Crashes after upgrading from 650Ti to 1660 Super Mini ITX OC

[Pinko]

Just upgraded from a GeForce 650Ti to a Gigabyte GeForce 1660 Super Mini ITX OC (went with mini version as tower isn't that big). I had previously removed my game list so installed something quick to test it out, Battlefield BC2, and it ran perfectly fine. With the 650Ti the GPU was at 100% on High settings in Task Manager. With the 1660 ITX it was at 0.1% on High settings. So thought all was good.

Decided to play Team Fortress Classic (old but great) and the game was juddering and I noticed in the Task Manager the GPU usage went between 40-60% consistently. This game came out in 1999 and lets just say you could play it using a brick for a GPU - the recommended GPU via Steam is a "32mb+ video card". After playing for about 2 minutes the PC crashed and restarted. Tried again, and same again, crashed and restarted.

I have GeForce Experience installed and all drivers are up-to-date. Did a clean install of the drivers. I haven't installed the Aorus Engine as I have seen posts suggesting it isn't really needed - I have no plan to push the GPU, I really just wanted something that would allow me to play some modern games at reasonable settings.

Wondering what could be causing an ancient game to crash but a more recent, albeit still dated, game to run fine. I'm currently re-downloading Doom to test that out but thought I'd see if there'd be any advice from more knowledgeable folk here.

The 1660 spec states 450W PSU, mine is 460W so thought it would be fine.

Appreciate any advice.

PC Specs are below. I have made no modifications to the PC bar this new GPU.

• Dell XPS8700 (bought 2013)
  
• Windows 10 (64-bit OS)
  
• Intel Core i7-4770 CPU @3.40GHz
• 12GB RAM
• Dell 0KWVT8 Motherboard (BIOS fully up-to-date)
  
• Dell D460 AM-02 PSU (I believe this is 460W)
• Dual monitors, both 1920x1080, 60Hz.

 

[Pinko]

At least it seems like it's the PSU and not the GPU at fault. Just out of curiosity, why might it be that the ancient game that would run on a 32MB GPU cause a crash, while BF:BC2 with a recommended 512MB card not cause a crash?

Guess I'll post in the power supplies forum as that's one thing I've never replaced before.

Cheers the advice.

 

[Pinko]

What frame rate does this old game run at? If the card is running away at 600fps it might be your issue there

Steam in-game FPS shows the old game runs at ~60fps and even dropped to 32fps at one point, while Battlefield BC2 runs at ~200fps. Seems odd... I used to play the old game on an old Windows XP laptop with sweet eff all of a GPU in it with zero problems. I'm pretty sure I played this game first using a Voodoo2 12MB GPU with no problems.

Check the new card is seated properly first and the cables are properly plugged in.

I'll re-seat it and double check all cables are plugged in. I had to use a 6pin to 8pin adapter so perhaps there's something wrong there.

 

[Pinko]

Just played Doom 2016 on Ultra everything for a bit over 20 minutes. It maintained 60FPS throughout - it didn't go to 61FPS or 59FPS, it stayed exactly at 60FPS throughout - and no crashes. The old game crashes within 2 minutes so there must be something odd happening that is restricted to this old game. Gonna download Wolfenstein II tomorrow and test that.

I'll re-check the seating in the morning.

 

[Pinko]

When watching a YouTube video earlier on how to setup HWiNFO to link with Afterburner I got a crash and restart. That's never happened this PC before watching YouTube.

Check the new card is seated properly first and the cables are properly plugged in.

Re-seated the card. Same juddering in the older game. Didn't play it long enough to allow a crash as I'm not sure it's good for the system to be crashing so much.

The game might just be too old for the drivers/gpu hardware. I had a very similar problem with C&C tiberium sun, just could not get it to run.

I can see there are a few posts about older games not being compatible with nvidia cards. Bummer if that's the case here.

what is the power consumption of the new graphics card ? and what does the dell PSU have on the 12v rail ?

id look there first it might only have like 300w on the 12v or something being an oem psu

upgrade it !

PSU states 385W on the 12v.

Power consumption so far:

• Doom 2016 on Ultra Settings: Averaged around 50W, max it went to was 70W when a tonne of enemies and guns were going off.
• BF:BC2 on Highest Settings: Averaged around 100W, max 118W
• TFC (old game): Averaged 25W, max 34W

 

[Pinko]

Definitely open to any ideas!

Ran Wolfenstein II on highest settings ("Mein Leben!") for 30 minutes and all good.
Ran an MSI Kombustor stress test for 30 minutes and all good.

However, I've also had two crash and restarts just browsing the internet, which has never happened this PC so that has to have something to do with the GPU (or relatedly to PSU).

Checking Event Viewer in Windows and there have been six critical events, all just in the two days that I've had the new GPU installed. The critical events are listed as "Kernel-Power. The system has rebooted without cleanly shutting down first. This error could be caused if the system stopped responding, crashed, or lost power unexpectedly." This seems more related to the restart than the actual event that caused it and there are no clear events occurring just prior to the critical events.

I'm a bit stumped. I might just replace the PSU anyway for sh!ts and giggles just in case, no harm I suppose.

 

[Pinko]

OK maybe £40

I'd happily double that to get something solid and reliable!

The dimensions of Corsair RM PSUs are 160x150x86 and are 20mm longer than what I have. It would be pushed right up against cables coming from USBs at the top of the PC.

Would a TX650M be overkill? Maybe enough with a TX550M?

Newegg has my system as using ~340 Watts.

 

[Pinko]

Okay, so just installed new TX650M.

Played old game again. Much smoother and same FPS as the old 650Ti. However, after about 20 minutes the game crashed and the PC restarted. Previously it would have been about 2 minutes in before the crash. So the crash has been 'delayed'.

Played Doom 2016 again just to check in case and played 30 minutes on ultra and no issues. Removed the 60fps cap and was getting >100fps on Ultra.

Can't be the PSU I guess. Can't be the GPU either if Doom plays OK.

Must just be the old game being incompatible.

Thing I don't get though is, in relation to the old game that crashes the PC:

• 650Ti and Dell 460W PSU, the old game gives ~70fps and never crashed
• 1660 Super Mini ITX OC and Dell 460W PSU, the old game would give ~30 fps and crashed
• 1660 Super Mini ITX OC and TX650M, the old game gives ~70 fps and crashed

So the new PSU did something given the fps difference between the two PSUs. It does though seem that this old game ain't gonna work with the new GPU... bummer.

 

[Pinko]

Have you checked the Nvidia Control Panel settings?

The game does not appear within the list of games in the control panel. Would there be a specific global setting I should check?

Imo just dont play the old game if thats the only game that causes it or try http://dege.freeweb.hu/dgVoodoo2/dgVoodoo2.html wrapper to emulate a old direct x for it.

Or try dling and installing dx9 redistro.. https://directx-9-0c.en.lo4d.com/windows incase of missing dx files for old games, u can install this fine on win 10 even if u use dx 12.

It's my go to de-stress game - it's old but lots of nostalgia and an active enough online group to keep it alive, so I'm more inclined to go back to 650Ti if I can't get this sorted and pop the 1660 in whenever I get a new game. Tend to play through modern games once, max twice, so won't be too much hassle.

I'm running through various checks etc. So will add the dx9 to the list and see how that goes, thanks for the steer.

Can you describe the nature of these crashes in detail?

Game Crashes
> There is no early indication a crash is on its way (e.g. no slowing down, increase in stuttering etc.)
> Game playing as normal, then freezes for approx. 3-5 seconds. Whatever was happening on screen stays on the screen (no colour changes, no flickering)
> Screen then goes black (not blue)
> Screen stays black for approx. 5 seconds
> I have dual monitors (Main screen = A, secondary = B). Screen A stays black while screen B gets the Dell logo and then the swirly circles start, then the windows login screen appears on screen A. This restart is pretty quick and takes the same time as a normal restart.
> When it last happened there were some off colours on the login screen and on web pages that disappeared when I did a 'proper' restart. I don't remember these being there when crashes occurred using the old Dell PSU but could be wrong - I think this colour change has only started since I installed the new PSU. See this image for example of these off colours (e.g. purple pixels all over windows login screen - this was taken with phone so the curvy lines are just because of the phone camera).

Crashes While Browsing (This Has Happened Twice Only)
> Was browsing internet.
> Dragged a browser window from Screen A over to Screen B
> Screens froze for at max 3 seconds
> Screens went black (not blue)
> Then same as above - black for approx 5. seconds, then Dell logo, swirly circles, login screen.

check windows event logs as well.

Maybe try a windows fix, incase files are corrupt. https://www.howtogeek.com/222532/ho...-system-files-with-the-sfc-and-dism-commands/

Ran sfc /scannow, and some corrupt files were found and replaced. Maybe that might do the trick, though the game ran perfectly fine on this PC (with corrupted files) using the 650Ti, so not 100% sure the corrupt files are to blame. Will see!

Will recheck logs and see whats thrown up.

 

[Pinko]

That sounds like a GPU stability issue.

Forgive my computer illiteracy if this is a stupid question - does that suggest there could be an issue with the GPU itself? As in that it's faulty?

I was wondering if the global settings are as they were?

You can have both GPUs in your PC at the same time if you have the slots. Just put the 650 Ti in the x4 slot (it should be open-ended). Have one monitor on each. To switch gaming GPUs just go into Display Settings and set the primary monitor.

I tend to agree. What's the airflow in the case like?

Unfortunately only one space for a GPU.

It's an XPS8700 from 2013, only one fan at the back but I've also kept the side panel off this whole time.

I ran a Kombustor stress test for 30 minutes (100% GPU load throughout) and no crashes - it hit 60 degree C during the test and stayed at that temperature. The GPU is the ITX version - so only one fan.

Meant to add the power consumption based on three games:

• Doom 2016 on Ultra Settings: Averaged around 50W, max it went to was 70W when a tonne of enemies and guns were going off.
• BF:BC2 on Highest Settings: Averaged around 100W, max 118W
• TFC (old game that is causing the crashes): Averaged 25W, max 34W

 

[Pinko]

Thanks for all the suggestions / queries. I'm checking on things in between work so will get to them all as I go.

New tests done (haven't had a chance to play the game this evening though so not sure if the DDU and re-install will do the trick). Plan to try tomorrow and make a point of capturing all Windows Events at the time of the crash:

• No memory errors using Memtest86 (4 passes).
• Case is pretty clean (that's probably subjective but it looks reasonably close to brand new inside). I periodically give it a blow out. Though maybe dust buildup underneath the CPU radiator that I can't see? I can see right through the CPU radiator 'blades' (not sure what they're called) to the motherboard beneath and no sign of dust.
  
• I didn't use DDU. Took the chance that since I was staying with a NVIDIA it would be OK. I just used it there to fully wipe drivers and re-install them though just in case. Just now after installing GeForce Experience I got a bunch of green pixels randomly over 'active' parts of the screen (e.g. on buttons you'd click on, on YouTube video thumbnails). Restarted PC and all green pixels are now gone -bit odd though?

Old tests done:

• Ran sfc /scannow: Corruption found and replaced/repaired.
  
• Kombustor stress test: 30 minutes, 100% GPU load, no crashes. GPU maintained 60 degrees C throughout.
• Kombustor benchmark (1920x1080, MSI-01) = 1960. Which seems around the same as other benchmarks listed on the Kombustor score index website.
  
• CPU temps: During stress test CPU temps on CoreTemp were in the high 50 degrees C I think. Don't think they went into the 60s.
  
• CPU temps: During Kombustor stress test the highest CPU temps were 55-57 degrees C.
• Played game in compatibility mode for XP:SP3 and Windows 7, and still crashed. Can't remember if I ran them as administrator - will try that also.

 

[Pinko]

Maybe pop by steam forums for the game as they might know more as they play it more than what we would and tell em ur woes on the crashing.

Yeah good idea, have a post up on a server specific forum. One person's running the game using a 2070 Super without issues. Given my 650Ti worked fine and a 2070 Super works fine, it does seem to be something specific to my 1660 / my system setup. Will keep troubleshooting and come back with anything of use.

 

[Pinko]

I'm perplexed. At point of giving up on this.

Only since using DDU and re-installing the drivers:

• I now CAN'T play Wolfenstein on max settings anymore. I loaded up the game and it told me the GPU ran out of memory and I had to reduce the settings. Prior to using DDU I had played the game for 30 minutes on maxed out setting and it was SMOOOOOOTH. I don't believe there have been any changes to the GPU settings in NVIDIA control panel / GeForce experience.
  
• I keep getting weird green flashing pixels on active elements on the screen every now and then. It's fixed by plugging in and plugging out the cable (Display port to DVI). But this was never happening before using DDU.
  
• There is also weird behaviour with these pixels. I have dual monitors. So say A display port on GPU goes to A DVI on monitor (A-A) and B display port on GPU goes to B DVI on monitor (B-B). The green pixels were appearing on A monitor but when dragging a window to B monitor the pixels were gone (drag back to A and pixels are there again) - so they were only on monitor A. I switched the cables around at the monitor end only - so A display port now going to B monitor (A-B), and B display port goes to A monitor (B-A). So if it is an issue with the A display port (or the cable) it should now be showing on B monitor, but it doesn't. The green pixels continue to show on A monitor. Which suggests the monitor is faulty - but again, this was not happening before using DDU and reinstalling the drivers.

Prior to using DDU, when first installing the new GPU, I just popped the new GPU in and rebooted and had none of the above behaviour at all.

List of things done without resolution:

• Verified game files on Steam.
• Tried compatibility setting for ALL previous version of Windows.
• Tried running as administrator.
  
• Ran sfc /scannow: Corruption found and replaced/repaired. Crashes still occur after repair.
  
• Ran DISM.
  
• Kombustor stress test: 30 minutes, 100% GPU load, no crashes. GPU maintained 60 degrees C throughout.
• Kombustor benchmark (1920x1080, MSI-01) = 1960. Which seems around the same as other benchmarks listed on the Kombustor score index website.
  
• CPU temps: During stress test CPU temps on CoreTemp were in the high 50 degrees C I think. Don't think they went into the 60s.
  
• CPU temps: During Kombustor stress test the highest CPU temps were 55-57 degrees C.
• No memory errors using Memtest86 (4 passes).
• Case is clean - looks reasonably close to brand new inside. Don't believe dust is an issue.
• Ran DELL boot up diagnostics - all clear.
• Reinstalled drivers three times. All this did was create the green pixel problem and make Wolfenstein no longer playable on max settings.
  
• Windows is fully up-to-date.
• Purchased new PSU (went from 460W to 650W).

When checking the event logs around these crashes all I get is the below. I've Googled and advice around a dump file not being created is generally because a clean-up type of software is installed (e.g. CCleaner) but I have no such software installed:

Error - volmgr - 161 - None
Dump file creation failed due to error during dump creation.

Critical - Kernel Power - 41 - (63)
The system has rebooted without cleanly shutting down first. This error could be caused if the system stopped responding, crashed, or lost power unexpectedly.

Error - EventLog - 6008 - None
The previous system shutdown at 15:02:37 on ‎31/‎10/‎2020 was unexpected.

Error - EventLog - 1101 - Event Processing
Audit events have been dropped by the transport. 0

 

[Pinko]

Hate when a result isn't posted back up so thought I'd post in case anyone else comes a cropper like this.

I found some similar experiences with Dell XPS8700 machines and the 16** series of NVIDIA cards.

The solution appears to be power management related based on the below four steps from another post:

1. Settings > System > Power & Sleep > Additional Power Settings > Set to High or Ultimate Performance.
2. In the sleep/turn off settings > set both to Never.
3. Under the advanced power settings > Ultimate performance > change plan settings > set PCIe Link State Power Management to off.
4. NVIDIA control panel set Global settings/Power Management to High Performance

I went to test each one individually and found that turning off the PCIe Link State Power Management has resolved the issue by itself. Have had no crashes since making this change.

I wonder if the older game required so little GPU processing that the PCIe was shut off which was causing the problem? Or was it shut off and then, for some reason, came back on that it caused instability? Either way turning off the PCIe Link State Power Management seems to have done the trick. This might also explain why modern games ran OK, perhaps the PCIe was never shut off for those games because the GPU was actively being used.