Sudden FPS Drops (3-7FPS) Like in a Loop & Stutters, Then Recovery

[3cl1ps3]

I have an old Windows 10 (latest service pack) rig that is going to be completely upgraded once the top two X3D models come up to the market in Q1 2025, maybe keep it as a second system. i7 6850K overclocked to 4.4GHz adaptive all cores, 32GB DDR4 RAM 3.2GHz from default XMP I believe 16-18-18-38 2T to 15-16-16-34 1T), 7900XTX, multiple NVMe and SSD's (all still at 100% and 99% remaining life), Essence STX II discrete audio card used in a 5.1 mode, Seasonic Prime TX 1000 PSU.

Unsure what may have caused this, but I started getting serious FPS drops from let's say 60-70 FPS to 2-5 recently. This happens in both Windows games and emulators (for example Xenia and RPCS3). Steps I've undertaken to troubleshoot and resolve this:
- System file check (sfc), no errors found.
- Loading BIOS optimised defaults. This resets everything to stock - CPU to 3.8GHz, RAM to 2100 or whatever it is, standard voltages, etc.
- The 7900XTX is undervolted and overclocked. Reset to default settings. No change.
- Removed GPU driver with DDU and re-installed from scratch. This has improved things a bit in the emulators, like before that dropping the FPS/stutters almost froze the screen, with the reinstall it still does that but more like a slide show dropping from let's say 70 to 2-5 FPS (no freezing), then returning back to 70.
- Found some Reddit and other threads about disabling Gamebar, including in regedit, which I've done with no change.
- Another "solution", apparently only referring to Nvidia cards in my experience, is disabling driver-specific HDMI sources in Sounds\Playback. Did that with no change too.
- Some posts said that RivaTuner monitoring and especially GPU power on display caused these stutters. Played with the settings and removed "GPU power" from MSI Afterburner monitoring, also completely closed it when gaming, no obvious change.

No FPS drops are observed when CPU and GPU are idle, it starts when there is load and they are always repeated, like almost at the very same intervals (30 to 50 seconds, every minute). Like 70-100 FPS, down to 3-9, then recover to the original level. From monitoring CPU almost never goes to 100%, in the emulators it is mostly from 35% to 90% (according to Task Manager), average and even under load most of the time it is at the ballpark of 70-80% max. GPU power also stays around and up to a maximum of 375W, most of the time less than 250. When I check Task Manager when playing games, although there are other programmes in the background they always use not more than 5% in total, even less.
---
There may have been FPS spikes/stutters before, but I believe something in my system must have changed recently. It may be a Windows update, a GPU driver change or something else, maybe the motherboard (Rampage V Edition 10) is showing its age and the BIOS is unstable? I am yet to try with a GPU driver from let's say 6 month ago and see if there will be any change. I was initially suspecting something wrong with my CPU (for the emulators predominantly), but I've done many benchmarks and CPU always stays at 100% with no drops, never throttling, by the way this is the same in gaming (CPU package is never above 76-78 degrees maximum in a very well ventilated case). Could it be something with the GPU ot RAM?

Device Manager shows no hardware errors so I will blame it on something software-wise, at least for now. Apart from the older GPU driver, I will make some GPU benchmarks and see if there is a GPU drop too. Judging by my online searches, unresolvable FPS drops seem to be pretty common including on new hardware, but in my case these are repeated, like they happen in a timed manner, like something is pushing the system to its limits or something else happens that drops performance to the bottom for 2-7 seconds, then recover.

Any more suggestions or help, I'd really appreciate it. Many thanks.

 

[3cl1ps3]

Thanks a lot guys, this is all very helpful. Not having too much available time at the moment so will first benchmark the GPU with a few tools to see of these repeated stutters happen the very same way as in gaming. Excellent suggestion about Event Viewer, should have done this before but was so certain it has something to do with the CPU. My previous GPU in this rig was 1080Ti and with the CPU overclock for years everything was rock solid until I've got the 7900XTX. The default voltage the latter came with was pretty high - 1150mV and it was also extremely power hungry - like under extreme load/benchmarks it will reach nearly and above 550W. Now it is 1044 with 250Hz+ on memory and if not wrong from default frequency, at least 200Mhz+ main overclock. Power came down by 100+W (if not more but can't go lower as with ray tracing power simply goes through the roof). So I've had to make changes to the adaptive CPU voltage and other secondary ones I've been using for years. It was not stable with the new graphics card, especially with the undervolt. Changes were slight and the system looked stable, no WHEA errors etc., the stable for years RAM at the modified timings also looked okay. But saying that and if this behaviour persists in the GPU benchmarks I may need to look again at the Vcore and VCCSA (Broadwell-E has very weak memory controller). Though at the BIOS Optimised Defaults and stock RAM settings with stock GPU settings I've seen these FPS problems too, which makes me think it could be a more complex and hopefully - a purely software issue. Like a driver one (thanks Ormy1), the Xonar Essence STX II has a main board and a daughterboard where I am getting all 7.1 analog connections from to my 5.1 system (plus optical out also connected as the speaker system supports DTS Live), or a GPU driver one. Onboard audio is disabled in BIOS, bluetooth and any other similar services are off, no bloatwhere and viruses as far as I am aware (will go through them again), all chipset and other drivers seem to be the latest ones (although the rig itself is now ancient). My xbox-type FlyDigi 2.4GHz controller dongle is however in, which makes me think that I am utilising every single USB-A port at the back and there are "if" limitations on the motherboard as far as connections go. Actually this may be the main thing to check now. If reducing the load on the backside USB-A ports changes the situation.

Thanks once again. I will share how it goes.

 

[3cl1ps3]

So yesterday I've bench'd the 7900XTX with Furmark 2 at both 1080p and WQHD and it stayed fully stable with fluctuations within the 20-25FPS range or less, up to 403FPS and solid 2.3/2.4 frame time at 1080p OpenGL and about 170 at the second benchmark at the same resolution. This is with my own undervolt and overclock, though I've raised the voltage just a tiny bit to go from 398-399 max to 403. More than that resulted in worsened performance.

Also checked and disconnected all USB ports at the back of the case. It looks like these influence lower SATA speeds or such on the PCI-e slots, but they seem to have no relation to the FPS fluctuations I am experiencing (although switching connections to different external devices may have improved the way the BIOS treats them and I did it with this exact purpose). For some reason also, AMD Adrenaline disappeared from the taskbar, i.e. it can be opened if I look for it but did not appear in the taskbar even after restarting. So after the benchmark testing I went quickly into a few games and experienced similar performance issues like before (although like a bit less pronounced), which makes me think that I will need to take another look at the CPU/system agent voltages. I may raise a bit the VCCSA and the Vcore while also raising the CPU cache frequency a bit and see if it makes any difference. MSI Afterburner and RivaTuner (on and off) don't seem to matter, but reading comments online I am looking into other methods to display live statistics (maybe AMD's own one or HWInfo + Rivatuner or else). And by the way, unsure why HWInfo during the benchmarks (but also in gaming) has always reported power lin the ballpark of like 100W more than Furmark itself. For example Furmark will report power of 440-470W while HWInfo will indicate 550-560W. The same with GPU clock - Furmark will report up to 2700Mhz (predominantly in the ballpark of 2350-2500) while HWInfo (including in gaming) will go up to 2900-3030MHz. Both report memory frequency of up to 2735 which is what I am expecting/it has been overclocked to (2750). I am still to check Event Viewer in more detail, but if I can't solve this completely will just wait for a few more months and upgrade fully. Still have some hope playing with the CPU voltages and yesterday's benchmarks with solid frame times and framerate (380-403 the whole time) make me think the GPU itself is fine. Forgot to say, I've made a full drivers scan again (Driver Booster free version) and it looks like I am ok. For years it reported updates to the chipset drivers available, but if I went for that the system won't boot at all so I have to revert to the ones I've had/system restore. This is why I don't believe a chipset driver is to blame here while the rest of the drivers seem to be up to date.