PRESSZONE

BLOG POST By Andrew Holdsworth 13/05/2026
BACK TO MENU
NVIDIA Elite Partner logo

RTX PRO SERVER - 2026 UPDATES

We reported on NVIDIA RTX PRO servers at their launch about six months ago, so we figured it’s about time for an update. Just a reminder that RTX PRO rackmount servers are a versatile, highly configurable GPU platform, providing best-in-class performance for visualisation and AI workloads. Let’s see what’s new.

New GPU Choices

First thing to mention is that there are now more GPU options - at launch there was only an NVIDIA RTX PRO 6000 Blackwell Server Edition card available. Now there is a NVIDIA RTX PRO 4500 Blackwell Server Edition card for projects that don’t require the awesome power of the flagship card.

GPU NVIDIA RTX PRO 6000
Blackwell Server Edition
NVIDIA RTX PRO 4500
Blackwell Server Edition
Architecture Blackwell Blackwell
CUDA Cores 24,064 10,496
Tensor Cores 752 (5th Gen) 328 (5th Gen)
RT Cores 188 (4th Gen) 82 (4th Gen)
Memory 96GB GDDR7 32GB GDDR6
ECC Memory
Memory Controller 512-bit 256-bit
MIG Instances 4 4

As you can see from the above table, the RTX PRO 4500 card offers a more modest solution alongside the behemoth RTX PRO 6000.

Embedded Networking

In the original blog, we mentioned that in addition to the standard networking option based around ConnectX-7 PCIe cards, there would be additional ConnectX-8 embedded versions, offering key advantages over traditional PCIe based networking. They promise to be faster than traditional GPU servers with PCIe switches due to ConnectX-8 being PCIe 6.0 based, rather than the PCIe 5.0 ConnectX-7 NICS. Furthermore, as networking traffic doesn’t have to go up or down to a PCIe switch, but directly from the GPU to the NIC, there will be significant internal latency reductions.

That said, CPU - GPU performance is still limited by the number of PCIe lanes available. The image below shows how the GPUs and NICs are connected in the new embedded version.

Diagram showing embedded ConnectX-8 networking inside an NVIDIA RTX PRO server

This new design also offers space-saving in the system chassis as NICs are now integrated into a separate motherboard, whilst also leaving space for a DPU PCIe card.

RTX PRO server internal view with PCIe slots for ConnectX-7 network cards and DPUs

PCIe slots for ConnectX-7 QSFP cards and / or DPUs

RTX PRO server with embedded ConnectX-8 QSFP ports and dedicated DPU slot

8x Embedded ConnectX-8 QSFP ports with DPU slot

Liquid Cooling

Recently announced was the NVIDIA RTX PRO 6000 Blackwell Server Edition liquid cooled PCIe card. It’s quite a mouthful, but it’s been made possible by embedding the networking as described above, offering more space in the chassis. It also should offer significant cooling cost savings.

NVIDIA RTX PRO MGX 6U server chassis designed for high-density GPU configurations
NVIDIA RTX PRO 6000 Blackwell Server Edition liquid-cooled GPU module

Look out for custom configurable 3XS RTX PRO Liquid Cooled Servers coming soon.

The most versatile workloads

NVIDIA RTX PRO Servers excel across diverse professional workloads, delivering optimised performance for every application, including visualisation and rendering, digital twins within Omniverse virtual environments, scientific computing & HPC, and generative agentic and physical AI.

RTX PRO server accelerating real-time visualisation and ray-traced rendering workloads

Visualisation and Rendering

Use technologies such as NVIDIA RTX and DLSS 4 to accelerate workloads including AEC & BIM, CAD & CAE, Media & Entertainment. Experience the very best in real-time ray tracing and AI-powered upscaling and frame generation.

Digital twin simulation running on an NVIDIA RTX PRO server using Omniverse

Digital Twins

Use software platforms such as NVIDIA Omniverse to build photorealistic simulations of real environments with accurate physics-informed characteristics. These are ideal for designing retail stores and manufacturing facilities, and training robots.

RTX PRO server supporting scientific computing and high-performance computing workloads

Scientific Computing & HPC

Use pre-configured NVIDIA frameworks to perform data analytics and complex simulations to extract meaningful insights from your datasets. Ideal for use cases including academic research, financial services, healthcare, space and earth science.

NVIDIA RTX PRO server used for AI training and inference workloads

AI

Use pre-configured NVIDIA frameworks and software platforms such as NVIDIA AI Enterprise to train and inference AI models. For the largest scale AI models used in projects such as generative AI, agentic AI, physical AI, and AI factories we recommend NVIDIA DGX appliances.

Why choose an NVIDIA RTX PRO Server from Scan?

NVIDIA chose Scan as one of its partners for RTX PRO Servers because of our deep expertise in building GPU-accelerated systems. Our in-house 3XS Systems division has been handcrafting PCs, workstations and servers for more than 30 years. We pioneered the AI dev box back in 2016, so we have a huge amount of experience in building highly reliable systems that deliver the most performance for your budget. Here are the key reasons to buy an RTX Pro Server from Scan.

NVIDIA Elite Partner

NVIDIA Elite Partner

Scan has been an accredited NVIDIA Elite Partner since 2017, awarded for our expertise in the areas of visualisation and AI.

AI Optimised

Workload Optimised

Our in-house team includes application specialists who optimise the configuration and software stack of each system to your requirements.

Trusted by you

Trusted by you

Scan 3XS Systems are trusted by thousands of organisations for their visualisation and AI needs.

7 Days Support

7 Days Support

Our technical support engineers are available seven days a week to help with any queries.

3 Years Warranty

3 Years Warranty

3XS Systems include a three-year warranty, so if anything goes faulty we’ll repair or replace it.