![]() I don't know if your ENB scale similiarly, but it also should On a sidenote, older ENB version had more customization like vram amount, occlusion culling. This is of course scaling for vanilla game only, i don't know if your ENB scale similiarly, but it also should.Īnd it shows that scaling between 1080p and 1440p for ENB is even higher than 1,4 i've assumed from vanilla tests! It's more like 1,5-1,6, so in theory i should get even more performance, when dropping to 1080p! But i don't get that! Even if the raw performance factor was 1,2, so only 20% faster than GTX 1070 i should still get 1,68 more FPS. So multiplying 1,4x1,4 factor i should get at least 1,9 more frames while using 5700 xt in 1080p vs GTX 1070 in 1440p. The 1,4 factor i've calculated for the same exact place as in video, while using my GPU -195 fps for 1080p/ 139 fps for 1440p =1,4. ![]() No it doesn't work like that, because we would be seeing 77% less performance in many game titles when switched from 1080p to 1440p, and we see 20-40% drop on average. Judging by the 1,4 factor between 1080p and 1440p is not true as i wrote in previous message, performance depends from number of pixels If you want good performance, then learn how to adjust quality of mod effects. The biggest change in balance possible for those videocards which are not well balanced, like very fast memory (high bandwidth, 1024 bit memory), but at same time slow gpu speed (less shader units, frequency, old architecture), then i could make more samples and less math for example, i did this for some past versions of the mod, but videocards live short time, so any generic algorithm is way better. ![]() Even if i would have this goal to increase performance with amd only cards or nvidia only cards without cheating with artificial delays, that would be few percents faster maximum and very dependent from certain model of videocard which use for testing. There is no way to make some algorithms amd friendly or nvidia friendly, code remains the same for any effect, it's not old days when developers used workarounds against certain drivers or used vendor hacks to have extra features or performance for each vendor, those times fortunately gone and i hope forever. If videocard would be something very simple, then performance would be 1.77 times lower, but it is not, doing many other things beyong just drawing pixels, except effects like ssao, detailed shadows, sun rays and maybe 4-6 more which run per pixel. So, it's 1.77 more pixels must to be rendered. ![]() Then, Judging by the 1,4 factor between 1080p and 1440p is not true as i wrote in previous message, performance depends from number of pixels, not the size of screen. This depends from what kind of code is executed mostly, the architecture of the game and how much it cpu dependent. 5700XT is not 40-50% slower than your card, i checked videos and special benchmarking site, it's 25-30% faster in average mostly by my opinion. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |