Drin

Hi there, sorry if this post is long, but I thought it was best to post as much info as I could to begin with, incase any of it helped. Also, this may help someone else out there in a similar situation (unlikely, but who knows!).


New to OC:N and thought I would share my OC attempts as well as seek some advice if possible, the majority of information I used came from this site to begin with so I thought it only right. First I'd like to say that the advice/attitudes I've seen on these boards is great, and that I hope to be able to contribute in a similar fashion some day

History/experience:

Have built roughly around 20 systems now for friends/family, so while no expert I know myself well enough around the inside of a case and some mobo's ^^. I've played around with settings in the bios before, achieving slight/decent overclocks, but never tried to push a system to its absolute limit before. Done some case mod work, so not affriad to drill holes, hack in mounts and whatnot, but not seasoned or anything. Also played around with after-market HSF's and the like, but never used a HDT variant before - infact I was never too great at applying TIM to begin with and HDT's seem trickier.

System rationale:

This system was built as a hold-over system able to handle the upcomming SC2 and current gen games until Diablo 3 comes out, with the proviso that I could hit 3.2ghz - putting it slightly above/behind an E8400 @ stock speeds by my rough calculations (please correct me if I'm wrong!). My brother and I had decided to upgrade our systems, and use the pieces leftover from our old systems to build a web-surfing PC capable of decent gaming @ lower res, and a media HTPC. While I decided to not run the risk with my brother and built him an entirely new decent Intel E8400 system, I thought I'd try to squeeze as much performance as I could from an old case and amd 5200+ brisbane I had, with the goal of oc'ing it by .5 of a Ghz. Memory, drives and keyboards etc were retained from the old systems, and the only thing purchased was an asus M3A32-MVP mobo, ati 4850, Xigmatek HSF and about 5-6 antec fans

Initial attempts/current results:

First things first, had to rotate the periplex in my side case panel because the fan wouldn't fit with the xigmatek, and secondly it was no longer directly above the GPU. Drilling a hole in the side of the case, and using some old aluminium from a can seemed to work

Looks a bit cheap, but decent enough for now, especially considering that it reminds me of the greatest softdrink ever known to mankind - pepsi blue! It'll have to go eventually though, seeing as all my internal lighting is green, which makes it seem a bit odd

Using durch's excellent guide in addition to other usual resources found within these forums resulted in my first stable oc to pass a 24h orthos test:

http://img399.imageshack.us/my.php?image=oc1mu4.jpg

Memory was not oc'ed on that attempt, but I got lucky and got a pretty good batch of corsairs, that I have not actually reached the limit for yet (have only tried up to 1080 Mhz).

temperatures were pretty fine even @ 1.52v vcore for the most part, but at 22 hours it started to hit 65 degrees in the hottest part of an average day, so I got worried and turned on the AC to see it through to 24 hours. Pretty happy, but deciding I needed to lower temps even further I went about redoing all my cabling in my case, as well as re-applying my TIM/remounting the xigmatek.

Cable management/airflow:

http://img133.imageshack.us/my.php?image=cablesae1.jpg

tim application guide

Results and Temps:

I decided to lower my OC to 3.12ghz from 3.137 for the time being, which allowed me to lower my vcore from 1.52 to 1.47.
Initial impressions were ok. My computer was idling as low as 31-33 degrees (32 average) after just turning on my pc, and after a couple of orthos tests/application work it seemed to idle @ ~34-35. Previously, it would idle @ around 36 after being turned on, 38 after some hours of stress testing.

However, when I tried to do a 24 hour stress run, my cpu started to hit a whopping 68 degrees 4 hours in before I decided enough was enough, and this was with a LOWER vcore. Ambient room temperature was probably about the same, give or take a few degrees. I'm thinking I screwed up the TIM/mounting, but how does that make sense considering it idles lower? That's easily over a 30 degree increase anyway you look at it, and I'm not sure if my cpu would have hit 70 degrees under load (it tends to go up very, very slightly as the stress test continues, even into night).

Even if I were to hit 3.2ghz without increasing my vcore above 1.6, the temperatures as they are now are probably far too high.

Questions:

1. Is lapping a Xigmatek s1284 worthwhile? Reading up on experiences people have had with the s1283, people seem devided on this issue. Some say its already smooth enough, while some people have claimed amazing drops of 8 degrees from lapping a s1283 . Never lapped before myself.

2. How much vcore can brisbane chips handle safely, ignoring temperatures? my personal upper limit is 1.55v, and while many say that anything over 1.5 is dangerous I've had some claim 1.6 is fine (on air to boot).

3. What temperatures would be reasonable given that during summer my ambient room temp should be around 28-29? On hot days it can hit up to 33-36 here, but if it started hitting 33 degrees + I'd probably throw the AC on. I'm aiming to keep my system below 65 degrees full load, any thoughts on this?

4. I've always believed that faster ram > tighter ram timings, is this correct? Atm I have a choice between running my ram @ 1040 + (could probably push 1080 easy) with 5-5-5-15 timings, or running it @ 667 speeds with 4-4-4-12.

5. Is there any easier method of figuring out what clock skews will help my system regain stability @ 3.2ghz other than trial and error? Having tested out my system @ cpu multi 12, FSB 267, vcore 1.55v, auto clearly does not work. While I have not managed a stable 24hr orthos test, different settings range from crashing instantly (most often the cause) to crashing after ~2 hours, twice.

6. Airflow in my case (minus psu fan) is the following: 2x front intake 80mm antec tricools @ 34 CFM, one side fan @ 34 CFM over the gpu, one rear exhuast @ 79, one top exhaust of 26 (set to low). Ignroing the psu, that gives an exhaust of 105, with an intake of 102. Considering my psu's fan is variable based on temp, should I just forget about trying to balance CFM and set my top exhaust to 34? (high). Should I remount my HSF so that the fins/exhaust faces my rear case fan? (currently its mounted to push exhaust upwards, with the hsf taking air in from below) Should I drill in additional holes and attach a 120mm fan to the top for exhaust?

7. How much variance should there be with brisbane cores between idling and under load? I know it would vary between cases and implementations, but a whopping 30-33+ degrees difference seems crazy to me. Have I mounted my HSF wrong? or applied TIM wrongly? My application method was to first apply some tim to fill the cracks, pressing down with a latex glove to ensure it was packed and smooth. Then cleaning off the excess, I applied three lines along the middle 'dividers' as such:

http://img100.imageshack.us/my.php?i...icationmi7.jpg

picture quality is low, but unfortuantely I did not set my camera's macro mode properly and thats how it turned out.

So I'm wondering if I can hit my goal of 3.2ghz, or should I accept that my current cpu will not go higher? Mobo and Ram seem to be fine, so I can probably rule them out as limiting factors. I know I've hit 1.35ghz stable @ 1.567 vcore, but upping it to 1.6v does not seem to regain any stability. I do not know if upping it past 1.6 is even viable, and should it be I'm almost certain my temperatures would be too high given my system's current state. I was about to give up hope, but playing around with the clock skews renewed my determination as I managed to run an orthos test for 2 hours, surf the internet while doing it and listen to some music. Unfortunately it did crash, and simply put when you factor in all the different fsb/multi combos to reach 3.2ghz, there would be thousands of combinations taking into account clock skews. Clock skews have a whopping 14 options per core alone. Not sure on the other cpu settings as well, I have cool n' quiet, auto xpress, cpu tweak disabled, but have not touched 3D now! or amd live etc etc.

Sorry for my long post, but hopefully it was at least interesting to some ^^. I was thinking I should give up, especially considering how well (and easy, good lord) my brother's intel system could OC, but the fun of an amd system lies in the challenge Any advice would be greatly appreciated and thanks in advance!

timw4mail

Have you tried relaxing and underclocking your memory speeds? I've gotten up to 3GHz on my X2 3600+ (Windsor) using such a technique.

__________________
AM2/AM2+ Club ATI Radeon 4870 Club .: ATI HD4800 series club :.
Phenom II 940 BE @3.75GHz Athlon 64 X2 3600+ @3GHz 4870 Validation 8800 GTS Validation

Dobbie

Check my sig rig for so similar specs to yours. Might help. Personally I haven't seen my temps go past 52 while Folding for over 24 hours @ 100% load

__________________

Drin

Thanks for the quick reply All oc attempts begin with me setting my memory speed to 200MHz, and timings to 5-5-5-18-26. After it passes a 24hr Orthos test I then play with the memory. Should I set loosen them even further? Can't remember if they will, but I'll def give it a try.

Will do, many thanks

alwaysAMD

You will need at least 1.4v if it's a brisbane, mind you. Also, don't be afraid to run your HT bus lower than 1000. My system is much more responsive with faster memory and a 900mhz HT than it was when I was at normal speeds. Faster ram > lower timings.

Drin

Righto was never afriad to drop below 1000, so I've played with 4x and 3x, but never 2x (would lower my HT bus to lower than 750). I read in Durch's guide that it could impact performance if it went below 750 but is this a rule or just common practice? I might try lowering the HT multiplier to 2x, lowering my cpu multi, and increasing my FSB to compensate.

going to give it a try, cheers!

Drin

Actually, after posting I noticed you have a xig s1283 lapped - I know the s1284 (first release) is its less efficient successor, but I believe they are more or less similar, given its increase of around ~3-4 degrees and extra heatpipe. Would you recommend me lapping my s1284? and 1.4v is no problem ^^ staying below 1.55v is a challenge for me as is

Dobbie

I might have missed it before, but what CPU cooler are you using?

__________________

alwaysAMD

Lapping actually only helps by a degree or 2, and i blame that on the xig's design. I tried lapping further and I think I may have wrecked it, which is why i'm using a CNPS9500 right now until I can get my hands on another one.

Drin

s1284 Dobbie, I would've picked up a s1283 but the store I usually go to only stocks the s1284.

And ahhh that makes sense, I had heard someone (on some other forums, I'll try to pull up the link). Most people say its not really worth it with the xigs as you said always, and I'm thinking thats the case. They're pretty smooth as is. As for that other person, most people said he probably mounted it wrong to see such a drop (although he denied it).

Hope that xig works better wish I had not settled for a s1284 myself!
Btw, running @ 3.2ghz atm, with my timings lowered to the lowest they'll go (6) and stable so far @ 5 minutes into orthas. Thanks guys, gotta hope it lasts 24 hours now (a long stretch). vcore is @ 1.58 tho ><