Originally Posted by gamervivek
It's the opposite if you look at the 1080p results. You wasted keystrokes in reiterating your point by taking my reply line by line.
What? No, it's not the opposite. The 980 system is clearly outperforming the 390X rather massively at 1080p with Gameworks on and off and that relative performance change between Gameworks on and off that you're using as the sole point of your argument occurs at all resolutions (showing that CPU limitations don't play a role in that discrepancy).
This completely invalidates any notion of the 980 somehow being the one with higher CPU overhead due to the Gameworks off results having a lesser advantage over the 390X. What you're doing is beyond cherrypicking, it's flat-out ignoring direct evidence to the contrary and arguing some vague point I never made. I commented on a CPU limitation and you're just making incomprehensible points that have nothing to do with it.
If you think that gameworks/tessellation takes less of a hit on the maxwell cards, well that's the crux of the issue and you conceded the point.
How is tesselation taking less of a hit on Maxwell cards the crux of the issue of AMD's CPU overhead... with Gameworks off? And what point did I concede, exactly? I simply demonstrated that your interpretation of the results as Nvidia somehow being the one with a CPU overhead issue is incorrect.
Secondly, if 390 is CPU bottlenecked and still takes more of a hit with gameworks, then it concedes the issue ever more.
As I said, I don't know what it is I'm conceding exactly, but your interpretation of the results once again eludes me. Benchmarks results, especially of framerate, are calculated as either averages or presented as minimums. The average FPS figure isn't a chart of peaks and lows, but a single number derived from all recorded data points including both GPU and CPU-limited points. Heavy CPU limitations affect the average, but don't comprise it in its entirety. Relative and absolute loads on both the GPU and CPU are constantly changing in video games.
The minimum isn't always in line with the actual worst frame-times experienced as it itself is an average of the slowest period of frame-pushing during the selected benchmark sequence. It doesn't denote irregular peaks and lows nor pure CPU restrictions on its own either. Gameworks features are specifically handled by the GPU. Additional demand placed on the GPU will affect performance. Even on a game where one system is experiencing frequent and heavy CPU restrictions, it is rarely 100% the case that every frame is CPU-limited therefore poorer GPU performance will still bring down the average framerate or even the minimum.
Thirdly, if it's taking that hit across all resolutions then it concedes the issue even further.
No, that hit across all resolutions invalidates your
point that rested purely on the occurrence at 1920x1080 as being indicative of the 980 hitting a CPU limitation instead of the 390X even though you blatantly disregarded the same occurrence at all (including more GPU-intensive) resolutions, disregarded the obvious reason (being that Maxwell performs the calculations better), and disregarded the 390X itself being the one heavily underperforming at the lower resolutions.
And we haven't even started looking at the Fury cards performance.
PClab covered it; CPU limitation, according to their results. Regardless, I'm not sure what you'd expect to see with the Fury X. It's had a lion's share of low resolution scaling indicating that it, being even more powerful than Hawaii, is even more likely to be impacted by AMD's factually-occurring CPU overhead issues.Edited by Serandur - 11/10/15 at 10:06am