Seems sane to me, people spend a lot of money to get a reliable work machine, which powercable won't burn or won't overheat, they probably need to use it 24/7 and possibly more than one in the same machine
Honestly, if it can't keep it's boost / performance, I'd not buy it as a pro card that's supposed to run for prolongued periods of time (hours / days).
But yeah, if it throttles to 50% of the clocks, 50% of the performance is not such a surprise, is it? o_O
The double memory capacity also takes away 10-15 Watts of power from the core power budget, compared to a 4090.
The result is actually impressive, workstation cards do not tend to be optimized for game workloads.
I don't see the big news here....- this is normal.
It was always like this...since GTX times. Their professional cards with lots of memory and cuda cores are not optimized for games and even if in theory they should be faster in reality they are slower because they use different drivers not optimized for games like others said here. - actually looking at some old comparison from quadro and GTX cards the results from the 6000 are impressive compared to the 4090!
The fully enabled block enables more CUDA cores which in turn increases parallel operations. In specific computational applications it will smoke the 4090 even at lower clock speeds and voltage. Its also capable of producing certifiable results. No one should be buying one of these intending it to be for gaming as a primary use case.
It uses the same 16-pin power connector, so not entirely certain what the point of this comment was.
I'm certain people installing an RTX 6000 will know how to properly connect a power cable.
It uses the same 16-pin power connector, so not entirely certain what the point of this comment was.
I'm certain people installing an RTX 6000 will know how to properly connect a power cable.
Guess it's like the next Titan but with training wheels.:D
Once nVidia have saturated the RTX 6000 pro market they'll start dumping these on us. Think of an amount you'd feel comfortable paying for a GPU and multiply that by 10.
ADA 6000 uses 16-pin Edit: Also, that second picture you posted is of a RTX 6000 from 2016, you can tell by the 4608 CUDA cores Edit 2: And it's clear that the first picture is of a A6000 from 2020, like i speculated earlier, as that is the same picture used for that GPU many places https://www.nvidia.com/content/dam/en-zz/Solutions/gtcf20/a6000/nvidia-quadro-a6000-closeup-og.jpg This is the RTX 6000 ADA which clearly shows the 16 pin https://i.ibb.co/zX5SB6J/image.png per, nvidias website https://www.nvidia.com/en-us/design-visualization/rtx-6000/