NV NVIDIA Prosumer Discrete GPU

NVIDIA RTX PRO 6000 Blackwell

Runs downloadable LLMs up to ~149.5B at Q4, ~292-348 tok/s on an 8B model.

NVIDIA RTX PRO 6000 Blackwell
96GB
VRAM
1792
GB/s bandwidth
149.5B
Max model (Q4)
$8,500
MSRP

What you can run

At 8k context, FP16 KV. Full-speed tok/s holds up to "max ctx"; beyond it the context spills to system RAM (CPU/PCIe offload) and drops to the degraded figure.

Model sizeFP16Q8Q4
3B 218-259.8 t/s
≤1233k ctx · then ~4.9-5.8 (slow)
436.1-519.7 t/s
≤1279k ctx · then ~9.7-11.6 (slow)
778.7-928 t/s
≤1299k ctx · then ~17.4-20.7 (slow)
7-8B 81.8-97.4 t/s
≤540k ctx · then ~1.8-2.2 (slow)
163.5-194.9 t/s
≤601k ctx · then ~3.7-4.4 (slow)
292-348 t/s
≤628k ctx · then ~6.5-7.8 (slow)
13B 50.3-60 t/s
≤371k ctx · then ~1.1-1.3 (slow)
100.6-119.9 t/s
≤451k ctx · then ~2.2-2.7 (slow)
179.7-214.2 t/s
≤485k ctx · then ~4-4.8 (slow)
32B 20.4-24.4 t/s
≤87k ctx · then ~0.5-0.5 (slow)
40.9-48.7 t/s
≤209k ctx · then ~0.9-1.1 (slow)
73-87 t/s
≤263k ctx · then ~1.6-1.9 (slow)
70B won't run 18.7-22.3 t/s
≤51k ctx · then ~0.4-0.5 (slow)
33.4-39.8 t/s
≤145k ctx · then ~0.7-0.9 (slow)
120B won't run won't run 19.5-23.2 t/s
≤60k ctx · then ~0.4-0.5 (slow)
400B won't run won't run won't run

Estimate: tok/s ≈ bandwidth ÷ (params × bytes/param) × efficiency (NVIDIA). Cyan = fits ~88.3 GB usable. Beyond "ctx", KV spills to system RAM (CPU/PCIe offload) (rough degraded speed shown).

What else you need for a full build

Minimum supporting components for a desktop built around the NVIDIA RTX PRO 6000 Blackwell. Guidance — exact needs depend on your case, other parts and how much you offload to system RAM.

Power supply
≥ 1000W · 80+ Gold
Headroom over the 600W card plus the rest of the build.
System RAM
≥ 128 GB · 256 GB recommended
At least the 96 GB of VRAM; 2× lets you CPU-offload models larger than VRAM.
Motherboard
1× free PCIe x16 slot (4.0/5.0)
A full-length x16 electrical slot; PCIe 4.0 is ample, 5.0 future-proofs.
CPU
Any modern multi-core
The GPU runs inference; more cores and memory channels only matter when offloading to RAM.
Storage
≥ 1 TB NVMe SSD
Weights are large — a 70B model at Q4 is ~40 GB; budget for a few.
PSU connector
12V-2×6 / 12VHPWR (or 3× 8-pin)
High-power NVIDIA cards use the 16-pin 12V-2×6 power connector.

Run popular models on the NVIDIA RTX PRO 6000 Blackwell

Videos about the NVIDIA RTX PRO 6000 Blackwell

RTX Pro 6000 LLM test

768GB of VRAM. 8x RTX PRO 6000 Blackwell. Liquid-cooled. 4U. Full review at StorageReview.com.

More hardware like the NVIDIA RTX PRO 6000 Blackwell

About the NVIDIA RTX PRO 6000 Blackwell

NVIDIA RTX PRO 6000 Blackwell is a discrete GPU from NVIDIA, 2025. For local genAI the numbers that matter are its 96 GB (how big a model + context fits) and 1792 GB/s bandwidth (generation speed). Plan for a ~1000W power supply.

Frequently asked questions

What size LLM can the NVIDIA RTX PRO 6000 Blackwell run?

With ~88.3 GB usable it runs up to ~149.5B at Q4 (~42B at FP16) at 8k context. Bigger spills into system RAM (CPU/PCIe offload) and slows sharply.

How many tokens/sec does the NVIDIA RTX PRO 6000 Blackwell generate?

Generation is bandwidth-bound. At 1792 GB/s an 8B model at Q4 runs ~292-348 tok/s (estimate).

Why does long context slow it down?

The KV cache for the context must also fit fast memory; once weights+KV exceed 88.3 GB it spills to system RAM (CPU/PCIe offload) and throughput collapses. Each row shows the max usable context.

What else do I need to run the NVIDIA RTX PRO 6000 Blackwell?

Besides the card: a ≥1000W 80+ Gold power supply, at least 96GB of system RAM (2× the 96GB VRAM is ideal for CPU offload), a free PCIe x16 slot, any modern multi-core CPU, and a ≥1TB NVMe SSD for model weights.