first OC

Hi guys I've had the system in my sig for about 3 years and I finally want to OC it because I have the time now. I want some help getting the best results I can out of my system

default:
CPU 200x11 = 2200mhz @ 1.35v
RAM 200mhz 2-3-3-6 @ 2.75v

I was able to push the CPU to 220mhz FSB lowering ram to 100mhz, multi to 4x, and HTT left at 1x @ 1.375v

This was basically the limit, because 225mhz wouldn't boot even at 1.45v - why is the limit for this chip so low? is it the PSU or did I just get a bad OC'ing chip? I was able to successfully boot with 220x11 = 2420mhz with an HTT multi of 5x(ram still at 100mhz)

Is there something I'm missing because I thought the average 3700+ was able to OC to like 2.6-2.7ghz ish

you lowered your ram and HTT too much. lower the HTT to 4x and your ram to probably 166 or so

Alright thanks I was able to get 250mhz @ 1.35v I'm going to try to push it a bit more i'll post back with what happens

EDIT:
Okay 250mhz is the max FSB speed. 255mhz would not boot even at 1.5v(I don't think I should go higher right?) I used a 6x multi and 3x HTT multi with the RAM at 166.5mhz.

So now what? Do I just see how high I can raise the multi and whatever? And before I go stress testing and whatnot is a 250x10=2500mhz or 245x11=2695mhz better? Or just in general is it better to have higher effective speed but lower FSB or the other way around?

EDIT 2:
CPU-Z seems to give a vcore reading of 1.504 that changes to 1.520 and back while in the BIOS it's set to 1.475v - what's the deal here?

the voltage is just how it is, some selections are acually a little lower than selected, or in your case a little higher, and its also how it gets read by an application like that.

nothing is going to be 100% accurate unless you take a multimeter to it and get the infor straight from the horses mouth sorta speak but its atleast close.

what do your temps look like at idle and load? you may be able to push a little bit more voltage into it

uhh i wouldnt, hes running atleast over 1.5v a little bit, they may not be 100% accurate reading voltage but they are accurate within a range and i think his is some where between that 1.5 and 1.52.

just because the temps are good doesent mean you have room to add more voltage

jmt391 said:

Alright thanks I was able to get 250mhz @ 1.35v I'm going to try to push it a bit more i'll post back with what happens

EDIT:
Okay 250mhz is the max FSB speed. 255mhz would not boot even at 1.5v(I don't think I should go higher right?) I used a 6x multi and 3x HTT multi with the RAM at 166.5mhz.

So now what? Do I just see how high I can raise the multi and whatever? And before I go stress testing and whatnot is a 250x10=2500mhz or 245x11=2695mhz better? Or just in general is it better to have higher effective speed but lower FSB or the other way around?

EDIT 2:
CPU-Z seems to give a vcore reading of 1.504 that changes to 1.520 and back while in the BIOS it's set to 1.475v - what's the deal here?

Click to expand...

Chances are it's not the CPU but the system clock that has reached a top end. If true, increasing CPU voltage won't help. (You did turn the HT frequency down to 3x at 250 MHz?)

CPU speed is the #1 goal in OC'ing so, to answer your question, 245x11 is better if you can run that fast. Your RAM should also be able to run that fast too - that XMS is great stuff! Set the RAM back to DDR-400, relax the timings a little, and push the vDIMM to 2.8v. My XMS is all 2-3-2-6 stock and for me, 2.5-3-2-6 is stable up to 245-250 MHz. (Since yours is 2x512 it should be good to 250 MHz at that vDIMM.) You might try 2.5-3-3-6 first and if that doesn't work drop back to 2.5-3-3-7 (all those at 1T - BTW).

For the vCore you usually reach a diminishing returns thing about there. The more vCore you need and push the hotter the CPU so the more vCore it needs to stay stable, which makes it hotter still, and so on. WC helps a lot but still, there's a limit. I run my Opty 180 at 1.55v (load temps of 43°C) for one month a year during the SETI races (it gives me an extra 100 MHz) but I don't do it year around - too much heat for a small gain.

It's really great to see someone OC'ing an s939 rig again! I'm still running the Opty 180 as my main rig and I've got four other, active s939 rigs around the house. Remembering how to OC these is not an issue - I still play with them off and on ...

My friend just gave me one of these CPUs to mess around with. I got it to 250 x11 easily. I'm now pushing it to 270 x11, but keep having memory errors, but I think I just got it figured out.

EDIT
Nope Something strange just happened when I was pushing to 270 x11. One of my HDD started clicking very loudly. Although I believe it was that. Anyway, if you are still interested in OCing this CPU look for a post sometime this week. I'll be showing off my work on the 3700+ San Diago, and the 3600+ x2 Manchester.

Here is my data:

2.5
http://img504.imageshack.us/my.php?image=25ye4.jpg
http://img504.imageshack.us/my.php?image=3700sandoced25kx7.png

2.7
http://img227.imageshack.us/my.php?image=27gr2.jpg
http://img293.imageshack.us/my.php?image=3700sandoced27br6.png

sorry guys I've had a bit of a roadblock with OC'ing, and it's called schoolwork. I will try making progress little by little but thanks for all your information about everything. I do not want to push the voltage past 1.5v because I feel like that would be unsafely high. I will take the advice on the RAM and try squeezing the timings down, only letting up if i HAVE to.

I'll slowly post back with progress - thanks again

PS - just one more question: what's a reliable stability test within prime95?(not the 24 hour torture test i mean for a quick FSB/HTT test and whatnot to see if i can squeeze some more speed out) I don't fully understand how to use prime95 effectively so I was thinking of using superpi 32m and then testing with the 24 hour torture test after I have the whole system at a potential OC (RAM + CPU).

I use Orathes to test my system and at standard I will run it for 2hrs straight.

jmt391 said:

sorry guys I've had a bit of a roadblock with OC'ing, and it's called schoolwork. I will try making progress little by little but thanks for all your information about everything. I do not want to push the voltage past 1.5v because I feel like that would be unsafely high. I will take the advice on the RAM and try squeezing the timings down, only letting up if i HAVE to.

I'll slowly post back with progress - thanks again

PS - just one more question: what's a reliable stability test within prime95?(not the 24 hour torture test i mean for a quick FSB/HTT test and whatnot to see if i can squeeze some more speed out) I don't fully understand how to use prime95 effectively so I was thinking of using superpi 32m and then testing with the 24 hour torture test after I have the whole system at a potential OC (RAM + CPU).

Click to expand...

A SuperPi run of 2M is a good filter for obvious problems. Anything longer is best tested with Prime95 even if it's only for a half-hour or so ...

But how do I know if superpi it fails? Will the program just crash out or is there an error message?

So far everything that booted into windows successfully passed the 32M test in superpi(or so I think - it didn't say anything other than finished calculation). I decided to go with 32M rather than 2M because I did not want to be disappointed in the long, dreaded, torture test.

Dolk - those are some impressive results and the fact that youre using an a8n-SLI is good because I have the same mobo just no SLI

So far I have calculated:
Max HTT: 1225mhz (superpi 32M)
Max FSB: 295mhz (superpi 32M)
CPU Max: 280x11 = 3080mhz @ 1.5v (superpi 32M)
RAM Max: currently testing 240mhz 2.5-3-3-6 @2.75v

Ah, it looks you're using the Easy 1, 2, 3 thread! Great start for OC'ing!

SuperPi will give you an error (usually a rounding error) if something goes wrong.

And, personally, I wouldn't push the HT Link much past 1100 MHz. Some boards will do it and some won't but the problems you'll run into if it's too fast are things like corrupted HDDs - just not worth pushing it, IMHO ...

QuietIce said:

And, personally, I wouldn't push the HT Link much past 1100 MHz. Some boards will do it and some won't but the problems you'll run into if it's too fast are things like corrupted HDDs - just not worth pushing it, IMHO ...

Click to expand...

Listen to QuietIce, PLEASE, I just corrupted one of my HDDs, thank god I have spares and was able to back it all up. It crashed right after I set my HTT from x3 to x4 while trying to push for 2.9 on the 3700+ SanD.

On a side note, I got it to 3.0! but I kept getting memory issues. For now I'm back to using my 3600+ x2.

When I got the 3700+ SanD up to 3.0, it became a bit power hungry. My board maxes at 1.5v (for good reasons), but in CPU-Z, the CPU was asking for around 1.55v.

I'll be posting my results for the 3700+ and my 3600+ x2 sometime today.

I am using the 1 2 3 thread haha I decicded that would give me a good idea of what I can expect out of my system. After that, I'll try different combinations with 3dMark 06 and see what gives me the best performance

hmm so then i guess i won't push the HTT past that then - its okay if its a little slower than 800mhz though right because it's not like its saturated anyway

If its a tad bit under 800, you should be fine. Make sure that it does not hit 600.

Easy formula to follow:
HTT Speed for 939
600 < HTT <= 1000

Alright I'll try to stick with that the best I can although I don't think it could hurt running the HTT at like 1050 or something if I had to could it?

BTW i maxxed the CPU check the results a few posts up - god knows if it's prime95 stable but I can hope haha

Running your HTT at 1050 isn't a terrible thing, but there could be some side affects. I'm not the one to ask what a high HTT could do to your system, but I have always found it in best practice to keep the HTT around 800.

Getting up to 280x11, that is amazing. Now can you make that stable? And what voltage do you need for that Frequency?

****EDIT****

I have a HUGE post that will be coming this weekend. I have to do a lot of my tests all over again because I failed at capturing my temps and SSing everything

1.5v for that but it passed the superpi 32M test so it isnt necessarily rock solid - we'll see when i run prime95 when all this is said and done

EDIT:
Dolk, that really sucks - at least you know the OC is stable so no sweat. just time. lots, and lots, of time haha

I had 2 939 amd chips that ran for 2 years with 1.6+v
3200+ and 4400+