P4 3.4EE has landed (in my Mach2 hehe)

[maverik-sg1]

BOARDY,

OK I am running stable at 258fsb @ 1.75v, early results are promising:

27202 in 3dmark2001se

31s in superpi 1m

over 30mins in 32m test (compared to 26m 19s for Prescott 3.2 @4.3)

Have not installed MBM yet but I can tell you under load in 'NORMAL' mode I am running at 42degress on the MACHII - I guess thats -35ish at CPU?

I checked the voltage to CPU using CPUZ (as I already had it on my PC), running superpi volts started at 1.744, dropped immediately to 1.712, once or twice over a ten minute period (I have a short attention span!!) it dropped to 1.696v.

I Guess this is called vdroop? OK where will I find the mod (piccies work well for me), how easy is it (I am crap with a soldering iron) can't I just adjust as volatge rail on my PSU (tell me which one supplies the juice to the CPU).

 

[boardy]

fatty - I had about 3.3v going thru it on a modded Abit NF7-S v2. Wasn't the most stable thing in the world mainly due to a crap PSU lol.

mav - See what I mean about the droop of the Vcore. That is what's making the clock unstable mostly. I'm putting a Vdroop mod on my P4P800-E if my soldering iron will play ball, if not it's going to a mates who's gonna do it for me lol. We shall see how that performs if I can get it done.

31s in PI is ok. I got 31s with this 3.4EE @ 230FSB I think it was but I was just messin about with stuff. Don't really use PI. I'll do a run of it in a min and post the result if I can be bothered. I've also taken my DDR Booster outa this machine and managed to push the FSB to 245 atm, stable!! I'm gonna try for 250 in a while, if I can I'll run PI.

For your Vdroop mod have a look on XtremeSystems and google it. I have seen one for your board before so I no it exists. Messing with the PSU rails won't really help the droop tbh. Best to find the mod and then find someone who can do it for you if you can't do it yourself.

Cheers

Boardy

 

[maverik-sg1]

Boardy,

I pushed the rails on PSU up to plus 5% ( on 3.3, 5 and 12v rails) and your right the difference it makes is minimal.

Not sure if I should push some more (7%?) just to minimise the effect it has now, but concerned how much overvolts the MOBO can take - advice?

If you are going to run superpi, I am not that bothered about 1m test, but u need to set aside 30mins (I use it to check stablility instead of Prime) and do the 32m supertest.

Both my Pressie @ 4.3 and 3.2ee @ 4.1 manage 31s so it's difficult to differentiatie between the two at this level. The BIG difference is in 32m and 26m 19s, with my pressie has a top score and difficult to beat even by 3ghz FX's, best fx53 score I have seen is 28m 0s.

I am interested to see how a Extreme Edition CPU @ 4.3 performs in that test as it is a real number cruncher.

Mav

 

[boardy]

I would say 7% wont make much more of a difference on the Vcore tbh. As you say I'm not sure how much the mobo can take overvolted so I'd be carefull with the rails. I havn't modded mine at all and it fluctuates from about 1.72v to 1.88v and it's set to 1.825v in the BIOS, so it definately needs the droop mod. Sits at 1.88v all the time so I no my board overvolts slightly aswell now.

Did a 1m test quickly and it came up as 33s, but that was on my custom desktop, so samurize, MBM, WinAmp playing and all my other progs are open which will have slowed it down no end.

I'll do the 32m test a little later when I have more time on my hands. Just trying to find an old bit of hardware to mess about with and practice my soldering on lol.

Cheers

Boardy

 

[Lars]

Boardy,

I pushed the rails on PSU up to plus 5% ( on 3.3, 5 and 12v rails) and your right the difference it makes is minimal.

Not sure if I should push some more (7%?) just to minimise the effect it has now, but concerned how much overvolts the MOBO can take - advice?

If you are going to run superpi, I am not that bothered about 1m test, but u need to set aside 30mins (I use it to check stablility instead of Prime) and do the 32m supertest.

Both my Pressie @ 4.3 and 3.2ee @ 4.1 manage 31s so it's difficult to differentiatie between the two at this level. The BIG difference is in 32m and 26m 19s, with my pressie has a top score and difficult to beat even by 3ghz FX's, best fx53 score I have seen is 28m 0s.

I am interested to see how a Extreme Edition CPU @ 4.3 performs in that test as it is a real number cruncher.

Mav

Well I think it depends a lot on memory performance also. I'm currently awaiting a OCZ 520W PSU, so I can adjust my 3.3V up. Currently my BH-5 memory stops @ 252 Mhz cas 2-2-2-5 @ 3.2V, so above that I have to use the 5:4 divider on memory. With my P4 3.4EE @ ~4.4 Ghz I get 29.05 with mem @ ~205 Mhz. I have a screenshot uploaded here http://www.overclocking.dk/billedbrowser.asp?id=7913&pic=8

 

[maverik-sg1]

Guys, just to let you know if you go over 5% you do get a slight improvement on the droop, but it knocks the AGP out of whack (mega unstable!!).

OK the RAM I have is OCZ latest Pc4400 @ 2.9v 2.5-4-4-4-8 and it runs great with my Prescott at 268 or 270 FSB in 1:1

I do not have any diveders on my MOBO for memory so I guess I just tell it's DDR333 instead of 400, to see if that makes a difference - advice here would be appreciated.

The OCZ Powerstream 520w I am using alos provides a stable system for my Prescott @ 268/270FSB.

Good overclock on your 3.4ee by the way lars, what cooling/volts have you got? My Pressie at 3.2 @ 270FSB can do superpi 32m in 26mins 19s, on a clean install, your Extreme does not perform so well, is that the hit you have to take using ram dividers?

 

[Lars]

Good overclock on your 3.4ee by the way lars, what cooling/volts have you got? My Pressie at 3.2 @ 270FSB can do superpi 32m in 26mins 19s, on a clean install, your Extreme does not perform so well, is that the hit you have to take using ram dividers?

Thanks. Yes, I suspect the main hit I get is from running the system with the 5:4 dividers. However it wasn't run on a fresh install either (4th mainboard with same installation actually). I don't bench too much, I just basically test new parts out when I get them to find max. stable clocks/settings and leave them there. I'm probably not the "typical" Mach owner in this respect

Actually though, I think the Prescott actually is faster in the 32M test clock for clock than the EE. It seems to be the other way around in the 1M test. The best I have gotten is 29 secs. in 1M @ 4463 Mhz, mem @ 5:4.