Question PC Restarts when under certain loads

[fnikol22]

Hello,
I have a pretty unusual setup: it's a Lenovo D20 workstation (two CPUs are present) with a gtx 1080 ti. My problem is that when the card is stressed with a specific type of load it causes a PC restart. What I mean by specific load is this: the card runs fine in the Battlefront II menu which causes almost 100% gpu load and 100% power draw. I can also play every game mode but in one of them (HvV) a reboot may occur - sometimes it does sometimes it does not - it is very rare, may not happen for 3h, may happen after 15 mins. In Shadow of Mordor the card uses 120-150W on average with spikes to 200W and there is only one action that causes a shutdown (interrogating a normal enemy) and again it does not always happen - from what I have seen if I wait one or two seconds before performing said interrogation after domating an enemy, it does not happen. However there are loads that always cause almost immediate restart, for example when using the 'test' option in OC tab in MSI Afterburner. I tested it with different power limits and 80% limit is stable, 90% may crash, 100% crashes always after few seconds. I also tested The Witcher 3 in 4k DSR (which uses 100% GPU power obviously) and it crashed 50% of the time however if it did not crash in the first 10 seconds it would not do it at all. I also played a lot of other games (from those more GPU intensive: FC5, Shadow of War, KCD, BF1, Apex Legends (with high FPS)) but never experienced a crash in them. My question is what could be the cause of those crashes - is it GPU, PSU, MOBO or something else (the system is quite old so it could be anything really)? I logged all the voltages that speedfan has access to and right before the restart they were stable. I obviously cleaned the whole PC with compressed air right after it happened (wasn't very dusty though). I should also add that even though the card is out of specs for this workstation (250W vs specs 216W) I was able to crosswire it with a bit of help from Lenovo forum - 6-pin on the card is delivered from one line while the 8-pin is from another. The card was bought used but I've tested it on another system and MSI Afterburner did not crash it. Previously I had a gtx 970 and it was fine. Stressing CPUs is OK too, all temps that HWiNFO displays are well within the specs. If you need any additional information just let me know. Thanks!

 

[Roland Of Gilead]

Hey,

Well, if you can start by listing your exact PC specs.

First off, I'd say PSU, seems like when your system load spikes, it's shutting down to protect itself. This can happen even with good PSU's. But if yours is more than 4 years old, and/or is of good or not so good quality would help determine a solution. When your playing games, the load can vary pretty wildly depending on whats on screen, how many players your up against (Multiplayer) all of which will increase your power draw on the CPU and GPU. Some 1080ti's can actually hit 300w if really pushed!

It's possible it's other issues, but ram errors, typically will show as a BSOD with a dump file, CPU errors would be freezing on screen and having to reset and GPU issues at first would be shown by on screen tearing/glitches or artifact. So that leans me towards your PSU.

 

[jeremyj_83]

Hey,

Well, if you can start by listing your exact PC specs.

First off, I'd say PSU, seems like when your system load spikes, it's shutting down to protect itself. This can happen even with good PSU's. But if yours is more than 4 years old, and/or is of good or not so good quality would help determine a solution. When your playing games, the load can vary pretty wildly depending on whats on screen, how many players your up against (Multiplayer) all of which will increase your power draw on the CPU and GPU. Some 1080ti's can actually hit 300w if really pushed!

It's possible it's other issues, but ram errors, typically will show as a BSOD with a dump file, CPU errors would be freezing on screen and having to reset and GPU issues at first would be shown by on screen tearing/glitches or artifact. So that leans me towards your PSU.

It is most likely the PSU since s/he modded it to work with the 1080Ti. "I should also add that even though the card is out of specs for this workstation (250W vs specs 216W) I was able to crosswire it with a bit of help from Lenovo forum - 6-pin on the card is delivered from one line while the 8-pin is from another."

 

[fnikol22]

Thanks for the replies. Yes, it does seem like most of the times the crash is caused by a spike. The PSU is certainly older than 4 years but seems like the quality is good ('seems' being the keyword as it is a custom PSU for this exact workstation so it's hard to come by some professional reviews). It's rated for 1060W, from which 900W are on 12V. So is there maybe a way to prevent a card from jumping in power consumption too much? Not ideal but still better than crashing
The rest of my specs:
2x Xeon X5675 @3.06~3.46GHz
Gigabyte GTX 1080 Ti Gaming OC
36GB of DDR3-1333
Samsung 860 EVO 500GB and 5x2TB Seagate ES.3 on RAID5

It is most likely the PSU since s/he modded it to work with the 1080Ti. "I should also add that even though the card is out of specs for this workstation (250W vs specs 216W) I was able to crosswire it with a bit of help from Lenovo forum - 6-pin on the card is delivered from one line while the 8-pin is from another."

Somebody on the Lenovo forum had a very similar situation - gtx 780 Ti (a 250W card) inside the same system (so same PSU), and reported that it worked perfectly after crosswiring. Basically what I do is I have two 12V lines for two PCIe cards both with 6+8pins and I use 6-pin from one line and 8-pin from another to supposedly distribute the load between both. In theory it sounds great but of course that also might be the problem - maybe the card draws too much power from let's say 8-pin and does not distribute it correctly - but then it would have to draw over 216W from 8pin which to my knowledge is over the specs for 8pin connector (150W?). Additionally the PC should crash during 100% GPU load in games yet it doesn't.

In the worst case scenario I will have to buy a new PSU, but one that would support this system most likely wouldn't be the cheapest so I want to be sure that the PSU is the culprit. Is there some kind of experiment I could do to further confirm or maybe rule out the PSU being faulty?

 

[rgd1101]

custom psu? what is the prebuild model?

 

[jeremyj_83]

Thanks for the replies. Yes, it does seem like most of the times the crash is caused by a spike. The PSU is certainly older than 4 years but seems like the quality is good ('seems' being the keyword as it is a custom PSU for this exact workstation so it's hard to come by some professional reviews). It's rated for 1060W, from which 900W are on 12V. So is there maybe a way to prevent a card from jumping in power consumption too much? Not ideal but still better than crashing
The rest of my specs:
2x Xeon X5675 @3.06~3.46GHz
Gigabyte GTX 1080 Ti Gaming OC
36GB of DDR3-1333
Samsung 860 EVO 500GB and 5x2TB Seagate ES.3 on RAID5


Somebody on the Lenovo forum had a very similar situation - gtx 780 Ti (a 250W card) inside the same system (so same PSU), and reported that it worked perfectly after crosswiring. Basically what I do is I have two 12V lines for two PCIe cards both with 6+8pins and I use 6-pin from one line and 8-pin from another to supposedly distribute the load between both. In theory it sounds great but of course that also might be the problem - maybe the card draws too much power from let's say 8-pin and does not distribute it correctly - but then it would have to draw over 216W from 8pin which to my knowledge is over the specs for 8pin connector (150W?). Additionally the PC should crash during 100% GPU load in games yet it doesn't.

In the worst case scenario I will have to buy a new PSU, but one that would support this system most likely wouldn't be the cheapest so I want to be sure that the PSU is the culprit. Is there some kind of experiment I could do to further confirm or maybe rule out the PSU being faulty?

General rule of thumb is don't modify anything with the PSU. The worst case is an electrial fire that takes our your apartment/home/etc... Also don't forget that PSU is a good 10 years old and you can tell by how the wattage is laid out. Modern PSUs will be able to do 99% of the rated wattage on the 12V rail.

 

[fnikol22]

custom psu? what is the prebuild model?

The PSU model is a DPS-1060AB A but is also easily found on sale as Thinkstation D20 PSU.

General rule of thumb is don't modify anything with the PSU. The worst case is an electrial fire that takes our your apartment/home/etc... Also don't forget that PSU is a good 10 years old and you can tell by how the wattage is laid out. Modern PSUs will be able to do 99% of the rated wattage on the 12V rail.

Yeah I know that but if a Lenovo staff member says that crosswiring should work and somebody before me did that and said it worked then I assume it is not that dangerous to try. And it does work to a certain degree - previously putting GPU under 100% load caused reboot because of exceeding current limit, now only certain load types and/or spikes restart the system.

 

[mdd1963]

Whether or not a modded PSU not equipped with correct 6 pin/8 pin PCI-e VGA PWR cables should work or could work or are advisable are all subject to debate...

What is not subject to debate is that your rig is hard resetting/power-off/rebooting under load...(classic symptoms of overloaded/straining under load/weak PSU...

I'd first try a new PSU of sufficient wattage.

(Be sure to replace any/all modular cables with the PSU, as even SATA power cables are often not cross-compatible...)

 

[Rjich]

I'd echo the intelligent statement above. There is no point debating anything past the funky PSU on this particular issue. The FPS is never import enough to risk safety.

 

[Roland Of Gilead]

Totally agree with mdd1963. 'Just because' this or that about the PSU, is almost irrelevant. Your system is re-starting because of the PSU. Anything older than 4 years and/or is not reputable (and I'd defo put a Lenovo 'Custom' PSU or any other resller of PSU's - Dell, Lenovo, HP etc in that bracket) will cause issues. A decent 850w PSU from a good brand will last you 5-10 years.

After 3 years typically, the PSU starts to degrade. It doesn't output the same power, the protections may fail (specially so with generic PSU's). They are made from cheaper quality parts (to keep costs low) and are usually rubbish. They WILL fail. Often, taking pieces of hardware with them.

Get a new PSU. Something like this:

PCPartPicker Part List

Power Supply: EVGA - SuperNOVA G3 850 W 80+ Gold Certified Fully Modular ATX Power Supply ($137.89 @ OutletPC)
Total: $137.89
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2019-05-15 13:24 EDT-0400

This will serve you well, has a 7 year warranty, and is a quality unit that will not let you down.

In the meantime, if you can borrow a similarly specced PSU as your own, and try it out to confirm the crashes stop. Then you have the answer.

 

[jeremyj_83]

Totally agree with mdd1963. 'Just because' this or that about the PSU, is almost irrelevant. Your system is re-starting because of the PSU. Anything older than 4 years and/or is not reputable (and I'd defo put a Lenovo 'Custom' PSU or any other resller of PSU's - Dell, Lenovo, HP etc in that bracket) will cause issues. A decent 850w PSU from a good brand will last you 5-10 years.

After 3 years typically, the PSU starts to degrade. It doesn't output the same power, the protections may fail (specially so with generic PSU's). They are made from cheaper quality parts (to keep costs low) and are usually rubbish. They WILL fail. Often, taking pieces of hardware with them.

Get a new PSU. Something like this:

PCPartPicker Part List

Power Supply: EVGA - SuperNOVA G3 850 W 80+ Gold Certified Fully Modular ATX Power Supply ($137.89 @ OutletPC)
Total: $137.89
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2019-05-15 13:24 EDT-0400

This will serve you well, has a 7 year warranty, and is a quality unit that will not let you down.

In the meantime, if you can borrow a similarly specced PSU as your own, and try it out to confirm the crashes stop. Then you have the answer.

An alternative to the EVGA would be the Seasonic Focus Plus.
PCPartPicker Part List

Power Supply: SeaSonic - FOCUS Plus Gold 850 W 80+ Gold Certified Fully Modular ATX Power Supply ($128.96 @ Amazon)
Total: $128.96
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2019-05-15 13:33 EDT-0400

 

[Roland Of Gilead]

An alternative to the EVGA would be the Seasonic Focus Plus.
PCPartPicker Part List

Power Supply: SeaSonic - FOCUS Plus Gold 850 W 80+ Gold Certified Fully Modular ATX Power Supply ($128.96 @ Amazon)
Total: $128.96
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2019-05-15 13:33 EDT-0400

Yeah, quality unit

 

[fnikol22]

Sorry for the late reply, I had very little time during the week.

Thanks for all the responses and suggestions. For now I reduced the power limit to 80% and it seems like it eliminated restarts or at least made them rare enough that I have not experienced one yet. However it's not a satisfying solution (obviously ). I guess I have no other choice but to buy a new power supply in the nearest future. I looked at both ones you linked but I'm not sure if they have the connectors my MOBO needs. SATA and VGA are obviously more than enough but each one of my 2 CPUs has a 10-pin connector on the MOBO - these PSUs have 8-pins for CPU. Maybe it is fine to just connect 8-pin to 10-pin but I want to make sure. Also my main motherboard power is a 24-pin whereas these PSUs have 18+10 pins. Maybe it is possible to use 18+6 or I simply just don't understand something so I would really appreciate your comments on that. I am including a photo of my motherboard layout:
https://imagizer.imageshack.com/img923/866/Ge91D5.jpg

Edit: I would lean more towards Seasonic's one as it's about 20% cheaper than EVGA in my country.