NVIDIA RTX PRO Blackwell GPUs Unveiled: What You Need to Know

NVIDIA held its much-awaited keynote at GTC 2025 this morning. This is when NVIDIA unveils its latest technology for the professional, enterprise, and data center markets. Many, myself included, watched eagerly for new GPUs to be announced. While all eyes were on Jensen’s presentation as he discussed data centers and new AI technologies, NVIDIA quietly announced the new RTX PRO Blackwell GPUs on their website.

These GPUs are based on the same Blackwell architecture as the GeForce RTX™ 5000 Series GPUs we recently reviewed. While not covered in the GTC keynote, the product pages are up and provide some tantalizing details. A whole range of GPUs were announced, covering servers, workstations, and laptops. We’ll leave off the laptop for now, and instead focus on the server and workstation cards:  

• NVIDIA RTX PRO™ 6000 Blackwell Server Edition
• NVIDIA RTX PRO™ 6000 Blackwell Workstation Edition
• NVIDIA RTX PRO™ 6000 Blackwell Max-Q Workstation Edition
• NVIDIA RTX PRO™ 5000 Blackwell
• NVIDIA RTX PRO™ 4500 Blackwell
• NVIDIA RTX PRO™ 4000 Blackwell

The first thing to note is that NVIDIA has added “PRO” to the product line name. In years past, the professional cards had “Quadro” branding, which was dropped a few generations ago. This caused some awkwardness as the consumer had a brand name, GeForce, but the professional did not. Sometimes, little things like this make a reviewer’s life better.

NVIDIA has specifications for these cards listed on their website, but the first thing that caught our eye was that all of the GPUs get a bump up in VRAM. The RTX PRO 5000 goes from 32GB to 48GB, and the 4500 moves from 24 to 32. But the real winner is the RTX PRO 6000, which jumps from 48GB to a whopping 96GB! This stat alone will get many people to upgrade. 

One of the details that was confusing for us at first is that there are three different RTX PRO 6000s listed: “Server Edition”, “Workstation Edition”, and “Max-Q Workstation Edition”. This is the first time we’ve seen multiple versions of one GPU, and Max-Q is a branding for their laptop cooling system, so seeing it on a workstation card is surprising. 

What are the differences between these three? The Server Edition is made for, well, servers. It has passive cooling and relies on the chassis airflow to keep it cool. Most standard desktop cases do not provide enough airflow to keep this card cool. And not just any rackmount system will do. It will need to be a chassis designed specifically for passibly cooled cards. These tend to be very loud and are best suited to climate-controlled server rooms. Those building servers for GPU rendering will likely be drawn to this card. 

The NVIDIA RTX PRO 6000 Blackwell Workstation Edition has a cooler design similar to that of its consumer-grade sibling, the GeForce RTX 5090. This cooler has two fans that pull air from below and push it through the card and out the top. NVIDIA has been iterating on this design for a while now, and it works exceptionally well. The biggest downside is that putting multiple in a system can be difficult as the bottom card will be blowing hot air onto the top card. The chassis’s cooling system and keeping the cards spaced out can mitigate some of this, but even in the best-case scenario, only two should be put into a desktop. Add to this the fact that each of these cards draws 600 watts. With two cards, that’s 1200 watts, and the biggest desktop PSU is 1600W, leaving only 400W for the CPU, storage, networking, etc. 

This is where the NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition comes in. There are two significant differences that are really exciting. First, the cooler design returns to the design last seen on the RTX A6000 released in 2020, a two-slot blower card that pulls air in from the front of the card and blows it out the back of the case. Second, the power has been cut from 600W down to a mere 300W. Decreasing the power reduces performance, but according to the spec sheets for the GPUs, there should only be a difference of around 12%. A small deduction in per-card performance would be outweighed by the fact that four GPUs would take the same amount of power as two slightly higher-performing GPUs. 

Naturally, we will want to test the real-world performance of these new GPUs in rendering applications like Redshift and V-Ray, as well as various AI models. Our R&D team will also look closely at power draw and cooling with multiple GPUs to ensure we can provide a stable product. Sadly, NVIDIA has stated that these new GPUs will not be available until April or May, so it will be a while before we can get some benchmark data. But this announcement has us very excited to see what we will be able to do with these upcoming cards!