Peaks

Distinguished
Sep 12, 2008
611
0
18,990
Hey,

Long time since I posted on this forum, I am having some trouble.

I have an i5 - 750 oced to 3.6Ghz with a Biostar Tpower i55 board. It is stable in Prime95 but I occasionally get BSOD's in games for a number of reasons, but the most common is because of an ati system driver (according to crash dumps).

I am not sure what the problem is and was wondering if anyone on here could help. I originally thought it was the on board PCI-e controller that was unstable so I tried setting the QPI freq to that of the Uncore but this didn't seem to help.

Volts:

Vcore - 1.240
VTT - 1.160
All others auto


thanks
 

Just a nickname

Distinguished
Sep 2, 2009
230
0
18,690
How long did you run prime?
Did you run it on small FFT (not the default one)?
Stable run is 8-12 hours from what I've read.

Also you can use multiple tress test like LinX. 20 loop when you increase clock/speed, 35+ loop for max stability test.


 

Peaks

Distinguished
Sep 12, 2008
611
0
18,990
thanks for the responses.

VDIMM is currently at 1.66v. But it was at 1.85v (stock for the board) and the problem was still occurring.

I ran prime for about 5 - 6 hours on blend test so I could also stress the mem, I thought that would also indicate problems with the CPU is there were any.
 

Peaks

Distinguished
Sep 12, 2008
611
0
18,990
Im guessing that its not that my QPI and uncore freqs were not the same then? Ok cool thanks, I will try all of these! :)

But..

Increase VDIMM? Really? My BIOS shows values higher than 1.85v in red. So is it safe to run mem on that high a voltage for constant use? Its the same with the Vcore. I know that i5's sometimes require even 1.4v for 4.Ghz, but this is 3.6GHz, didn't think it would need to be that high?
 

Peaks

Distinguished
Sep 12, 2008
611
0
18,990
ok so I am safe to run my DDR3 with anything up to 2v? But that would be pretty hot? I will try but I think the VDIMM is ok, its pretty good RAM.

My temps are currently 72 degrees under prime load with Akasa Nero heatsink.


Should I only be worrying about SmallFFT test? I thought blend would be more important to test the link between mem and cpu.
 
IMO add a little more voltage to the VTT, try 1.18. There's some wierd thing I've read that VTT and VDIMM must be .5V or less different, which actually seems to be true.

It's quite possible it's the RAM tho, but if it needs more than 1.7V then you're much better off just lowering the speed or raising the timings than increasing the voltage.

What's the current base clock, what's the RAM speed and timings set to, and did you use turbo boost or not? I also doubt you'll have to disable EIST, C States etc, but it can cause issues so if you want, get stable with them off but try them turned on after that because they are good features.
 
Oh and as for testing, I find Intel Burn Test to be the most complete and quick test to run. High stress setting, 10 passes should give an idea of stability

Also, that ATI driver crash is generally not associated with the CPU stability. They've reported having fixed the issue in more recent driver releases, so you should definitely try 10.10 if you're not. It can also be a random issue between the drivers and windows, and sometimes a fresh windows install will fix it.
 

Peaks

Distinguished
Sep 12, 2008
611
0
18,990
Hey, here is some more info about my current setup besides the volts I mentioned above.

The baseclock is 180, multiplier x20.
The mem is at 1440,timings:
CAS 9
tRCD 10
tRP 10
tRAS 25

I thought goin above 1.7v for the vDimm was too high, I have been running blend test on prime for 6 hours now and all is fine, surely that means the mem is fine?

Hmmm I will try a new instal of windows also,currently running most 10.10 ati drivers.
 
When running prime make sure you tell it to check for rounding errors(under the advanced menu in prime).

Try Small FFTs it hits the cpu good and hard(without much memory testing, thats what memtest is for :)). At least over night.

NEVER go over 1.65 memory(1.66 generally gives 1.65 on most boards)
 

Peaks

Distinguished
Sep 12, 2008
611
0
18,990
Yea my bad my sig is out of data, I have updated it now.

Ok I will try Small FFT with rounding errors. In the past tho, I have just run small FFT tests and found the CPU to be stable but then when running a blend test I find prime to fail.

@ Wolfram23: I also have turbo boost disabled aswell as all C-States and Intel Virtualisation Tech (whatever that is).
 

Peaks

Distinguished
Sep 12, 2008
611
0
18,990
Hey guys,

No response for a while as I have been carrying out all the advice that you guys have given me.

First off here are some shots of the memory and SPD tabs from CPU-z:





Here again is my current settings:

BCLK = 180
Multi = x20
Vcore = 1.232v
VTT = 1.160v
VDIMM = 1.166v
All rest is on auto.


With these settings I have run small FFT and blend tests on Prime both for 6 hours, 20 tests with memtest and 10 tests with Intel Burn using both standard and maximum test levels, all of which passed. So I guess I am stable.

I still think that because this BSOD only occurs in games that it might have something to do with the integrated PCI-E controller. To get this stable I guess I need to increase Vcore?

The only thing left to do now is what Wolfram23 suggested which is a fresh install of windows 7 and latest ATI (sorry AMD :p) drivers.

On a side not, does anyone think I would be able to improve my mem speed/timings? Based on the CPU-Z screen shots above I would have thought that I could get things tighter, although I have read that the difference in real world performance between CL9 and CL8 is minimal, its all about getting to CL7.
 
With a memory voltage of 1.65(1.66 on some boards) you can get at lest 8,8,8,24.

You may be able to get it to 7,8,7,24. You will only be able to test that by setting it and running memtest86(or memtest86+) overnight.

Does this problem occur if you revert to stock settings?