LinX stable, start game = crash

[BTRY B 529th FA BN]

I've ran 20 passes of LinX without fail, opened up UT3 and just as I click to join a server my machine will crash. I set all my bios to default/stock clocks and joining a server works. :hmm:

 

[zagood]

From stock, start overclocking step by step. There may be one particular setting that's making it trip, not the actual overclock itself. Something like you forgot to lock PCIe frequency.

 

[Spikesoldier]

crash = BSOD? reboot? hard lock? no display?

 

[MJinZ]

LinX stable? So what?

Prime stable?

3DMarks Stable?

OCCT Stable? Furmark Stable?

Then you can consider getting game stable.

 

[BTRY B 529th FA BN]

crash = BSOD? reboot? hard lock? no display?

Total reboot, no bsod etc.

The PCI-e is locked at 100

 

[Thor86]

LinX stable? So what?

Prime stable?

3DMarks Stable?

OCCT Stable? Furmark Stable?

Then you can consider getting game stable.

+1

This just tells me never to use LinX as a stability test.

I've always used Prime95 stability testing, and have never had gaming issues with stable overclocks.

 

[SHAQ]

And 20 passes isn't much unless it is 64bit with max memory. Try 50 passes this time.

 

[OILFIELDTRASH]

Man I run 100 passes Linx then the real test to me is 24 hours prime95 stable. I've had rigs pass linx several hundred passes and not pass a 24 hour prime95 run.

 

[BTRY B 529th FA BN]

People say the same thing about Prime95 as I have also experienced stable p95 large and small ffts' and then failed LinX

 

[Maximilian]

Just give it a bit more voltage. You can run prime95 and linX until your CPU is obsolete or you can do your 20 passes then play games and see what happens... i do the latter, much easier and less of a PITA.

I did the same as you, ended up with a crash, gave my cpu a bit more voltage and its fine now Been fine for weeks.

 

[ashishmishra]

This same exact issue happened to me in Crysis, it turned out that one of my power supply rails was crapping out. Bad PSU design, one of the rails was powering both CPU power connector as well as a PCIe Connector. RMAed the OCZ GameXstream 850W (4 Rails) got a PCP&C 860W Turbocool (Single Rail) = no more crashing in-game.

BTW OCZ has since rectified the design defect and none of the CPU rails power graphics as well.

 

[VirtualLarry]

This same exact issue happened to me in Crysis, it turned out that one of my power supply rails was crapping out. Bad PSU design, one of the rails was powering both CPU power connector as well as a PCIe Connector. RMAed the OCZ GameXstream 850W (4 Rails) got a PCP&C 860W Turbocool (Single Rail) = no more crashing in-game.

BTW OCZ has since rectified the design defect and none of the CPU rails power graphics as well.

Interesting. I bought two 850W OCZ GameXStream PSUs, reconditioned, from svc.com a while back. So you tell me that they have rail problems? Just great...

 

[EarthwormJim]

+1

This just tells me never to use LinX as a stability test.

I've always used Prime95 stability testing, and have never had gaming issues with stable overclocks.

How did you come to that conclusion? Linx is the single best CPU stress tester. It's a better test than Prime95 small FFT or large FFT (for the CPU). My old Core 2 Quad would pass Prime 95 large FFT for 24hrs+ but fail LinX in after 40 passes.

Linx doesn't stress test your memory subsystem too well though. Recently playing with my Core i7, at 4ghz I could pass LinX, but Prime 95 blend I'd fail. I just had to bump my QPI voltage up a bit, the memory subsystem was the source of the error. The reverse is more common though, passing blend but failing LinX.

The real way to go is to do multiple stress test utilities (not at the same time though), since they all stress your system in their own unique way. Linx for the CPU then Prime95 large/small FFT, Prime95 blend for CPU and Memory, HCI memtest for memory. If your system can pass all those, you're pretty much golden (a system could still produce errors in video encoding even after passing those though).

And honestly 20 passes of Linx is not enough in my opinion. A good overnight test is what I'd go with, so 100 or more passes. An unstable overclock may make an error say 1% of the time, so if you only do 20 passes, that 1% chance may not manifest itself.

 

[EarthwormJim]

Just give it a bit more voltage. You can run prime95 and linX until your CPU is obsolete or you can do your 20 passes then play games and see what happens... i do the latter, much easier and less of a PITA.

I did the same as you, ended up with a crash, gave my cpu a bit more voltage and its fine now Been fine for weeks.

There's no pain in the ass, just do the tests overnight.

 

[ashishmishra]

Interesting. I bought two 850W OCZ GameXStream PSUs, reconditioned, from svc.com a while back. So you tell me that they have rail problems? Just great...

Well it is not a "rail problem" per se. Look at the sticker on the side of your PSU, it will tell you the connectors that are powered by each rail. The older units split one of the rails between the EPS 12V connector and a PCIe connector. The newer units have a different rail allocation, more specifically the PCIe Connectors DO NOT share a rail with a CPU power connector.

If you happen to run a heavily overclocked CPU (like my Core i7 @ 4.0) and a high end GPU (Like my GTX 285), one of the rails that is sharing the EPS and PCIe connectors apparently runs into its max current limit and shuts the power supply down.

Hope that makes sense.....

EDIT: Similar to the OP's situation the crash (reboot) didn't really happen in game, but just as the menu loaded. I did not see the crash when CPU was stock, and interestingly I didn't have the problem with a GTX 260 (with i7 @ 4.0).......but had a similar issue when I was running a 9800GX2 with (Q9550 @ 3.8) (except it was rebooting in-game rather than menu)

 

[TekDemon]

It sounds largely like a power issue to me, where increasing the voltage to the CPU is likely to make the problem worse and not better.
Try seeing if a slightly lower overclock using a lower voltage going to the CPU makes it more stable. Like drop it by a few hundred megahertz but also drop the voltage to see if it becomes stable. You might find that even at speeds/voltages where the CPU won't pass LinX it'll boot into the game just because the power draw isn't as large.
It might not necessarily even be the PSU though-it could be the motherboard not dealing well with the increased power draw, even if you don't see any obviously blown caps there might be bad caps. Keep in mind that the video card does draw power from the motherboard even if it also uses external power connectors.

So, what model and how old is the motherboard in this machine? What kind of power supply and how old is it?

 

[Spikesoldier]

Total reboot, no bsod etc.

The PCI-e is locked at 100

sorry, i had this response typed out yesterday but didnt hit the submit button until now.

usually a hard reboot like that is cpu related

in turn, it could be an insufficient power supply that cant muster serving up max cpu oc plus the vga card on top of that. i run furmark concurrently with p95 for a true worst-case thermal and electrical beatdown.


the first thing i would try though if you can, is to do a new windows install and see if it is still acting up. for the next part, i am assuming you are running a LGA1366 i7 and a enthusiast grade vga card

before you go through the trouble of trying a new PSU, i would give these quick bios switches a shot (one at a time):
pci-e bus to 101MHz
+ more vIOH to the X58 (your pci-e contoller rests here!)
+ little bit vtt
+ little bit of vcore
lower QPI multi

 

[BTRY B 529th FA BN]

Well, I put everything back to stock. Ran p95 blend with Furmark. It was running fine until I clicked ATI-tool to kick up my fan speeds on my 1900XTX and it hung/froze. Nothing is reacting like ctrl+Alt+delete etc..

 

[BTRY B 529th FA BN]

Tested again after uninstalling ATI-tool. Ran Memtest HCI, p95 blend, & fur-burger mark and it crashed. I'm thinking it's the motherboard. Second guess is it's the Seasonic PSU

 

[aka1nas]

It's usually the PSU when it does a total reboot under heavy load like that. It's not surprising if your CPU OC was taking it right to the edge of what the PSU can reliably deliver and then you try to load the GPU on top of it.

 

[BTRY B 529th FA BN]

It's usually the PSU when it does a total reboot under heavy load like that. It's not surprising if your CPU OC was taking it right to the edge of what the PSU can reliably deliver and then you try to load the GPU on top of it.

And if it's not the PSU what would be your second guess?

TBH: i'm not sure it's the PSU because I did the same thing with my i7 920 + Asrock setup (p95 + HCI + Furmark) and my wattage load reader was showing 430-470w draw from the outlet. This machine has got to be using less than that and it's a 650w psu. Board maybe?

 

[EarthwormJim]

And if it's not the PSU what would be your second guess?

TBH: i'm not sure it's the PSU because I did the same thing with my i7 920 + Asrock setup (p95 + HCI + Furmark) and my wattage load reader was showing 430-470w draw from the outlet. This machine has got to be using less than that and it's a 650w psu. Board maybe?

Or just your overclock, maybe your memory timings were too aggressive, or the memory was running too fast. Maybe your board didn't like the FSB you were running.

Have you tried a setting that was 5 or 10% slower than your overclock?

It's highly unlikely it's the PSU, it's 650 watt and it's a Seasonic. It has plenty of amperage.

Hard reboots with a blue screen usually indicate a memory subsystem instability. If one or two cores were unstable, you usually don't get a full blue screen.

 

[ashishmishra]

If you can borrow Enermax Revo 950W from your other rigs to test, that would certainly the put the possibility of it being a psu issue to rest.

 

[BTRY B 529th FA BN]

Or just your overclock, maybe your memory timings were too aggressive, or the memory was running too fast. Maybe your board didn't like the FSB you were running.

Have you tried a setting that was 5 or 10% slower than your overclock?

It's highly unlikely it's the PSU, it's 650 watt and it's a Seasonic. It has plenty of amperage.

Hard reboots with a blue screen usually indicate a memory subsystem instability. If one or two cores were unstable, you usually don't get a full blue screen.

The timings are very loose 6.6.6.18 at 400fsb or 800Mhz

There is no overclock on anything (cpu, gpu, ram, pci-e, etc)

No blue screens, just rebooting.

It's liquid cooled, the cpu has never been abused. The ram is stable at 500fsb (extensive testing) & I've tested the hell out of the subsystem. And I have an active fan over the NB HS

I have a UD3P on the way to cross check the mb.

 

[aigomorla]

can u check to see if your pump is on?