2500k 4.6ghz IBT stable, fails Prime

PaulyD

PaulyD

PaulyD

Soldato
Joined
24 Mar 2008
Posts
4,654
Location
High Wycombe
I usually advocate Prime blend as the best for a stability test and like to get nothing less than 8 hours out of it before I am content with stability.

Not unfamiliar with overclocking - clocked Q6600 3.6ghz @ 1.344v, Q9650 4.4ghz @ 1.392v and 920d0 4.2ghz @ 1.288v previously but am getting stumped with sandybridge stability at the moment!

Current clock I have is 4.6ghz using turbo, using DVID in bios so no LLC set and using offset to increase voltage. Settings below from memory:

Voltage: Normal (allow DVID)
Offset: +0.060v
VTT: 1.05v
System Agent: Auto
CPU PLL: Auto (set this as got CPU PLL Overvoltage on, so thought Auto would be needed for this to take effect? Probably wrong here and two different things)
DRAM: 1.5v
RAM: set to XMP Profile1 so correct timings, RAM perf set to Standard but was on Turbo previously with same outcome.

The above passes IBT with AVX instruction set at max memory (~6900mb RAM) for 50 passes - no problem. Temps get to mid 70's and voltage on load is between 1.336v and 1.344v - understand low volts for this clock but nothing particularly impressive overall.

Prime blend however will BSOD with 0x124 error code after a few hours. From what I have researched this is typically vcore on SB. I have yet to whack the voltages right up just as a test to see if this is the case but was wondering peoples thoughts on the following. Hits same voltage but temps just get below 70c on hottest core

Why no instability in IBT at this clock after 50 runs of max mem?

Ideas on tweaking VTT "just" in case its this? Same for system agent voltage?

Any help/input appreciated :)
 

PaulyD

PaulyD

PaulyD

Soldato
OP
Joined
24 Mar 2008
Posts
4,654
Location
High Wycombe
OLDPHART said:
When you ran IBT the correct way is to set the number of threads for your CPU not Auto, try running IBT with the number of threads and see if it passes.
Click to expand...

Running on all threads, not auto, task manager also shows 4 cores being used

8 Pack said:
124 is a VTT voltage error usually can also be Vcore related. Increase VTT to 1.15v ish. then try Vcore. I would also set internal PLL over volting to enable.
Click to expand...

CPU PLL already enabled (mentioned in OP)

Neil79 said:
1.05vtt is to low, anything upto 1.2vtt is safe :), on our gigabyte boards we use 1.2v vtt
Click to expand...

From further research today and some feedback on another board it was looking more like VTT being the culprit rather than vcore so will be doing some further testing tonight but should be able to nail it soon :)

Cheers guys
 

PaulyD

PaulyD

PaulyD

Soldato
OP
Joined
24 Mar 2008
Posts
4,654
Location
High Wycombe
Justintime said:
Latest prime 27.4 uses AVX instructions and is much much more brutal on the IMC etc.. than LINX/IBT etc.. VTT is a likely culprit.
Click to expand...

Noticed that but didn't think it would be more brutal than IBT in general. Guess IBT can be used to quickly find vcore issues and then use IBT to find IMC/VTT issues

Justintime said:
Possible but i've seen mostly VTT related issues with the newer versions of prime, had to revise a few overclocks when that came out.
Click to expand...

Interesting to know, believe its yourself that has mentioned a few times at the amount of rigs you build (for clients?) so good feedback from someone who has clocked a lot of different builds

wazza300 said:
yeah,ive noticed though on gigabyte boards they tend to need more vtt and dram,but on this asus ive only needed to up the dram voltage,vtt is stock on 1.05v and thats with 16gb @2133mhz
Click to expand...

Thats the reason I also believed the stock 1.05v would be fine for 8gb @ 1600mhz.

Didn't get round to testing last night but will bump the VTT up and see how it goes
 

Justintime

Justintime

Justintime

Soldato
Joined
24 Jul 2006
Posts
8,876
Location
Hoddesdon, London, UK
PaulyD said:
Interesting to know, believe its yourself that has mentioned a few times at the amount of rigs you build (for clients?) so good feedback from someone who has clocked a lot of different builds
Click to expand...

lol i know, have to mention it sometimes to sort of backup my position, makes me feel a bit like i'm showing off but i mess with a fair amount of hardware. Yes i build systems on the side and its part of my job for the specalist side of things at work (video editing, music etc.. but those aren't overclocked :) )
 

PaulyD

PaulyD

PaulyD

Soldato
OP
Joined
24 Mar 2008
Posts
4,654
Location
High Wycombe
Justintime said:
lol i know, have to mention it sometimes to sort of backup my position, makes me feel a bit like i'm showing off but i mess with a fair amount of hardware. Yes i build systems on the side and its part of my job for the specalist side of things at work (video editing, music etc.. but those aren't overclocked :) )
Click to expand...

Not showing off, just proves you can back it with experience which is important :)

I have built and clocked a few systems in my time but Sandy is new to me and from what research I done seems to have its completely own quirks, which to be honest, all platforms do. Far easier on 1155 than messing with straps and dividers on 775 though!

dougal1331 said:
I'd say that for 4.6gig, Vcore should be nearer 1.36- 1.38V...
Click to expand...

All chips will be different and i will only raise the vcore if its the last resort. I prefer to test and find the absolute lowest for my chips. Seeing as I can pass Max IBT 50 times and get a few hours into Prime blend makes me think this won't be completely vcore related.
 

PaulyD

PaulyD

PaulyD

Soldato
OP
Joined
24 Mar 2008
Posts
4,654
Location
High Wycombe
Bump of VTT did the trick, not quite a full pass of Prime blend but good enough for me, especially when it would BSOD after a couple of hours. I tried dropping vcore one nudge and it BSOD on the small FFTs so definitely the lowest I can go with this :)

46primestable.jpg
 
You must log in or register to reply here.
Back
Top Bottom