OC Help w/DK X48 and E8400

[Zayla74]

More Problems I made a post a couple days ago asking about temps and voltages trying to OC my E8400 to 3.6 but seems I may have other problems but im not sure.
im running vista 64bit, DFI LP DK x48 MOBO, 2x2GB Corsair XMS2 pc6400 dhx,and a E8400.
Current Settings are 333/800 380 FSB CPU 1.2V according to CPUZ. DFI Guardian shows 1.19

For 3.4 Ghz:
I set at 333/667 (4-4-4-12 timings and 780mhz mem) Orthos ran for 12 1/2 hrs all OK.
I reset too 333/800 and set FSB to 380 (letting RAM run at 5-5-5-18 910mhz) and was ok for 15hrs.
However, when using the Prime95 64bit version It will fail at either setting after 10min or sometimes 45min. Somewhere in between never same time. It give a Hardware Failure Rounding error. Either of these 2 settings same result.

This is the same error I would get trying to reach 3.6 400/800 ratio.

I wanted to rule out RAM so I done 2 tests with MEMtest .. one at 4-4-4-12 timings running at 780mhz and a second at 5-5-5-18 timings running at 910 mhz. Both Memtests ran 8hrs or a little more without error.

Another thing I found odd... is at the settings I have now, which is a 380 FSB, as I said Orthos shows all ok after 15 hrs, Prime 95 64bit will fail within 10-45 min.
However ... If i increase the cpu voltage .. Prime 95 fails almost immdiatley.

When I have tried 3.6 even Orthos would give an error after a little over a hour, so I may have a voltage issue here. Prime 95 64bit does same tho it may run 1 min or it may run up to 45min. Never same time.

Im wondering if one of the other voltages may possibly be off or if maybe Prime 95 64bit is just bugging out on me.

Voltages I have now:
CPU: 1.2 FSB VTT: 1.1 NB Core:1.26 VDimm: 2.11

Im also showing 38C idle and 53C max at full load according to DFI Guardian ... (48C Max using real temp) at current settings.So Temps seem ok...

With a Vcore of 1.28 tho it gets much higher ... DFI Guardian reporting as high as 63C :-/
Real temp seems consistantly 5C lower ...

Would these Voltages seem normal For a E8400 @ 3.4Ghz up to 3.6Ghz ... Or would some be off somewhere?


Also: My Ram is rated at 2.1 V so I set it there. Just curious tho if running at 4-4-4-12 timings and 780mhz-800Mhz as rated is better than say dropping to 5-5-5-18 and running at 910-940 or so?

 

[Nenu]

I'll try and sum that up so its easier to go through.

3.4Ghz clock runs stable in Orthos no matter what the ram config.
3.4GHz clock does not run stable with Prime95 64bit.

Memory doesnt appear to be the problem as it passes Memtest.
Increasing CPU voltage makes Prime fail quickly instead of 10 to 45 mins normally.

At 3.6GHz, Orthos fails after 1 hr, Prime95 fails in under 45 mins.

Vcore = 1.2
Temps are well within spec

When running 64bit code you are using parts of the CPU that arent used with 32bit software.
Some of that silicon looks like it cant run quite as fast, unfortunate.
If you are going to run 64bit code, you may not be able to get as high overclocks.


Can you explain more about when raising the CPU voltage, it makes Prime fail faster?
This doesnt make sense as the CPU is nowhere near its maximum safe voltage and is not running too hot so higher voltage will not be detrimental.
That is unless your PSU is flaking out, in which case try another PSU to confirm.

Disregarding the raising Vcore issue...
It looks like you have hit the limit of your chip at 1.2V (with your cooling).
I wouldnt expect you to be able to get much more than 3.6GHz at 1.2V.
(I'm running 4.13GHz at 1.4V)

Raise the voltage more, up to 1.365 Volts if you wish to remain safe and keep temps at least 25C below TjMax (TjMax = 95C using Realtemp).
ie in Realtemp keep distance to TjMax more than 25C.
Those are the safe rules, if want to push it you can step outside those bounds but its best to read up others experiences first.


btw dont bother making your memory run as fast as possible.
There is a performance penalty when you dont run the ram 1:1 with the FSB so you need to run the ram seriously fast to overcome the penalty.
Leave memory running 1:1 with the FSB and reduce the voltage to the maximum it needs to remain stable (after you have the system completely stable first!).

 

[Zayla74]

Thanks for the info. I was very curious about the RAM issue, I dropeed back to a 1:1 ratio.

As far as the voltages go ...

I done a bit of research about them, and what others are running at. It would seem the CPU VTT setting many are running at 1.2-1.3. Mine is set at 1.1.
The NB core I seen many others running at 1.35 - 1.55 mine is set at 1.26.

So I bumped up the CPU VTT to 1.2 (my guardian shows 1.18) and bumped my NB to 1.34 from 1.26 it was initially.

with 1.24 CPU core .. and 400 FSB 1:1 ratio Prime 95 64 bit ran 50 min then i got a rounding error.

I went back to bios and bumped the Vcore 1 notch .. ((My board doesnt have a direct Vcore setting .. just a special add ... default setting is 1.1))

When I get into windows Prime 95 failed with a rounding error after 30 seconds.
Reboot and give Vcore another bump and its running now ... Tho only 10 min.

That what I meant aboput voltages going up and failing faster .. just odd .. but by the same token errors in prime95 do not always happen at same time :-/

Quick question tho ... is a CPU VTT @ 1.2 and NB Core @ 1.32 safe??

I do not know much about these settings but i seem top be getting farther under Prime than B4. 40min was highest it would run before at higher vcore.

As far as PSU, I just put in a new one along with the new RAM I got. Its a Corsair 650W ... should be enough or at leeast hope so :-/ Its 650W with a single 52A 12V rail ..

Also, Im wondering if I should get stable in Orthos and let it be or try for Prime 95 stable ... Orthos will stay stable for hours but I cant get Prime to be. Just irritating .. . Just wanted 64 bit stable since running vista 64 .. but i dunno

 

[learners permit]

Post a screenshot of your desktop displaying smart guardian, memset and cpu-z's main and memory tabs. That should help us see where things are.

 

[Zayla74]

CPUZ Main:

As I posted before defaul the NB Core was 1.26 and i upped it to 1.32
CPU VTT default was 1.1 and I upped it to 1.2 in bios.

At this setting ... DFI Guardian is peaking at 64-65 C at times But Realtemp is showing a max of 60C. Staying around 58C for most part ... Is this OK?

Not sure if this is normal .. but im certain im not doing something right

 

[Zayla74]

WOW,,, Sorry the pics were so big

Also, if it matters ... at idle CPUZ shows 1.29 Volts to CPU. Vdroop im guesiing why lower under laod

 

[Zayla74]

Well, once again after 45 min (give or take) Prime 95 gets a rounding error.

It seems to be giving me same error around same place each time from 1.25V to 1.3V @ 400FSB.

I can get thru 3dmark06 .. but im guessing thats nbo sign at all of stabilty.

Oddly ... I drop back to 380 FSB (3.4GHZ) and Prime 95 gives rounding error at that speed as well. Same voltages even.

No clue whats going on here It seems only the 64 bit version does this. I can get stable 380FSB 3.4Ghz 1:1 w/Orthos ... but havent tried to much w/Orthos @400 FSB.

Any ideas?

 

[learners permit]

Memset will display some important secondary timings and current performance level that can make or break an O/C so it would be helpful to see them too. I'll give everything a closer look l8r as I'm at work now.

 

[Zayla74]

Thanks learners. I dont have that program, but i will get a copy and post that info for ya.

I reverted back to a 380 FSB, 1:1 and 4-4-4-12 Timings for the time being.

I did take a few different approches yesterday all with same results.

i tried dropping to a x8 multi and a 450 FSB and 1.28V to cpu.
All fired up and went great but as always ...
64 bit of Prime gives rounding error 45 min or so in. Everything I have tried seems to give the error at this point .. when i think round 4 starts .. the 894 itterations i think it is ...

Would just trusting Orthos be a good call? I can get it to pass for hours?

Im just wanna be able to pass the 64 bit version being as i have vista 64 ...

 

[Zayla74]

Ok, Im beggining to think my whole problem here is simply Vista 64bit OR 45nm CPU + Prime 64bit for some reason.

I bumped up a bit and at all stock voltages... except RAM and CPU, I got to 390 FSB easily. 400 as well tho not totalty stable. (seems to be just a vcore isssue but havent tested much therre)

However .. at 390 fsb and 3.5 Ghz I ran through OCCT for the 2hr test OK. Also, I ran Orthos 8 1/2 hrs no problem. (first run w/orthos failed after 20min. 1 bump in vcore got me 8hr stable)

So .. this morning i ran through 3dmark06, and Vantage and got through tem with decent scores.


BUT ... I cant get through Prime 95 64 bit still. either less than a minute, 2 min, or as much as 45 min.

So, To be sure, I ran one pass of memtest85+ ... 1 stick at a time and passed.
I then set all clocks and voltages to STOCK.
run Prime 64 bit and fail in less than a minute.

When I say it fails the 64 Prime Prime Im refering to the blend test BTW.

I find it odd That this is the same way it acts at any FSB/Voltage.

I hope its not my board or chip thats bad But if they were would it not show problems somewhere at some point?? I mean ... passing OCCT and Orthos , and system itself seems stable.

Anyhow, I know others have said elsewhere the 64bit works for them .. but is it possible it has something to do with 64bit vista + 45nm CPUs?

Any help appreciated.

 

[Nenu]

Its possible some part of the 64bit silicon in your chip is out of spec/damaged.

 

[Zayla74]

If that is the case ... which I was afraid might be ...

If I run through the small fft tests, and there was a 64bit area of the cpu thats damaged or out of spec .. Would the errors not show up there?

 

[Nenu]

Anything could happen from wrong results, application crash to complete OS crash.
It depends on what has failed or is out of spec.