Question System Lag/Stutter - CPU/Motherboard/PSU ?

[CelticPharaoh]

Hi everyone,

I have been trying to diagnose a system lag issue with my system and I'm at a point now where I can't figure out what may be causing the issue. Basically, I have noticed that at certain points when gaming or OS use, whenever I would load a program or app, there would be a lag or stutter and then everything would return to normal.

So far I have tried the following:

• Temperatures
  • All temperatures reported normal in HwInfo
    • Idle 40c-50c. Average load operation 65-70. Maxing out at 78c but never sustained (Ryzen is fun...)
    • Idle 40c core, 50c memory junction. GPU core around 70c-74c under full load, memory Junction around 90c-92c under load.
• BIOS
  • Tried using BIOS optimized defaults
  • Turned off Security Module / TPM
• RAM
  • Removed 1 stick of RAM, leaving only 1 module in A2 (motherboard specification)
  • Replaced that 1 stick of RAM with the one I removed
  • Tried slots A1 through B2
  • Replaced entire kit with a 4x8GB 1R Kingston 3600Mhz RAM (Kingston HyperX Predator RGB 32GB (4x8GB) DDR4 PC4-29000C17 3600MHz Quad Channel Kit (HX436C17PB4AK4)
  • Replaced GPU with a friend's ASUS ROG Strix RTX 3090
• Drivers
  • Updated GPU drivers to latest
  • Downloaded latest drivers direct from vendors (e.g.: AMD for Chipset, Realtek for LAN, Intel for LAN/WiFi etc.)
    • Did a fresh Windows reinstall after downloading them
• Firmware/BIOS
  • BIOS Updated to latest (4201)
  • NVME drives firmware updated using Samsung Magician
• OS
  • Tried basic startup, loading only essential services
• PSU was RMA'ed as suspected power issues
• Moving games between SSDs and HDD
• Double and triple checked all my PSU connections to motherboard, GPU to insure fully inserted and secured
• Tried connecting GPU using this wiring diagram reference. (PCIe 1 on a 1x 8-pin, 1x 8-pin unused, to 12-pin PSU connection, PCIe 2-3 on a 2x 8-pin to 12-pin PSU connection)
  • 

So I've explored virtually every avenue except replacing CPU and motherboard, neither of which I am able to currently swap in/out at the moment. Aside from that, I have a minor concern with the PSU though it may be normal. Whenever the GPU is under-load, reaching approximately 390w-400w (I am using stock, normal BIOS on my GPU) the PSU seems to momentarily click. It only really happens once when subjected to load and I haven't heard it in games that have lower GPU utilization. As an experiment, I tried limiting power in EVGA Precision to 83%, matching NVIDIA's TDP, which I believe is 350w, the clicking also didn't occur there but games didn't run as smooth. It wasn't a slideshow but it kind of felt equivalent to a car engine, bouncing on the rev limiter. Though if this clicking is an issue, perhaps I need a larger 1200w PSU even though I read from a few sources that 1000w should be ample but I recently saw YT videos and read about GPU transient spikes being 2x or even 3x their TDP! Which would mean the GPU could be hitting 800w-1200w momentarily.

There are 2 more things I want to try with the PSU but I am doubtful if they'll make any difference from what I've read. Be quiet! supplied 2x 8-pin PCIe cables instead of the usual single 8-pin cables. Now they're not exactly like the daisy-chain ones but they feed into a single, 12-pin connector, essentially meaning there are 2x 8-pin cables coming out of a single 12-pin connector on the PSU side. But I went ahead and ordered their single 8-pin cables as I had read that the 3080/3090 prefer dedicated 8-pin connectors. Though from what it looks like, its basically the same 2x 8-pin cable that plugs into the 12-pin connector but being a single cable coming out of some pins but not all 12 like a 2x 8-pin. At the very least, it will make cabling neater in the case but I am skeptical if it'll make a functional difference to the system. The other idea is to use the jumper to switch the PSU from multi-rail to single-rail mode. I would REALLY prefer not do this for safety reasons as well as protecting components but if it will make the system run smoother, it might be a risk I'd be willing to take.

So aside from all this, can anyone identify any possible causes of this? I am tempted to contact AMD and arrange an RMA with them as I said CPU and motherboard are the last 2 things I haven't swapped out but I would rather contact either AMD or ASUS and arrange an RMA than to wastefully arrange an RMA with AMD when the issue is motherboard related or vice versa.

Thanks for reading!

 

[COLGeek]

High-end Nvidia GPUs can be very picky in terms of PSUs. I would think you are okay in this regard. Is the list of components in your sig block for this system?

When you built this rig, did you perform a full, clean install of Windows (not a recycled install from a previous config)?

 

[CelticPharaoh]

High-end Nvidia GPUs can be very picky in terms of PSUs. I would think you are okay in this regard. Is the list of components in your sig block for this system?

When you built this rig, did you perform a full, clean install of Windows (not a recycled install from a previous config)?

I have read that yep. OCP tripping on some PSUs but read from a source that the be quiet! 1000w+ PSUs should be okay. I referenced the PSU tier list as well.

Yep, my sig is my system and I've done at least 2 Windows reinstalls at this point

 

[COLGeek]

I would start from a default baseline. Reset the BIOS to its default state (not the optimized setting, pure default), including memory running at 2133 (no-XMP) and test from there. Have you already done this? No tweaks, nothing. Just pure baseline default settings.

 

[CelticPharaoh]

I would start from a default baseline. Reset the BIOS to its default state (not the optimized setting, pure default), including memory running at 2133 (no-XMP) and test from there. Have you already done this? No tweaks, nothing. Just pure baseline default settings.

I have tried this yep. Both with my current RAM kit, Team Group 2x16gb dual rank 3600mhz at non-XMP/DOCP and at 3600Mhz, as well as with a Kingston 4x8gb single rank kit, non-XMP/DOCP, DOCP 3000Mhz and DOCP 3600Mhz.

All the things I've tried are in bulleted list in my original post

 

[COLGeek]

When you installed Windows, was the target SSD the ONLY storage device installed?

 

[CelticPharaoh]

When you installed Windows, was the target SSD the ONLY storage device installed?

I had my other 2 drives in there, an NVME and a HDD so 3 storage drives and then an optical DVD-RW drive.

My primary NVME drive is partitioned, 256gb for OS and apps and the rest of the space as another data partition. Each time I installed Windows, I delete the OS partition and the ones Windows previously created, Recovery and System/EFI and then create a new partition from scratch from the unallocated space left behind, creating 3 partitions, a new recovery, system/EFI and the actual OS partition for storing Windows and other apps.

One thing I was curious about when investigating my drives in the Windows Disk Management tool, all drives are basic drives and not dynamic. Should NVME SSDs be set as dynamic or they work fine in either mode?

 

[COLGeek]

When installing Windows, it is best to only have the target storage device installed. Then, connect other storage devices after Windows is setup.

For your other question, see here:

https://learn.microsoft.com/en-us/w...nt/change-a-dynamic-disk-back-to-a-basic-disk

Basic is the better choice for nearly all common scenarios.

 

[CelticPharaoh]

When installing Windows, it is best to only have the target storage device installed. Then, connect other storage devices after Windows is setup.

For your other question, see here:

https://learn.microsoft.com/en-us/w...nt/change-a-dynamic-disk-back-to-a-basic-disk

Basic is the better choice for nearly all common scenarios.

Thanks for the information and link.

I don't think its a driver issue and I would rather not remove my second NVME SSD, connected through the chipset, as that would mean I would need to replace the thermal pad and I don't have any spare with me at the moment. Disconnecting the HDD could be possible but looking at my partitions in Disk Management, they look correct (Disk 1 is my NVME under the GPU connected to the CPU)
https://drive.google.com/file/d/1-NLWfbmh_hEuxSOMC1BbMAtGCZCzt5w7/view

I uploaded a video here showing a little bit of what is going on with some explanation. Any advice would be appreciated
https://drive.google.com/file/d/1Bzkv5ZSZeT5pA06T-aqIBYVXDQNIbt1x/view?usp=share_link