My new PC is cursed!

[Grall]

It would have to be one hell of a vibration to affect the PCIe connector adversely, especially if your PC has those hooks that are common on PCIe16x sockets that grab hold of the graphics board to hold it down.

I can't vouch for how much your WD drive should or should not vibrate - my own WDs, two 1TB Raptors and one 3TB Green series drive hardly vibrate any, but the Raptors have very low rotating mass (only two wafer-thin 2.5" platters), and the Green is a low-speed unit, so therefore inherently less...vibratory...in nature. The Green you can hardly feel at all when it's spinning, it's kinda creepy really.

 

[Npl]

It would have to be one hell of a vibration to affect the PCIe connector adversely, especially if your PC has those hooks that are common on PCIe16x sockets that grab hold of the graphics board to hold it down.

Sure thing... usually. I dont think its the reason why is broken, rather something like bad solders, but it might aswell be the reason for the irregularity.

I can't vouch for how much your WD drive should or should not vibrate - my own WDs, two 1TB Raptors and one 3TB Green series drive hardly vibrate any, but the Raptors have very low rotating mass (only two wafer-thin 2.5" platters), and the Green is a low-speed unit, so therefore inherently less...vibratory...in nature. The Green you can hardly feel at all when it's spinning, it's kinda creepy really.

Its a Green drive and I stayed loyal to WD exactly because of their low noise and vibration.
(Anything in my build is supposedly inaudible outside games, picked fans, coolers, case, rubber grommets especially for this reason - the WD is the single thing standing out and its indirect noise through vibration)

 

[Npl]

Hmm, seems to be some combinations of memory and pcie express speeds are causing this.
Memory alone is stable up to 2133MHz (ran Prime95 for over a day), but PCIe v3 speeds are only possible if the memory bus is at 1333MHz (else the driver faults and sluggish fps).

Dropped down to PCIe v2 and running the ram at 1600MHz for now, gonna see if this finally is the solution. So far its guessing and the issue might crop up in a couple days again =(

 

[Npl]

Knocking on wood... never had the issue since then.
Raised Memory to 1866MHz and keeping PCIe at v2, gonna see how that fares.

 

[Albuquerque]

I ran into stability issues on my "beast rig" when I first put it together (i7-3930k @ 4.5Ghz, 32GB ram @ 1600/1T, 7970 Dual-X, six SSD's in RAID0) -- problems where individual stress testing would go all day every day, but combined load testing like games would tank the box after varying amounts of time.

What I found helpful was to run Prime95 in with huge FFT's (so it uses lots of ram) and then simultaneously run FurMark in stress mode. I could crash the box repeatably every time after that.

Turned out that I needed a tad more voltage for the CPU IO pins and the ram itself. This was with all four channels dual-populated (eight total sticks), which I'm sure was the true root cause of needed the extra bit of voltage bump. After the tweaking, I was able to run P95 Large FFT's and FurMark for eight hours straight, and since then I've had no issues.

 

[Mize]

Okay, my turn to seek some assistance. I have an i7-2600k rig (P67 chipset) that can Furmark+Prime95 for days without issue at 4.5 GHz. It will, however, randomly reboot at idle every 36-48 hours. The idle reboots don't seem to correlate with anything else. HWMonitor will show some funny business on Vcore at idle (min will register like 0.2v after a few hours) so I suspect that Vcore drops too low at times. Thing is I'm already using way more offset than I need to keep this thing stable at 4.5GHz. I can run at that freq with about 1.33v but I've already bumped my offset to run at 1.37GHz to minimize the idle reboots (nominally 1.08v at idle not counting the mysterious drops).

So one question is which component would allow Vcore to drop too low? My PSU is a Maxrevo 1350 which is presumably one of the best, but this same rig was once stable at idle and PSUs are about the only component that shows substantial aging over a few years. The rig is also on a crappy AC line as it has two freezer/refrigerator compressors on it. I've got a line-filter and UPS to minimize the AC drops, but those are relatively new. I suppose the poor AC could have damaged the PSU stability over a couple years...still neither my media server nor my dedicated firewall PC have issues...

Fortunately it's still a competent gaming rig since this only happens at idle

Ideas?

 

[homerdog]

Try removing the overclock. That CPU is fast enough on its own anyway.

 

[Mize]

Try removing the overclock. That CPU is fast enough on its own anyway.

Sure, but at idle it's already running at 1.6GHz so it's not likely clock that's causing idle reboots. Like I said, it's a new behavior, but I don't mind stock-clocking as a test.

 

[Albuquerque]

Since it's an SNB-"K" series, are you just upping the multiplier? Or are you adjusting the BCLK strap?

Edit: Nevermind, only the SNB-E's got the BCLK strap option.

Someone over on [H]ardForums had a similar problem to what you describe, specifically with the P67 chipset. It may be related to the Intel RapidStorageTechnology driver suite. The person in the forum link ended up turning of AHCI mode, but I don't think that's necessary.

http://hardforum.com/showthread.php?t=1593414

 

[Mize]

Since it's an SNB-"K" series, are you just upping the multiplier? Or are you adjusting the BCLK strap?

No BCLK change (100), just a 45 multiplier and an offset on Vcore with C3/C6 states disabled.
Lowering LLC and phase control actually improved idle stability (much longer between idle reboots) which indicates a switching issues of sorts. I could disable all power saving/turbo/etc. and just leave it at full clock and full Vcore, but that's not a fun option to me.

 

[Mize]

Okay, I put my multiplier back to Auto for homerdog's benefit (why 3.5GHz is "fast enough" compared to 4.5GHz, I'm not entirely clear about) and we'll see if I can get past the 48 hour mark for idle reboots. Note that the idle clock and Vcore are identical to 4.5GHz.

 

[homerdog]

I think even at stock that CPU should run games very well. My 3770K runs at stock and it rocks every game at 60+fps (I got it for free else I would have gone with the regular 3770). It's not much faster than your CPU.

Of course I'm assuming you aren't doing other CPU intensive stuff, in which case there is no such thing as fast enough

 

[Mize]

Nah, not really. I occasionally transcode stuff, but typically I just game at 3800x1920...I return to Civ V an awful lot and later in the games the computer AI can take a fair bit of time...

 

[Davros]

whats bclk ?

 

[Grall]

So one question is which component would allow Vcore to drop too low?

BIOS (UEFI?) issue very likely, as it seems to be the component that dictates voltage regulator behavior, even with windows running. If there isn't an option to set minimum voltage in the bios/uefi setup then I'm not sure how to prevent this happening - other than start disabling power saving options, which you seemed unwilling to consider. You might want to try contacting the support department for your motherboard vendor and ask for assistance, perhaps they know what to do... (*big laugh*)

whats bclk ?

"Base clock", I believe. The oscillator from which several other important system/CPU clocks are derived, including DRAM, QPI/DMI, PCIe and SATA clocks.

 

[Mize]

Hey Grall. Yes, BIOS is the first suspect but I've been through every P67 guide and tweaked and tweaked. There is no Vcore minimum setting. You can either set a Vcore that never changes or use the VID from the CPU (essentially a vcore as a function of frequency) and introduce either a positive or negative offset. So at idle let's say VID is 1.05v and at 3.5GHz VID is 3.3v you can then introduce an offset of 0.05 so that at idle Vcore *should* be 1.1 and at 3.5GHz Vcore will be 3.35v minus any Vdroop (so let's say it will really only reach 3.33v.

The problem is that Vcore seems to mysteriously drop at times when at idle. Right now HWMonitor claims that it has seen a Vcore of 0.12v.

So is this a flub in the 12v rail? A motherboard capacitor? A faulty CPU?
I have no clue.

 

[Grall]

Could be that vcore readings is a red herring and the crashes are really because of something else; sometimes these hardware monitoring softwares get things wrong and give bad output data. Have you tried more than one monitor program and verified that you see the same pattern of voltage drops?

Hardware issues are often tricky to pin down without changing things. If you suspect PSU at all, then switch it out. You have more than one rig at your place, don't you? If you have modular connectors you could also try just moving the CPU aux power cable to a different output rail and see if you notice a change. Input voltage should not be able to drop your vcore that easily, there's pretty massive caps inside high-end high-wattage PSUs to accommodate dips in input line voltage, and at idle your system would not draw much power. If harddrive spindles etc were able to suck the caps dry to drop your vcore (unplug any unneeded drives while testing as an experiment! ), you should see other components than the CPU start acting out too at times methinks.

 

[Lightman]

I put my money on Motherboard - sorry

Either you're experiencing BIOS bug connected with CPU VRM regulation or simply VRM's are dying.

Still, changing PSU seems to be the easiest option to try first! Good luck!

 

[Mize]

Yeah. MB seems the most likely. Too bad 1155 MBs haven't dropped in price - if anything they've climbed in price. The thought of swapping out my motherboard right now is just ugh. I've even been postponing my 6-month water-change on my cooling loop...sigh.

 

[Mize]

what are some other tools for monitoring and (preferably) logging system voltages?
Speedfan doesn't detect my temps correctly and the log interval is too long...just want to see if anything else will find a similar Vcore min to HWMonitor.