How to build a 1500 Euro trashbin! (Could use some help.)

Hello OCF!

posted this on the nvidia forums at first when i suspected the gfx to be the problem:

Complete new PC with these specs:

ASUS P6X58D Premium, Socket-1366
1TB Western Digital Caviar Black 64MB SATA III (x2)
Intel Core i7 950, 3.06GHz
Corsair 12GB (3x4096MB) 1600MHz VENGEANCE (9-9-9-24)
KFA2 GeForce GTX 580 1536MB
Corsair HX 750W 80+
Lian Li Lancool PC-K56,
Samsung SH-S223C 22X DL

Everything is set to auto in BIOS since im pretty much a novice of bios management, suspect the ram settings might be wrong.

No hardware errors or anything that would seem to be off and no signs of damage on anything.

Installed win 7, all the drivers and everything seemed fine, but when i tried playing games (League of Legends and Darkfall so far.) ive been having sub-par performance (LoL dipping between 20-120 fps, Darkfall being pretty random but about the same as my 4yo pc.) and several problems including screen going black and resetting drivers, massive amounts of artefacts and 4 full hangups.

Thought it was software related so reinstalled everything, tried beta drivers etc no change. Put in my old ATI 4xxx card but had the same amount of artefacts, which rules out the gfx card for that.

Put the 580 back in and installed Win Xp x64 and are currently running it, no artefacts but the performance is still ****-poor and ive had one bluescreen about nva4_disp.dll or something (so far only when ive had the case side on which would mean overheating i guess, currently testing a few days without side.) and two times with textures tearing and using textures from somewhere else (portrait textures for minimap, character texture for the terrain etc.)

Ran a couple of tests, furmark for 25 min, case open, 83*C no problems at all. Ran the CPU stability test to rule out that, no problem whatsoever. Ran 3dMark 06 and got 21022 points, which is what, 5000 below the WORST online records with my setup. Ran it this morning again after reinstalling DirectX and get 20740...

i checked the KFA2 website: http://www.kfa2.com/gtx580.html

"Minimum 600W or greater system power supply (with a minimum 12V current rating of 42A" (Asked a friend to look up mine and apparently it have 60A+ something)
and "Overclocked products as standard offering"

Sent them a message to see if they know anything about it.

The thing is though that with Win Xp i dont have the artefacts anymore, just generally "crap" performance (and so far the crashes/bluescreen only occur when i have the case side on, figure its overheating.) as a persistant effect.

Sadly i dont have any other PSUs to test with around, except for another 750w Corsair, but many years older version and probably not the same specs.


ill look into new bios and vbios but i do start to suspect the motherboard (Dont want to randomly send RMAs to the shop i bought from however cause they charge 25euro for each part that they cant find anything wrong with. Greedy bums.)

Aright so i updated my bios, turns out there was one update you had to do manually for some reason, now im getting a buckload of wierd windows errors and applications not running as they should, did a 3dmark 06 test again and am up at 21646. Better but just seems completely random to be honest.

Probably do another try on Win 7 tonight with the proper installation cycle. Cant find anything about a new vbios though.---

Win 7 installed, back to artefacts, whoopde****dedo.

TL; DR Version:
I suspect something is wrong with the motherboard or the RAM settings, anyone have some insight or settings i should/could test before i send RMAs? (check system up top.)

Thanks in advance.

Is the 83*C CPU or GPU temp?

I would run Prime95 or Ortho (assuming the 83*C isn't the CPU) for at least 12 hours to help rule out cpu problems

Run Memtest86+ for a couple full loops to rule out memory issue.

IMHO it sounds like a heat or memory issue.

jmh547 said:

Is the 83*C CPU or GPU temp?

I would run Prime95 or Ortho (assuming the 83*C isn't the CPU) for at least 12 hours to help rule out cpu problems

Run Memtest86+ for a couple full loops to rule out memory issue.

IMHO it sounds like a heat or memory issue.

Click to expand...

GPU temp during stress testing on Furmark for 25 min, apparently mine should run below 85 to work good, though i am worried that with my case side on it is much more unstable.

Did i mention the insane amounts of program crashes and random errors im getting ?

Ah and also afaik the ram is probably not running on the timings listed, dont really know how/what to change for it however.

as long as the mobo is set to auto it should get the correct timing from the SPD. i am not familar with that mobo so i really can't give you any direction on how to check.

what temp is your cpu running at?

i forgot that cpuz shows ram timings. Download THIS, run it, and go to the memory tab

jmh547 said:

i forgot that cpuz shows ram timings. Download THIS, run it, and go to the memory tab

Click to expand...

DRAM freq 534.6Mhz
CL 8
tRCD 8
tRP 8
tRAS 20
Command Rate 1T

no idea if you need any other info.

CPU temp is around 38-40c with no particular load, windows and some random programs. around same with mobo and HDDs. during load of 10 min in darkfall CPU is around 50, each core vary 2-4+- gpu at 73C, normal worktemp. mobo and hdds unchanged. (all according to EVEREST)

Restart your computer, hit the delete key when the computer posts to enter the bios. Hit the right arrow key once the bios comes up and hit the down arrow until you see DRAM Frequency, change it to 1600Mhz and then scroll down again until you find DRAM Voltage, make sure it is set to [1.5] once you confirm this hit F10 to save and exit. Once you have done this check CPU-Z again

I recommend downloading and runningMemtest86+ over night just to make sure the memory is ok. After that run Prime95 for 12 hours minimum.

jmh547 said:

Restart your computer, hit the delete key when the computer posts to enter the bios. Hit the right arrow key once the bios comes up and hit the down arrow until you see DRAM Frequency, change it to 1600Mhz and then scroll down again until you find DRAM Voltage, make sure it is set to [1.5] once you confirm this hit F10 to save and exit. Once you have done this check CPU-Z again

I recommend downloading and runningMemtest86+ over night just to make sure the memory is ok. After that run Prime95 for 12 hours minimum.

Click to expand...

QPI/DRAM Core Voltage or DRAM Bus Voltage?

There is a buckload of others too but i guess you dont mean them, like DRAM DATA REF, CTRL REF etc.

DRAM Bus Voltage.

I believe QPI = VTT which you shouldn't need to touch unless you plan to overclock.

jmh547 said:

DRAM Bus Voltage.

I believe QPI = VTT which you shouldn't need to touch unless you plan to overclock.

Click to expand...

Oh nice, this looks better:

DRAM freq 801,4Mhz
CL 9
tRCD 9
tRP 9
tRAS 24
Command Rate 1T

Was the voltage set at 1.5? If it was lower that may have helped with the problem.. of not then you still have some digging to do.

jmh547 said:

Was the voltage set at 1.5? If it was lower that may have helped with the problem.. of not then you still have some digging to do.

Click to expand...

Everything had [Auto] so hard to say, gonna keep testing and running stuff and see what happens.

Right, this didnt really help anything im afraid. No performance issues fixed and just now the pc died again. sounds die, screen goes black, yet the power is still running in the pc, fans and lights are still on.

When i turned it back on it said overclocking failed f1 to setup bios etc. Remember my old pc could just randomly say that when i was booting too.
Still running the same settings though.

Dont get it at all.

So no other news, games still dropping in fps, low performance and the stability is ****.

Gonna try to check with memtest and prime95 to see if i can find something out.

Sounds like a good start, but from your symptoms I would say most likely a CPU RAM MB problem. I would make sure all memory modules are fully socketed and the cpu as well as checking the the MB CPUsocket contactors for damage. I would also confirm HSF is appliying proper pressure and flush against the CPU face when you re-assemble. also to the Forums.

TheNewbie said:

Sounds like a good start, but from your symptoms I would say most likely a CPU RAM MB problem. I would make sure all memory modules are fully socketed and the cpu as well as checking the the MB CPUsocket contactors for damage. I would also confirm HSF is appliying proper pressure and flush against the CPU face when you re-assemble. also to the Forums.

Click to expand...

Im not sure i get what you mean, fully socketed? you mean that they are properly inserted? HSF? Im a bit new to this.

Ill make sure to re assemble everything again and check for damages.

Also i heard that my MB might not be compatible with my RAM brand at that exact freq, But is this something that would give me these kind of problems, bad performance, randomly dying pc and generally strange errors and graphical issues?

Had my eyes on another MB before but got this as a gift and the giver was a pc service company saying it had the same and better specs.

Any ideas?

Right. So now i have parts all over the place.

Any more ideas before i randomly send back parts starting with the motherboard?

Try leaving only 1 RAM stick in your 1st slot and see how this goes.
Also if you have a mate who could lend you another motherboard to test the components in there, that would be a good thing.
If everything works on another motherboard or it works good with only 1 RAM stick than its motherboard RMA time.

That RAM is most likely not on the manufacturers Qualified Vendor List (QVL). Very few 3 DIMM 12GB kits are tested on motherboards. If you have set the speed, timings, and DRAM Bus Voltage manually... then the next step is to add QPI/DRAM Voltage (also known as CPU VTT on other motherboards). I would suspect that auto is not capable of dealing with 12GB of memory, I have seen this before.
I would try bumping the voltage to 1.275V, but you may need more than that... each chip is different. The main concern in adding additional voltage to compensate for the Ram is going to be the increased Core Temperatures. If you are still using the stock HSF (Heatsink Fan) that came with the CPU, then you may not be able to add enough voltage to allow the CPU to deal with the Ram requirements. You may want to invest in an aftermarket HSF anyway, as it will be much quieter (and perform much better) than the stock unit. They are not ridiculously expensive, a decent one can be had for $25-$35.

Your plan of action.
1. Enter the BIOS and set the QPI/DRAM Voltage to 1.275V (make sure the Ram is set to spec as well)
2. Boot into windows
3. Run a stress test program. I like IBT (Intel Burn Test) as it's quick and effective. Set it to at least Very High or Maximum stress level and run 10 passes or so.
4. Open a temperature monitoring program. I like HWMonitor
5. While test runs, watch your core temperatures. If they start surpassing 85C, then shut the test down. You have 2 options:6. If the temperatures do not get that high, and the program fails or you get a BSOD then you can increase the voltage a little bit more, and try again.
7. If the test passes without a hitch, then try your programs that were not running correctly before.

Come back and report your findings and we can continue further. One thing you should not do, is run Furmark on a GPU. That test is simply brutal, and is known as a GPU killer. Your 580 actually looks for that program, and as soon as it is started, it downclocks (in an attempt to save itself) Do not run it. If you want to check for artifacts, simply run your favorite game or a 3D benchmark.

Mario1 said:

Try leaving only 1 RAM stick in your 1st slot and see how this goes.
Also if you have a mate who could lend you another motherboard to test the components in there, that would be a good thing.
If everything works on another motherboard or it works good with only 1 RAM stick than its motherboard RMA time.

Click to expand...

I have been testing this and other stuff over the past week, Each day i swapped ram in the slots, 1st day was 1 in 1st, 2nd day 1 in 2nd slot etc.

I am now up to 2 in slots 1 and 2.

1st day i didnt experience that much of the artifacts and the fps seemed pretty stable, but all of a sudden BSOD.
2nd and 3rd day I ran with alot of jumpy fps (going from 20-120 in LoL.) and a couple of BSODs plus a few freezes that would last up to 4 sec.

Sadly i recently moved halfway across the country so checking with a buddy for another Mobo would be hard.

nzaneb said:

That RAM is most likely not on the manufacturers Qualified Vendor List (QVL). Very few 3 DIMM 12GB kits are tested on motherboards. If you have set the speed, timings, and DRAM Bus Voltage manually... then the next step is to add QPI/DRAM Voltage (also known as CPU VTT on other motherboards). I would suspect that auto is not capable of dealing with 12GB of memory, I have seen this before.
I would try bumping the voltage to 1.275V, but you may need more than that... each chip is different. The main concern in adding additional voltage to compensate for the Ram is going to be the increased Core Temperatures. If you are still using the stock HSF (Heatsink Fan) that came with the CPU, then you may not be able to add enough voltage to allow the CPU to deal with the Ram requirements. You may want to invest in an aftermarket HSF anyway, as it will be much quieter (and perform much better) than the stock unit. They are not ridiculously expensive, a decent one can be had for $25-$35.

Your plan of action.
1. Enter the BIOS and set the QPI/DRAM Voltage to 1.275V (make sure the Ram is set to spec as well)
2. Boot into windows
3. Run a stress test program. I like IBT (Intel Burn Test) as it's quick and effective. Set it to at least Very High or Maximum stress level and run 10 passes or so.
4. Open a temperature monitoring program. I like HWMonitor
5. While test runs, watch your core temperatures. If they start surpassing 85C, then shut the test down. You have 2 options:

1. Buy a better CPU cooler
2. Return your Ram and get 6GB set
​

6. If the temperatures do not get that high, and the program fails or you get a BSOD then you can increase the voltage a little bit more, and try again.
7. If the test passes without a hitch, then try your programs that were not running correctly before.

Come back and report your findings and we can continue further. One thing you should not do, is run Furmark on a GPU. That test is simply brutal, and is known as a GPU killer. Your 580 actually looks for that program, and as soon as it is started, it downclocks (in an attempt to save itself) Do not run it. If you want to check for artifacts, simply run your favorite game or a 3D benchmark.

Click to expand...

Thanks im gonna try this when i get back tonight, Also ive already been planning to buy a new HSF and an SSD disk, only been reluctant to order before i know whats wrong. Gonna get one now then though.

ill post again when ive gone thru the step.

Woah, yeah i badly need a new HSF, the test had barely started and the cpu cores shot up to almost 100.

Any tips on one suitable for my model ?

EDIT: Looks like a Noctua NH-D14 is nice, however it seems its way too large for my RAM coolers to fit in. Any good alternatives?