Home / Enterprise Solutions / RTX PRO 6000 8-GPU Server
RTX PRO 6000 Blackwell · 96GB GDDR7 · For AI inference
NVIDIA RTX PRO 6000 Blackwell 8-GPU servers for AI inference, in Egypt & the UAE.
Eight NVIDIA RTX PRO 6000 Blackwell GPUs with 768GB of GDDR7, two 128-core AMD EPYC 9754 processors and 3TB of memory in one rack server: built to run AI in production for many users, and to render, simulate and generate video.
768GB of GPU memory
8 × 96GB GDDR7 with ECC, and up to 1.8 TB/s per GPU.
256 AMD EPYC cores, 3TB RAM
2 × AMD EPYC 9754 with 24 × 128GB DDR5-4800 memory.
800 Gb/s of networking
Four dual-port 100GbE QSFP56 network cards.
24-month NBD server warranty
Plus a 12-month replacement warranty on the GPUs, with burn-in testing and support from our engineers.
What it’s for
Built to run AI for your whole business.
Most GPU time goes into running models, not training them. The RTX PRO 6000 does that at scale, and handles graphics too.
LLM inference and AI agents
Serve chatbots, CX agents and copilots to many users; a 70B model in FP8 fits on a single GPU.
AI video and image generation
Text-to-video, image-to-video and AI presenters, with 4 NVENC and 4 NVDEC engines per GPU.
Rendering and visualisation
188 fourth-generation RT cores per GPU for ray-traced rendering and digital twins.
Simulation and engineering
120 TFLOPS of FP32 per GPU for engineering simulation and scientific workloads.
Document AI and OCR
Run AI OCR and document models at scale, alongside our OCR Scanners.
One server, many teams
Up to 4 MIG instances of 24GB per GPU, so one server serves several teams.
Why the RTX PRO 6000
AI and graphics in one server-grade GPU.
96GB of GDDR7 per GPU
ECC memory on a 512-bit interface, with up to 1.8 TB/s of bandwidth.
Blackwell Tensor Cores
752 fifth-generation Tensor Cores and up to 3.8 PFLOPS of FP4 per GPU, for fast, efficient inference.
Multi-Instance GPU
Split each GPU into up to four isolated 24GB instances.
PCIe 5.0 x16
Fast PCIe 5.0 connections between every GPU and the AMD EPYC processors.
Media engines
4 NVENC and 4 NVDEC engines per GPU for video AI, transcoding and streaming.
Graphics built in
188 RT cores and four DisplayPort 2.1 outputs for visual workloads.
Specifications
RTX PRO 6000 Blackwell 8-GPU server specifications.
The standard configuration of our RTX PRO 6000 Blackwell 8-GPU server, with NVIDIA’s published GPU specifications. We can adjust processors, memory, storage and networking to your workload.
Server: RTX PRO 6000 Blackwell 8-GPU Server
| GPUs | 8 × NVIDIA RTX PRO 6000 Blackwell 96GB GDDR7 |
|---|---|
| Total GPU memory | 768GB GDDR7 |
| Processors | 2 × AMD EPYC 9754: 128 cores / 256 threads each, 2.25GHz, 256MB cache |
| Total CPU cores | 256 cores / 512 threads |
| System memory | 3TB: 24 × 128GB DDR5-4800 load-reduced ECC (LRDIMM) |
| RAID | Hardware RAID controller |
| Boot drives | 2 × 960GB SATA SSD, read-intensive (RAID 1) |
| Data drives | 4 × 7.68TB U.2 NVMe PCIe 4.0, read-intensive (RAID 5, about 23TB usable) |
| Networking | 4 × dual-port 100GbE QSFP56 PCIe low-profile NICs (8 × 100GbE) |
| Power | 4 × 3,200W Platinum power supplies |
| Warranty | 24-month Mega Build Up next-business-day (NBD) warranty on the server; 12-month replacement warranty on the GPUs |
Configurations can be adjusted to your workload. The final specification is confirmed on your quote.
GPU (per card): RTX PRO 6000 Blackwell
| Architecture | NVIDIA Blackwell |
|---|---|
| CUDA cores | 24,064 |
| Tensor Cores | 752 (5th generation) |
| RT Cores | 188 (4th generation) |
| GPU memory | 96GB GDDR7 with ECC |
| Memory interface | 512-bit |
| Memory bandwidth | Up to 1.8 TB/s |
| FP32 | Up to 120 TFLOPS |
| FP4 AI performance | Up to 3.8 PFLOPS* |
| Multi-Instance GPU | Up to 4 instances @ 24GB |
| Encoders / decoders | 4 NVENC, 4 NVDEC |
| Max power | Up to 600W |
| Display outputs | 4 × DisplayPort 2.1 |
| Interface | PCIe 5.0 x16 |
* Theoretical FP4 performance with sparsity. Figures are NVIDIA’s published peak specifications.
How we deliver
From first call to a running system.
01
Consult & size
We review your models, users and data, and recommend the right configuration.
02
Source factory-new
New systems with genuine NVIDIA GPUs and parts.
03
Configure & burn-in
Built, updated and stress-tested before it leaves us.
04
Deliver & install
Delivery, racking, cabling and power-on by our engineers.
05
Support for life
Warranty, upgrades and expansion as your AI grows.
Compare our AI systems
From your desk to a full AI cluster.
For development
NVIDIA DGX Spark
128GB unified memory
On your desk · models of up to 200B parameters
For inference
RTX PRO 6000 Blackwell 8-GPU Server
768GB GPU memory
8 × 96GB GDDR7 · AI video, rendering and simulation
You are here
For training
NVIDIA H200 8-GPU Server
1,128GB GPU memory
8 × 141GB HBM3e · NVLink
Is the RTX PRO 6000 Blackwell good for AI training?
It can fine-tune models and train smaller ones well. For training large models the H200 8-GPU server is the better fit, with more memory per GPU (141GB vs 96GB), about three times the memory bandwidth and NVLink bridges. Most businesses use the RTX PRO 6000 to run models in production.
How many users can one server handle?
It depends on the model size, the context length and the response time you need. Tell us your models and expected users in the form below and our engineers will size the server for you.
Can one server run several models at once?
Yes. With 768GB of GPU memory, and up to four isolated MIG instances per GPU, one server can host several models or teams at the same time.
How much power does it need?
Each GPU can draw up to 600W, so eight GPUs can draw up to 4.8kW before the rest of the server, which runs on four 3,200W Platinum power supplies. Plan for dedicated circuits, two power feeds and cooling sized for the load. We confirm the exact figures for your configuration.
Do you deliver and install in Egypt and the UAE?
Yes. We deliver, rack, cable and power on the server at your office, data centre or colocation facility in Egypt or the UAE.
What warranty and support do I get?
The server comes with a 24-month Mega Build Up next-business-day (NBD) warranty, and the GPUs with a 12-month replacement warranty. Our own engineers handle diagnosis and on-site support. See our warranty services.
How do I get a price?
Answer three quick questions in the form on this page. An engineer replies within 24 working hours with a recommended configuration and price.
Get a quote
Configure your RTX PRO 6000 server.
Answer three quick questions and an engineer replies within 24 working hours with a recommended configuration and price.
- No obligation, no sales pressure
- Sized by an engineer, not a script
- 24-month NBD server warranty, 12-month GPU replacement
Step 1 of 3
Talk to an AI infrastructure engineer.
Call, WhatsApp or email our team and we’ll reply within 24 working hours.
