RTX PRO 6000 Blackwell,
all three editions
One die, three cards, 96 GB each, and they are not interchangeable. The cooler decides how many you fit, and one cable decides whether you get 600 W or 450.
- EU contract and invoice
- Licences sized with the hardware
- Engineering, if you want it
- Warranty and support
The three editions
Same memory, different everything else. Pick by chassis and by how many cards you need in it, not by the name.
| Edition | Cooling | Power | MIG / vGPU | Fits |
|---|---|---|---|---|
| Workstation | Two axial fans, breathes through the chassis | 600 W | MIG, no vGPU | One card comfortably, two at most. The first card’s exhaust feeds the second, so density is poor. |
| Max-Q | Blower, half the power limit | 300 W | MIG, no vGPU | Up to four per workstation, at about 12% less FP32. |
| Server Edition | Passive, needs chassis airflow | 300 to 600 W, configurable | MIG and vGPU | Rack servers only. Roughly 11% fewer tokens/s than Workstation on generation. |
All editions are available to order. Lead time is confirmed with every quote.
Cards alone are a fine order
You do not need a licence bundle or an engineering contract to buy from us. If you need the cards and nothing else, in whichever edition your chassis takes, we quote exactly that, invoiced in the EU, with quantities and lead time confirmed in writing with the quote.
Not just a box on a pallet
One counterparty in Vienna. No cross-border purchase to arrange, no import to clear on your side.
Where a card needs a licence to do the job you bought it for, it is in the quote, not discovered three months later.
Our partner Vixen.UNO designs, deploys and supports it. Under the same contract, or not at all. Your call.
Manufacturer warranty handled through us, with support alongside the team who runs the system.
Which one is right for you
In 48 GB a 70B model fits only with aggressive quantisation and a trimmed cache. In 96 GB the same model sits in Q4 with a working context. GamersNexus logged a 928% gain over an RTX 5090 on Llama 3.3 70B; that is purely the effect of 32 GB being too little.
Read the full analysis →The 16-pin 12V-2×6 connector’s SENSE pins tell the card what the cable can deliver, and the card sets its own ceiling from them. NVIDIA’s product brief for the Server Edition lists a 600 W and a 450 W mode, set by how the cable is strapped: on a 450 W cable the card reports Max Power Limit: 450.00 W and nvidia-smi -pl 600 changes nothing. Use a 600 W cable only where the server vendor supports 600 W in that slot.
Blackwell splits into four 24 GB instances, two of 48, or one of 96. Only the Server Edition got vGPU, licensed per concurrent user for graphics and per GPU for compute. Without a licence a vGPU runs 20 minutes at full speed; then a graphics vGPU is degraded, and degraded further after a day, and a compute vGPU drops to idle-level compute, as NVIDIA’s current client licensing guide describes it.
Read the full analysis →RTX PRO 6000, in short
Which edition should I order?
Is the Server Edition slower?
Do I need a vGPU licence?
Can I just add these to my existing servers?
Why no prices?
From our engineering blog
The full spec table NVIDIA never published in one place: bandwidth, cooling, MIG, vGPU and power caps.
Read the article →96 GB GDDR7 vs 48 GB GDDR6, FP4 and MIG that Ada never had, and where Ada still wins.
Read the article →Per-card numbers, licences per user for graphics and per GPU for compute, and where the setup breaks.
Read the article →Check price and availability
We reply within one business day