OCing 2.4C M0 - New to Intel

[The Coolest]

Hey guys, I recently switched from my XP-M on NF7 to the rig #1 in my sig. The reason for this is that I needed a system that shows up as "SMP" in windows cheaply, and P4 does just that.
Anyway. I'm using the Zalman heatsink at max fan speed with AS5 on the CPU.
I've got the FSB to 271MHz, so its running at 3257MHz right now, memory is at 5:4 divider. Voltage is 1.575v in the BIOS, under load with dual F@H MBM5 reports average of 1.54v (default is 1.525v), NB voltage is set to the default 1.55v in BIOS, MBM5 reads it at 1.6v. Temps read in MBM5/Abit EQ are 50C CPU under 2 instances of prime. At these settings I got over 17 hours of dual prime95 run till I had to stop it.
I tried 3.3GHz with both 1.55v and 1.575v in BIOS and both times prime95 failed exactly after 3.5 hours. I just have a feeling this chip might have some more headroom in it.

What else do you think I can do to squeze a bit more out of this chip. Maybe its the motherboard that's limiting me? I've heard somewhere that this particular board reports temps 10C higher than they're really are, is this true?

[The_man27]

you can up the cpu volts to 1.6 but I wouldn't go past that
up the NB voltage (that helps in stability)

[-_{MoW}_-Assasi]

like i said over msn, its ur choice :)

[hUMANbEATbOX]

that mobo shouldn't limit you, some people ran that thing up past 300fsb.

my sl6z3 takes 1.6v in bios to do 3.45ghz, it droops to (about) 1.56v under load. if i run 1.675v in bios/1.625ish load, i can bench at 3.6, but not prime stable.

to give you a good idea as to if your board is reporting cpu temps correctly, see what it does at 100% stock. i would imagine around 40-45 would be "normal". keep in mind these m0 chips are cut down EE's, they won't be the coolest running chips on the planet (mine can easily hit upper 60's at speeds in my sig, if i don't turn up my fans when i game/encode), but that shouldn't really hamper your oc TOO much (mine can stay stable even at 69c, not forever, but long enough to get through sp32mb).

that's a nice chip you got there, i would imagine you should be able to get 3.4 totally stable after all is said and done. did you get it from the classifieds or something? if so, what does the previous owner say it can do?

[The Coolest]

Well at stock speed MBM5 did report around 40c dual prime.
So should I increase the vcore to 1.6v and should I even bother to touch the NB voltage?

I got this setup from the Anandtech classifieds forum. The previous owner wasn't a heavy OCer, and he had this chip clocked at 3GHz.

[The_man27]

yes, don't hesitate to give it 1.6v
increase the NB only if prime fails or something, but it doesn't hurt your system anyway so I say do it

[The Coolest]

Ok thanks. I'll give this a shot

[The Coolest]

Ok. I put up voltages...
FSB: 277MHz (for 3.33GHz)
NB voltage to 1.65v (MBM5 read 1.7v)
CPU voltage to 1.6v (dual prime95 averaged at 1.56v in MBM5)
failed prime95 after 3 hours. So I'm guessing its the chip afterall, I've been running small FFTs which basically isolates the CPU from anything else that could cause instability in a system...
Any ideas are more than welcome. Thanks for all the help so far guys.

[hUMANbEATbOX]

if its at 1.56v under load, that's not really too far off from its stock setting.

i would try 1.6-1.675 in bios, just MAKE SURE to check what your vcore is at idle. most mobos will overvolt at idle, and undervolt at load. if you set it to 1.65v in bios, and it sits at 1.67v idle, and goes down to 1.62v at load, then you should be safe. that should get you over the 3.33ghz hump.

i have run these voltages on my chip many times over the years, my chip is still going strong, i can still max it out at 300fsb with 1.675v in bios to bench with.

[The Coolest]

Well... I think it was my memory. I just ran 222MHz FSB 1:1 at 2.95v to RAM and it failed memtest86+ after almost 2 hours.
I'm now memtesting it at 230MHz and the max vdimm this board gives (3.2v). I really hope this will help me get more out of this CPU, if 230MHz is stable then at 5:4 ratio I should be able to crank up the FSB to upto 287-288MHz which is ~3.45GHz. So we'll just have to wait and see.

[The Coolest]

Ok, 3 hours memtest stable at 230MHz mem. now testing 277x12 (~221MHz on RAM).
Gonna let prime run over night and see how it'll work.

[The Coolest]

FSB: 277MHz (for 3.33GHz, memory ~222MHz 5:4 divider)
RAM voltage to 3.2v (MBM5 read 3.28v)
NB voltage to 1.6v (MBM5 read 1.66v)
CPU voltage to 1.575v (dual prime95 averaged at 1.54v in MBM5)
failed prime95 after 1 hour.
I guess it wasn't the memory that caused Prime95 small FFTs to fail, 'eh? Any one got any ideas of what else I could do except go crazy with voltage? I do want to keep this chip for a while...

[The Coolest]

With some help from Gautam I found that a lot of 2.4Cs max out at around these speeds, and from the looks of it are northbirdge limited. Unfortunatly the max voltage selectable for the NB in BIOS is 1.65v, which reads 1.7v in MBM5. So I took the stock heatsink off today, cleaned of the thermalpad it had on it and applied some AS5, put it back on and I'm giving it max volts the mobo can.
I'm try to get 3.3GHz stable again, as I don't think its the CPU that is limiting me.

FSB: 275MHz (for 3.30GHz, memory ~220MHz 5:4 divider)
RAM voltage to 3.2v (MBM5 read 3.28v)
NB voltage to 1.65v (MBM5 read 1.7v)
CPU voltage to 1.575v (dual prime95 averaged at 1.54v in MBM5)
Been priming for the past half hour, wish me luck :)

[Jimbob7]

Trying running the ram 3:2, then CPU volts to 1.675 / 1.700 in bios that should get you around 1.65 in MBM5. NB 1.65 is highest that board gives if I remember right. With those settings you should find that CPU's max speed. You may not want to run those volts for your CPU long term, but it should help you find what sort of speed you can get from say 1.62v and takes the ram out of the posable prime95 errors.

[The Coolest]

Well... From preliminary testing, 3:2 wouldn't even POST at stock and slightly OCed settings so i dunno.
Either way after performing what I mentioned in my previous post, I'm still priming for 5+ hours. That's compared to previous 3 and a half.
I think that the NB is what's limiting me right now, not the chip. Its real shame there are no VMods for the NB...

[Jimbob7]

No post for 3:2 thats rather odd. Do you have lastest bios etc? Eitherway, you could prob give the CPU more volts. I doubt its the NB, I as many others ran AI7's at 290+, so the board should have some headway.

[The Coolest]

Yeah I'm begining to think so too...
3300 passed 6.5 hours of prime. 3350 failed after 3.5 hours, I lowered the NB voltage and I'm running prime again to see if it fails after the same time again

[The Coolest]

Ok. I tried 279MHz (3.35GHz) FSB at both 1.6v and 1.65v NB voltage while staying with 1.575v BIOS on the CPU, and both failed prime95 after around 3 hours 30 minutes.
I think that before I applied the AS5, the NB might've been one of the limitations.
Now I upped the vcore to 1.625v in BIOS (1.58v in MBM5 under 2xPrime95) and I'm going to run Prime95 overnight and see if it breaks the 3.5 hour Prime limit. If it does, then I'll just prime it at 275MHz (3.3GHz) at 1.575v and if it checks out after 24 hours I'll just have to settle for that...

[The Coolest]

Update:
Ok Prime95 has been running for almost 9.5 hours. I can't decide, whether to try to clock it up higher at the current vcore setting (MBM5 shows 1.58v average and 1.63v peak, for Prime95, I usually run F@H 24/7, so with F@H the average will be higher by 0.01-0.02v)
Or see if I can get it 24 hour prime95 stable at 3.3GHz and 1.575v and settle for that?

[pscout]

If you are going to be running 2 folding instances on that puppy, I would suggest a folding test. You may get it prime stable, but don't bet on it being folding stable at that OC. One of my 2.4's would run p95 at 300, but it produces EUE's dual folding if I run it above 285.

I have a 3 others that i had to drop 5-10 fsb to run 2 folding instances stable (no EUE's). They are all in the 260-270 range now.

Running 1 folding instance, you might be ok ... when qmd's stopped I added second instances to these rigs and had to drop the fsb on them.

All of them are running at 1.6V i think.

[The Coolest]

Interesting.
Well I guess you learn something new every day. Is there anything in particular I have to do to perform this folding test you're talking about, or just let it run and see if it produces bad WUs?

[pscout]

after you have a couple of WU's downloaded ... stop the folding instances.
If you already have dual folding instances setup you don't need new WU's downloaded.

Then backup the WU's by copying the WORK subdir and queue.dat file to a backup subdir in each folding directory. MAKE sure the console is stopped before you copy its WU.

Disable your lan either by unplugging or disabling the interface through network connections in windows (i prefer the latter ... saves messing with cables).

Then start up your folding instances. If they eue, you can just delete the work subdir and queue.dat and copy the backups back into the folding directories.
Since your lan is disconnected, the EUE's will not be sent in. You can just begin folding again from the point you took the backups.

just make sure the folding console is stopped before you move the work and queue.dat files.

You will probably find your temps will be higher daul folding than running dual p95's.

[The Coolest]

Hmmm, thanks. This creates a bit of a problem as that is my main rig, and if I kill the connection I won't have any kind of internet at all. I'll see what I can do about it tho. Right now I'm at 5.5 hours of prime.

[pscout]

You can enable/disable the lan during your test ... the proability of it EUE'ing if connected for short periods is relatively low ... and if you run the folding instances as visible consoles you should be able to react quickly enough to not turn in an EUE.

Overnight is best or when you are not using the rig.

[raven]

What else do you think I can do to squeze a bit more out of this chip.

I have a very similar set up on my main rig at home (see sig). The only difference is that I am running 2 x 256 BH-6 instead of one 512 stick, but none the less, I am maxing out at 3.3GHz as well (prime stable). What everyone has been suggesting have been valid points, but I would just like to add a bit. To do this, I was wondering if you could just answer a few simple questions about your set up.

What Game Accelerator settings (aka GAT (http://www.computerpoweruser.com/editorial/article.asp?article=articles/archive/c0405/11c05a/11c05a.asp&guid=)) are you using? These settings can be found in the BIOS where you set your memory timings. Since an AI7 uses an 865 chipset (which does not include PAT (http://www.beyond3d.com/reviews/asus/p4c800/)), Abit introduced GAT as a way of helping to bridge the performance between an 875 (canterwood) and an 865 (springdale). The most stable settings have been found to be:

Auto
Auto
Auto
Disabled
Disabled

How much airflow do you have in your case? Making sure that everything is at least partially cooled is key to overclocking.
What happens when you try running your memory at a 3:2 ratio with the FSB at 270? There could just be a limit to how far you can underclock your RAM, because not being able to POST at 2.4GHz with a 3:2 ratio is odd.
Also, do your voltage settings go significantly below than what you have set them in the BIOS when you use AbitEQ? If so, your overclock could be PSU limited as fluctuating voltages have never been good for overclocking.


So before you decide to throw in the towel, give the above a try, and then let us know how things go.

I achieved a stable overclock of 3.3GHz on my machine by lapping my XP-90 heatsink, applying AS5, installing an after market NB cooler, and making sure that I had sufficient air flow in my case. All of this was done at stock voltages (except for the RAM which I run with 3.0 volts) and I have been happy since. Many 2.4 MO chips have gone past a stable overclock of 3.3GHz, so there is still a chance that you can get yours up there too.


Raven

[hUMANbEATbOX]

about 3:2, some ram just does NOT like this devider AT ALL. i believe CH-5 is a culprit, it can have a hard time running below stock sometimes when in 3:2

1.63v is more than acceptable. mine has been sitting at that (when idling) for about 3 years now. no degredation in my chip whatsoever so far.

gl, i still think with the correct vcore, 3.4ghz totally stable should be a cinch.

[The Coolest]

Raven:
I've got GAT set to all Auto, I tried playing with it and it just wouldn't POST or go beyond POST screen with any GAT setting on.
Airflow has always been a problem in my case. Right now I have a 80mm fan blowing onto the video card and mobo, I have a 80mm fan that pulls air out of the case at the back, and the PSU fan.
The NB was cleaned with Acetone then rubbing alcohol and cleaned from the crappy thermal pad resedue, same with the NB heatsink. I applied a real thin layer of AS5 and there seems to be pretty good contact.
Regarding the 3:2 setting, it just doesn't POST. I've tested my RAM and it should be good for upto 230MHz at least.
I'm not using AbitEQ as I really hate it. From MBM5 the vcore is pretty stable at around 1.54-1.55v, 12v line is 12.3 at idle and around 12.1v at full load on 2xPrime. All other rails are fine. I know this PSU isn't great, but that's all I have right now, and will have to do until my enermax is either fixed or needs to be replaced.

Basically, at 3.35GHz and 1.625v it ran Prime95 over 10 hours until I stopped it. at 1.575v it constantly crashed prime after 3 and a half hours.
Right now as I type this, I'm priming at:
FSB: 275MHz (for 3.30GHz, memory ~220MHz 5:4 divider)
RAM voltage to 3.2v (MBM5 read 3.28v)
NB voltage to 1.6v (MBM5 read 1.66v)
CPU voltage to 1.575v (dual prime95 averaged at 1.54v in MBM5)
And so far I'm over 9:20 on one instance and just over 9 hours on the second.

[hUMANbEATbOX]

seems like you've got everything nder control.

but about the GAT, make sure to set the last 2 options to "disabled", not auto. turning those things on at stock speeds works fine, but most find with OC'ing, it just causes problems.

i would even go so far as to start over, and try again with them disabled.

[The Coolest]

Isn't GAT all about memory? I guess I will have to play with this a bit more next week when I might have some time, for now, I'll leave it at Auto if it passes prime for 24hours.
For better performance, and if RAM can handle it, wouldn't it be better to set CPC to Enabled, though?

[raven]

Yes, GAT does have to do with memory, but also it is very important to the success of your overclock. By setting GAT to the settings mentioned in my last post, you will be able to maximize your overclock and your memory bandwidth. Two years ago I had performed multiple tests with varying settings, and those settings where the ones which produced the best overclock with the most bandwidth.

Also, just to improve air flow, I would suggest that you, if you can, install a fan in the front of your case. Also try to put this fan as far down as it can go on the case so that it takes in the coolest air possible (hot air rises, cold air drops). Currently you have negative air pressure in your case which is not good because less air, means less substance to transfer heat to, which means your components stay hotter. Negative air pressure is caused by having more fans blowing out than in, so by sticking in an intake fan, it should help things out. (but I'm sure you already know all of this and would have done it if you could have). Also, I am pleased with the fact that you have a side intake because this directly blows air onto the main components of your system, keeping it as cool as possible.

Your PSU seems to be running stable, so it should be ok, but you might want to get a third opinion on your VCore from either another program such as AbitEQ (I know you dislike it, but it is meant to work with that Abit board), or some other form of voltage reader since you say your BIOS is set to 1.65 and MBM is reading it at 1.55. Those two readings do not sound too convincing, so I would just double check on that.

Two more things I can suggest for you to try if you have more time would be to:
Lap the CPU heatsink while cleaning it with isopropyl (97%+ isopropyl). After having done this many times, I can say from experience that lapping does get a lot of extra gunk off, and gives you a cleaner and smoother contact surface with the CPU. This intern should decrease your temps by a few degrees. Maximus Nickus has written a great cooling guide which can be found HERE (http://www.ocforums.com/showthread.php?t=70679). It should be able to help you to increase efficiency of total system cooling.

Try to put your stick of RAM into a different slot. Sometimes this helps with your overclock, as some people have reported in the past, so giving it a try wouldn't hurt. By doing this, people sometimes notice more stable overclocks (CPU & Memory related), and it might be able to help you squeeze out an extra few MHz out of your system.


If you decide to try any of these things out, be sure to keep us updated to your status, as I am sure others will find this thread useful in the future.

[The Coolest]

Yes, GAT does have to do with memory, but also it is very important to the success of your overclock. By setting GAT to the settings mentioned in my last post, you will be able to maximize your overclock and your memory bandwidth. Two years ago I had performed multiple tests with varying settings, and those settings where the ones which produced the best overclock with the most bandwidth.

Ok thanks, next time I reboot I will try it.

Also, just to improve air flow, I would suggest that you, if you can, install a fan in the front of your case. Also try to put this fan as far down as it can go on the case so that it takes in the coolest air possible (hot air rises, cold air drops). Currently you have negative air pressure in your case which is not good because less air, means less substance to transfer heat to, which means your components stay hotter. Negative air pressure is caused by having more fans blowing out than in, so by sticking in an intake fan, it should help things out. (but I'm sure you already know all of this and would have done it if you could have). Also, I am pleased with the fact that you have a side intake because this directly blows air onto the main components of your system, keeping it as cool as possible.

Well to be honest my side panel is off, so I don't have any kind of air pressure in the case I'm guessing. If only I had a clear side panel I would probably cut out the fan grills on this case as they're really restricting and put more fans in.

Your PSU seems to be running stable, so it should be ok, but you might want to get a third opinion on your VCore from either another program such as AbitEQ (I know you dislike it, but it is meant to work with that Abit board), or some other form of voltage reader since you say your BIOS is set to 1.65 and MBM is reading it at 1.55. Those two readings do not sound too convincing, so I would just double check on that.

I think you have misunderstood. Right now the voltage in the BIOS selected is 1.575v, 1.54v average is read by MBM5 (AbitEQ does give the same reading).
At 1.625v with 2xPrime95 I'd get average of 1.58v.

Two more things I can suggest for you to try if you have more time would be to:
Lap the CPU heatsink while cleaning it with isopropyl (97%+ isopropyl). After having done this many times, I can say from experience that lapping does get a lot of extra gunk off, and gives you a cleaner and smoother contact surface with the CPU. This intern should decrease your temps by a few degrees. Maximus Nickus has written a great cooling guide which can be found HERE (http://www.ocforums.com/showthread.php?t=70679). It should be able to help you to increase efficiency of total system cooling.

Try to put your stick of RAM into a different slot. Sometimes this helps with your overclock, as some people have reported in the past, so giving it a try wouldn't hurt. By doing this, people sometimes notice more stable overclocks (CPU & Memory related), and it might be able to help you squeeze out an extra few MHz out of your system.

Have you lapped your heatsink, and what kind of results did you get? If so what grit did you start with, and what grit did you finish with?
What kind of heatsink is it? I have the Zalman in my sig, and the finish on the thing is damn good from what I could see.
One more thing about cooling, have you ever consider to put heatsinks on the MOSFETs of this board? It seems pretty hard as most of them are located at real tight places, and near caps and such..

Thanks for the tip on the RAM. Right now its running in the first slot, as from what I understand that should be pretty good place for RAM.
And are you actually saying that with moving the memory around I can get higher FSB even if the RAM is not being the limiting factor?

If you decide to try any of these things out, be sure to keep us updated to your status, as I am sure others will find this thread useful in the future.

Sure will, thanks for your time and a great deal of helpful tips and info!

[raven]

Well to be honest my side panel is off, so I don't have any kind of air pressure in the case I'm guessing. If only I had a clear side panel I would probably cut out the fan grills on this case as they're really restricting and put more fans in.


I agree. Fan grills and filters reduce the amount of air flowing in/out be a great deal. Since your side is off right now, I doubt you are worried about dust getting into yoru system, but if you ever are, you should try to keep a positive pressure of air inside your case. This way, it will help keep the dust from slipping in through the crack of your pci slots or and fanless opening in your case.


I think you have misunderstood. Right now the voltage in the BIOS selected is 1.575v, 1.54v average is read by MBM5 (AbitEQ does give the same reading).
At 1.625v with 2xPrime95 I'd get average of 1.58v.

Yes, I misunderstood that. In that case, that drop in voltage isn't something that you want to see from an overclocked system. That drop might be a sign of a weak PSU which, might be limiting your overclock. From reading your sig, it looks like you are waiting for another PSU to come in. Hopefully that Enermax should stabalize your voltages and maybe increase your overclock by a few MHz.



Have you lapped your heatsink, and what kind of results did you get? If so what grit did you start with, and what grit did you finish with?
What kind of heatsink is it? I have the Zalman in my sig, and the finish on the thing is damn good from what I could see.
One more thing about cooling, have you ever consider to put heatsinks on the MOSFETs of this board? It seems pretty hard as most of them are located at real tight places, and near caps and such..

I have lapped many heatsinks, including the one I use in my 2.4MO rig. At first I used a heatsink that Maxvla sent me a few years ago, but the first thing that I did was lap it. I also did the same thing to my current heatsink which is the Thermalright XP-90 (http://www.thermalright.com/a_page/main_product_xp90.htm). I cannot give you any comparisons from a non-lapped and a lapped heatsink for this CPU, but I can say that by lapping my heatsink which is currently on my 1.8 P4 @ 2.4, it did bring the temps down a few degrees.

When lapping, I started with very low grit paper so that I could rough out as much as I possibly could. I would then proceed to use a higher grit papter, and then an even higher one untill I reach a paper with over 2000+ grit (or so I think. My lappying paper is at home so unfortunately I cannot just check it for you...). You would probably find grit of 2000+ in the automotive section of a hardware store as people tend to use it to polish paint off cars and such. I have read other threads and forums which have people using super high grit paper to lap their heatsinks but I haven't done this myself. Although, I must point out that after every lappying job I would take a Q-tip or paper towel (because kleenex/tissue leaves residue particles) and soak it in isopropyl and then rub it over the heatsink. You would be surprized to see how much gunk there actually is in those microscopic grooves of the cooper/aluminum.

One other thing to remember about heatsinks is that when you are done lapping, avoid touching the base of the heatsink with anything including your fingers, as your fingers have many oils which would reduce heat transfer and contact to the CPU. Yes, it seems very miniscule, but it is very important.

Though I haven't thought of putting heatsinks on the MOSFETs, to be honest, I haven't even had time to think about that yet. Some some people have gotten more stable overclocks with MOSFET heatsinks, but with the overclocks that the both of us are aiming at, it doesn't seem too nessesary as this board can handle much more then what we are trying to overclock to. If you were thinking of raising the voltages more though, you might want to think about giving the MOSFETs some cooling. Having good air flow in your case helps with this as with the more air you push through your case, the more heat can be dissipated. I have 2 front intake fans, 1 side panel intake, 2 rear exhaust, and a PSU exhaust which gives me plenty of MOSFET cooling while using stock voltages.


Thanks for the tip on the RAM. Right now its running in the first slot, as from what I understand that should be pretty good place for RAM.
And are you actually saying that with moving the memory around I can get higher FSB even if the RAM is not being the limiting factor?

Yes I am. If not a higher overclock, then at least a more stable overclock and/or more stable RAM at tighter settings. This was the case with my CPU. I found my CPU limit of 3.3GHz. This is rock stable with my RAM in slots 1 and 3 (running dual channel), but not completely stable when in slots 2 & 4. This might not be the case for every board, but it is my belief that it is worth a try.

[The Coolest]

Well Prime95 stable for 25hours. (25 CPU hours) at the above said settings.
This will have to do for now. I will play with it more when I have some extra time


Edit:
...You will probably find your temps will be higher daul folding than running dual p95's.

And BTW, with folding the CPU is 3-5C cooler and the PWMs are 5-6C cooler than with dual prime...

[pscout]

I think it may depend on the types of WU's. I haven't done any recent measured tests on a p4 HT ... my observations were most likely made on a 920 and with QMD's folding which are the most stressful WU's mem bandwidth wize. My last build was back in Jan.

My comment on dual folding stability still stands. But I usaully prove my stability by folding for a day rather than with p95. I use p95 but only through the complete first test in torture test (small fft's) and helfway through the second. After that i find my stability is usually there for folding. And so most often my stability tests produce useful folding results. But I don't try to get every last mhz out of the fsb since i need a buffer to allow for ambient fluctations, and the inevitiable dust buildup over time.

I haven't caught up on your whole thread, but I assume you have 2 sticks of mem in there now ... cuz you can't really complete the overclock without it.

You seem to have things nailed pretty well with the posssible exception of case cooling.

Hopefully we get some DGromacs and even some QMD's to fold soon ... the current workload mix tends to favour AMD rigs i think.

Happy folding !

[The Coolest]

Unfortunatly currently I'm stuck with only one stick of memory. But I put up some old stuff for sale here locally and if someone buys it off me, I'm gonna get me a set of 2x512MB TCCD sticks. That's probably the best I can get for not a lot of money, and still be able to run at 1:1 ratio.
I'm pretty sure that right now the real power of this rig is being killed by single channel memory which is also running on a divider...