It's good that you guys decided to cover this topic and spread the awareness, because it's really unknown fact that additional sensor polling actually cost cpu performance👍
Ha, cameo! Nice video bud, was a nice touch adding ShadowPlay recordings as well as ReLive. I always knew ShadowPlay recordings had a bit of an impact, but ReLive having no impact is a very nice surprise.
This has been a developing topic the last couple of days and I wanted to look into it myself a bit more and share my findings with you all. Share your thoughts and experiences down below.
I have Afterburner, RTSS, HWInfo64 using a lot of sensors and just got myself a 9800X3D going from Intel. I did some tests now and everything will be off during actual gameplay until this is resolved. HWInfo64 also make the games stutter. Many thanks for the Video, and obviously many thanks to DannyZ as well.
I'm on holidays far from home and don't spend that much time messing around on my PC, but I came across this topic and cannot wait to go back and test it. I thought I was going paranoid but felt more stutter since I upgraded form my 5909X to a 9800X3D but apparently this might be it. Thanks for another great video and the dedication you put.
Great video, appreciate your testing! Gonna test it for myself in Cyberpunk PT and Alan Wake PT on my 7700x/4080 system. Might do RT Ultra on Cyberpunk and High settings on Alan Wake 2 too, just to see what impact (if any) that has. Figured it'd be interesting to see how a lower tier CPU/GPU combo performs with Afterburner on vs off. Have a few theories on that, but they're all wildly speculative so I'll wait til I have some actual data.
@@TerraWare Yeah, the question I have is whether or not starting from a lower base framerate means those 1% and .1% lows aren't as dramatic in terms of percentage difference. That's my hypothesis, but it's not exactly a controlled experiment lol.
I had a similar issue a while back, I commented about it on Mostly Positive Review's video. It has made me a bit paranoid about the power overlays since then. I have also seen some outlets reporting that Nvidia cards have had worse 1% lows and frametimes in some recent games like Stalker 2 so I wonder if some outlets are running into this same issue without knowing. More people should be aware of this. Maybe something Nvidia can look into fixing at some point since afterburner is so necessary on Nvidia cards for all the functions such as undervolting/OCing and the overlay.
It's not just afterburner tho. Windows, steam, msi, nvida app...all of these hook into low level apis in your system...then you got hwinfo, xbox game bar, and probability rtss. Just with no extra software and everything default, we have 3 software suites recording gameplay alone
It's a shame the last time I recorded gameplay was back when I had an Nvidia card, as it seems like the ReLive does a great job in terms of performance.
Yes it's pretty good. On the XTX anyway. May have to do a comparison with 6800XT to see how they differ and what the upcoming 9700XT will be or whatever stupid name they are going to call if it's not that.
What a nasty bug. I recently upgraded my system with a focus on maximizing lows. I didn't get the results I expected but it was still a noticeable difference so I figured it was fine. Power consumption is something I track as well and I'm looking into undervolting my GPU to improve upon it. Tragic. I guess I just run multiple tests for the same benchmark.
Interesting! The one thing I had noticed when running game capture on both gpu's however through OBS was the XTX was limited vs the 4080. At 1440p using a high capture was no problem. At 4k using a high capture the XTX would crash. I had to lower the XTX to use a lesser quality capture at 4k to stop crashing. I then started to use relive for 4k capture and that was fine at its highest capture but noticed the file size was much lower compared to OBS. I guessed relive was just running a lesser quality capture as the control settings on capture are limited. Shadowplay is also limited on capture but my file size was generally larger so I assumed yet again I was using a more taxing capture on the 4080 than the method used on the XTX... Ideally I decided running benchmarks whilst using internal capture was a bad idea as the performance hit was sometimes hard to equate so kinda decided to end it until I set up a dedicated capture system..
@TerraWare have you ever looked into the impact of SAM? I wonder if that would have any impact on something like this? I'm gonna assume you have it enabled..
could you test nvidia app performance overlay and see if this also has a similar impact, this would help narrow down if the NVAPI is at fault or if rivatuner statistics overlay is at fault for polling to much
9800x3D paired with rtx 4090- 3D MARK Night Raid without MSI and NVIDIA APP 90566 points, while using both i got 85230 even if i disable POWER PERCENT! got rid of them
a saw i video where a youtuber (idk if i can write his channel, i'll don't do that)) said this started to happen after nvidia launched the new experience app then msi tried to fix it but the problem still persist, would be interesting to see a test on old driver before new app design and the old version of msi on screen display if you have time ps. 7900xtx have a lot of l3 cache maybe it help to mitigate the latency
Great video, thanks. Is this also broken when using HWiNFO64 (with Rivatuner) for the power and something extra like tracking the core speed / usage etc? Would be nice if you could test that as well.
While I haven't tested that going over what the creator of MSI AB said it has something to do with having to make API calls to Nvidia that can effect things. I'd imagine it wouldn't make much of a difference but this is just speculation. GPU Power isn't so bad. Percent Power is quite heavy and I'd make sure that is disabled.
@@TerraWare I gave it a try earlier with CapFrameX for capture and my Rivatuner overlay (using hwinfo64 for some sensors) gets worse 1% / 0.1% lows by 20-30 FPS (in some games it's less - the 20-30 is in high fps games) when I turn on hwinfo for the sensors. I turn it off and I get back those lows. Seems everything is broken. I am not using the MSI Afterburner sensors at all. The test was done on Cyberpunk sprinting for a minute or so same location. Curious if you get the same issue. 9800X3D / 4090 for reference.
As much as I'd like to I don't have any Intel CPU's. I don't have anything against them just so happened that Ryzen upgrade path was a bit easier to upgrade to. I have heard from people I trust that these same issues are smaller on Intel which may be due to ecores. Just what I heard.
It only happens on AMD CPU? My mouse is 1K polling rate. I did watch a couple videos they didnt mention it was a Ryzen specific thing but I don't know anything about it.
@TerraWare 8k polling is hard on every CPU. But x3d seems to particularly not play nice with it. 1k, 2k and even 4k should be fine. But also it's somewhat game depends. Older games are a big no, no matter what hardware.
well if you add only power% off and power on with shadowplay capture that would be great. We need to know which one is more impact on performance. %power or power. thanks for the video.
@@TerraWare please do. Honestly im not sure what you tried lmao. The issue was about gpu sensors in afterburner and you started testing relive vs shadowplay ,I dont get it :D
@@vyathaen I tried what I set out to do but since Shadowplay was also impacting the 0.1% lows I pointed that out as well because I felt it was relevant. Than tested all that again on Radeon out of curiosity.
una 4090 lavora meglio coni ntel, esmpio 14900k. nessun problema con questi programmi saluti. ho avuto anche amd con 4090 ma non avevo la stessa fluidità. ciao
While you are at it, please test the performance difference between mouse polling rate 1000hz vs 250hz and Nvidia app with filter overlay on and off. There is so many fps killer that we often neglect or not aware to which invalidate all the benchmarks in the world. I'm surprised AMD or Nvidia not even addressed anything about the issue to consumers or to public wicis stewpid in my opinion because otherwise we coulda have made better decisions. 🤧🥲
It's good that you guys decided to cover this topic and spread the awareness, because it's really unknown fact that additional sensor polling actually cost cpu performance👍
Indeed. Everyone ads their discoveries and the bigger picture starts to reveal itself.
Ha, cameo!
Nice video bud, was a nice touch adding ShadowPlay recordings as well as ReLive. I always knew ShadowPlay recordings had a bit of an impact, but ReLive having no impact is a very nice surprise.
Yeah a ReLive was a nice surprise. Probably deserved its own video
This has been a developing topic the last couple of days and I wanted to look into it myself a bit more and share my findings with you all. Share your thoughts and experiences down below.
I have Afterburner, RTSS, HWInfo64 using a lot of sensors and just got myself a 9800X3D going from Intel. I did some tests now and everything will be off during actual gameplay until this is resolved. HWInfo64 also make the games stutter. Many thanks for the Video, and obviously many thanks to DannyZ as well.
It's not just monitoring...msi is garbage now. It's so buggy I wont install it any more.
Thanks for the video.
I'm on holidays far from home and don't spend that much time messing around on my PC, but I came across this topic and cannot wait to go back and test it. I thought I was going paranoid but felt more stutter since I upgraded form my 5909X to a 9800X3D but apparently this might be it.
Thanks for another great video and the dedication you put.
Thanks, let us know what you find.
that's why I love this channel
Great video, appreciate your testing! Gonna test it for myself in Cyberpunk PT and Alan Wake PT on my 7700x/4080 system. Might do RT Ultra on Cyberpunk and High settings on Alan Wake 2 too, just to see what impact (if any) that has. Figured it'd be interesting to see how a lower tier CPU/GPU combo performs with Afterburner on vs off. Have a few theories on that, but they're all wildly speculative so I'll wait til I have some actual data.
Power Percent seems to be what costs the most. Everything else that I use anyway isn't that big of a difference.
On Radeon it doesn't seem to matter.
@@TerraWare Yeah, the question I have is whether or not starting from a lower base framerate means those 1% and .1% lows aren't as dramatic in terms of percentage difference. That's my hypothesis, but it's not exactly a controlled experiment lol.
I had a similar issue a while back, I commented about it on Mostly Positive Review's video. It has made me a bit paranoid about the power overlays since then. I have also seen some outlets reporting that Nvidia cards have had worse 1% lows and frametimes in some recent games like Stalker 2 so I wonder if some outlets are running into this same issue without knowing. More people should be aware of this. Maybe something Nvidia can look into fixing at some point since afterburner is so necessary on Nvidia cards for all the functions such as undervolting/OCing and the overlay.
I can see people not being aware. I really wasn't.
It's not just afterburner tho. Windows, steam, msi, nvida app...all of these hook into low level apis in your system...then you got hwinfo, xbox game bar, and probability rtss. Just with no extra software and everything default, we have 3 software suites recording gameplay alone
I saw that dude’s video and was hoping you’d cover this issue. What a crazy glitch.
It's a shame the last time I recorded gameplay was back when I had an Nvidia card, as it seems like the ReLive does a great job in terms of performance.
Yes it's pretty good. On the XTX anyway. May have to do a comparison with 6800XT to see how they differ and what the upcoming 9700XT will be or whatever stupid name they are going to call if it's not that.
What a nasty bug. I recently upgraded my system with a focus on maximizing lows. I didn't get the results I expected but it was still a noticeable difference so I figured it was fine. Power consumption is something I track as well and I'm looking into undervolting my GPU to improve upon it. Tragic. I guess I just run multiple tests for the same benchmark.
intresting ... 👍
What about the amd or Nvidia metrics??????
Interesting!
The one thing I had noticed when running game capture on both gpu's however through OBS was the XTX was limited vs the 4080.
At 1440p using a high capture was no problem.
At 4k using a high capture the XTX would crash.
I had to lower the XTX to use a lesser quality capture at 4k to stop crashing.
I then started to use relive for 4k capture and that was fine at its highest capture but noticed the file size was much lower compared to OBS.
I guessed relive was just running a lesser quality capture as the control settings on capture are limited.
Shadowplay is also limited on capture but my file size was generally larger so I assumed yet again I was using a more taxing capture on the 4080 than the method used on the XTX...
Ideally I decided running benchmarks whilst using internal capture was a bad idea as the performance hit was sometimes hard to equate so kinda decided to end it until I set up a dedicated capture system..
It can be a bit taxing on .1% lows especially but am currently waiting on HDMi 2.1 card
@TerraWare have you ever looked into the impact of SAM?
I wonder if that would have any impact on something like this?
I'm gonna assume you have it enabled..
Nvidia has so much more CPU overhead than AMD, Hardware Unboxed talked about this years ago
could you test nvidia app performance overlay and see if this also has a similar impact, this would help narrow down if the NVAPI is at fault or if rivatuner statistics overlay is at fault for polling to much
it had a big impact on performance for me atleast... ~10% every game
9800x3D paired with rtx 4090- 3D MARK Night Raid without MSI and NVIDIA APP 90566 points, while using both i got 85230 even if i disable POWER PERCENT! got rid of them
a saw i video where a youtuber (idk if i can write his channel, i'll don't do that)) said this started to happen after nvidia launched the new experience app then msi tried to fix it but the problem still persist, would be interesting to see a test on old driver before new app design and the old version of msi on screen display if you have time
ps. 7900xtx have a lot of l3 cache maybe it help to mitigate the latency
Great video, thanks. Is this also broken when using HWiNFO64 (with Rivatuner) for the power and something extra like tracking the core speed / usage etc? Would be nice if you could test that as well.
While I haven't tested that going over what the creator of MSI AB said it has something to do with having to make API calls to Nvidia that can effect things. I'd imagine it wouldn't make much of a difference but this is just speculation.
GPU Power isn't so bad. Percent Power is quite heavy and I'd make sure that is disabled.
@@TerraWare I gave it a try earlier with CapFrameX for capture and my Rivatuner overlay (using hwinfo64 for some sensors) gets worse 1% / 0.1% lows by 20-30 FPS (in some games it's less - the 20-30 is in high fps games) when I turn on hwinfo for the sensors. I turn it off and I get back those lows. Seems everything is broken. I am not using the MSI Afterburner sensors at all.
The test was done on Cyberpunk sprinting for a minute or so same location. Curious if you get the same issue. 9800X3D / 4090 for reference.
This was picked up by frame chasers year and years ago
Would you be able to do a benchmark correct without gpu pwr so that the results come out correctly 9800x3d vs 13700k/14700k or 13900k? Great work👍
As much as I'd like to I don't have any Intel CPU's. I don't have anything against them just so happened that Ryzen upgrade path was a bit easier to upgrade to.
I have heard from people I trust that these same issues are smaller on Intel which may be due to ecores. Just what I heard.
Better recording quality with greater FPS loss. The ideal is to film as you did in this video or use an external capture card.
True. Am waiting on a capture card.
X3d also shits the bed with high polling rate mice, idk if its with keyboard also but 8k mouse will make it stutter hard.
It only happens on AMD CPU? My mouse is 1K polling rate. I did watch a couple videos they didnt mention it was a Ryzen specific thing but I don't know anything about it.
@TerraWare 8k polling is hard on every CPU. But x3d seems to particularly not play nice with it. 1k, 2k and even 4k should be fine. But also it's somewhat game depends. Older games are a big no, no matter what hardware.
Does Intel's PresentMon tool also affect performance while monitoring power?
Bro my 1% and 0.1% show me under 10fps why something wrong?
What about CapFrameX?
I believe it makes the same API calls.
well if you add only power% off and power on with shadowplay capture that would be great. We need to know which one is more impact on performance. %power or power. thanks for the video.
Percent Power has the most impact, GPU Power is quite small. Check 3:07 Left is with both Power% and GPU Power OFF, right has only GPU Power ON.
@@TerraWare sorry i was sleepy when watching video thanks for the video again. i will keep my power on and power percent off.
@@ismailornekol It's ok.
does using RTSS overlay instead of Afterburner (disable all sensors in afterburner) resolve the issue?
Not sure. Didn't try that.
@@TerraWare please do. Honestly im not sure what you tried lmao. The issue was about gpu sensors in afterburner and you started testing relive vs shadowplay ,I dont get it :D
@@vyathaen I tried what I set out to do but since Shadowplay was also impacting the 0.1% lows I pointed that out as well because I felt it was relevant. Than tested all that again on Radeon out of curiosity.
Can I use the amd Radeon software or will have the same issue ?
You should be fine. It's what I am using to capture Radeon footage. At least with my card there doesn't appear to be much of a cost.
una 4090 lavora meglio coni ntel, esmpio 14900k. nessun problema con questi programmi saluti. ho avuto anche amd con 4090 ma non avevo la stessa fluidità. ciao
Nvidia fan boi makes a video... jesus
okay i see the afterburner... bye iam out...this video can not be good...
Oookk
While you are at it, please test the performance difference between mouse polling rate 1000hz vs 250hz and Nvidia app with filter overlay on and off. There is so many fps killer that we often neglect or not aware to which invalidate all the benchmarks in the world. I'm surprised AMD or Nvidia not even addressed anything about the issue to consumers or to public wicis stewpid in my opinion because otherwise we coulda have made better decisions. 🤧🥲
Optimum did a review of an 8k polling mouse and it had a massive hit to performance. More than any benefit the higher polling could provide.
there is already a temp fix for the nvidia app in the current update. overlay is off by default now. they are working on it.