NoMoreQuarantine wrote:In an effort to make a predictive ranking of GPU performance, I took the specs of all of the desktop GPUs since 2016 for both AMD and NVIDIA from here
https://www.techpowerup.com/gpu-specs/ and placed them in tables sorted by FP32 TFLOPS at boost frequency. The calculation for the TFLOPS is TFLOPS = FP32 * Frequency * 2. The last two columns show watts/TFLOPS & dollars/TFLOPS at boost frequency. The prices are the lowest prices for a new video card I could find on Amazon or Newegg. This ranking assumes that memory bandwidth has negligible impact on FAH performance.
Please let me know if you see any errors in my methodology. If there aren't any error, I think this is about as good as I can do from theoretical standpoint and I'll focus on gathering real-world data which should supersede these tables in the future.
First let me preface this by saying I'm really not trying to crap on your efforts. To the contrary, I appreciate that you are trying to find good data to help people in picking GPUs for the task at hand, and their personal preferences/desires.
But in doing so, I think you are not seeing that there seems to be no single performance metric (I could be wrong, maybe one of the FAH benches?) that indicates folding performance. All the charts showing FP32 operations are great for any use that relies mostly on FP32 operations. But the charts
https://docs.google.com/spreadsheets/d/ ... tput=html# in this spreadsheet show that FP32 is not the answer for folding metrics. I would say overall that having read quite a few threads, looked at data, etc since joining here there is not one clear cut data point that really helps in calculating PPD returns. If there was (and once again, there might be) such a data point everyone would use it. But I'm not sure there is, other than looking at data and overall averages of actual PPD returns. From there, everyone could make decisions on their choices and preferences that suit their own needs and desires.
Until proven otherwise, I think the metric that will have the most impact is going to be based on real world returns. I've been around computer geekery long enough to know that you can't always look at charts and determine the "best" for much of any particular task.
As stated before, I appreciate your efforts. I think they could help a lot of people. But I think it's a task that would take a lot of time and effort, and probably take direct user input in a similar format to the link from Overclockers. Even beyond that, the maintenance on such a spreadsheet would be crazy. It might make sense to see if it could be made without pricing and such input, but the formulas in place. That way anyone using it could input current pricing and go from there.
Oh, and one error I noticed just because it's a card I'm considering. The 1650 Super is actually a TU116 card.
