News Nvidia RTX Pro 6000 up close: Blackwell RTX Workstation, Max-Q Workstation, and Server variants shown

[Admin]

Nvidia will offer three variants of its RTX Pro 6000 GPU. All three have the same base specs, with 24,064 CUDA cores, 188 SMs, and 96GB of GDDR7 memory, but the design and power constraints can be very different.

Nvidia RTX Pro 6000 up close: Blackwell RTX Workstation, Max-Q Workstation, and Server variants shown : Read more

 

[klatte42]

Is there no information about memory, or is it just the same as a 5090?

 

[JarredWaltonGPU]

Is there no information about memory, or is it just the same as a 5090?

There's a whole bunch of information about the memory... that I only put in the initial announcement (which is linked in the first sentence). I've repeated that here now, with some additional information. Cheers!

 

[bit_user]

Probably the most interesting change vs. previous generations is going above the 300 W barrier. To my knowledge, that's a first for their workstation cards.

 

[Li Ken-un]

Probably the most interesting change vs. previous generations is going above the 300 W barrier.

I’m going to stick with 300 W since that is still an option.

For the Max-Q Workstation Edition, the TGP gets capped at 300W. Half the power will naturally mean lower typical boost clocks for a lot of workloads, though

…that also means not rolling the 🎲 die on potential connector meltage.

 

[JarredWaltonGPU]

I’m going to stick with 300 W since that is still an option.


…that also means not rolling the 🎲 die on potential connector meltage.

Workstations don't typically skimp on materials quality. I suspect a large number of melting connectors have been more about contamination of the metal connections (along with some incorrectly inserted connectors). Have there been any reports of melted 16-pin connectors from servers? Because they're using the same connector there as the consumer cards, but I've heard basically nothing about problems from that sector.

Crazy to think about the servers shown where there are eight RTX Pro 6000 cards lined up. 4800W right there! LOL

 

[jp7189]

In servers and workstations cable routing and air flow tend to be more engineered/controlled. Technically, the 16AWG wire used on the 16pin connectors can go up to 18 amps if you're willing to have a 90C tolerance. That's 1296 watts, and plenty of safety factor IF all parts of the path are designed for it (high temp insulation, good cable routing, air flow over every part of it)

 

[Loadedaxe]

…that also means not rolling the 🎲 die on potential connector meltage.

In the server market these issues dont arise. You have IT people that actually know what they are doing and take the time to install things properly and have the proper power and cooling.

The cable melting issue were due to user error and power supplies that were not adequate.

 

[Li Ken-un]

In the server market these issues dont arise. You have IT people that actually know what they are doing and take the time to install things properly and have the proper power and cooling.

You see though… I’m not an IT team. I’m another individual installing a GPU in a typical desktop tower form factor, who likes neatly organized cables, and so has a collection of customized cables that are snipped to just the right length.

 

[bit_user]

I’m another individual installing a GPU in a typical desktop tower form factor, who likes neatly organized cables, and so has a collection of customized cables that are snipped to just the right length.

Wow, I'm impressed!

I once bought a cable sleeving kit to put sleeves over my PSU cables (before this was standard, on higher-end PSUs), but it was such a PITA that I vowed never to do it again. Luckily, the industry embraced the style and I didn't have to.

In that machine, I also tried installing special noise-dampening foam on the inside of the case (like the DynaMat stuff sold to car stereo enthusiasts), but I can't say it made much difference. It's better to simply start with a case that's designed to be quiet. Also, I positioned my PC farther away from where I sit.

 

[Loadedaxe]

You see though… I’m not an IT team. I’m another individual installing a GPU in a typical desktop tower form factor, who likes neatly organized cables, and so has a collection of customized cables that are snipped to just the right length.

And there are a lot of people just like you, I used to be one, takes too much time though and I lost interest as no one saw my pc except me and I dont need to impress myself.

But most that do, they usually do it correctly. Then there are some not so technically inclined. That isn't the fault of Nvidia. And they definitely need a crack in the knees to knock them down a notch, but this isn't it.

Still, your original comment was the "rolling the dice" on server gpu being installed that the average consumer is not going to buy, those that do buy, they know how to RTFM. 😉

 

[bit_user]

takes too much time though and I lost interest as no one saw my pc except me and I dont need to impress myself.

This. At home, almost nobody sees my PCs except me. I did spend an inordinate amount of time on a SFF build, last year. A lot of that was just on cooling, to try and keep the fan at low RPMs as much as possible. I didn't have to use such a small case, but I really liked it and didn't realize how much trouble it'd be.

At my job, all the IT equipment is standard issue (Dell) and there's no budget (nor interest) for mods. I did recently add a fan controller to a desktop that was thermally-throttling, but that's another story.

But most that do, they usually do it correctly. Then there are some not so technically inclined. That isn't the fault of Nvidia. And they definitely need a crack in the knees to knock them down a notch, but this isn't it.

I did once install a Nvidia card in a Dell workstation, at work. I can say that if it had been > 300 W, that would've definitely been an issue. So, it's nice that they've retained that option.

 

[JarredWaltonGPU]

This. At home, almost nobody sees my PCs except me. I did spend an inordinate amount of time on a SFF build, last year. A lot of that was just on cooling, to try and keep the fan at low RPMs as much as possible. I didn't have to use such a small case, but I really liked it and didn't realize how much trouble it'd be.

At my job, all the IT equipment is standard issue (Dell) and there's no budget (nor interest) for mods. I did recently add a fan controller to a desktop that was thermally-throttling, but that's another story.

I did once install a Nvidia card in a Dell workstation, at work. I can say that if it had been > 300 W, that would've definitely been an issue. So, it's nice that they've retained that option.

Not a Dell Workstation, but at one point I was in IT and we had about 150 Optiplex desktops from Dell. Discovered that there was a bug that would cause them to hard reset if you used Photoshop and did just about anything remotely strenuous (like a magic wand select).

It was a known issue, and we didn't need Photoshop on most of the PCs, but it was super irritating that Dell knew about the problem and had a fix (new motherboard revision!), but that most of the PCs were just left in place because they weren't running Photoshop. What other apps might later cause the same PCs to crash (it was a memory problem)? Who knows!

Never cared much for Dell after that...