PDA

View Full Version : Barton 2500+ overclocking problems


reelbigfish
18 Jul 2003, 8:00pm
Hey guys. I recently purchased a Barton 2500+, Vantec Aeroflow cooler and Shuttle AN35N-Ultra motherboard. I'm running it with Geil PC3200 RAM. It runs fine at all stock speeds. When I overclock it to 10x200 it works fine. It also works fine at 10.5x200, which are 3000+ speeds. At this speed my 3dMark 2001SE score is 15937. When I change the multiplier to 11x200, which is 3200+ speeds and a 100Mhz boost in clock speed, the score drops to 14581. It's faster, yet drops over 1000 marks. Any ideas as to why this is happening? Sandra says the clock speed is higher, and the memory bandwith is still the same. I have no idea.

TheLostSwede
18 Jul 2003, 8:39pm
Very strange and another lesson learned that the synthetic benchmarks isnt worth too much really.

Maybe that 11th mutli isnīt showing itīs real value somehow?

Try to set it to 9.5 and run a 3d. The difference between 1 or 1.5 in multi is about 1000 points if iīm not mistaken. Edit, thats more of a guess, but try it. I wanna see if that 11th maybe in reality is something else.

Omega65
18 Jul 2003, 9:37pm
Athlons required Low Latency memory to achieve their full potential.

If your memory timings are Cas2 at FSB333 (166mhz) and Cas3 at FSB400 (200mhz) you'll experience a drop in performance.

Make sure you're running at least Cas2 -3-3 timings at both speeds (Cas2-2-2 is better) and retest

reelbigfish
18 Jul 2003, 10:54pm
both tests were run on a 400Mhz FSB so the memory and FSB were clocked the same with the same timings. The only thing that changed was the muliplier. I will bump it down a .5 mult to see what happens now.

TheLostSwede
18 Jul 2003, 11:21pm
Omega65 said
Athlons required Low Latency memory to achieve their full potential.

If your memory timings are Cas2 at FSB333 (166mhz) and Cas3 at FSB400 (200mhz) you'll experience a drop in performance.

Make sure you're running at least Cas2 -3-3 timings at both speeds (Cas2-2-2 is better) and retest

True but it dont justify a thousand point loss on a 100 mhz faster cpu.

GHoosdum
18 Jul 2003, 11:40pm
Maybe this is a demonstration of the bias that synthetic benchmarks typically show toward the Intel processors... maybe Intel paid them to show a drop in performance for the XP3200+ so that people would buy the P4 instead.

reelbigfish
19 Jul 2003, 12:23am
ok, so I retested and here is what I got:
11x200 - 14562
10.5x200 - 15989
10x200 - 15620
9.5x200 - 15288

very strange. I will have to re-run all the sandra tests to see whats up. Also, my mem timings are CAS 2.5-3-3 @ 400Mhz.

TheLostSwede
19 Jul 2003, 1:53am
I would guess that 11 multi actually might be 8.5.
Can you run a 3d on 8.5?, im curious :P

TheLostSwede
19 Jul 2003, 1:54am
GHoosdum said
Maybe this is a demonstration of the bias that synthetic benchmarks typically show toward the Intel processors... maybe Intel paid them to show a drop in performance for the XP3200+ so that people would buy the P4 instead. ;D

muddocktor
19 Jul 2003, 2:17am
RBF, I"ve seen somewhat the same phenomenon with a Tbred 2100 I used to have in my Asus board with folding. I had been folding with it at around 2300 MHz and I decided to bump the speed up some more and got it somewhere in the neighborhood of 2360 and the folding times got longer, not shorter. I theorize that you are getting a hair past your max overclock at that temp and vcore and it's getting enough tiny errors or something in it's calculations to where it actually processes slower than at a lower clock speed. Bump it back up to 11 X 200, but bump the vcore up some more and see if that helps. You could also pull the side cover off your case and put a box fan blowing at it to help cool the proc off too.

reelbigfish
19 Jul 2003, 2:24am
I will try that mudd. I benched at 8x200 and got a a score of 14615. So thats basically what I get at 11x200. I will now bump up the vcore.

reelbigfish
19 Jul 2003, 2:36am
ok, so I bumped the vcore up to 1.7v. I tested twice. the score.....

13637. Yes, thats right, another 1000 points lower! :wtf:

:banghead:

muddocktor
19 Jul 2003, 4:25am
OK, back down the vcore and see if you can get some extra cooling on it, like with an extra fan blowing into your hsf. It sounds like it's heat related, because it got worse scores with the higher vcore than before. That's a nforce2 board, right? Does your bios have a setting that enables clock throttling at a certain temp, like the Asus, Abit and Epox boards do? Maybe you are seeing some clock throttling due to heat.

a2jfreak
19 Jul 2003, 7:25am
BIOS update?

Video driver update?

Also, you said stock voltages. If I'm not mistaken, Geil requires a higher VDD than stock to run PC3200. I'd up the VDD a hair and make sure you're running the most recent BIOS, most recent video driver and most recent 3DMark2001 patch.

Also, did you verify the speed of the CPU with something like WCPUID or CPU-Z?

TheLostSwede
19 Jul 2003, 11:46am
By this time, he would get errors in 3d already if it was unstable.
If this was common, i would have seen this before but i havenīt.

reelbigfish
19 Jul 2003, 12:09pm
well guys, thanks for the help so far. Right now I'm headed out for the weekend. when I return I will attempt to tame this beast. I appreciate all the help so far.