Piggybacking as you may find it useful. I built https://flopper.io and aim to do this differently that OP. We started with datasheets and individual GPUs. We're trying to get as many as we can.
Currently trying to work to have a plugin of similar to post back to get accurate benchmarks.
Yes, we're trying to get those ingested. We have quite a few but you raise a good point that inference at home is rising!
Part of the problem we have is some of those desktop GPUs don't have power and flops data. We're toying with the idea of writing a CLI to run/publish these but obviously a lot of work.
These three are measured against the old model of Llama 2 7B, but it also means the numbers are without any per-model quirks that reduce compute and memory BW usage, thus more or less generically applicable.
Currently trying to work to have a plugin of similar to post back to get accurate benchmarks.
Take a look if you're interested!