NVIDIA RTX PRO 6000 Blackwell Server Edition
NVIDIA: specifications, comparisons and alternatives.
The NVIDIA RTX PRO 6000 Blackwell Server Edition is a Blackwell GPU for servers, intended for AI inference, rendering and visual work. According to the manufacturer, it has 96 GB of GDDR7 memory with ECC, 1.597 TB/s of bandwidth and power of up to 600 W. It uses PCIe Gen 5 and supports MIG.
NVIDIA
- 96 GB
- Memory
- GDDR7 (ECC)
- Memory type
- 1,6 TB/s
- Memory bandwidth
- 600 W
- Power (board power or TDP)
You can see the current price and availability with the partner as soon as a link has been set up.
How are the points calculated?
Clouwino points are an arithmetic summary of the specifications the manufacturer states, on a scale of 0 to 100. It is not a benchmark and not a judgement on quality or suitability: a product with fewer points may suit your situation better. Products are compared with other products of the same type. For each property we count how far a product sits between the lowest and highest value in the group:
- FP32 compute (for: consumer), carries heavy weight
- Memory (for: consumer), carries heavy weight
- Memory bandwidth (for: consumer), carries heavy weight
- Compute units (for: consumer), carries light weight
- Boost clock (for: consumer), carries light weight
- Memory (for: professional and data centre), carries medium weight
- Memory bandwidth (for: professional and data centre), carries medium weight
- FP16/BF16 (for: professional and data centre), carries heavy weight
- FP8 (for: professional and data centre), carries medium weight
- FP64 (for: professional and data centre), carries medium weight
If too much data is missing, we do not show points. Points are relative: when a new product is added, they can shift. Read more at how we compare.
Specifications of NVIDIA RTX PRO 6000 Blackwell Server Edition
All specifications the manufacturer lists, per item. Want to see another product next to it? Pick one below.
| Specificatie | NVIDIA RTX PRO 6000 Blackwell Server Edition |
|---|---|
| Overview | |
| Clouwino points | Not stated by the manufacturer |
| General | |
| Architecture | Blackwell |
| Manufacturing process | Not stated by the manufacturer |
| Form | PCIe 5.0, dual-slot FHFL (lucht) / single-slot FHXL (vloeistof) |
| Launch | Not stated by the manufacturer |
| Compute | |
| FP32 compute | 120 TFLOPS |
| FP64 | Not stated by the manufacturer |
| FP16/BF16 | Not stated by the manufacturer |
| FP8 | Not stated by the manufacturer |
| FP4 | Not stated by the manufacturer |
| Performance notes | The page states FP4 4 PFLOPS, FP8 2 PFLOPS, FP16/BF16 1 PFLOP, TF32 234 TFLOPS without saying whether this is dense or with sparsity (no footnote). PFLOPS converted to TFLOPS (unit only). FP64 not stated. Because NVIDIA does not indicate whether these figures are dense or sparse, these fields have been left empty (FP4 4 PFLOPS, FP8 2 PFLOPS, FP16/BF16 1 PFLOPS according to the page). |
| Memory | |
| Memory | 96 GB |
| Memory type | GDDR7 (ECC) |
| Memory bandwidth | 1,6 TB/s |
| Power and connections | |
| Power (board power or TDP) | 600 W |
| GPU interconnect | PCIe Gen 5; no NVLink stated |
| Multiple instances | MIG: up to 4 fully isolated instances |
Note: Figures per GPU. Max power consumption 'up to 600 W (configurable)'. Memory bandwidth 1,597 GB/s converted to TB/s. Process node and launch date not on the page.
Alternatives to NVIDIA RTX PRO 6000 Blackwell Server Edition
Products with similar specifications from the same overview. This is not a ranking and not buying advice.
- NVIDIA H100 SXM80 GB · HBM3 · 3,35 TB/sCompare with NVIDIA RTX PRO 6000 Blackwell Server Edition →
- NVIDIA H200 SXM141 GB · HBM3e · 4,8 TB/sCompare with NVIDIA RTX PRO 6000 Blackwell Server Edition →
- NVIDIA B200 (HGX B200)180 GB · HBM3e · 8 TB/sCompare with NVIDIA RTX PRO 6000 Blackwell Server Edition →
- NVIDIA Rubin (per GPU)288 GB · HBM4 · 22 TB/sCompare with NVIDIA RTX PRO 6000 Blackwell Server Edition →
- NVIDIA L40S48 GB · GDDR6 (ECC) · 0,86 TB/sCompare with NVIDIA RTX PRO 6000 Blackwell Server Edition →
- AMD Instinct MI300X192 GB · HBM3 · 5,3 TB/sCompare with NVIDIA RTX PRO 6000 Blackwell Server Edition →
Form factors and uses of the RTX PRO 6000 Server
This Server Edition is a PCIe 5.0 card that comes in two forms: dual-slot FHFL for air cooling and single-slot FHXL for liquid cooling. That means it fits in servers not built specifically for SXM or OAM.
NVIDIA positions it for a mix of tasks, such as AI inference and visual workloads. You can find an explanation of the difference from an ordinary graphics card in graphics card or data centre GPU. MIG makes it possible to divide the GPU into up to 4 fully isolated instances. An NVLink connection is not mentioned. The process node and launch date are not on the page.
That gives flexibility when choosing the server platform, provided the chassis supports the chosen form.
The large 96 GB of memory is particularly useful if you want to load several tasks or larger models onto one card, although usage depends on your software.
Memory, power and the compute figures
96 GB of GDDR7 with ECC and 1.597 TB/s of bandwidth determines how large the models or scenes are that you can load. The power is up to 600 W and configurable. You can read more about memory in what is VRAM.
FP32 is 120 TFLOPS. For lower precision, the page lists FP4 at 4 PFLOPS, FP8 at 2 PFLOPS, FP16/BF16 at 1 PFLOPS and TF32 at 234 TFLOPS, but the manufacturer does not say whether these are dense or with sparsity. We therefore quote them as stated and not as a basis for comparison. FP64 is not stated by the manufacturer.
For the lowest precisions, the label is therefore important: without that information, you should not set these numbers against values that do have a label.
Choosing: cooling and platform
The first decision is the cooling. The dual-slot FHFL version is for air, the single-slot FHXL version for liquid. Check whether your server can handle the chosen form and the power of up to 600 W.
- Check whether your platform offers PCIe 5.0.
- Without a stated NVLink, communication between GPUs runs via PCIe.
- Take a look at the guide to renting or buying a GPU if you are torn between the cloud and your own server.
Also ask a provider about the exact version, because air and liquid variants have a different enclosure.
You can see the current price and availability at the partner via the button on the page.
Frequently asked questions
Are the FP4 and FP8 figures dense or with sparsity?
That is not known. The page lists FP4 at 4 PFLOPS and FP8 at 2 PFLOPS with no footnote on sparsity. So do not simply compare these figures with values for other GPUs.
Which form factors is the RTX PRO 6000 Blackwell Server Edition available in?
According to the data, it exists as PCIe 5.0 dual-slot FHFL for air cooling and as single-slot FHXL for liquid cooling. The maximum power is 600 W and configurable.
Does this GPU support MIG?
Yes, MIG is possible with up to 4 fully isolated instances per GPU. This lets you share the card between different tasks or users. That is handy when several smaller tasks run on one card at the same time.
Last updated on
