System forced restart every 5-20 minutes of gameplay

[zodiac678]

Hello,

 

I've played 7DTD for a couple years now and have had no significant issues.  I have a new PC and since I installed 7DTD I have crashes every 5 to 20 minutes of gameplay (decent variation).  The error is that I get a black screen and my computer reboots during gameplay.  This does NOT happen with other games.  I have tried a number of things, including windows Event Viewer, the logs in the 7DTD folder, updating all current drivers, verifying integrity of game files, updating Windows and lowering my overclock on my RAM from BIOS to defaults.

 

System:

RTX 3090

Ryzen 5950x

32 Gigs RAM

2TB SSD

ASROCK x570 Phantom Gaming X Motherboard

1200W PSU

 

Any suggestions? Thanks!

Edited July 7, 2021 by zodiac678 (see edit history)

[SylenThunder]

System crashing is a PC issue, not a game client issue. In your particular case, it is most likely the PSU reacting incorrectly to the power draw from the GPU. This has been noted to occur with several brands of PSU's on the RTX 30-series cards. The issue is only triggered a couple of games that have a high draw on both the CPU and GPU side of the system. 

 

Jayz or Linus brought this up in one of his recent videos where they talked about benchmarking PSU, and the troubles they were running into with the 30-series setups. They only ran into a couple of titles that pushed the system enough to trigger it, but basically the issue is that the PSU thinks the GPU is trying to draw way more power than it actually is, and it triggers a failsafe shutdown.

[Fox]

When you built your PC, did you test your ram using the free software MemTest86 to see if there was any errors on the sticks? This too is a common occurrence, especially with this game as it uses a lot of ram whereas so many other games will sneak by without any issues at all because it doesn't use enough to find the errors in the ram.

 

I'm also curious on the brand / model of your PSU. 1200w is more than enough for your hardware, but if you got a cheapy model, then SylenThunder could be right about it causing your issues.

[SylenThunder]

1 hour ago, Fox said:

but if you got a cheapy model, then SylenThunder could be right about it causing your issues.

There were a couple of big-name brands that had the issue as well.  Trying to locate the video now. I saw it briefly mentioned in another video they were doing when a PSU they were testing literally exploded off the side of the set.

Found where Jayz brought it up some in their AMD video a few weeks ago. Thor was the PSU they were using there that kept having the OCD protection trip on the 30-series cards.  Still looking for the vid that had the details though, because they listed a few brands that they tested and found this problem on.

Edited July 7, 2021 by SylenThunder (see edit history)

[Fox]

Of course. All brands have several bad models, even Seasonic. A lot of the time though, the prices come down very quick to go with those bad models, hence cheapy model.

 

And ya, I heard about Gamer's Nexus having a PSU explode on them (I think it was a very crappy PSU though) and Jay having issues with 30xx series cards too.

Edited July 7, 2021 by Fox (see edit history)

[Jugginator]

39 minutes ago, Fox said:

 

And ya, I heard about Gamer's Nexus having a PSU explode on them (I think it was a very crappy PSU though) and Jay having issues with 30xx series cards too.

Yeah I saw a video where he had a psu blow up on him.

 

You're also right about memory too. Sometimes you hit the silicone curse. Could be unstable with the current frequency/timings, I would run memtest twice or just lower the frequency more and/or loosen the timings. Could be the cpu not liking the ram too (i.e. the memory controller). I  have dealt with that personally a few times.

[Fox]

51 minutes ago, Jugginator said:

You're also right about memory too. Sometimes you hit the silicone curse. Could be unstable with the current frequency/timings, I would run memtest twice or just lower the frequency more and/or loosen the timings. Could be the cpu not liking the ram too (i.e. the memory controller). I  have dealt with that personally a few times.

Sometimes it's just partially faulty no matter what timings / clock settings you have, and sometimes the contacts are just dirty (unlikely in this particular scenario though). Not too long ago, I had it happen to me with some brand new Corsair Vengeance ram sticks. Both were ever so slightly faulty to the point where the memtest would only come up with errors once or twice every several tests, but was enough to mess with me in gaming. Swapped out the ram and problem went away completely.

Edited July 7, 2021 by Fox (see edit history)

[Jugginator]

7 minutes ago, Fox said:

Sometimes it's just partially faulty no matter what timings / clock settings you have, and sometimes the contacts are just dirty (unlikely in this particular scenario though). Not too long ago, I had it happen to me with some brand new Corsair Vengeance ram sticks. Both were ever so slightly faulty to the point where the memtest would only come up with errors once or twice every several tests, but was enough to mess with me in gaming. Swapped out the ram and problem went away completely.

 

Yeah sometimes computer building can be a royal pain. I got a pair of gskill ripjaws (16gb each) and my cpu or mobo doesn't let them run at stock xmp timings for 3200 (well they can, but with a 20 percent chance of crashing in 7 days which is a no no lol, tho both can run 3200 solo stable but I need 32gb). Lowering the frequency to 3133 or whatever is in that range and reducing/tightening the timings to 16 18 18 18 32 makes it stable and runs faster anyway.

 

I could run stable at 3200 but I'd have to up the voltage which I find unnecessary anyway for the barely nominal speed increase (granted i can lower the CAS timing and make the rest tighter with 3133 and stock voltage, so maybe it's faster anyway lol).

[Beelzybub]

Sounds power related. I'd check the 4 & 8 pin cpu connectors on the nw corner of your mb. Also check the 3090's 12(?) pin to 2x 8 pin dongle thingy.

[The_Great_Sephiroth]

One reason I spec my system to work. 1.2GW PSU? That's what, three or four cards in SLI type power? I ran the 6950X, 128GB RAM, 1TB NVME, 1TB WD Blue Sata, and four 1TB WD Blacks (RAID10), and an Ae5 and Blu-Ray burner at 750W with two 1070's in SLI. Yes, the SATA disks are mechanical. Also has two 150mm fans and six 120mm fans.

 

That said, has anybody considered heat? He's running an AMD CPU. Nothing wrong with that, but they run hotter than their counterparts across the isle. I have dealt with several situations where cheap cooling didn't provide enough airflow for an AMD build.

[SylenThunder]

44 minutes ago, The_Great_Sephiroth said:

That said, has anybody considered heat? He's running an AMD CPU. Nothing wrong with that, but they run hotter than their counterparts across the isle. I have dealt with several situations where cheap cooling didn't provide enough airflow for an AMD build.

Good point. I haven't had any issues at all with heat on my 3900X, but the 5950X's almost require an oversized heatsink and really good case airflow. You pretty much have to have either a NH-U12A or similar cooler, or a custom loop with a decent capacity tank and large rads. (Your typical Corsair Hydro or equivalent isn't going to cut it.)

[The_Great_Sephiroth]

I normally disagree and can prove many things Linus does wrong (I actually work in the field and have for decades), but he did a GOOD video on why you shouldn't liquid-cool unless your scenario really needs it. In fact, in the average room (air-conditioned, etc) air-cooling cooled his systems better than water cooling. I have seen this to be true in every case I have come across also. The only exception is very hot, dry environments, like a desert. Or I suppose you overclock to the extreme instead of buying the correct parts for the job, but that's a whole other conversation.

 

Anyway, maybe the OP can check or record his temps right before the crash/restart?

[Fox]

Speaking of checking connectors and whatnot... I do hope the OP has his/her 3090 using 2 separate cables to feed power to his/her video card. Just because there's 2 8-pin connectors per cable doesn't mean you should use both, especially on a 3090.

[SylenThunder]

31 minutes ago, Fox said:

Speaking of checking connectors and whatnot... I do hope the OP has his/her 3090 using 2 separate cables to feed power to his/her video card. Just because there's 2 8-pin connectors per cable doesn't mean you should use both, especially on a 3090.

Yeah, you definitely don't want to be using the same rail for that.

[The_Great_Sephiroth]

I have seen large power supplies (850W like my dad bought, against my advice) that are single-rail. In fact most nowadays appear to be going single rail. No idea why.

[Vampirenostra]

11 hours ago, The_Great_Sephiroth said:

I have seen large power supplies (850W like my dad bought, against my advice) that are single-rail. In fact most nowadays appear to be going single rail. No idea why.

manufacture expenses?))

[zodiac678]

Thank you everyone for the replies -- I really appreciate that.  I considered many of the issues you all suggested but will try some things that you have mentioned.

 

PSU is: 1200 Watt - Thermaltake Toughpower Grand RGB - 80 PLUS Platinum, Full Modular

 

I went overkill on wattage in case I wanted to do something ridiculous later and I was worried about spikes in the 3090 power draw.

 

I haven't tested MemTest86 but I actually considered it during research for this issue and have it on a USB stick right now and will test it if I continue to have issues.  I lowered my memory overclock but I will explore some more changes to the memory and the timings.

 

I'll check my power connections again to the GPU as well.

 

I've checked my CPU and GPU temps (although VRAM can become very hot and I was not sure for those temps the DDR6 can get very hot) and did not notice anything very high.  I'll try to check again and have some more reliable numbers.

 

If you all think of anything else please let me know.

 

Thank you!

 

 

[The_Great_Sephiroth]

Good luck! My only other advice is to run all your hardware STOCK for a bit. See if that helps. I never over/under-clock anything. My preference, but it has served me well over the decades.

[Beelzybub]

33 minutes ago, zodiac678 said:

If you all think of anything else please let me know.

An easy way to run a memory test is to just type "mdsched" in Windows search bar. It can take a long time, but there are settings you can change to make it quicker.

[SylenThunder]

20 minutes ago, Beelzybub said:

An easy way to run a memory test is to just type "mdsched" in Windows search bar. It can take a long time, but there are settings you can change to make it quicker.

And always run it two or three times. I can't even count the number of times a stick passed on the first pass, but was eventually found faulty later.

[The_Great_Sephiroth]

Exactly why you run Memtest86+, not the Windows junk.

[Star69]

I have pretty much the same rig as OP and have zero issues. My psu is 1000 watts Corsair platinum and power doesn't seem to be an issue. The 5950x runs about 53 degrees after an hour of playing 7 days. I'm air-cooled with a Noctua fan set up and have a huge case with plenty of case fans helping out.

 

This was my very first amd build and didn't realize that you needed to be carefull of your ram speed with amd cpu's. I bought 4200mhz ram when, in fact, the max speed for the 5950x is 3800. If I try to go past 3800mhz, my system becomes quite unstable and either constantly reboots or won't even leave setup. Leaving it at 3800 and the system runs great. So maybe take a look at your ram speed?? Just a thought....

[zodiac678]

So far, I ran Memtest86 about 4 runs and no errors came up with all RAM inserted.  I am pretty sure that people just do 1 at at time to eliminate and see which one has errors.  If I am mistaken about that let me know.  I could try more passes.

 

I checked my 8pin connectors to the GPU and I have 3 8 pin connectors, 2 were split so I had 2 from PSU to 3 on the GPU.  I tried connecting 3 8 pins from PSU to 3 8 pins to GPU and it still crashes.

 

I don't think my temperatures were excessive for GPU or CPU but again I don't have proper VRAM readings I'll have to have those up when I test again.

 

Any other thoughts?

[Beelzybub]

1 hour ago, zodiac678 said:

Any other thoughts?

I think memtest86+ will tell you which stick is bad. Could be wrong.

 

I would do stress testing with prime95 and furmark to see if I could repro the crash outside the game.

How are you monitoring temps? hwinfo is good.

What speed and timings are you running your ram at?

Did you check your power connections to the cpu?

[zodiac678]

Okay.  So I had flashed my bios I thought but I saw my bios were pretty old date so I checked to see if anything was new and they were still pretty old bios.  I flashed new latest bios and so far no problems!  Can't believe I didn't have the latest ones after all this checking everything.  So far the longest I've gone without crashing.  Thank you everyone for the help I really appreciate all of it.