lizardpeter wrote: ↑14 Sep 2021, 22:07
Eonds wrote: ↑14 Sep 2021, 20:38
The testing which has been done by youtubers is not accurate enough to measure the difference between the two.
If this is the case, where do you get your information from? Do you test it yourself? If so, you should create some kind of database with your info or upload your results. That would be awesome.
I do a lot of research, I inspect the PCB's, I read documentation from the companies themselves, (i cant say this part but it's driver related inspection), overclocking capabilities, general architecture overview, security on the GPU, etc. You can draw a very reasonable conclusion as to which GPU has lower latency simply based on the architecture, OC, how much tiling takes place before a frame is completed, what type of power savings/how they affect latency, etc. I post a lot on my twitter/discord server of documentation/important information regarding all of these things. Another thing to consider is what you're able to disable, enable, and modify. I could take the easy way out and say simply because of the higher core/memory clock that AMD has lower latency, but it's much more complicated than that. TBRD implementation (nvidia mostly over using it hurting latency) is for people who jerk off about FPS. AMD is more open source than NVIDIA, and is much more open about everything. I think testing with scopes and probes is important but that's very expensive + you need a custom setup to do it properly to get accurate results. So yes testing is important, but I personally can't do it accurately without spending thousands of dollars, and we can draw a well supported conclusion that modern AMD GPU's have lower latency. That's my conclusion and if there's anything that i'm missing/skipping over that could turn the tides, please lmk.
Edit: I forgot to mention default VRAM timings is also important in this context + the ability to tune them + driver latency.