X2 4400+ aggressive OCing

[IamTHEsnake]

I've been waiting for this new computer for a long time and have been very anxious to get it. Now that I have it, I am just as anxious to see just how far I can OC it! So my plea to the OCers of Anandtech, is to help me reach my goal of 2800mhz. I've been doing some OCing and have managed to get the CPU to 2700MHZ @ 1.475 prime stable and am pushing for 2800Mhz. Now @ 2800Mhz I can do all of windows fine, cinebench(multi-cpu test), and superpi 1M @ 31s . But as soon as I try to start prime it fails at the first test saying that it rounded to 4, not the 4.99999975... etc...

Currently here is what I have tuned...

DFI Ultra-d NF4 Mobo.

HSF is Zalman 7700 - CU @ Max speed

CPU @ 2800mhz
vCore @ 1.55v
Core Temp, idle @ 38 , Load @ 56
11x multi have also tried 10x with 255mhz. no go.
4x HTT
280mhz FSB

OCZ 2x1024MB EL Platinum
Mem @ 210mhz
Divider @ 150Mhz
vDimm @ 2.80V
ram timings 2.5 - 3 - 3 - 7
The option right above the timings is ENABLED and the rest of the options below in the timings thing are on AUTO.

PSU is no problem.

Help me make this clockspeed possible!

 

[Dadofamunky]

Cut back a bit on the FSB. See if you can get Prime stable on a lower speed (see my sig). Without Prime stable for a day it's not really a good OC. Once you've established that, begin experimenting with increased voltages and different dividers. I'll defer to others who are actually using DFI boards, which rock as far as I know.

 

[Furen]

lower your ldt/memory divider to take both out of the equation. Then just feed it enough voltage to maintain 2.8GHz. I wouldnt hesitate to cut back 50-100MHz for stability and lower heat, X2s generate a lot of heat at higher clockspeeds and voltages, so dont think of it the same way you would think of a single core cpu.

 

[Bona Fide]

What Dadofamunky said is right. Unless you can't get it stable, it's not a good OC anyway.

Btw...270x11 does not equal 2800...

And also, drop your FSB down to around 260 and run Prime. If it's stable, at least you have a solid point there. Now work your way up. The DFI board should give you all the voltage you need, for both your CPU and RAM. So no worries there.

 

[Shimmishim]

make sure you are running 2 instances of prime to test each core...

 

[Amaroque]

Originally posted by: Shimmishim
make sure you are running 2 instances of prime to test each core...

To fully utilize a dual processor machine, you must run two copies of
prime95.exe. Run one copy of prime95 as normal. Run the second
copy of prime95.exe with the -A1 command line argument. Make sure the
second copy also has the "Start at Bootup" option set.

Dual processor machines can also improve performance by setting
processor affinity. Use the Advanced/Affinity dialog box to do this.

You can also change the affinity with the Task Manager. Just set one instance to "CPU 0", ant the other instance to "CPU 1"

 

[IamTHEsnake]

Yeah the 270 was a typo, sorry bout that!

But anyways last night I ran some benchmarks for stability @ 2.6Ghz with a somewhat lower voltage. I am going to slowly work my way up from there. Prime ran fine all night with the vCore@ 1.4V. Then just for fun I ran 3dmark 05 with an OCed 7800gtx @ 450/1350 and got 8986

First, what is the max safe vCore for the cpu?

Second, what if I up the chipset and HTT link voltage up as well?

 

[Furen]

you shouldnt need to up the ht/chipset voltage unless you go really high on the htt frequency (320+ high, I'd guess).

 

[Vegitto]

Try just as you were (280x10) and run two instances of CPU Burn-In for 24 hours. That'll do the trick. Turn error protection off.

 

[IamTHEsnake]

What, in sandra?

 

[Duvie]

There is a newer version of prime95 available...out on 7/05/05......version 24.13...may have fixed any issues...believe to have fixed the blend test problem with taking too much memory and causing thrashing....

What I find interesting is th efact that one instance of FH loads each core but does not exceed a total of 50%...Why not just peg one core like my P4 w/ HT used to do???

 

[Amaroque]

I've noticed that too, but with different programs. I don't know why it's reported like this. Maybe it's just the way AMD reports CPU usage.

 

[Vegitto]

Originally posted by: IamTHEsnake
What, in sandra?

YHPM.

 

[jpeyton]

Well I can't even get my 4400+ to load Windows at 2.8GHz. Does 2.7GHz at 1.55V, using the same motherboard and same heatsink/fan as you're using. Loads Windows at 2.75GHz, but exhibits the same problems with Prime95 that you're having.

BTW, are you stress testing with Small or Large FFTs?

What I do is just download the ZIP copy of Prime95, extract copies to two folders, and run each one. In the "Advanced" menu of Prime95, you can set the processor affinity. I set one copy to 0 and the other to 1. Then I usually run Large FFTs for a few hours if I'm still tweaking, and at least 24 hours if I'm done tweaking and testing system stability.

I think you should face the realistic possibility that it won't do 2.8GHz without a good water cooler, if not something more exotic. 2.7GHz is nothing to sneeze at; I've seen people with 4400+ reporting everywhere from 2.55GHz to 2.8GHz on air, with a lot of 2.7GHz overclocks.

 

[IamTHEsnake]

2800mhz was the highest I tried last night and it loaded into windows just fine each time. I managed to run benchmarks such as cinebench, super pi 1m, and 3dmark01. All worked except prime95 which failed immediately.

In prime I was using Large FFTs and yes you ccan have dual instances of prime.

I guess I have one of the better clockers, but like any madman, I want more.

haha.

BTW I noticed that as I ran 3dmark01, with the vid card @ stock, the score to go down further and further the higher the cpu clockspeed went up. Is this because I was using dividers and a lowered htt multi?

 

[JohnAn2112]

Originally posted by: IamTHEsnake
I've been waiting for this new computer for a long time and have been very anxious to get it. Now that I have it, I am just as anxious to see just how far I can OC it! So my plea to the OCers of Anandtech, is to help me reach my goal of 2800mhz. I've been doing some OCing and have managed to get the CPU to 2700MHZ @ 1.475 prime stable and am pushing for 2800Mhz. Now @ 2800Mhz I can do all of windows fine, cinebench(multi-cpu test), and superpi 1M @ 31s . But as soon as I try to start prime it fails at the first test saying that it rounded to 4, not the 4.99999975... etc...

Currently here is what I have tuned...

DFI Ultra-d NF4 Mobo.

HSF is Zalman 7700 - CU @ Max speed

CPU @ 2800mhz
vCore @ 1.55v
Core Temp, idle @ 38 , Load @ 56
11x multi have also tried 10x with 255mhz. no go.
4x HTT
280mhz FSB

OCZ 2x1024MB EL Platinum
Mem @ 210mhz
Divider @ 150Mhz
vDimm @ 2.80V
ram timings 2.5 - 3 - 3 - 7
The option right above the timings is ENABLED and the rest of the options below in the timings thing are on AUTO.

PSU is no problem.

Help me make this clockspeed possible!

Turn down your HTT multiplier from 4x to 3x if you're going above 250 FSB.

 

[IamTHEsnake]

Really? Even if the mem is using a divider?


I will Get to some more OCing tonight. Will update soon.

 

[GuitarDaddy]

Yep, HTT multi has nothing to do with memory

 

[IamTHEsnake]

But won't dropping the HTT multi too low, start lowering performance?

Ex 4x250 = 1000mhz
3x250 = 750mhz

I don't understand why lowering the HTT will help me.

But I will give it a try..... right now.

 

[GuitarDaddy]

The total HT speed makes almost zero difference to performance, but anything much over 1000 can cause system instability. Thats why over 250mhz you should drop it to 3x

 

[IamTHEsnake]

no go for prime again with HTT @ 3x.

However superpi 1M ran at 31s.

vCore is 1.55v and clock @ 2800mhz.

FSB 280 x 10

Ram divider at 140mhz and 2-3-2-5.

Should I up the voltage a bit more?

 

[Pabster]

No.

1.55V is pushing it already. Unless you'd like a X2 keychain...

 

[GuitarDaddy]

I agree, thats too nice a chip to risk it. 2.8ghz is like having a pair of FX57's under the hood be happy

 

[Furen]

weee, x2 keychain.

I've heard of some people having problems with certain memory dividers, you might want to think about that as a possible source of instability (though I'd say it's most likely the clockspeed).

 

[IamTHEsnake]

yar