TPU shows it faster by 20% at 1080p, but the gap widens as resolution goes up. The 1080 is probably bottlenecked at the backend at higher resolutions.
Edit: dumb autocorrect
TPU shows it faster by 20% at 1080p, but the gap widens as resolution goes up. The 1080 is probably bottlenecked at the backend at higher resolutions.no it is 30-35% faster.
Its just cpu bottlenect at 1080PTPU shows it faster by 20% at 1080p, but the gap widens as resolution goes up. The 1080 is probably bottlenecked at the backend at higher resolutions.
Edit: dumb autocorrect
AFAIK, Forza 6 Apex runs faster on NVidia hardware as well.
But how does that console overhead translate to PCs? It doesn't. I tried the FH3 demo after the parallel rendering patch, and the framerates could easily hit triple digits at 1440p maxed settings if I remember. So basically, the overhead associated with lower frametimes on consoles doesn't mean squat on PC, provided your machine is reasonably modern.
Supposedly NVidia supports fully bindless now remember? I don't know if it's been tested yet, but NVidia updated their resource binding tier from tier 2 to tier 3 a few months ago with a driver up date.
Actually a whole lot different, as NVIDIA cards don't suffer as badly @1080p, they stay ahead of AMD GPUs.
FuryX having better minimums doesn't really prove anything, since it's slower than 980Ti, and the same thing couldn't be replicated on any Vega card, could be just an anomaly in the test.
The behavior of Forza 6 is the same as Forza 7 @4K though.
*snip*
DF was pre-"multithreaded" patch. His claims refer to post-patch.I didn't say it did. There isn't a whole lot of rigorous data regarding FH3 but how did your framerates 'easily' hit triple digits at 1440p when a Titan X (pascal) dropped frames at 1080p on ultra settings w/ 4x MSAA according to DF's testing ? (there was even a 8x MSAA option)
I didn't say it did. There isn't a whole lot of rigorous data regarding FH3 but how did your framerates 'easily' hit triple digits at 1440p when a Titan X (pascal) dropped frames at 1080p on ultra settings w/ 4x MSAA according to DF's testing ? (there was even a 8x MSAA option)
I guess they do but I'd be interested to see what their GPU fetch performance with structuredbuffer is compared to a constant buffer. I'd also like to know the GPU fetch performance when the total amount of constant buffers goes over the 64KB limit and more importantly I'd want to know if it also causes a flush. To my knowledge Nvidia has never raised the constant buffer limits in OpenGL or Vulkan over 64KB even after the update ...
Yeah that test was before the parallel rendering patch, which nearly doubled performance in many cases. Here's a video of pre and post patch performance. As you can see, the uploader manages to be in the triple digits through most of the run post patch, and he's using a GTX 1080. Before the patch, he would get dips under 60 FPS at times.
*snip*
I guess we'll have to see. I'm surprised the folks over at Beyond3d haven't tested it yet to see if it's legit.
Huh, barring the fact that the tester in the video benchmarked different areas compared to DF I'd be interested to know what the performance is in more urban settings packed with vehicles or how the more foliage dense areas perform in a woodlands like setting ...
But Turn 10 Studios does have a valid reason when turning down multi-threaded rendering in relationship to user experience ...
Which is an anomaly because it couldn't be replicated with a Vega 56 or even 64, it's probably a bug in the test. By your standards FuryX is smoother than Vega as well. More evidence that this was an anomaly. FuryX can't hold a candle to GP102.Fury X doesn't only have the best minimums but it also has the best frametimes (lowest 1% of frametimes) so it has the smoothest experience out of all the setups ... (Fiji is SMOOTHER than GP102!
Nvidia driver 387.92, released today, claims to give a 15-25% performance improvement depending on the GPU.
It was pretty obvious this was coming. I'm interested to see how the benchmarks play out with this included.
This is exactly what I said would happen. The Nvidia supporters were up in arms saying the game was poorly coded. Wonder what they would say now.
To be fair, some driver optimizations are done to account for poor coding. We have no idea what Nvidia did to get this performance gain.
I emailed the article's author and asked him to re-test with the new driver and include the results with the original numbers.
Not impossible but I find it unlikely that Microsoft would write such bad code.
Not impossible but I find it unlikely that Microsoft would write such bad code.
This is exactly what I said would happen. The Nvidia supporters were up in arms saying the game was poorly coded. Wonder what they would say now.
This is the second "game ready" driver for Forza 7. In my tests (with the "old", 1st game ready driver) the RX 580 was 50% faster than the GTX 1060. The RX 550 was 2x faster than the GT 1030.Nvidia driver 387.92, released today, claims to give a 15-25% performance improvement depending on the GPU.
It was pretty obvious this was coming. I'm interested to see how the benchmarks play out with this included.
I thought it's a Microsoft Store exclusive game - or can Steam forums discuss non-Steam games now?Go look at the Steam forums. Lots of people are having performance issues with the game, AMD users included.
I thought it's a Microsoft Store exclusive game - or can Steam forums discuss non-Steam games now?
Yeaa. The same principle could be said for fps like bf series.DF was pre-"multithreaded" patch. His claims refer to post-patch.
Anyway, the method used by most reviewers to test GPU and CPU performance in racing games is completely retarded.
GPUs should be tested on an empty track but with high geometric detail on the track itself. CPUs should be tested with lots of AI in non-ideal conditions. Then people wouldn't have to scratch their heads about something like Project Cars 2 in wet conditions with AI getting it's FPS cut in half, because the AVX used for particle effects is limiting GPU performance in this case.
Similar behavior is observed in Dirt 4/Dirt Rally Landrush/Rallycross compared to rally stages in those games.
So now GTX 1080 = Vega 64 , while the GTX 1060 stomps all over the FuryX and RX 580. GTX 1080 Ti isn't tested with the new driver just yet though.
https://www.computerbase.de/2017-09/forza-7-benchmark/2/#diagramm-forza-7-2560-1440
So now GTX 1080 = Vega 64 , while the GTX 1060 stomps all over the FuryX and RX 580. GTX 1080 Ti isn't tested with the new driver just yet though.
May have something to do with the game's crappy CPU optimization. It hammers two cores, and barely uses the others. FH3 had the same problem until it was fixed in a patch nearly a year later, so it's surprising that FM7 kept the same CPU usage model that lead to stuttering and poor performance in FH3.
It was rather clear anomaly. I guess NVIDIA may be able to squeeze more out (judging by 1080p results).Vega's moment in the performance spotlight ended awfully quick.