Join for quick access. Enter your email below and we'll send confirmation plus sign you up to our newsletter.
By submitting your information, you confirm you are aged 16 or over,
have read our
Privacy Policy
and agree to the
Terms & Conditions. Geographical rules apply.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
You are now subscribed
Your newsletter sign-up was successful
Join the club
Get full access to premium articles, exclusive features and a growing list of member rewards.
An account already exists for this email address, please log in.
Nvidia RTX Pro 6000 Blackwell GPU spotted with 24,064 CUDA cores, 96GB GDDR7, and 600W — 11% more cores than RTX 5090
Over two years after the RTX 6000 Ada GPUs launched, leaks have finally emerged surrounding Nvidia's next-gen Blackwell workstation offerings, courtesy of a few shipping manifests and web scrapers. Cargo records from NBD show that Nvidia has been shipping two new workstation GPUs for testing and validation. Details and specifications of one of these GPUs were discovered on LeadTek's website and subsequently extracted, as shown by Harukaze on X.
From the looks of it, Nvidia is dubbing its next-gen workstation cards with a new "Pro" label. The flagship reportedly continues to fall under the "RTX 6000" brand but with an added "X" identifier; is the ProViz equivalent of Super/Ti? Historically, Nvidia hasn't been consistent with its workstation nomenclature, so it's hard to say what this suffix represents.
Shipping data from NBD indicates that Nvidia is seemingly working on two new GPUs: the RTX Pro 6000 X Blackwell and the RTX Pro 6000 Blackwell. We've extracted all the specs we could from the given information, including preliminary data from LeadTek. Both GPUs reportedly carry 96GB of GDDR7 memory, suggesting a 512-bit interface coupled with thirty-two 24Gb (3GB) ICs in clamshell mode, where two ICs share a single 32-bit memory controller.
The non-X variant reportedly features 188 SMs equating to 24,064 CUDA cores or a 97.9% enabled GB202 die. LeadTek's data mentions a 600W TGP, over twice what the RTX 6000 Ada mandates but enough to be fulfilled by a single 12V-2x6 power cable. Further details from NBD draw up what we believe are Blackwell GB20X dies for workstations. The kingpin RTX Pro 6000 X Blackwell seemingly carries the GB202-870 die, which is expected to feature all 192 SMs.
Workstation GPUs under Nvidia's RTX lineup are tailor-made for professional applications. However, in most scenarios, 96GB of VRAM is overkill unless you're looking for training or inferencing AI locally. While you wouldn't typically use these GPUs for gaming, the added memory can be helpful in tasks involved with game development, content creation, ProViz, and computer-aided design.
Depending on where you live, the RTX 6000 Ada retails for between $6,000 and $8,000. We expect a similar MSRP for its Blackwell equivalent. Given their niche use case and reduced market demand, hopefully, these GPUs will not be struck by the same shortages that plague their GeForce counterparts. With GTC just around the corner, it's reasonable to assume that Nvidia will debut these GPUs at the event.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Hassam Nasir is a die-hard hardware enthusiast with years of experience as a tech editor and writer, focusing on detailed CPU comparisons and general hardware news. When he’s not working, you’ll find him bending tubes for his ever-evolving custom water-loop gaming rig or benchmarking the latest CPUs and GPUs just for fun.
This is all well and good for a workstation GPU. They doubled the chips with a clamshell design, then used 3 GB GDDR7 modules to give another 50% boost.
Nvidia might launch a 96GB RTX 4090 as weak supply limits market growth
But is there any credibility to this recent 4090 96 GB story (even if it's an aftermarket mod instead of a real product)? It seems like it would require a memory maker to manufacture 4 GB GDDR6 modules.
This is all well and good for a workstation GPU. They doubled the chips with a clamshell design, then used 3 GB GDDR7 modules to give another 50% boost.
Nvidia might launch a 96GB RTX 4090 as weak supply limits market growth
But is there any credibility to this recent 4090 96 GB story (even if it's an aftermarket mod instead of a real product)? It seems like it would require a memory maker to manufacture 4 GB GDDR6 modules.
I'm also curious what the vBIOS situation is for these. Does the stock vBIOS just support any size and layout of GDDR6X? I didn't think the signature checking for 40-series had been cracked to allow vBIOS editing, and I doubt Nvidia would be signing 48GB or 96GB versions.
I'm kinda surprised both have 96GB, I figured the lower model would have 64GB.
I'm actually happy to see 600W. Ada clockspeeds were gimped at 300W which made the 4090 faster at some tasks which I feel should never happen when you're spending that much extra money.
I do hope they put 2x power connectors for added safety factor, but somehow I doubt it.
