major issues with whole system, need advice

[grumperfish]

For software monitoring, download the latest beta version of HWiNFO64 (set to sensors-only) instead of using HWMonitor. HWiNFO is generally much more accurate.


I wouldn't bother running Prime95 to test for CPU stability at stock clocks until you get the RAM situation sorted. It's likely either you're seeing bad contact with the RAM as Diggrr pointed out, improper RAM voltages/timings/ or you just have one or more bad sticks. Try testing one stick at a time individually with memtest for at least 2 runs each, and make sure your DRAM voltages and timings are set to the RAM-specified settings in BIOS.

What's the exact model number of the RAM? If they are 1.65V sticks you'd be running them way below spec at 1.54V (if they're 1.5V sticks, 1.54V would be OK though). Reducing the RAM speed isn't a fix either, as they should be stable at stock speeds. What do your QPI/DRAM ratios look like in BIOS? Also, what is your QPI/VCCIO voltage set to? Auto, or manual?

 

[Neuromancer]

Also note that for memtest testing..

Bad sticks of RAM will fail at any speed or timings. So even letting mem multi and timings on all auto and running memtest will let you know if you have bad memory. Make sure you have uncore and QPI set to auto as well.

 

[CURTO]

For software monitoring, download the latest beta version of HWiNFO64 (set to sensors-only) instead of using HWMonitor. HWiNFO is generally much more accurate.


I wouldn't bother running Prime95 to test for CPU stability at stock clocks until you get the RAM situation sorted. It's likely either you're seeing bad contact with the RAM as Diggrr pointed out, improper RAM voltages/timings/ or you just have one or more bad sticks. Try testing one stick at a time individually with memtest for at least 2 runs each, and make sure your DRAM voltages and timings are set to the RAM-specified settings in BIOS.

What's the exact model number of the RAM? If they are 1.65V sticks you'd be running them way below spec at 1.54V (if they're 1.5V sticks, 1.54V would be OK though). Reducing the RAM speed isn't a fix either, as they should be stable at stock speeds. What do your QPI/DRAM ratios look like in BIOS? Also, what is your QPI/VCCIO voltage set to? Auto, or manual?

Also note that for memtest testing..

Bad sticks of RAM will fail at any speed or timings. So even letting mem multi and timings on all auto and running memtest will let you know if you have bad memory. Make sure you have uncore and QPI set to auto as well.

****S SAKE! i totally replied to you guys like an hour ago but i mustnt have pressed submit. actually i remember now, i previewed it and then did something else and came back and assumed i replied. argh jesus thats annoying. ok well anyways basically i said:

- that is an awesome program thanks!
- i already ran prime for 6.5hrs and it was fine
- i did have my ram under volted (theyre 1.65v bangers) but thats fixed now, thanks
- ratio is 2:10 according to cpuz, is that what you wanted? theres no ratios displayed in my bios.. my last gigabyte board had em but yeah, not asus
- i have my qpi voltage the same as the vcore since thats the way ive been taught previously on this forum. correct me if im wrong (atm its at 1.25v)
- yes qpi and uncore and auto
- ram timings are set to the stock timings as written on the modules

oh and also 3.5 passes, no errors.. weird



i guess now im gonna focus on the gpus, giving them a couple tests n ****. thanks for the help so far.. i dont get why the ram is suddenly ok but?

 

[CURTO]

well i ran 3dmark and checked out some other results with similar specs to mine and it seems as if im getting the right score for my build. so yeah it seems as if everything runs fine now, hopefully i dont get any more problems while playing games and whatnot. thanks everyone for their help especially Diggrr, i dunno what id do without this forum

 

[grumperfish]

- i have my qpi voltage the same as the vcore since thats the way ive been taught previously on this forum. correct me if im wrong (atm its at 1.25v)

This isn't really correct, as the QPI voltage can be independent of vcore, however as long as it's no more than .5V below RAM voltage to keep up with the memory controller (which you've got covered) and somewhere between 1.2V and ~1.35V (due to overclocking) it's fine.

Not sure why the RAM's working fine now - probably a good idea to do an extended memtest run to be safe.


Yup, I was thinking about Gigabyte boards where the QPI/VTT setting can be independent.
I meant that QPI/VTT could be less than vcore; doesn't have to be exact 1:1 with vcore (although QPI/VTT shouldn't be higher than vcore as stated). If you're running at 1.4V vcore, running at 1.4V VTT/QPI is probably unecessary (depending on the chip), potentially dangerous to the IMC, and would just result in higher temps than with the vcore at 1.4V on its own (which with Nehalem is gonna pump out a ton of heat by itself).
VVV

 

[Neuromancer]

QPI=VTT when you have separate settings in BIOS (ASUS does not) is recommended never higher than Vcore. VTT for 24/7 1.35 is recommended max. I think 1.45-1.5 is recommend max for the chips.


as for

i did have my ram under volted (theyre 1.65v bangers) but thats fixed now, thanks

That will do it EVERY time as well

X58 had a really forgiving IMC (I suppose due to the triple channel?) It was so forgiving I was able to run mine with a dead stick of ram in and thought the board was just dropping a channel

Makes disagnosis a bit harder since no other platform would even POST with the dead stick in (P55, P67 and AM3)


Anyway glad you are all sorted

 

[Diggrr]

Glad the other two stepped in actually, I haven't OC'd in so long I don't do anything other than set voltages on my ram anymore. *blush*

Glad it's working now, and yeah, I'd run an extended memtest run too, just to make sure this isn't one high-point in the random trouble roller-coaster.

Have a great time!