference and video workloads in compliant regions.

VRAM24 GB GDDR6
Bandwidth300 GB/s
InterconnectPCIe Gen4
Available2023

L20

China market variant of the L40. 48GB GDDR6 for AI inference and graphics. Meets US export controls while delivering high performance for data center workloads.

VRAM48 GB GDDR6
Bandwidth864 GB/s
InterconnectPCIe Gen4
Available2023

L40S

Datacenter GPU optimized for AI inference and graphics. Successor to A40 with 2× performance.

VRAM48 GB GDDR6
Bandwidth864 GB/s
InterconnectPCIe Gen4 (64 GB/s)
Available2023

RTX 4000 Ada

Entry-level Ada Lovelace workstation GPU. 20GB GDDR6, single-slot design for compact workstations.

VRAM20 GB GDDR6
Bandwidth360 GB/s
InterconnectPCIe Gen4 (64 GB/s)
Available2023

RTX 4500 Ada

Mid-range Ada Lovelace workstation GPU. 24GB GDDR6 for professional rendering and AI workloads.

VRAM24 GB GDDR6
Bandwidth432 GB/s
InterconnectPCIe Gen4 (64 GB/s)
Available2023

RTX 5000 Ada

Mid-range Ada Lovelace workstation GPU. 32GB GDDR6 for professional rendering and AI workloads.

VRAM32 GB GDDR6
Bandwidth576 GB/s
InterconnectPCIe Gen4 (64 GB/s)
Available2022

L4

Low-power, single-slot GPU for AI inference and video processing. 72W TDP — no extra power connector needed.

VRAM24 GB GDDR6
Bandwidth300 GB/s
InterconnectPCIe Gen4 (64 GB/s)
Available2022

L40

Datacenter GPU for visual computing and rendering. Lower AI performance than L40S.

VRAM48 GB GDDR6
Bandwidth864 GB/s
InterconnectPCIe Gen4 (64 GB/s)
Available2022

RTX 6000 Ada

Workstation GPU for professional AI, rendering, and graphics. 48GB GDDR6 with ECC support.

VRAM48 GB GDDR6
Bandwidth960 GB/s
InterconnectPCIe Gen4 (64 GB/s)
Ampere11 GPUs

Mainstream (2020-2022). HBM2e memory, MIG support, NVLink 3. The most widely-deployed AI GPU in cloud today.

Available2022

A800 SXM4

China export variant of the A100 SXM4 80GB. Meets US export controls with reduced interconnect bandwidth. Built on the Ampere architecture for AI training and HPC workloads in compliant regions.

VRAM80 GB HBM2e
Bandwidth2.0 TB/s
InterconnectNVLink 3 (400 GB/s)
Available2021

A10

Mid-range AI inference and graphics GPU. 24GB GDDR6 with low power consumption.

VRAM24 GB GDDR6
Bandwidth600 GB/s
InterconnectPCIe Gen4 (64 GB/s)
Available2021

A100 PCIe

Standard PCIe Gen4 form factor (no NVLink). 80GB HBM2e. 300W TDP — no multi-GPU NVLink interconnect.

VRAM80 GB HBM2e
Bandwidth2.0 TB/s
InterconnectPCIe Gen4 (64 GB/s) — no NVLink
Available2021

A100 SXM 80GB

SXM4 form factor, 80GB HBM2e variant. Full NVLink 3 (600 GB/s). 2× VRAM of original A100 for larger models.

VRAM80 GB HBM2e
Bandwidth2.0 TB/s
InterconnectNVLink 3 (600 GB/s)
Available2021

A16

Multi-instance GPU for VDI (Virtual Desktop Infrastructure). 4 GPUs in one board for virtualization.

VRAM64 GB GDDR6 (16 GB × 4 GPUs)
BandwidthTBD
InterconnectPCIe Gen4
Available2021

A2

Entry-level AI inference GPU. 40-60W TDP — lowest power datacenter GPU. Single-slot, passive cooling.

VRAM16 GB GDDR6
Bandwidth200 GB/s
InterconnectPCIe Gen4 (64 GB/s)
Available2021

A30

Mid-range AI and HPC GPU with HBM2 memory. Lower power than A100.

VRAM24 GB HBM2
Bandwidth933 GB/s
InterconnectNVLink 3 (600 GB/s)
Available2021

RTX A4000

Entry-level Ampere workstation GPU. 16GB GDDR6, single-slot design for compact workstations.

VRAM16 GB GDDR6
Bandwidth448 GB/s
InterconnectPCIe Gen4 (64 GB/s)
Available2021

RTX A5000

Mid-range Ampere workstation GPU. 24GB GDDR6 for professional rendering and AI workloads.

VRAM24 GB GDDR6
Bandwidth768 GB/s
InterconnectPCIe Gen4 (64 GB/s)
Available2020

A40

Datacenter GPU for visual computing. 48GB GDDR6 with ECC for professional workloads.

VRAM48 GB GDDR6
Bandwidth696 GB/s
InterconnectPCIe Gen4 (64 GB/s)
Available2020

RTX A6000

Flagship Ampere workstation GPU. 48GB GDDR6 with ECC for professional AI, rendering, and graphics.

VRAM48 GB GDDR6
Bandwidth768 GB/s
InterconnectPCIe Gen4 (64 GB/s) / NVLink Bridge
Turing3 GPUs

Budget inference (2018-2020). GDDR6 memory, low power. The most affordable AI GPUs for light inference workloads.

Volta1 GPU

Legacy (2017). First GPU with Tensor Cores. Mostly replaced by Ampere/Hopper but still in use.