Right, the 5090 isn’t scaling linearly. Performance doesn’t scale linearly with # of GPUs in SLI/nvlink either.
I’m unimpressed. 33% better performance over 4090 for 27% more watts is a joke, in my personal opinion.
4090 at least had 49% better performance vs a 3090 in the most generous benchmark results but using 29% more watts
3090 was a 23% performance increase over 2080 Ti using 40% more watts 💀
The writing has been on the wall since the 2000 series that NVIDIA can’t hit marketing perf uplift targets without substantially increasing power usage. I am just not in their target market because I am not interested in any GPU that uses more than 300W stock.
The only thing I’m impressed by is that they can cool a 575W card in a 2 slot form factor, except they’ve disabled the junction temp sensor for the 5000 series, so I would say “we’ll see” what the temps are like, but actually we won’t. You’re not allowed to see!
Processors don’t usually scale linearly so it’s still impressive.
Right, the 5090 isn’t scaling linearly. Performance doesn’t scale linearly with # of GPUs in SLI/nvlink either.
I’m unimpressed. 33% better performance over 4090 for 27% more watts is a joke, in my personal opinion.
4090 at least had 49% better performance vs a 3090 in the most generous benchmark results but using 29% more watts
3090 was a 23% performance increase over 2080 Ti using 40% more watts 💀
The writing has been on the wall since the 2000 series that NVIDIA can’t hit marketing perf uplift targets without substantially increasing power usage. I am just not in their target market because I am not interested in any GPU that uses more than 300W stock.
The only thing I’m impressed by is that they can cool a 575W card in a 2 slot form factor, except they’ve disabled the junction temp sensor for the 5000 series, so I would say “we’ll see” what the temps are like, but actually we won’t. You’re not allowed to see!
https://videocardz.com/pixel/nvidia-has-removed-hot-spot-sensor-data-from-geforce-rtx-50-gpus