Maximus X Apex NVME causing frame time issues (Sam 960 Evo)

[SMB_92]

Hi folks,

I've been experiencing stuttering issues since I got my 8700k and Apex board, but only on games on my NVME Samsung 960 Evo drive (500gb). I have it installed on slot 0 of the DIMM.2 addin. I was able to eliminate every possible (at least known) other causes of stutter during the last 2 weeks of investigation, but it only dawned on me to move the games to another drive today lol (namely a sata 850 evo 1tb.)

The frame time issues are gone on the 850 evo. I never had a problem with this on the Z270 rig I was running prior. Hence why I am posting here. FYI I am using afterburner to monitor the frame times.

I can't find any solid info on where the DIMM.2 slot is routed via PCI E but Crystal disk shows good read write but magician shows about half IOPS or less. I have 1 1080ti in the first slot, a wireless card in PCIX4_3 (Asus ac1900 card) and all 4 sata occupied.

Any ideas what causes this problem with the DIMM.2? I have reseated the drive, updated firmware and driver (using Sammy driver) to no avail. Load times are also much longer in the NVME despite the good reading, and he monitor is showing temperatures as good (think the highest I saw was 60c).

[Menthol]

I have no issues with a 960 Pro on my Apex, 60c is a little warm and may be throttling, I have a 90mm fan on my DIMM module keeps the 960 very cool, 27c idle but that probably isn't your issue
Both M.2 slots on the DIMM.2 card run off the PCH pcie lanes
Is the OS a fresh install on this board, you didn't transfer drive and OS from another build or a cloned OS instead of fresh install?
Can you return the drive for replacement or refund and get a new drive?

[SMB_92]

The drive would still be under warranty, and my primary install is on a transferred system (I let windows 10 do it's jazz after installing, removed all the old leftover stuff myself and reinstalled all the new drivers). I do have a couple other oddities, Aida64 shows L2 cache as being half the bandwidth the online reviews show, and the ring bus although set to max of 4.8ghz never gets to 4.8 (always hovering somewhere under) but I am using speedstep and stuff so suppose that's nothing. Haven't noticed any issues at all apart from this.

Suppose I could try a fresh OS, see if that does anything. I had wondered if there was some other power saving features unique to the NVME (disabled it all for regular drives) that I couldn't get hold of. Or perhaps latest security patches for meltdown or spectre have this effect? Do you happen to have those active? I've fine tuned my OC as well to see if there was any effect, managed to shave a few mvolts in the process, but I've not touched PCH related stuff (that I am aware of)

Suppose I'll try the new OS real quick, not much left to do but bite that bullet. I'll post back the outcome of that.

[Carlyle2020]

Hello there,

testing lower temps by blasting the DIMM with air and doing a check is less time consuming than a fresh install.
60c on the temp sensor did worry other overclockers. They said the internal controllerchip would be far hotter than the sensor which ought to be positioned on the Memory Chip somewhere.


Regards
Carlyle

[SMB_92]

I seem to have not posted the above post earlier as I was at work. Anyhow apologies if double post is frowned on.

I've just tested the frame times again with new cooling solution as above and new OS on a 860 drive. Same result. Max temp recorded on drive (sensor 2, controller) is now 57 with those fans. I have even swapped the slot it was in.

As this issue did not occur on my MSI Z270 rig, and that was the heat shield known as heat trap lol, and issue only occurs on NVME, I am suspect on the board or bios. I don't know enough about NVME to say if the drive can cause this somehow.

[P-Dub]

What are you using to measure frame times? Is it something obvious that you can see while gaming?

The reason I ask is because I recently got a X Hero/8700K/960 Pro NVME build and it would randomly pause the frame during a game and stutter the sound for a split second. It turned out to be the front panel USB port(s). It's possible that a power option change to USB selective suspend might have fixed the issue, but just removing whatever was plugged in definitely fixed the issue for me.

My NVME drive also gets up to the mid to high 70's and sits around 50-60 Celsius. I'm talking about the controller (temperature 2 in hwinfo), of course.

[SMB_92]

I was monitoring both sensors thru HWinfo, one never went higher than 40ish the other was in the 60s, but I didn't actually test a full saturated load. I did this and after a while I got the second sensor to read 105c. That was shocking (10 pass crystal test).

I found an old Gskill ram cooler, took the 50mm fans off it and mounted them to the dimm.2 with the included bracket. I ran extensive tests with before I went to bed, the max temp being recorded now by the second sensor was 68c. The other stayed 40ish at most.

I have heard 70c makes it throttle. However I cannot see how this would cause frame time spikes as it does, it shouldn't be throttling in such a way that would cause that at least in my mind it shouldn't. I didn't get to test the frame times again before bed, but I'll be testing this first thing when I get home.

I have a 250gb Samsung 860 evo ready to install a new OS for testing as well. As a last measure as well, I changed the slot the drive was, so now it is closer to the Ram side was fans concentrated over the top. I will purchase a heatsink for it also (probably ek)

[SMB_92]

I was monitoring both sensors thru HWinfo, one never went higher than 40ish the other was in the 60s, but I didn't actually test a full saturated load. I did this and after a while I got the second sensor to read 105c. That was shocking (10 pass crystal test).

I found an old Gskill ram cooler, took the 50mm fans off it and mounted them to the dimm.2 with the included bracket. I ran extensive tests with before I went to bed, the max temp being recorded now by the second sensor was 68c. The other stayed 40ish at most.

I have heard 70c makes it throttle. However I cannot see how this would cause frame time spikes as it does, it shouldn't be throttling in such a way that would cause that at least in my mind it shouldn't. I didn't get to test the frame times again before bed, but I'll be testing this first thing when I get home.

I have a 250gb Samsung 860 evo ready to install a new OS for testing as well. As a last measure as well, I changed the slot the drive was, so now it is closer to the Ram side was fans concentrated over the top. I will purchase a heatsink for it also (probably ek)

[SMB_92]

Using Afterburner for frame times. I have disabled all power saving not related to CPU speedstep, including LPM, ASPM, USB suspend, but my woes were confirmed when my 850 drive did mot have the same problem with the same game.

I decided to talk to Samsung, and they seem to think it is a faulty drive, but suggested I try a secure erase and over provision. The drive was only up to 380gb and I didn't go any further for performance sake. I am about to try this disk again after secure erasing, will report back.

[SMB_92]

Secure erase did not solve the issue. I played Skyrim Special edition and the frametimes where okayish, spikes to the 30s and 40s, but still spikes where the 850 had none.

Seems to me it's an issue with IOing the drive or something like that. I don't see how the drive could be faulty when everything else shows up good I.E the benches are as I'd expect