Why PureGPU?
An infrastructure built for scale, reliability and best-in-class hardware and software.
Enterprise Network
Fast, reliable, and redundant. All PureGPU infrastructure is connected to the internet via our 100+ Gbps multihomed network. All of our datacenter locations have diverse fiber paths and redundancy at every level. Every PureGPU service is connected to the internet with:
- 10Gbps Uplink
- Premium, Multihomed Peering
- 99.999% Network Uptime
Datacenter Reliability
Mission critical ready. PureGPU machines are hosted only in true SSAE 16 certified tier 4 datacenters. Unlike competitors, our infrastructure is backed by A+B power, UPS generators, redundant cooling, and 24/7 biometric security at the physical level. Our goal is to ensure that our GPU infrastructure is ready for mission critical use.
Top Tier Hardware
We build all of our GPU machines with the latest and greatest supporting hardware. Our current iteration of GPU servers include:
- Latest Gen AMD EPYC Genoa CPUs
- RAID10 PCI 5.0 NVMe Storage
- High-Speed DDR5 ECC RAM
Software
Our custom management and provisioning platform is designed to make your deployment pipeline as simple as possible. All GPU instances are provided with:
- KVM Virtualization with PCI passthrough
- Provision with Dockerfile or one of our premade templates
- API Access for Management
- Robust Monitoring
RTX 4090 GPU servers
Powered by AMD EPYC 7763 @ 3.7 GHz
RAID1 NVME SSDs
DDR4 RAM
Up to 4 NVIDIA® RTX™ 4090 GPUs. Experience the ultimate in performance, efficiency, and AI-powered graphics with the RTX™ 4090, featuring 24GB of GDDR6X memory and the latest Ada Lovelace architecture, and running at full PCIe 4.0 x16 speeds.
CPU
16 CoresSSD
400 GBRAM
48 GBvRAM
24 GBHourly
0.58/h1-month
0.55/h6-months
0.49/hCPU
32 CoresSSD
800 GBRAM
96 GBvRAM
24 GBHourly
1.17/h1-month
1.10/h6-months
0.99/hCPU
48 CoresSSD
1200 GBRAM
144 GBvRAM
24 GBHourly
1.75/h1-month
1.64/h6-months
1.48/hCPU
64 CoresSSD
1600 GBRAM
192 GBvRAM
24 GBHourly
2.33/h1-month
2.19/h6-months
1.97/hCPU
96 CoresSSD
2400 GBRAM
288 GBvRAM
24 GBHourly
3.50/h1-month
3.29/h6-months
2.96/hCPU
128 CoresSSD
3200 GBRAM
384 GBvRAM
24 GBHourly
4.67/h1-month
4.38/h6-months
3.95/hRTX 4090 GPU servers
Powered by AMD EPYC 9334 @ 3.9 GHz
RAID10 NVME SSDs
DDR5 RAM
Up to 4 NVIDIA® RTX™ 4090 GPUs. Experience the ultimate in performance, efficiency, and AI-powered graphics with the RTX™ 4090, featuring 24GB of GDDR6X memory and the latest Ada Lovelace architecture, and running at full PCIe 4.0 x16 speeds.
CPU
16 CoresSSD
1024 GBRAM
80 GBvRAM
24 GBHourly
0.58/h1-month
0.55/h6-months
0.49/hCPU
32 CoresSSD
2048 GBRAM
160 GBvRAM
24 GBHourly
1.17/h1-month
1.10/h6-months
0.99/hCPU
48 CoresSSD
3072 GBRAM
240 GBvRAM
24 GBHourly
1.75/h1-month
1.64/h6-months
1.48/hCPU
64 CoresSSD
4096 GBRAM
320 GBvRAM
24 GBHourly
2.33/h1-month
2.19/h6-months
1.97/hRTX 4000 ADA GPU servers
Powered by Intel Xeon E5-2699 V4 @ 3.6 GHz
RAID10 NVME SSDs
DDR4 RAM
Up to 4 NVIDIA® RTX™ 4000 Ada GPUs. Unlock superior rendering, AI, graphics, and compute capabilities with the RTX™ 4000 Ada, equipped with advanced RT, Tensor, and CUDA® cores, and 20GB of GDDR6 memory, ideal for cutting-edge professional workflows, operating at full PCIe 4.0 x16 speeds.
CPU
11 CoresSSD
512 GBRAM
60 GBvRAM
20 GBHourly
0.18/h1-month
0.17/h6-months
0.15/hCPU
22 CoresSSD
1000 GBRAM
120 GBvRAM
20 GBHourly
0.36/h1-month
0.34/h6-months
0.31/hCPU
33 CoresSSD
1500 GBRAM
180 GBvRAM
20 GBHourly
0.55/h1-month
0.51/h6-months
0.46/hCPU
44 CoresSSD
2000 GBRAM
240 GBvRAM
20 GBHourly
0.73/h1-month
0.68/h6-months
0.62/hRTX 6000 ADA GPU servers
Powered by AMD EPYC 9334 @ 3.9 GHz
RAID10 NVME SSDs
DDR5 RAM
Up to 4 NVIDIA® RTX™ 6000 Ada GPUs. Experience unparalleled rendering, AI, graphics, and compute performance with the RTX™ 6000 Ada, featuring advanced RT, Tensor, and CUDA® cores, and 48GB of GDDR6 memory, perfect for today’s demanding professional workflows, running at full PCIe 4.0 x16 speeds.
CPU
16 CoresSSD
1024 GBRAM
128 GBvRAM
48 GBHourly
1.17/h1-month
1.10/h6-months
0.99/hCPU
32 CoresSSD
2048 GBRAM
256 GBvRAM
48 GBHourly
2.33/h1-month
2.19/h6-months
1.97/hCPU
48 CoresSSD
3072 GBRAM
384 GBvRAM
48 GBHourly
3.50/h1-month
3.29/h6-months
2.96/hCPU
48 CoresSSD
4096 GBRAM
512 GBvRAM
48 GBHourly
4.67/h1-month
4.38/h6-months
3.95/hUse Cases
AI, Machine Learning, Deep Learning.
Accelerate your AI projects from machine learning to deep learning with our powerful compute solutions. Researchers and data scientists can efficiently train complex models, analyze large data sets, and innovate faster.
Generative AI Models.
Harness the power of our GPU cloud to train and deploy cutting-edge generative AI models. From text generation and image synthesis to audio production and beyond, our high-performance compute infrastructure supports the most demanding generative tasks, enabling researchers and developers to push the boundaries of what's possible in artificial intelligence.
Video Encoding and Transcoding.
Transform and optimize your media with the latest iterations of NVIDIA NVENC™. Ideal for streaming solutions that require high-efficiency video encoding and live streaming capabilities.
Virtual Desktops, Simulations.
Remotely access a Windows or Linux desktop experience and run all your demanding applications on the latest graphics hardware with high performance from any internet-connected device.
Not finding what you need?
Contact usAvailable worldwide
Hosted in Central U.S.
How can we help?
Get in touch with our sales and support teams for support, or product questions.
1209 Mountain Rd Pl NE, Suite 4282
Albuquerque, NM 87110
+1 (505) 375-0027
Frequently Asked Questions
Unlike most competitors, PureGPU servers are hosted only in true enterprise data centers. All of our datacenter locations feature robust power, cooling, and network infrastructure, with redundancy at every level. Additionally, all data centers are secured at the physical level with biometric access.
We’re always looking for new opportunities. If you don’t see what you’re looking for on our website, just submit a ticket here with your requirements.
All of our offerings are delivered in the form of KVM virtual machines. GPUs are exposed to the virtual environment via PCI passthrough. This enables native performance with no overhead for GPU bound tasks. CPU cores are dedicated and pinned to your virtual machine, so you can utilize 100% of the CPU cores you pay for.
If you need to render an application on a display connected to the GPU, or otherwise have a non-CUDA workload, some will use the preferred graphics card set in Windows' graphics settings for the specific application. Otherwise, you will need to contact our support to receive a GPU with a dummy display adapter connected.
We offer a variety of pre-made Linux and Windows templates with GPU drivers pre-installed. Customers can also upload their own ISO to install any custom operating system of their choice.