RTX PRO 6000 Blackwell Server Edition 96GB
96 GB GDDR7 per slot, passive. Eight in a 4U is 768 GB of VRAM for $69,184 at the 8-unit tier.
In stock. Lead time 72 hours.
Volume pricing
| Units | Discount | Unit price |
|---|---|---|
| 1 unit | List | $9,200 |
| 2 to 7 units | 3% off | $8,924 |
| 8 to 31 units | 6% off | $8,648 |
| 32+ units | 10% off | $8,280 |
Unit price at 1: $9,200
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
- 1.6 TB/s
- FP8
- 3.6 PFLOPS
- FP4
- 7.2 PFLOPS
- Board power
- 600 W
- Form factor
- PCIe dual-slot
- NVLink
- No
- Architecture
- Blackwell
- GPUs
- 1
- Per GB of VRAM
- $95.83
- Availability
- In stock, ships 72h
- Sale
- Buy online
- SKU
- RTX-PRO-6000-SERVER
What it is for.
Workloads, sizing and what ships in the box.
96 GB per slot for $96 per GB.
96 GB of GDDR7 per dual-slot card, passive cooling, 600 W, 3.6 PFLOPS FP8 and 7.2 PFLOPS FP4. Eight in a 4U is 768 GB of VRAM for $73,600 in cards at list. The lowest $/GB in our datacenter tier. In stock, ships in 72 hours. $9,200 each, 3% off from 2 units, 6% from 8.
Good for
- Llama 3.1 70B at FP8 (84 GB with headroom) on one card.
- gpt-oss-120b at 4-bit (70 GB) on one card.
- Llama 3.1 405B at FP8 (486 GB) across 8 cards (768 GB) with tensor parallel over PCIe.
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
- H200 NVL 141GB, when you need HBM bandwidth over capacity per dollar.
- L40S 48GB, for light inference and video in the same racks.
Passive cooling. Best $/GB in the datacenter tier.
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.