my new upgrades have arrived,now i need your help please

[micovwar]

hi there

a couple of days ago i ordered my new upgrades consisting of an AQXEA xp2500 and 2x 512mb twinmos 3200 ram from cpu city and theyve just arrived this morning,and now i need ure help please

im totally new to overclocking so i dont realy have a clue other than upping the fsb on how to get a succesful overclock,,my upgades are going into my sn41g2 which as we all know is an nforce2 board (supposedly great for overclocking),,so what ide like from u guys is a surefire way to get a succesful overclock on my new processor (which is supposed to be a great clocker) some people have had results of upto 2.4ghz using this particular processor but to be honest i dont realy wanna push it that far as its in such a small case and as we all know shuttles are pretty hot at the best of times, ide be more than happy with 2.0ghz or maybe a little higher if possible without having to adjust the core voltage
so if any of u guys could spare me some of ure overclocking expertise and tell me what i need to be doing to reach my goal ide be forever thankful,,i realy donno nothing about ram timings or anything like that or even what the aggressive/optimal settings do,,im pretty much just starting from basics so please help me through this

thnx in advance

[Knoxville]

first off you need to make sure you've got the newest sn41 bios as the older revisions didn't really like overclocking.

unfortunatly you've got no multiplier adjustment so you're gonna be stuck with the standard multi (11 iirc)

personally i'd up the fsb to 180 first and leave the ram on its standard settings, if its stable at that try 190, if its still stable at 2ghz then try upping it gradually to 200fsb

once you've done that and got it running stable (a task in itself)

then you can try tightening the ram timings

[Austin]

Yeah you should find the XP2500+ comes with 1.65v as std, you should find it's happy UNDERVOLTED to 1.50v at that speed and would run VERY cool too. Basically I'd be VERY surprised if you have any problem running it at 2.0ghz as it should do so with less than 1.60v let alone the 1.65v it comes with as std. SO check what voltage it uses by default, I'd suggest either leaving it alone if it's 1.60v-1.65v or set it to this if it's different, 1.65v is normal for all AthlonXP of XP2000+ and higher. The higher voltages obviously help to reach higher speeds at the expense of heat and every 0.05v above 1.65v adds risk to the CPU.

TwinMOS PC3200 is designed for use up to 400mhz, you should find your BIOS has the RAM set to sync with the FSB (eg 1:1, 4:4 etc), that's best. Leave the RAM timings on AUTO / SPD, as said you want to find your maximum stable o/c and then try tightening them if you wish.

sn41g2 is based on the original nForce2 which is still very good but could have some problems with 400FSB esp on Barton (proabably fixed via BIOS anyway). However nForce2 with onboard gfx (GF4MX) don't really do 400FSB but should easily do 333FSB and probably get a good 380FSB I should imagine. If using the old onboard gfx then be sure to use 2 sticks of RAM in each channel for Dual Channel mode as this takes you from a lowly GF4MX420 to a GF4MX440.

Hopefully the latest BIOS will allow for better o/c'ing and possess voltage adjustments which the early BIOS' lacked IIRC. I'd be sureprised if you didn't have multiplier adjustment.

So basically I would suggets you find what FSB you can run at by upping it gradually and testing thoroughly for stability each time (Prime & 3Dmark). Ideally you want to keep your CPU speed close to 1.83ghz so as to only invlolve the FSB to find the highest it can go (stable). Once you find out lose about 4mhz to ensure you aren't too near the bleeding edge of stability. Then simply try upping the multiplier 1 step at a time with your new max FSB. When you find the max stable setting it's time to either back off a notch for the long term or apply a little more voltage and see if that helps you to get decently higher. When you have your max stable multiplier with your max stable FSB you can either leave it there or play around find the exact combination for optimal o/c, such as 11x200=2.20ghz VS 11.5x197=2.27ghz VS 13x180=2.34ghz. Just rem it's a good idea not to be too close to the absolute max o/c. You could also see how far you can get undervolting if you want to run as quietly as possible.

[micovwar]

thnx 4 the info guys,

austin as far as i can see in the bios (updated to newest) there are no multipliers or voltage adjustments but there are fsb options,so how would i get around the voltage,ide love to be able to undervolt it if i could as i want the quietest coolest running system that i could get from this chip,but judging on its default voltage im assuming it may be as hot as my xp2400 if not maybe hotter (which i realy wanna avoid)

also i think my only real options for overclocking this system is the fsb option,,how far do u think i shud get roughly (im to paranoid to try lol) and where can i get some good benching progs like prime and this other 1 (which i cant remember but i know it has a 5 in the name somewhere) which is supposed to be a realy good torture test for cpu,s?


knoxville i realy donno anything bout ram timings etc,so wouldnt know what i was doing even if it slapped me in the face,,do u think that it could reach 200fsb on its default voltage?

thnx in advance

[micovwar]

omg omg omg

just a quick update fellas,,ive now upped the fsb to 11x192 and its posting as an xp3200 and booting into windows and in system properties is showing 2.10ghz but im scared to push it any further incase i break it !!

wot do u guys recon i shud do,,im itching to go for 11x200fsb but will i need to add more voltage?

i havent tested or benchmarked it yet as im still salivating and foaming at the mouth,but i must say these xp2500,s rock

will report more later,wish me luck

thnx again guys

[micovwar]

well i eventually went with 11x200 and was rewarded with an absolutely fantastic 2.21ghz,i havnt fully tested it with prime95 or anything so im not 100% on its stability but i done some testing with sisoft sandra and it seemed stable with that,,is there a test which will guarantee my system is stable with this overclock as now that ive got it i wanna keep it (also saves at least £300 on the xp3200) this chip is an absolute bargain!!

[CrapshoT]

got my 2500xp running sweet at 11.5x205 fsb = 2.358g
thats at 1.7v runnin at 42C idle and 47C full load witha volcano 7 hsf
my 2500xp is a AQXDA not the special E code lol
but im well happy with it for £60 \m/-_-\m/

[Anders]

Sweet, sounds like you’ve got a gooden, just remember next summer, when the sun is out you should keep an eye on the temps. Happy days.

At least a couple of days worth of SETI or something that will use all CPU time available will check for you. Of course, you should all ways go on a marathon gaming session just to be sure

On the more boring side, it’s always good to check with Memtest86 (all tests), this is just a RAM checker, but it will show up problems with other hardware as well. Not done much with “Prime95” but it seems to have a good reputation.

Ooh, not sure anything can be guaranteed. If you want a long(er) run from parts then keeping temperatures down is important. Overclocking is likely to reduce the life span of parts.

[micovwar]

actually i recon i spoke to soon,when i further tested @ 2.2ghz lastnite it failed prime95 within litterally minutes and also when i used sisoft sandra my computer basically jus kept freezing all the time so i slowed it down to 11x192 but yet it failed on this too,but i fully tested @ this speed yesterday and it was fine,so i set it back to default settings and speeds of 1.83 ghz and tested it again using prime95 and its still giving me errors within seconds of starting the test (although i did manage to get almost an hour into the test @ 2.1ghz) so any ideas whats happened?

[Austin]

It's a shame there aren't multiplier or voltage controls in the BIOS and I don't think you'd find any dips on the mobo either unfort. You can force the voltage higher or lower or even alter the multiplier by modding the bridges (cutting or joining) but I would seriously avoid that, it doesn't leave much room for error, is risky and isn't very 'changable' either. Neither the stock 11x multiplier nor 1.65v voltage should limit your o/c'ing much at all anyway.

Do you know what your default voltage is? Most probably 1.65v which is what the XP2400+ is most likely to use too, so with the BartonXP2500+ at 1.83ghz and the TbredB XP2400+ at 2.00ghz you should find the Barton is a little cooler if anything and they should both be about the same speed (but Barton has more headroom).

You should have remembered the golden rule when o/c'ing ... use small steps and test thoroughly each time before making it stick or going higher, also best to step back a little from the absolute max. Just because you can POST or run a few things in Windows does not mean you're stable and there is then some component risk when going higher than the true max stable setting. So long as you avoid voltage increases above the core norm (1.65v for TbredB & Barton) there is virtually NO RISK at all in o/c'ing so long as you abide by the golden rule.

To test for stability you should definitely run at least 10mins of Prime95 torture test (don't multi-task) and also run at least 1 complete run (all tests) of 3Dmark2001SE paying particular attention to the nature test (which is the most stressful). Prime will ensure your CPU is solid as a rock no matter what you throw at it while 3Dmark will stress all of your system components in a close way to the most stressful thing you can use your PC for (gaming). Memtest86 (or other RAM testing apps) should not be needed so long as you are not running it faster than it's rated for and leave the timings set to AUTO / SPD. IMHO it's best to find your maximum CPU+FSB o/c and then look towards the RAM (esp timings). I do question the validity of Memtest (and other similar apps) as Windows handles things very differently to a DOS type environment. You can easily pass those tests and yet fail to boot into Windows. Sandra is useful for checking synthetic speed increases, for stressing certain aspects of your PC as well as diagnostics too.

As to things seeming Prime unstable even at stock I'd simply try it a day later after your first cold boot. Most proabably nothing to worry about at all.

[micovwar]

hi austin an thnx,

how long is the reccomended test which should tell me my processor is stable using prime95 (i read somewhere that if u pass 6 hours u can gurantee its stability) as i left both shuttles testing with prime95 before i went to work and when i came in my sk41g was still testing but the sn41g2 (overclocked) had failed after 45 minutes so i was just wondering as u said "at least 10 mins with prime95" is there an approximate time limit to ensure its stable?

the default voltage on the xp2500 is the same as the xp2400 @ 1.65v and i have to say its realy 1 helluva chip for £70 i had no problems at all hitting xp3200 speeds (i had read that a few others with the same machine couldnt get past 195fsb) and to know i have 2.2ghz at my fingertips using 11x200fsb is a fabulous feeling but im kinda gutted that it fails in prime95 as ide love to be able to run it at 2.2ghz but even still its failing @2.1ghz and 2.0ghz,,i have my ram running in sync with the processor is this right or should i use aggressive?

the only problem is that i dont understand the ram timings yet as im new to it all iirc my current ram timings are 8 3 3 and for some reason look strange to me (i heard lower was better) the ram im using is 2x 512mb twinmos 3200 cas 2.5 400mhz and i heard it was great for overclocking so should i use different timings and how will i tell the difference between the timings performance wise?

i havnt tested with 3dmark 2001se yet but im gonna download it now an do it,the fact that i seem to be failing prime95 does this mean ile fail 3dmark too?
what does it tell me if i get through 3dmark but still keep failing prime95 roughly within the first hour of testing,would my cpu be stable enuff for gaming or wud it be hit & miss (sometimes no probs & other times blue screens etc) ?

thnx in advance

[micovwar]

ok i ran 3dmark2001se and payed attention to the nature scene as u said austin my max framrate was about 90fps and it never went any lower than 48fps,,all tests ran perfectly fine and my score was a nice supprising 11803 (using the xp2400 i usually got between 9500 and 10000) those pfs were only on the nature scene,every other test was pretty much flying between 100-200fps

so based on this does it tell me anything good?
my current cpu speed atm is 11x191 = 2.1ghz,im gonna try playing a game of wolfenstien and try a couple of others too

thnx again

[micovwar]

well i spent alot of time lastnight gaming and i had no probs whatsoever no crashes/lockups or anything (but thats not to say it wont happent sometime 2day)
im pretty much stuck with 2.1ghz for the time being as i dont wanna push it to hard (although i aint tryed gaming etc @2.2ghz yet) i was thinking of putting 3dmark2001se on a loop today so itle run all the time im at work and maybe have sandra running in the background to stress the cpu as much as i can to see if it would shut the system down or anything

[Austin]

Sorry, been real busy. Basically Prime is a good tool to overtest your CPU, it should stress it a LOT more than you are ever likely to as a home user ... if it can run Prime (I'd say 30min is plenty) then it should handle anything you throw at it in the real world. Prime only tests the Integer (IIRC) but it's still a very good test to help ensure stability and locate your o/c limits.

Avoid aggressive RAM timings, leave them AUTO/SPD and definitely in sync ... esp so when finding your o/c limits. Lower timings are faster as it means there's less time given for the RAM to refresh before it starts performing the next task. Diffs are very small (1-2%) between the slowest and fastest, generally it simply isn't worth the hassle esp for those on Dual Channel setups and it takes a long time to test ... consequences can be dire if it's running a little too tight. TwinMOS PC3200 is very good stuff but I'd keep it no higher than 400mhz anyway.

If you're not satisfied that you're stable at a given speed (eg 2.1ghz) then you are best to work out the limitation. Often it's heat so ensure your HSF, case fan(s) and ventilation are working optimally and well placed. If you're sure it isn't heat then upping the voltage (if you can) can often give you that little bit more stability but it will give off a lot mroe heat and above the core norm carries a little more risk with each +0.05v.

3Dmark2001SE will test your whole system and not just the CPU as Prime does, it stresses your CPU far less than Prime but in a real world way. It emulates the most stressful thing most home users will do with their PC ... gaming. You also get FPS and other scores and can see hwo you fare with similar systems or what diff a new gfx card / CPU may make. It's only good as a general guide but still very useful, esp for stress testing.

Once you decide on your max o/c it is wise to back off a little, so if things seem a bit too hit and miss at 2.2ghz then 2.1ghz is certainly wise. When you do decide on your final o/c that's when you want to do some serious stress testing. I'd suggest running 3Dmark loops while you're at work as you suggest and also see how long Prime can run for. It is very useful (esp in Prime) to open the Windows Task Manager (CTRL+ALT+DEL or System Monitor) as this tells you many things, I suggest you try the 'Performance Tab' and also use a Windows App to monitor your temps (see if any instability is heat related). Sw from the actual manu of your mobo, preferably specific to your exact mobo is best, check they are close to what your BIOS reports but it's all only a rough idea. Most interesting to see how much your CPU temp rises and to get an idea if your case is adequately venting the hot air.

[Knoxville]

what timings are you using your ram at now?

it should be able to handle 6-2-2-2 easily, as standard cas 2.5 2700 will handle those timings at 333fsb

I prefer to use 3dmark2001 for testing stability, if it'll loop the test 10 times without crashing its a stable o/c imo.