RTX 5090, dedicated.
A single-GPU RTX 5090 machine, 32 GB of GDDR7 on NVIDIA's Blackwell architecture, on an x86 host. Listed at $0.69 per hour, billed per minute, the same terms as every other GPUwerk tier.
No capacity available right now. Join the waitlist: when a machine is free, we hold it for the first person in line with at least one hour of credit and email them. This page describes a tier we are testing demand for, not hardware currently installed in the fleet.
Specifications
Per node. Every GPUwerk instance, on any tier, is one whole machine to yourself.
- GPU
- RTX 5090, Blackwell architecture
- Memory
- 32 GB GDDR7
- Host
- x86 host configuration listed in the console
- Tenancy
- One customer per machine, never shared
We have not measured tok/s or training throughput on this tier yet, since no unit is in the fleet. We publish figures only after we run them ourselves; see the DGX Spark fleet benchmarks for how we report them once we do.
What would you run on it?
32 GB of fast VRAM is the consumer sweet spot for a single model that fits without quantization gymnastics.
Image and video generation
Diffusion pipelines that want raw VRAM bandwidth over unified memory capacity, the case where a discrete GPU usually beats a Spark.
Mid-size model serving
A 13B to 32B-class model at Q4 to Q8 fits comfortably, with room for a reasonable batch size.
Fast iteration
Consumer-tier CUDA compute for a single developer's fine-tuning or eval loop, where the 128 GB of a Spark is more capacity than the job needs.
Billing
- Rate
- $0.69 per hour, listed
- Billing
- Per minute, against a prepaid credit balance
- Fees
- No egress fees
- Commitment
- None
Prices are in USD, excluding applicable tax. Full billing mechanics, including what a stop versus a terminate does, are on the pricing page.
FAQ
Is the RTX 5090 tier available now?
No capacity is available right now. We list the tier and its rate so you can join the waitlist; if a machine becomes free, we hold it for the first person in line who has at least one hour of credit, and email them. Nothing on this page claims a 5090 is installed or booked today.
What's the host configuration around the GPU?
The host configuration (CPU, system RAM, storage) is listed in the console at deploy time, once the tier has capacity. We are not publishing numbers for a machine that is not yet in the fleet.
How does billing work on this tier?
The same as every GPUwerk tier: per minute, against a prepaid credit balance, in USD, excluding applicable tax, with no egress fees and no commitment.
32 GB not enough? Compare it against a Spark.
128 GB of unified memory versus 32 GB of fast VRAM: see where each one wins, or check current rates on the pricing page.
Join the waitlist