P4 2.4C m0 on IC7 - Need Some Help with OC

[Rhombuss]

Here's my setup:

2.4C m0 stepping /w Alpha PAL8942
IC7 with 1.9 BIOS
2 x 256MB OCZ Rev. 2 PC3200

CPU FSB = 267MHz
CPU/MEM Ratio = 5:4
MEM FSB = 214MHz

Memory running at 2.5-3-3-7

VCore = 1.575V (undervolting to around 1.55 V)
VDDR = 2.7V (in hardware monitor, shows as exactly 2.7V)
VAGP = 1.55V

-Hyperthreading and APIC enabled
-AGP/PCI ratio fixed
-All GAT options on "Auto"
-Temperature hardware monitor hovers between 40 (idle) to around 56 (load)

Currently prime stable at 3.204GHz. Going between 267-280 FSB (3.204 to 3.36 GHz) results in instant Prime95 crashes, but Windows appears to be stable. Going between 280-285 FSB (3.36-3.42 GHz) results in crashes either before, during, or after WinXP loads up. I'm trying to hit 3.45 to 3.5 as a target, so what else should I do within safe bounds? What's a safe limit to push the VCore, and does Hyperthreading make overclocking more difficult are higher FSBs?

I've run memtest with my RAM at 1:1 up to 220, and when I push my CPU above 3.204 GHz, I set the CPU/RAM ratio back to 3:2, so I'm positive the RAM isn't limiting here. It's most likely how much voltage the CPU is getting.

Just wanted to get one more thing clear. The Abit boards undervolt the VCore, so does that mean that the VCore that you set is undervolted and the voltage detected by the BIOS hardware monitor is the correct one? Because my hardware monitor always shows about 0.02V less than what I initially set.

Or is the BIOS hardware monitor ALSO not showing the correct voltage, and that the actual voltage is even less than what the monitor shows? So then technically there is no way of knowing the real voltage without connecting a voltmeter?

 

[Technonut]

All GAT options on "Auto"

Set the last two to disabled.

Set your VDIMM to 2.8V, the VAGP to 1.6V, and see how far you can get with the RAM @ 3:2. I would not push the VCore past 1.65V. (BIOS)

I have been through 5 2.4C M0 processors, and only had one rock solid @ 3.42GHz. The rest topped out around 3.3GHz. The best one out of those 4 ran @ 3.3GHz / 1.525V. (Will run 3.36GHz with increased VCore) I am seriously thinking about keeping that one for everyday use since I have had to increase the VCore to 1.65V with the one that does 3.42GHz. I would rather run one @ 3.3GHz / default VCore....

EDIT: I did try to disable Hyperthreading at a couple of points during testing the processors... It really did not make a difference in the maximum OC... Once the stable limit was found, that was it...

 

[Rhombuss]

Thanks for the info Technonut.

I tried those settings, and can post at 285FSB (3.4 GHz) at 1.625V, but crashes once it gets to the blue Windows loading screen. I upped the VCore to 1.65 and it gets into Windows fine, but fails Prime95 instantly and succumbs to random crashes. It still seems that 267FSB is my limit for complete Prime95 stability.

A few questions though - what is the purpose of bumping the VAGP to 1.6V? That, and how much credibility should be placed in a Prime95 torture test? I know that the longer it runs without errors, the more stable the system is up complete stability with no errors. But if it fails instantly, but "appears" to run everything else perfectly fine, even Sisoft CPU/MEM benchmark loops, is it really that much of a concern?

 

[Duvie]

The raise in vagp can help as it is tied into the vcore of the NB chipset....

The other concern maybe the heat of the system....The monitor you are using only shows "sys" which is not the NB chipset but a point on the boad that measures ambient air temperature in the case....The cpu is the cpu ...The pwm is not the nb chipset or a power temperature but a temperature taken around the power mosfets on the mobo....

Try to keep the sys temps below low 30's for load....Keep PWM temp around 40 and no more....I got rid of my alpha cause it could not cut it anymore, what are your temps being reported as???


I would look into placing some artic silver on the NB core and see if that can help a bit.....



Prime95 should not be the solve determination of stability on a board cause it cannot test and stress all areas of the system...However when it gives you an area yuo should consider it.....I recommend using older version 22 or make sure you are running version 23.8 as the older version 23 had some bugs....Overalll in my opinion the version 22 is bettar as the ppl at prime95 are getting the program too complicated and optimised for different procesors they have started to enter quirks and bugs....It didn't changing but like always they mess up something that wasn't broke....

What they need to do is make it so I don't have to run 2 instances of it to get 100% loading....Are you runnin 2 instances as well???

Update: Also my board undervolts closer to .045v as It says 1.625v in bios and reads 1.58v in the monitor as well as in cpu_Z_116.....

Pleaese read us the voltages being deliered at idle and load from the monitor program..Also waht is the specs on the PSU??? You PSU may not be cutting it....Don't tell me another antec true power as I have seen more of those of late not cutting it in this range...My enermax has a much higher amps on the 12v rail and I think that is a reason I only drop to 11.88v undr load and 12.04 under idle at 3.5ghz....

 

[Rhombuss]

Temps for the CPU are around 42 idle and 56 on full load. I'm assuming the Abit is still using the on-die thermal sensor, and not the external sensor, so these temperatures aren't the same as what most people are used to seeing.

I never bothered running dual instances of Prime95, since it isn't even stable with a single instance. The BIOS monitor undervolts to around 0.025V, but the Abit monitor program shows pretty closely the same voltage as the BIOS voltage option. I'm running an Enermax EG365P-VE, and the voltages on each of the rails is pretty close, within 0.03 V of each rail, so that shouldn't be a problem.

 

[Technonut]

Well, you seem to have the bases covered.... Hell, I even fitted a BH7 Northbridge HS retainer to secure a copper NorthPole NB cooler to my IC7-G when testing my M0s. You should try applying some Arctic Silver to your NB HSF... It may help some.

These SL6Z3 processors are just like any other... Luck of the draw. I have seen quite a few posts on forums about some doing great, and some that are struggling to do 3.2GHz. (Or less) Personally, I am getting tired after going through 5 of them for the "golden one".

I think I will be satisfied (For awhile at least ) with the one that does a 900MHz OC @ default VCore. I could throw my watercooling gear back together and try to reduce the VCore required on my 2.4C M0 @ 3.42GHz, but don't really think it is worth the hassle.

Good Luck..

EDIT: As far as Prime95 is concerned... I take it seriously. My idea of stability is being able to run ANYTHING without crashes or errors. Since P95 uses known good results in testing, if it finds an error, something is not right... 2+2 does not = 6...

Just for the helluvit, try running some PCMark2004 and do some encoding with TMPGEnc... I find either will show any OC'ed weakness quickly.

 

[Duvie]

Originally posted by: Rhombuss
Temps for the CPU are around 42 idle and 56 on full load. I'm assuming the Abit is still using the on-die thermal sensor, and not the external sensor, so these temperatures aren't the same as what most people are used to seeing.

I never bothered running dual instances of Prime95, since it isn't even stable with a single instance. The BIOS monitor undervolts to around 0.025V, but the Abit monitor program shows pretty closely the same voltage as the BIOS voltage option. I'm running an Enermax EG365P-VE, and the voltages on each of the rails is pretty close, within 0.03 V of each rail, so that shouldn't be a problem.

Ok...you don't follow instructions...I want to know the sys and pwm temps as well...i can read and saw your cpu temps mentioned above...I also want a list of the voltages at load, so you will need to find an app that loads 100% of cpu (prime95 one instance only loads 50%, as well as superpi loads only 50%)....Run tmpgenc or combination of say superpi32mb and another app to try to get heat up there....There is no way your voltages only fluctuate .03v with a 365 enermax (350watt) at load....The amps on the 12v are no where near as high as mine on the 465 and I fluctuate around .16v...

The fact is you haven't likely even seen the high temp since you can't run 2 instances of prime95 (a big heat stressor).....


frankly you haven't ruled much of anything out...


One follow up...In both of my p4c setups 2.6c@3.24ghz and 2.4@3.5ghz having HT shutoff in bios did not allow me to oc any further on my setup....HT should likely take a bit more power and increase temps a slight bit more since in fact you are running the chip more thoroughly and completely then you may otherwise with HT off...I notice about 1-2c under load so not much....I didn't notice any voltage on my 12v line but in theory it is likely it should irk a bit more...

 

[Technonut]

It may have something to do with the PSU, but even on my IC7-G with Tight Voltage Regulation, once the OC ceiling was hit, that was all she wrote... I tested before with the window in my office open when it was cold. The PWM temps as well as all other temps were low.. did not matter that much in my case. But it does not hurt to try different things.

Duvie: what kind of OC increase did you get with HT disabled? It really did not affect my OC's one way or the other with the M0s.

EDIT: BTW, the above screenshot was taken under P95 load wth the VCore set @ 1.60V, and the VDIMM @ 2.8V...

EDIT: Here is a current WHD pic @ 3.3GHz / 1.525V (BIOS) with my IC7-G. Dual P95 running over an hour in a hot room. Note the PWM temp... System is solid as a rock...

 

[Rhombuss]

Originally posted by: Duvie

Originally posted by: Rhombuss
Temps for the CPU are around 42 idle and 56 on full load. I'm assuming the Abit is still using the on-die thermal sensor, and not the external sensor, so these temperatures aren't the same as what most people are used to seeing.

I never bothered running dual instances of Prime95, since it isn't even stable with a single instance. The BIOS monitor undervolts to around 0.025V, but the Abit monitor program shows pretty closely the same voltage as the BIOS voltage option. I'm running an Enermax EG365P-VE, and the voltages on each of the rails is pretty close, within 0.03 V of each rail, so that shouldn't be a problem.

Ok...you don't follow instructions...I want to know the sys and pwm temps as well...i can read and saw your cpu temps mentioned above...I also want a list of the voltages at load, so you will need to find an app that loads 100% of cpu (prime95 one instance only loads 50%, as well as superpi loads only 50%)....Run tmpgenc or combination of say superpi32mb and another app to try to get heat up there....There is no way your voltages only fluctuate .03v with a 365 enermax (350watt) at load....The amps on the 12v are no where near as high as mine on the 465 and I fluctuate around .16v...

The fact is you haven't likely even seen the high temp since you can't run 2 instances of prime95 (a big heat stressor).....


frankly you haven't ruled much of anything out...


One follow up...In both of my p4c setups 2.6c@3.24ghz and 2.4@3.5ghz having HT shutoff in bios did not allow me to oc any further on my setup....HT should likely take a bit more power and increase temps a slight bit more since in fact you are running the chip more thoroughly and completely then you may otherwise with HT off...I notice about 1-2c under load so not much....I didn't notice any voltage on my 12v line but in theory it is likely it should irk a bit more...

Both of the other temps average around 30 degrees, and don't change much (1 or 2) under load. I never bothered running dual instances of Prime95 because it's not even stable running a single instance - get my point? Once I get a single instance running perfectly stable above 3.2GHz, then I'll attempt a dual instance. And about the voltage fluctuation, I mean 0.13V, not 0.03V - my bad.