RTX PRO 6000 Blackwell 96GB
The biggest VRAM you can put under a desk. 96 GB GDDR7, 24k CUDA cores, runs 70B models in a tower.
In stock. Lead time 72 hours.
Volume pricing
| Units | Discount | Unit price |
|---|---|---|
| 1 unit | List | $8,500 |
| 2 to 7 units | 3% off | $8,245 |
| 8 to 31 units | 6% off | $7,990 |
| 32+ units | 10% off | $7,650 |
Unit price at 1: $8,500
Request a bulk quote
The full sheet.
Figures from the manufacturer datasheet. Price is an indicative street price, shown as from.
- VRAM
- 96 GB
- Memory
- GDDR7
- Bandwidth
- 1.8 TB/s
- FP8
- 3.8 PFLOPS
- FP4
- 7.6 PFLOPS
- Board power
- 600 W
- Form factor
- PCIe dual-slot
- NVLink
- No
- Architecture
- Blackwell
- GPUs
- 1
- Per GB of VRAM
- $88.54
- Availability
- In stock, ships 72h
- Sale
- Buy online
- SKU
- RTX-PRO-6000
What it is for.
Workloads, sizing and what ships in the box.
96 GB under a desk.
The workstation RTX PRO 6000. 96 GB of GDDR7, 1.8 TB/s, 3.8 PFLOPS FP8 and 24k CUDA cores, with an active cooler for towers. Runs a 70B model at FP8 on one card. 600 W, Max-Q 300 W version on request. 3 year warranty handled by us. In stock, ships in 72 hours. $8,500, about $89 per GB.
Good for
- Llama 3.1 70B at FP8 (84 GB with headroom) in a tower.
- Qwen2.5 32B at FP16 (77 GB) for LoRA fine-tunes.
- gpt-oss-120b at 4-bit (70 GB) on one card.
Fit rule: FP16 weights take 2 bytes per parameter, FP8 1 byte, 4-bit 0.5 bytes, plus 20% headroom for KV cache at short context.
Pairs with
- RTX PRO 5000 Blackwell 48GB, for a second, quieter seat.
- RTX PRO 6000 Blackwell Server Edition 96GB, when the tower becomes a rack.
Active cooling. Max-Q 300 W version on request.
How much VRAM per box.
Card count is capped by slots and by the chassis power budget.
Parts bought alongside.
Same workload, a different trade between VRAM, bandwidth and price.