GPU MODELS / RTX 6000D
RTX 6000D cloud GPU pricing
3 listings across 2 providers. Cheapest published rate: $1.290/GPU/hr at CloudGPU. Sorted by per-GPU hourly price. Every row links to the provider's own published price.
WHAT IS THE RTX 6000D
The RTX 6000D is the China-market Blackwell workstation/server card: the RTX PRO 6000's GB202 die with 19,968 CUDA cores, but 84 GB instead of 96 GB to sit under export thresholds. It is the largest Blackwell card legally sold into China, which makes it the default high-end answer there for VRAM-heavy AI work.
MEMORY BANDWIDTH
1,568 GB/s
TENSOR COMPUTE
GB202 Blackwell, 19,968 CUDA cores, fifth-gen Tensor Cores
INTERCONNECT
PCIe 5.0, server edition
WHAT IT IS GOOD AT
Large-model inference
84 GB holds a 70B at FP8 or a quantized 100B-class model on a single card - the reason it exists.
Fine-tuning and multi-card servers
1,568 GB/s and 600 W server cooling support multi-card inference nodes; PCIe, not NVLink, limits scale-up.
Rendering and simulation
The full Blackwell RT stack with workstation-class memory - a rendering flagship where it is sold.
WHO SHOULD NOT RENT THE RTX 6000D
Buyers outside its market: the global RTX PRO 6000 carries 96 GB; the D variant exists for compliance, not preference.
Training clusters needing NVLink domains - this is a PCIe card.
THE RTX 6000D MARKET, BY THE NUMBERS
2 providers publish 3 RTX 6000D configurations on the panel today. Published per-GPU hourly rates run from $1.290 (CloudGPU) to $2.42 (GPUBrazil), a 2x spread between the cheapest and the most expensive published rate for the same chip. The median listing sits at $2.09/GPU/hr. 3 rows were reported in stock by the provider at last observation; the rest are published catalog rates. All rates are on-demand - committed-use and reserved pricing is excluded.
RTX 6000D PRICE TREND, DAILY PANEL SNAPSHOTS
Across every RTX 6000D row on the panel, the daily floor held flat at $1.290/GPU/hr over 10 days of tracking. Snapshots run daily; the full history is public in the repo.
RTX 6000D PROVIDER NOTES
CloudGPU · 1 config · 1 in stockSets the panel floor for this model.
from $1.290/hrGPUBrazil · 2 configs · 2 regions · 2 in stock2 of 2 configs reported in stock at last check.
from $2.089/hrBUYING RTX 6000D: WHAT THE PANEL SAYS
At the current floor, one RTX 6000D running around the clock costs about $942 a month (730 hours of on-demand arithmetic). Compare the alternatives below before committing - the same budget often buys more than one chip class.
RTX 6000D VS THE ALTERNATIVES
Same GB202 silicon and CUDA count, 84 GB versus 96 GB GDDR7, same 600 W. The D is the export-compliant trim; inside China it is the top of the legal stack, outside it the standard 6000 is the better card.
Compare live rates: RTX 6000D vs RTX PRO 6000
RTX 6000D PRICING FAQ
What is the cheapest RTX 6000D cloud GPU right now?
The cheapest published RTX 6000D rate on the panel is $1.290 per GPU-hour at CloudGPU (RTX 6000D x 1, Not published). Every price links to the provider's own published rate.
How much does a RTX 6000D cost per hour?
Across 2 providers tracked today, published per-GPU hourly rates for RTX 6000D run from $1.29 to $2.42, with a median of $2.09. Committed-use and reserved rates are excluded; everything shown is on-demand.
How many providers rent RTX 6000D GPUs?
2 providers publish 3 distinct RTX 6000D configurations on Compute Cafe. 3 of those rows were reported in stock by the provider's live endpoint at last observation.
Are RTX 6000D prices going up or down?
Over the 10 days Compute Cafe has tracked the panel, the RTX 6000D floor is holding flat: $1.290 to $1.290/GPU/hr (+0.0%). Daily snapshots are public in the repo, so the trend is auditable.
What does a RTX 6000D cost per month?
At the current floor of $1.290/GPU/hr, a RTX 6000D running 24/7 costs about $942 per month (730 hours of on-demand arithmetic on the cheapest published rate). Committed-use contracts can price lower; the panel tracks on-demand rates only.
Why does the RTX 6000D exist?
Export rules cap memory bandwidth and performance density for the China market. NVIDIA trim VRAM from 96 to 84 GB to land under the threshold while keeping the full GB202 compute die.
What can 84 GB hold?
A 70B model at FP8 with context to spare, a quantized 100B-class model, or large multi-app professional workloads. It is the most VRAM available on a single China-market Blackwell card.
See also alternatives to the RTX 6000D · cheapest GPU-hour per model · inference chips · training chips · by location
CloudGPU
GPUBrazil