GPU Mart

About GPU Mart

GPU Mart is the GPU hosting brand of Database Mart LLC, a US hosting company that has been running dedicated servers and VPS infrastructure since 2005. The GPU line launched in 2021 and now spans 25+ NVIDIA SKUs and more than 3,000 active GPUs, all hosted in SOC-certified data centers inside the United States. As a Database Mart brand, GPU Mart is covered by the company's published SLA (effective August 2026): 99.9% network and platform uptime, with service credits worth 3x the downtime beyond that allowance. It is worth being precise about what GPU Mart is, because it is not structured like RunPod, Vast.ai or Lambda Labs. GPU Mart sells GPU server hosting — a GPU VPS or a dedicated bare metal machine that you keep — rather than a per-second on-demand GPU cloud. Most of the catalog is billed monthly, and hourly billing is available on a subset of cards. That model has real consequences: you get a full server with generous CPU, RAM and storage bundled in, Windows or Linux, and no noisy-neighbour contention on the dedicated line, but you do not get per-second billing or instant teardown on the flagship GPUs. The catalog splits into two lines. GPU VPS covers RTX 5060 8GB up through RTX Pro 6000 96GB, with 16–32 vCPU cores and 28–84GB RAM per slice. Dedicated GPU servers run from an RTX 2060 12GB entry box up to H100 80GB, and include multi-GPU configurations (2× RTX 4090, 2× RTX 5090, 3× A5000, 3× V100, 3× A6000, 4× A6000, 4× A100 40GB). The enterprise dedicated tier ships with 256GB RAM, a 36-core dual E5-2697v4, and 240GB SSD + 2TB NVMe + 8TB SATA — storage and memory that on-demand GPU clouds normally meter separately. GPU Mart is also one of the few hosts publishing hourly rates on the RTX Pro Blackwell workstation cards and on GeForce RTX 4090, cards that are still hard to rent by the hour anywhere else. Support is 24/7 live chat, advertised at under 5 minutes average response and roughly 24,000 tickets and chats per month across the Database Mart estate.

GPU Mart Pricing

GPU Mart runs two separate pricing models. Verified against gpu-mart.com on 8 September 2026, three weeks after our initial review — the catalog moves, so figures below carry that date. Published hourly rates on 6 SKUs, all bookable immediately today: • GeForce RTX 4090 24GB GDDR6X (Dedicated Enterprise) — $0.76/hr, promo check current pricing • NVIDIA A100 40GB HBM2 (Dedicated Enterprise) — $1.11/hr, promo $0.88/hr • NVIDIA RTX Pro 5000 48GB (GPU VPS Advanced) — $0.48/hr • GeForce RTX 2060 12GB (Dedicated Professional) — $0.277/hr, promo $0.22/hr • NVIDIA RTX Pro 4000 24GB (GPU VPS Advanced) — $0.27/hr • NVIDIA RTX A4000 16GB (GPU VPS Professional) — check current pricing, promo $0.15/hr The RTX 5090 was removed from hourly billing since our last check on 18 August — it is monthly-only now. The A100 80GB remains hourly-capable per GPU Mart's FAQ, with no published rate; check with sales before planning around it. Monthly plans, month-to-month price first and the 24-month prepaid tier second: • NVIDIA H100 80GB — $2,599/mo, $2,099 at 24 months • NVIDIA A100 80GB — $1,699/mo, $1,559 at 24 months • NVIDIA A100 40GB — $799/mo, flat across every term • GeForce RTX 5090 32GB (Dedicated) — $599/mo, $479 at 24 months; $539/mo flat on the GPU VPS plan • GeForce RTX 4090 24GB (Dedicated) — $549/mo, $409 at 24 months • GeForce RTX 2060 12GB (Dedicated) — $199/mo, flat across every term • RTX A4000 16GB (GPU VPS) — $149/mo, $119 at 24 months • RTX Pro 4000 24GB (GPU VPS) — $249/mo, $189 at 24 months • RTX Pro 5000 48GB (GPU VPS) — $449/mo, $359 at 24 months • RTX Pro 6000 96GB (GPU VPS) — $799/mo, $649 at 24 months GPU Mart's own pricing page still lists quarterly as the shortest term for the RTX Pro VPS plans (4000, 5000, 6000), but GPU Mart has confirmed the month-to-month prices above directly with us, so we show those instead of the on-site quarterly rate. Read the "starting at" prices carefully. GPU Mart advertises figures such as "RTX Pro 6000 from $0.66/hr", "RTX 4090 from $0.56/hr", "H100 from $2.91/hr" and "GPU instances from $0.02/hour". Those are the 24-month prepaid monthly price divided by 720 hours, not rates you can book — and H100 and RTX Pro 6000 cannot be rented by the hour at all. On a 730-hour month the H100 plan is about $3.56/hr equivalent at the month-to-month price, or $2.88/hr if you prepay two years. Both are above what RunPod, DataCrunch or Lambda Labs charge on demand, so for bursty H100 work an on-demand provider is cheaper. GPU Mart's case is the always-on server with bundled CPU, RAM and storage, and the cheap entry hourly tiers. Longer terms cut the monthly rate by roughly 10–25%, and GPU Mart states that percentage-off promotions remain valid for the lifetime of the service.

Key Features

GPU VPS Line

Virtualized GPU slices from RTX 5060 8GB up to RTX Pro 6000 96GB, with 16–32 CPU cores, 28–84GB RAM and 240–400GB SSD. A cheaper entry point than dedicated hardware, with a dedicated IPv4 included.

Dedicated Bare Metal GPU Servers

Single and multi-GPU dedicated machines up to H100 80GB, with no hypervisor between your workload and the GPU. The enterprise tier bundles 256GB RAM, 36 cores and 240GB SSD + 2TB NVMe + 8TB SATA.

Hourly Billing on Six GPUs

RTX A4000, RTX Pro 4000, RTX 2060, RTX Pro 5000, RTX 4090 and A100 40GB are all bookable by the hour immediately. A100 80GB is hourly-capable by quote, with no published rate. Useful for short evaluations before committing to a monthly plan.

Full RTX Pro Blackwell Lineup

RTX Pro 6000 96GB, RTX Pro 5000 48GB, RTX Pro 4000 24GB and RTX Pro 2000 16GB — professional Blackwell cards with large VRAM, available where most GPU clouds stock only data-center parts.

Windows GPU Hosting

First-class Windows Server support alongside Linux, which matters for GPU-accelerated rendering, CAD, video encoding and Windows-only tooling that Linux-first on-demand clouds cannot run.

24/7 Live Chat Support

Human support around the clock with an advertised sub-5-minute response time, backed by 20+ years of Database Mart operations and roughly 24,000 tickets and chats handled per month.

Best Use Cases

✓Always-on inference endpoints — a monthly dedicated server beats hourly billing once utilisation approaches 24/7
✓Windows GPU workloads — rendering, video encoding, CAD and GPU-accelerated Windows applications
✓Stable Diffusion and ComfyUI on RTX 4090 24GB, rentable hourly at $0.76/hr for burst image generation
✓Fine-tuning on A100 40GB at $1.11/hr hourly, or on a monthly plan for multi-week training runs
✓Data-sensitive workloads that require US-only hosting on dedicated single-tenant hardware
✓Storage-heavy pipelines — the 8TB SATA + 2TB NVMe bundle avoids paying separately for dataset storage
✓Budget development and CI boxes on RTX A4000 at $0.25/hr or RTX 2060 at $0.277/hr

Compare Providers

Provider	Price (A100)	GPUs	Regions	Best For
GPU Mart	$1.11/hr (A100 40GB, hourly)	25+ SKUs	1 (US)	Always-on servers, Windows, bare metal
RunPod	$1.39/hr (A100 80GB)	200+	31	Bursty on-demand and serverless
Lambda Labs	$1.29/hr (A100 80GB)	10+	3	AI/ML tooling and reputation
Vast.ai	$1.89/hr (A100 80GB)	500+	40+	Lowest prices, variable reliability

Frequently Asked Questions

Can I rent GPU Mart servers by the hour?

Yes, on six SKUs, all bookable immediately: RTX A4000 check current pricing, RTX Pro 4000 $0.27/hr, RTX 2060 $0.277/hr, RTX Pro 5000 $0.48/hr, RTX 4090 $0.76/hr and A100 40GB $1.11/hr. A100 80GB is hourly-capable by quote with no published rate, and H100, RTX 5090 and RTX Pro 6000 are monthly plans only.

Does GPU Mart offer an SLA?

Yes. As a brand of Database Mart LLC, GPU Mart is covered by Database Mart's published SLA, effective August 2026: 99.9% network and platform uptime, with service credits worth 3x the downtime beyond that allowance, capped at the monthly recurring charge. Credit requests must be submitted within 8 business days of an incident.

Why does GPU Mart advertise GPU instances from $0.02/hour?

That figure is the entry 2GB GPU VPS on a 24-month prepaid term, with the monthly price divided by 720 hours. It expresses the effective cost of a long commitment rather than an hourly rate you can book. The same applies to the $0.66/hr RTX Pro 6000 and $2.91/hr H100 figures in GPU Mart marketing. The genuinely hourly entry point is the RTX A4000 at check current pricing.

Is GPU Mart cheaper than RunPod or Lambda Labs?

Not on price per GPU-hour for H100 — $2,599/mo month-to-month is about $3.56/hr equivalent, or $2.88/hr if you prepay two years, while RunPod charges $1.99/hr and Lambda $2.49/hr on demand. GPU Mart wins at the entry tiers, where check current pricing buys an RTX A4000, and on long-running servers, where 128–256GB RAM, 36 CPU cores and up to 10TB of storage are included rather than metered.

Where are GPU Mart data centers located?

All GPU Mart servers are in the United States, in SOC-certified facilities with redundant power. GPU Mart has not published the specific city or state, and there are no EU, UK or APAC locations, so latency-sensitive or data-residency workloads outside the US are better served elsewhere.

What is the difference between GPU VPS and a dedicated GPU server?

A GPU VPS is a virtualized slice with a passed-through GPU and 16–32 CPU cores, cheaper to start. A dedicated GPU server gives you the whole physical machine with no hypervisor overhead, far more RAM and storage, and the option of multi-GPU configurations up to 4× A100 40GB.

How long does provisioning take?

GPU Mart quotes 20 to 40 minutes for Linux and roughly 30 minutes to 2 hours for Windows. That is much slower than the seconds-to-minutes provisioning of on-demand GPU clouds, and worth planning around if you spin resources up and down frequently.

Who operates GPU Mart?

GPU Mart is a brand of Database Mart LLC, a US company founded in 2005 that also sells VPS and dedicated server hosting. The GPU line launched in 2021 and now covers 25+ NVIDIA models and 3,000+ active GPUs.

✓ Pros

✗ Cons

Popular GPUs

Visit GPU Mart