Is my PSU dying?

[Sm00thSm0k3]

So to cut a long story short, roughly 3+ weeks ago I dropped my EVGA GTX 1070 for a MSI GTX 1080 Gaming X.

Ever since the upgrade I've been experiencing some stability issues in games like Witcher 3 (which I'm playing a lot lately).
The game either just randomly closes without any errors/warning messages or the game would crash with Windows giving me a "Witcher 3 has stopped working" error.

My system is as follows:

3770K
ASRock Z77 Formula OC
24GB GSkill 2400Mhz RAM
MSI GTX 1080 Gaming X
Samsung 840 Pro and 750 Evo SSD's
Corsair HX750 (probably around 5 years old)

Since I never experienced these problems with the 1070 I thought at first it must be a software issue i.e. the updated drivers or Witcher 3 that doesn't like the new drivers so I did the following:

- Installed and tested 4 previous driver versions (yes I do a proper driver cleanup with DDU in safe mode before installing other drivers).
- Spent several hours searching for and trying just about all of the Witcher 3 crash fixes, workarounds etc that are available on interwebs.

All of this time and effort amounted to naught though and the crashes kept persisting.

A few nights ago I thought I'd run ROTR for a while to see if I could replicate the problem in another game and to my surprise after 3 or 4 minutes ROTR also crashed to desktop with a "Rise of the Tomb Raider has stopped working" error.

I then shifted my focus to hardware:

- Reset BIOS to defaults (already have the latest version so an update was out of the question)
- Did a RAM test with Windows Memory Diagnostic - I'm not sure how good this tool is but at the time I didn't feel like going through the shlep of loading Memtest86 on a bootable flash drive etc. but I'll try to get around to it this weekend.

As far as temps go, my CPU tops out at around 65 degrees (in W3 and ROTR) while the MSI 1080 sits at around 73/74 at full load. AFAIK these temps are well within acceptable operating norms so I doubt thermals are my issue.

It was then that I thought that perhaps my +/- 5 year old PSU was starting to die on me however the usual symptoms of a failing PSU are not there because in my experience, dying PSU's will usually cause random system resets or random system hangs or lockups which I wasn't experiencing.

I then did the following:

- Ran the Unigine Heaven benchmark @ 1440p 3 times - no crashes
- Ran Furmark @ 1440p 3 times - no crashes

Maybe I'm wrong but surely if my PSU is the culprit and after +/- 5 years can't feed stable power on the 12V rail anymore I would also experience crashes and/or stability issues when doing GPU stress tests?

I'll try to run some additional stress tests like Prime95 and OCCT over the weekend to confirm that my CPU and RAM are stable but other than that I'm a bit stumped.
What's most frustrating is I that cannot seem to replicate the crashes for instance, Witcher 3 could crash whether I'm running around Novigrad (where my GPU/CPU/RAM is working hard) or even when I'm browsing the inventory or playing a game of Gwent where the hardware is doing minimal work.

Any help or suggestions would be much appreciated.

 

[Chad29]

Try testing another PSU to verify that it is indeed your PSU

Sent from my HUAWEI Y560-L01 using Tapatalk

 

[Toxxyc]

Have you formatted or re-downloaded the game(s) yet? PSU issues will cause a power cut, in other words an immediate shutdown or reboot of the system, not a Windows error. The fact that you're getting a Windows error says that it's a software issue.

 

[Spidey8085]

When i used to OC my gpu too far, my games would crash, and since you didn't have these issues with the 1070, it might be a issue with power delivery to the 1080, try using different power connectors if your psu has extra?



Also, find a friend and put your gpu in his pc, see what that does.

 

[Sm00thSm0k3]

Try testing another PSU to verify that it is indeed your PSU

Yeah that would be the obvious choice but I unfortunately the two spare PSU's I have aren't beefy enough to reliably test with.
It will probably be less of a hassle to just buy a new PSU than trying to arrange a good quality test PSU from someone living around my area.

No harm in asking though....are there any Carb members living in the West Rand who would be willing to borrow me a PSU to test with?

Have you formatted or re-downloaded the game(s) yet? PSU issues will cause a power cut, in other words an immediate shutdown or reboot of the system, not a Windows error. The fact that you're getting a Windows error says that it's a software issue.

Agree 100% regarding PSU.
No I've not tried reformatting yet...that was going to be my last resort.

 

[SCHUMI4EVER]

Definitely not the PSU. However I have no clue what it might be.

Gigabyte motherboards have a setting called Memory Initialization Mode which from what I understand affects RAM timing settings. The options are performance (or speed, can't remember exactly which), auto and reliability. Now I have on occasion noticed an increased number of crashes when that setting is set to performance (it also warns you that might be the case). Have however not noticed a difference between them in actual game performance so these days I leave it on Auto or reliability if I am trying to hunt for game crash issues. Perhaps check for such a setting on your board?

 

[Sm00thSm0k3]

When i used to OC my gpu too far, my games would crash, and since you didn't have these issues with the 1070, it might be a issue with power delivery to the 1080, try using different power connectors if your psu has extra?
Also, find a friend and put your gpu in his pc, see what that does.

No harm in trying different PCI-E power connectors I guess.

If all my other diagnostic attempts fail this weekend I'll slap the 1080 in my Brother in-laws rig next weekend and see what transpires.

 

[Sm00thSm0k3]

Definitely not the PSU. However I have no clue what it might be.

Gigabyte motherboards have a setting called Memory Initialization Mode which from what I understand affects RAM timing settings. The options are performance (or speed, can't remember exactly which), auto and reliability. Now I have on occasion noticed an increased number of crashes when that setting is set to performance (it also warns you that might be the case). Have however not noticed a difference between them in actual game performance so these days I leave it on Auto or reliability if I am trying to hunt for game crash issues. Perhaps check for such a setting on your board?

Well since I did reset the BIOS, everything is basically now back to "AUTO".
Now that I think about it, the only thing I did change back after the BIOS reset was the XMP profile so maybe I should disable it again and see what happens....maybe one of the RAM sticks is dying and becomes unstable at 2400Mhz?

 

[Chad29]

Well since I did reset the BIOS to defaults, everything is basically now back to "AUTO".
The only thing I did change back after the BIOS reset was the XMP profile so maybe I should disable it again and see what happens....maybe one of the RAM sticks is dying and becomes unstable at 2400Mhz?

That sounds more likely than it being a PSU issue

Sent from my HUAWEI Y560-L01 using Tapatalk

 

[Chad29]

Try gaming with individual sticks perhaps?

Sent from my HUAWEI Y560-L01 using Tapatalk

 

[souljazk]

Check event log @ time of crash.

With the spare PSU's - connect one to GPU, to power it on, short any black wire on the 24pin to the green wire on the 24pin. Then play games as normal and see. Keep the current PSU powering the rest of the system.

 

[The King]

If you had no problems with the 1070 its possible the 1080 maybe the problem

Trying running the 1080 @ stock reference clocks on the core if it still crashes then
try dropping the mem speed on the card and test again.

I had a card that did something similar ran fine in stress tests but crashed in games after some time, I reduced the mem clock by 100Mhz made it stable.

If it turns out to be faulty vram then RMA the card.

 

[SCHUMI4EVER]

Well since I did reset the BIOS, everything is basically now back to "AUTO".
Now that I think about it, the only thing I did change back after the BIOS reset was the XMP profile so maybe I should disable it again and see what happens....maybe one of the RAM sticks is dying and becomes unstable at 2400Mhz?

Yeah definitely reverse that XMP profile. I remember what the actual technical terms were for what I was talking about, it's the CAS latency. Basically if your RAM is clocked beyond stock (XMP is an automatic "overclocked" setting basically, so stock is whatever the RAM sits at without XMP) and has timings that are too tight then it can cause instability which affects only software rather than the entire system as one would usually see from unstable setting. The setting on my motherboard adjusts the CAS timing between tight, loose and variably controlled/stock.

 

[Sm00thSm0k3]

If you had no problems with the 1070 its possible the 1080 maybe the problem

Trying running the 1080 @ stock reference clocks on the core if it still crashes then
try dropping the mem speed on the card and test again.

I had a card that did something similar ran fine in stress tests but crashed in games after some time, I reduced the mem clock by 100Mhz made it stable.

If it turns out to be faulty vram then RMA the card.

Will definitely try this. Thanks.

Probably one of the easiest tests one can do to either identify or rule out the GPU being the culprit.
Hoping it's not the GPU though...don't want to be stuck without a proper GPU for God knows how many weeks.

 

[Sm00thSm0k3]

To everyone who replied to my thread last week (@The King, [MENTION=28839]souljazk[/MENTION], [MENTION=39163]Chad29[/MENTION], [MENTION=29712]Toxxyc338[/MENTION], [MENTION=22315]BakaSpidey8085[/MENTION], [MENTION=38509]Mast3rBlast3r[/MENTION]) I thought you might be interested in an update to my quandary.

So over the long weekend I managed to do a bit more fiddling/testing and went about the following:

Swapped out PSU - crashes continued (at least I eliminated the PSU as being the possible culprit).
Disabled XMP and set all RAM setting to "auto" - no difference, crashes persisted.
Underclocked both the core and mem of my GPU by 100Mhz - crashes continued.
Ran Prime95 for 5 hours - no problems.
Ran MemTest for 4+ hours - no errors found.

At this point I was almost convinced that it must be my GPU however the only thing I hadn't yet tried was testing my RAM individually.
Now for the record I originally had 2 x 4GB sticks of G.Skill Ripjaws Z 2400Mhz RAM together with 2 x 8GB sticks of G.Skill Trident X 2400Mhz RAM installed (the speed, voltages and timings of both kits are identical).
Here I'll refer to the mobo slots as 1,2,3,4. The 2 x 4GB sticks were in slots 1 & 3 and the 2 x 8GB sticks were in slots 2 & 4.

For the purposes of testing I removed the 2 x 4GB the modules leaving just the 2 x 8GB sticks (lets call them stick A and stick B) in slots 2 & 4 - Didn't run memtest but in-game crashes persisted.

I then left stick A in slot 2 and removed stick B from slot 4 - Not only did the crashes persist but I experienced a hard lockup for the first time; had to hit the reset switch to recover the system.

I then removed stick A and put stick B in slot 2 - System completely stable. I ended up gaming for 3+ hours completely uninterrupted on Sunday morning. No crashes or problems of any sort.

I began to suspect that stick A might be dodge and but just to be to be sure removed stick B and put stick A back in slot 2 again and ran Memtest - Within minutes I got 2 errors during Test #7 (Moving inversions, 32 bit pattern).

At this point I though I was seeing some light at the end of the tunnel and thought "okay great, I've found the problem; it's a faulty RAM stick; I'll send it back to Wootware to get RMA'd"

To be 100% certain that stick A was indeed faulty I decided test stick A in the other RAM slots as well.
Due to the size of my CPU cooler (and the rather large heatsink on the Trident X RAM) I'm unable to put put stick A or B in slot 1 which leaves me with slot 3 & 4 as alternative slots with which to test. This however should be more than sufficient.

Now this is however where I started pulling my hair out....

When I put stick A in slot 3 or 4 and run Memtest I get no errors but I still get in-game crashes.
When I put stick B in slot 2, 3 or 4 no Memtest errors and no in-game crashes.
Out of interest sake I removed stick B then put 1 of the 4GB sticks in slot 2 and ran Memtest again - no errors.
I then went one step further and put the other 4GB stick in slot 4 as well (dual channel) - no Memtest errors and was gaming issue free for a good couple of hours on Monday.

So to summarize, Memtest only finds errors on stick A when stick A is in slot 2 however I continue to experience instability issues even if stick A is in slot 3 or 4.
Any other RAM I install in slot 2 works just fine.

I even went as far as to install stick A in another PC (my HTPC) and ran Memtest but once again no errors were found.

All I can say is WTF?!! This makes it extremely difficult to determine whether the issue is the RAM or the slots on the board.
Surely if stick A is defective then Memtest should spit out errors regardless of which slot or PC the RAM is installed in?
If I can't replicate or reproduce the Memtest errors in multiple slots or on a different board then there's no point in even trying to RMA the RAM.

If it's indeed slot 2 on my board that's faulty then it would make sense that any other RAM installed in that slot should also be problematic and give errors or cause stability issues?

Not sure what else to do at this point.

 

[Spidey8085]

To everyone who replied to my thread last week (@The King, [MENTION=28839]souljazk[/MENTION], [MENTION=39163]Chad29[/MENTION], [MENTION=29712]Toxxyc338[/MENTION], [MENTION=22315]BakaSpidey8085[/MENTION], [MENTION=38509]Mast3rBlast3r[/MENTION]) I thought you might be interested in an update to my quandary.

So over the long weekend I managed to do a bit more fiddling/testing and went about the following:

Swapped out PSU - crashes continued (at least I eliminated the PSU as being the possible culprit).
Disabled XMP and set all RAM setting to "auto" - no difference, crashes persisted.
Underclocked both the core and mem of my GPU by 100Mhz - crashes continued.
Ran Prime95 for 5 hours - no problems whatsoever.
Ran MemTest for 4+ hours - no errors found.

At this point I was almost convinced that it must be my GPU however the only thing I hadn't yet tried was testing the RAM individually.
Now for the record I originally had 2 x 4GB sticks of G.Skill Ripjaws Z 2400Mhz RAM together with 2 x 8GB sticks of G.Skill Trident X 2400Mhz RAM installed (the speed, voltages and timings of both kits are identical).
Here I'll refer to the mobo slots as 1,2,3,4. The 2 x 4GB sticks were in slots 1 & 3 and the 2 x 8GB sticks were in slots 2 & 4.

For the purposes of testing I removed the 2 x 4GB the modules leaving just the 2 x 8GB sticks (lets call them stick A and stick B) in slots 2 & 4 - Didn't run memtest but in-game crashes persisted.

I then left stick A in slot 2 and removed sticked B from slot 4 - Not only did the crashes persist but I experienced a hard lockup for the first time; had to hit the reset switch to recover the system.

I then removed stick A and put stick B in slot 2 - System completely stable. I ended up gaming for 3+ hours completely uninterrupted on Sunday morning. No crashes or problems of any sort.

At this point I began to suspect that stick A might be dodge and but just to be to be sure removed stick B and put stick A back in slot 2 again and ran Memtest - Within minutes I got 2 errors during Test #7 (Moving inversions, 32 bit pattern).

At this point it looked like I was seeing some light at the end of the tunnel and thought "okay great, I've found the problem; it's a faulty RAM stick; I'll send it back to Wootware to get RMA'd"

To be 100% certain that stick A was indeed faulty I decided test stick A in the other RAM slots as well.
Due to the size of my CPU cooler (and the rather large heatsink on the Trident X RAM) I'm unable to put put stick A or B in slot 1 which leaves me with slot 3 & 4 as alternative slots with which to test. This however should be more than sufficient.

Now this is however where I started pulling my hair out....

When I put stick A in slot 3 or 4 and run Memtest I get no errors but I still get in-game crashes.
When I put stick B in slot 2, 3 or 4 no Memtest errors and no in-game crashes.
Out of interest sake I removed stick B then put 1 of the 4GB sticks in slot 2 and ran Memtest again - no errors.
I then went one step further and put the other 4GB stick in slot 4 as well (dual channel) - no Memtest errors and was gaming issue free a good couple of hours on Monday.

So to summarize, Memtest only finds errors on stick A when stick A is in slot 2 however I continue to experience instability issues even if stick A is in slot 3 or 4.
Any other RAM I install in slot 2 works just fine.

I even went as far as to install stick A in another PC (my HTPC) and ran Memtest but once again no errors were found.

Now this makes makes it extremely difficult to determine whether the issue is the RAM or the slots on the board.
Surely if stick A is defective then Memtest should spit out errors regardless which slot or PC the RAM is installed in?
If I can't replicate or reproduce the Memtest errors in multiple slots or on a different board then there's no point in even trying to RMA the RAM.

If it's indeed slot 2 on my board that's faulty then it would make sense that any other RAM installed in that slot should also be problematic and give errors or cause stability issues?

Not sure what else to do at this point.

This is an interesting thread:

http://www.tomshardware.com/forum/327791-30-memtest-lying


You have an anomaly on your hands!! Use the 2x4gb sticks and pretend the other 2x8gb sticks don't exist




I wonder if that HTPC of yours is using a cooler that is causing some incidental airflow over that ram stick? I wonder if this issue is not related to overheating on that single stick...

 

[Sm00thSm0k3]

This is an interesting thread:

http://www.tomshardware.com/forum/327791-30-memtest-lying

You have an anomaly on your hands!! Use the 2x4gb sticks and pretend the other 2x8gb sticks don't exist

Lol well I have no choice but to do that until such time that I can get to the bottom of this however I can't have 16GB of good (well it used to be) RAM lying around doing nothing.

I wonder if that HTPC of yours is using a cooler that is causing some incidental airflow over that ram stick? I wonder if this issue is not related to overheating on that single stick...

My HTPC is as basic as it comes...stock i3 with stock intel cooler.

 

[souljazk]

Rule for mem test - test for 24hours. Iv had a few RAM modules that would only give error in memtest after ~23hours. I could repeat this result any time I ran the test over 23hours. Id tunr all cooling off except CPU cooling, let that sucker get hot...

Do you get a BSOD dump? Is BSOD dump enabled?


Regarding the RAM models having identical timings/volts etc - the sticker is just a semi guarantee that in most situations/with most hardware the RAM CAN run @ the timings/volts as per the sticker. You may find that once installed, the RAM is not running at those volts/timings - just something to look into.


"I then removed stick A and put stick B in slot 2 - System completely stable. I ended up gaming for 3+ hours completely uninterrupted on Sunday morning. No crashes or problems of any sort."
;
"I began to suspect that stick A might be dodge and but just to be to be sure removed stick B and put stick A back in slot 2 again and ran Memtest - Within minutes I got 2 errors during Test #7 (Moving inversions, 32 bit pattern).

At this point I though I was seeing some light at the end of the tunnel and thought "okay great, I've found the problem; it's a faulty RAM stick; I'll send it back to Wootware to get RMA'd"

To be 100% certain that stick A was indeed faulty I decided test stick A in the other RAM slots as well.
Due to the size of my CPU cooler (and the rather large heatsink on the Trident X RAM) I'm unable to put put stick A or B in slot 1 which leaves me with slot 3 & 4 as alternative slots with which to test. This however should be more than sufficient.

Now this is however where I started pulling my hair out....

When I put stick A in slot 3 or 4 and run Memtest I get no errors but I still get in-game crashes.
When I put stick B in slot 2, 3 or 4 no Memtest errors and no in-game crashes.
Out of interest sake I removed stick B then put 1 of the 4GB sticks in slot 2 and ran Memtest again - no errors.
I then went one step further and put the other 4GB stick in slot 4 as well (dual channel) - no Memtest errors and was gaming issue free for a good couple of hours on Monday.

So to summarize, Memtest only finds errors on stick A when stick A is in slot 2 however I continue to experience instability issues even if stick A is in slot 3 or 4.
Any other RAM I install in slot 2 works just fine."



I'd repeat what Iv put in bold+underlined for atleast 6 hours of gaming and 72hours of use.

I think somewhere in your anomaly is a issue with the dual channel setup on the CPU. AFAIR the mem controller is on the CPU.