[SOLVED] MSI RTX 3080 low-ish Time Spy score in comparison to others?

[turbot1984]

As above, now seen you have a 650W psu which might be the blocker. I reckon my system is pulling close to 600W overclocked, should be a bit less with your Ryzen? I would still go for 850W quality psu to get some pukka juice to your Trio.

 

[R00m99]

Guys, had the exact same problem and solved it.

look here:

https://www.reddit.com/r/RTX3080/co...urce=share&utm_medium=ios_app&utm_name=iossmf

 

[Therick04pp]

I have now concluded my tests and indeed setting my power management back to balanced decreases the score to 16k, putting it back to Ultimate Performance but with some adjustments to stop it running at 100% all the time has now increased to post over 18k results. I'm so happy that I have found the issues, as simple as it was. Please refer to my latest edit in the original post for more info

I know this is an old thread, but having read most of it, I wanted to jump in for anyone who happens to stumble on this thread with similar starting numbers as the OP. I was having similar results and found that changing Nvidia control panel to "Prefer maximum performance" netted me about 300 more points in Time spy. Also, changing "Texture Quality" from "quality" to "high performance" netted another 350 points and I have yet to find any discernable difference in graphics by using performance texture quality, over quality. My Graphics score is sitting at 18,936 with the 3080 trio x. 52c max temps, custom fan curve (about 80%). +175 core clock and +450 memory, stable in time spy for me with this result. Also running the 9700k @ 4,8 all core. +200 core or +500 memory (either one) and time spy freezes for me.

 

[ikinson]

I have a 3080fe and changing windows power plans didnt make any difference for me.

 

[Duncan Reay]

I know this is a relatively old thread but after pulling my hair out over some serious frame rate dips in World War Z, I discovered changing to High Performance Power Plan fixed it. None of my friends could replicate it either. Then I found this thread and hallelujah.

I know you guys have all been on about Time Spy and said it doesn’t affect games but boy it makes a big difference to World War Z and I wonder if other game engines are affected. I was running 1440p Ultra with average FPS around 130 FPS but the 99% values were sub 50 FPS at times creating real jerky stuttering (FYI you have to use the Nvidia performance overlay as the in game benchmark just doesn’t report these low FPS numbers for some reason). See the image linked below with the comparison of Balanced vs High Performance in the World War Z benchmark. The green numbers are the Nvidia performance overlay which picks up the drops and the white ones are the in-game benchmark which doesn’t. The final figures from the benchmark in the balanced run are all lower but it never shows those low <60 FPS numbers. The GPU usage is lower too in the Balanced run suggesting the CPU is dragging the numbers down by starving the GPU.

I have a Gigabyte Z390 Auros Pro (latest Bios F12L set to defaults except for XMP memory) also and a Gigabyte RTX 3080 Eagle OC with Intel i7 9700K (not overclocked). I haven't tried disabling Intel Speed Shift in the Bios but I am reluctant to as the link below to an AnandTech articles shows some significant gains in many apps from its ability to shift the clock speed so quickly. This must be some issue with the Balanced plan and the Gigabyte Z390 Auros Pro as so many of us have this setup. For now, I am using High Performance with CPU min set to 5% and PCI set to moderate savings.

https://www.anandtech.com/show/9751/examining-intel-skylake-speed-shift-more-responsive-processors

I will run some tests with Time Spy when I get a chance but the massive stuttering and frame drops in WWZ makes me want to go test other games.

 

[Sarge2255]

I know this is an old thread, but having read most of it, I wanted to jump in for anyone who happens to stumble on this thread with similar starting numbers as the OP. I was having similar results and found that changing Nvidia control panel to "Prefer maximum performance" netted me about 300 more points in Time spy. Also, changing "Texture Quality" from "quality" to "high performance" netted another 350 points and I have yet to find any discernable difference in graphics by using performance texture quality, over quality. My Graphics score is sitting at 18,936 with the 3080 trio x. 52c max temps, custom fan curve (about 80%). +175 core clock and +450 memory, stable in time spy for me with this result. Also running the 9700k @ 4,8 all core. +200 core or +500 memory (either one) and time spy freezes for me.

Changing graphics setting in Nvidia Control Panel is just lowering settings, so your score is a fake score.

It's like doing a benchmark on medium and comparing it to people on high.

 

[Therick04pp]

Changing graphics setting in Nvidia Control Panel is just lowering settings, so your score is a fake score.

It's like doing a benchmark on medium and comparing it to people on high.

Hahahahaha. Yup, it's fake. You caught me... Completely phony. Anybody that would set their GPU to maximize performance with no detectable negative effects to your rendering, is a fraud! Hahahaha.

My god. Peace out.

 

[BallistixOnZ490]

Its a valid issue.

People use these benchmark scores to compare cards, not just brag and compete on leader boards.

Artificially inflating your scores using reduced driver settings misleads people that spend ages googling these scores to find a suitable upgrade or replacement.

Everyone already knows that reducing driver settings improve performance, its exactly what they say they are for. Do it for your games if you want, but its completely pointless doing that to try and get more points in a benchmark.

The only way benchmarks work is if all settings are at a constant between different set ups.

 

[Duncan Reay]

Sorry for interrupting the conversation on benchmarking etiquette, but I wated to pull it back to the issue at hand with the performance in the default Balanced power plan. I thought the revelation that it affects games would have more impact since this thread has only talked about benchmark scores. I know you guys are keen to increase your Time Spy scores but I want to know my gaming isn’t affected by Gigabyte Z390 bugs as that’s my priority. Can anyone else test this World War Z poor 99% FPS in the balanced power profile mode? It was a free gave away on Epic in March last year and I assume some of you have it.

Do any of you see other games affected by this? I wonder how many games I have played since getting my RTX 3080 at launch without realising this CPU bottleneck bug in the Balanced power plan. I was getting stuttering in Ryse: Sone of Rome too and maybe it was this issue. The CPU must be constantly trying to down clock or park cores in balanced mode, even in games for some reason which it doesn’t in my friends gaming rigs. We must have enough evidence that its the Z390 Aorus Pro related to have Gigabyte investigate this? I saw @GIGA-Man from Gigabyte on the thread early on but he's gone quiet.

Also, what are your thoughts on the possible issue of disabling Intel Speed Shift in the bios I talked about above in my post? It may fix it but comes with a cost to other Windows application performance if it reverts to old school C-states to save power over the newer Speed Shift system according to Anandtech.

 

[Duncan Reay]

OK I have run Time Spy and have some interesting results. See the screen grabs of the results screen below but as expected the score in balanced power plan is down on the score with the high-performance power plan, same as most of you. What is more interesting, and I am surprised no one else noticed or shared this is the graph below the scores showing CPU clock. The balanced power plan is constantly trying to down-clock the CPU while High Perf keeps it at max clock, which is what we have suspected all along relating to disabling Intel Speed Shift to fix the balanced plan. Its blatant proof that the balanced power plan and likely our Gigabyte Auros Z390 Pro has a major bug we absolutely need Gigabyte or Microsoft to acknowledge.

So here is another image of two screen grabs showing the Nvidia performance monitor overlay. I would highly recommend everyone use this monitor in your games to see if this issue exists in them. The big red flag is the 99% FPS number utterly tanks in balanced mode. So, the 12% drop in your reported Time Spy scores does not paint the full picture. This 70% drop in minimum frame rate is the culprit for the lowered scores, not just a general reduction in performance. It’s a fundamental killing of the 99% FPS due to the CPU being throttled like crazy leading to stutter which is much worse in games than just a lowering of average FPS.

So, what now?? Do I or we submit this to Gigabyte support since @GIGA-Man has abandoned us, or just live with it and use high performance mode? Take a minute to consider all the other Gigabyte Z390 Auros Pro and RTX 30XX owners (is it just present with the RTS 30XX setup?) who don’t even realise the default power plan is killing their game performance. They need an official fix more than we do.

 

[Duncan Reay]

OK I am an idiot . I inserted images into my above posts from Google Photos and of course they showed on my browser where I was logged into Google but not anyone else’s. Long story short, my posts which would not have been very useful without the images of the performance numbers etc now have the images added and working.

Hopefully it will clarify what I have analysed, and you can take away that this is a gaming issue as well as a benchmark issue that we need a fix for, if not for all those gamers out there who don’t know why their performance is poor and stuttery and will still be using the default balanced power plan.

 

[Sarge2255]

OK I have run Time Spy and have some interesting results. See the screen grabs of the results screen below but as expected the score in balanced power plan is down on the score with the high-performance power plan, same as most of you. What is more interesting, and I am surprised no one else noticed or shared this is the graph below the scores showing CPU clock. The balanced power plan is constantly trying to down-clock the CPU while High Perf keeps it at max clock, which is what we have suspected all along relating to disabling Intel Speed Shift to fix the balanced plan. Its blatant proof that the balanced power plan and likely our Gigabyte Auros Z390 Pro has a major bug we absolutely need Gigabyte or Microsoft to acknowledge.

So here is another image of two screen grabs showing the Nvidia performance monitor overlay. I would highly recommend everyone use this monitor in your games to see if this issue exists in them. The big red flag is the 99% FPS number utterly tanks in balanced mode. So, the 12% drop in your reported Time Spy scores does not paint the full picture. This 70% drop in minimum frame rate is the culprit for the lowered scores, not just a general reduction in performance. It’s a fundamental killing of the 99% FPS due to the CPU being throttled like crazy leading to stutter which is much worse in games than just a lowering of average FPS.

So, what now?? Do I or we submit this to Gigabyte support since @GIGA-Man has abandoned us, or just live with it and use high performance mode? Take a minute to consider all the other Gigabyte Z390 Auros Pro and RTX 30XX owners (is it just present with the RTS 30XX setup?) who don’t even realise the default power plan is killing their game performance. They need an official fix more than we do.

There is nothing wrong with Balanced Mode! That is what it is supposed to do.

Hi performance more is fine as well! It is supposed to do that.

 

[Jack_Eao]

There is nothing wrong with Balanced Mode! That is what it is supposed to do.

Hi performance more is fine as well! It is supposed to do that.

Gigabyte would disagree with that when they released a bios update with the following fixes:
4. Fix CPU Vcore and power behavior

With the evidence provided in this thread alone its obvious that no, that's not how it should work at all.

 

[Duncan Reay]

Gigabyte would disagree with that when they released a bios update with the following fixes:
4. Fix CPU Vcore and power behavior

With the evidence provided in this thread alone its obvious that no, that's not how it should work at all.

Yes, I completely agree. My friends with various Asus and MSI motherboards have no issues running in the default Balanced mode with no effect on game performance. Yes, it should down clock in desktop when in light use, but in games, Windows and the hardware, even in balanced mode, should know to max out in a full screen game.

 

[Duncan Reay]

Wow, so check out the brand-new article below from PC Gamer where Microsoft are holding up their hands up about "this" issue causing low frame rates, I think. It sounds the same even if it’s to do with a recent Windows update from March 2021 which this thread predates.

Quote from the article: "The workaround is to set your power plan to 'High Performance'....."

https://www.pcgamer.com/uk/microsof...-game-mode-that-results-in-lower-frame-rates/

 

[raven74]

God bless you, i know it is an old thread but it helped me, gigabyte still has problems with intel speed shift, i9 10900k, aorus z490i, struggling for 3 months now with low gpu scores (Asus 3090 tuf 18k-18.3k graphics scores), found this thread, disabled speed shift and now 20k+ graphics scores in time spy. Made the acount specially to thank you.