OC peak?

[Calzinger]

After a long while of research as to the reason why my computer would never post after raising the fsb over 225 or 230, I finally figured out that you have to lower your LDT multiplier to 1x or 2x to raise the fsb over 230 without having post issues. As strange as that is, whatever, my system posts now.

I've currently have my system stable at 250x10, 1x LDT, 1:1, vcore 1.5v, 3-8-4-4. The temps idle around 38C, not above 41C under load.

When I raised the fsb to 260, windows wouldn't boot. I didn't try too much to fix that, just went straight to 255.

When I raised the fsb to 255, windows booted but I constantly got a semi-BSOD. The screen would just freeze as it was, and there would be that broken record sound. It seemed to happen under tremendous load. I raised the vcore to 1.55v; still had the same problem.

Then I tried 280x9 and everything was pretty stable. That was a pretty relief since I was still running 1:1 and I only have ddr550. But once I tried 285, the system wouldn't post at all, even after numerous restarts.

My cousin has the DFI Ultra-D I think; he was able to hit 309x9 1:1, but I'm not sure if it was completely stable.

Surely I'd love to get higher than 2.55ghz. I don't have a DFI, but think I can push out some more? So far, I think I did the basic OC'ing. Now I need to find a limit.

I plan to run a consistent stable OC'ed speed for some rendering that I do; the more the better.

 

[gaara]

Where'd you get the idea to lower the LDT multiplier to 1x? You have to lower it yeah, but only to 4x and 3x once you start getting over 250HTT

1.55vcore still has a bit of headroom left, especially since you're running water...I'd say you should be able to get to 2.6-2.7 and that'll be your limit which is pretty much peak for a Winchester I'd say

Remember to keep testing stability with prime95 every time you increase your frequency...simply booting into Windows isn't good enough, you need to 100% stress you processor

 

[Calzinger]

Where'd you get the idea to lower the LDT multiplier to 1x? You have to lower it yeah, but only to 4x and 3x once you start getting over 250HTT

Yeah I know, it should end up at around 2000, right? So 4x LDT with 250fsb should be fine (250 * 4 * 2). But apparently, many MSI boards just won't post unless the LDT multiplier is lowered to 2x or 1x. http://forum.msi.com.tw/index.php?topic=78841.0. I couldn't post at 220 with 4x LDT or even 3x LDT at same fsb.

The temperatures seem to be no problem (currently 36C idle, not above 41C load). Think the stability problems at 255 and the no-windows boot at 260 is a result of lack of voltage?

I usually do a superpi (1M) after each boot for stability. Current time is 46 seconds.

 

[gaara]

Try setting LDT voltage to 1.4v and chipset voltage to 1.7v...this might resolve the LDT ratio problem but it should help yield the highest overclock you can get...I'm not sure if the MSI BIOS has those options but if you can get to them you should change them

I would also recommend you have memtest availible and run that before you do anything else...your timings seem loose enough but you may need to tweak your vdimm or your A64 values...I just want to rule out memory being an issue

Vcore will indeed settle stability problems, on AMD64s temperature isn't really a problem anymore as they're all pretty energy efficent...you don't wanna go past 1.7v though or you'll kill your memory controller, I'd personally stop at around 1.65v

Your SuperPi results seem really low though =/

 

[Calzinger]

I was able to get 260 on the FSB. I pushed the vcore up to 1.601 (1.55v * 103.3%) and windows booted no problem, but as you said, that isn't enough to test stability. Sure enough, I tried a superpi only to get an error on each one ("NOT CONVERGENT IN SQR05" in particular). Of course, Prime95 faced errors as well.

How do you suggest I get stability out of this? Push the vcore even higher?

According to the bios settings, I should have a vcore of 1.601v. But the bios settings and CPU-Z are showing me otherwise, ~ 1.475v. These are inaccurate eh? I know that the rail readings are inaccurate, but is the vcore wrong as well? So when you say I shouldn't push beyond 1.65v, is that to what the bios is set to or to what the bios/CPU-Z tells me?

I wasn't able to find LDT voltage. But I did find the chipset (1.5v I believe) and ram voltage (AUTO). Can the RAM be the problem? I thought it was supposed to perform flawlessly until 275 1:1.

 

[gaara]

According to the bios settings, I should have a vcore of 1.601v. But the bios settings and CPU-Z are showing me otherwise, ~ 1.475v.

Huh? So what you're saying is you're manually setting your vcore in your BIOS to 1.6v, and then saving and rebooting and CPU-Z Is reporting it at 1.475vcore? I honestly don't know what the problem with that is...CPU-Z has almost never been wrong when it comes to getting core specifications and you can verify this with central brain identifier therefore I'm thinking your problem is a BIOS/CMOS issue

I would suggest you revert back to stock settings and figure out what is going on with your BIOS settings as you're getting the wrong voltage readoffs...I don't think you'd be able to get a Winnie up to 2.6GHz with little overall stock vcore increase and you sure as hell don't want to be finding your maximum stable overclock with bad voltage readings

Memory...auto sometimes isn't the best setting, although it is true that your memory is rated DDR550 you might still need to tweak your vdimm and bump it up to 2.8-2.9v...although I'm thinking MSI max vdimm is 2.85v

1.7v is the general AMD64 maximum vcore...the memory controller is built into the core itself as you may know therefore that acts as a restriction for how much vcore you can pump through the processor...it's been known to fry past 1.7v in most cases, especially on the newer 90nm fabrication

 

[Calzinger]

In the BIOS, I set the "Adjust CPU VID" to 1.55v. Then I set the "CPU Voltage" to 3.33%. After setting "CPU Voltage" to 3.33% (from nothing), the system booted at 260. But in the BIOS H/W monitor and in CPU-Z, I get a vcore reading of around 1.475v. Is the vcore reading usually accurate? Obviously the rails are not, but is the vcore reading in the bios/CPU-Z correct on most boards?

It fiddles around 1.472v, 1.456v, 1.488v.

I will be emailing MSI regarding the vcore issues.

What is usually the default vdimm? Once this vcore issue can be resolved, I plan to set the vdimm to 2.8v and try to push the CPU as far as the RAM will cooperate (275 I guess), increasing the vcore as necessary for stability but not to exceed 1.65v. Does this sound good?

 

[gaara]

I honestly have no idea what's going on with your vcore and that'd probably be an issue to take up with MSI...if might have something to do with your BIOS revision so you may wanna do some digging with that but in all honesty I have no clue

vdimm should be stock 2.6v however since you're running DDR550 I suspect you'll need to bump up your vdimm

 

[C0RR0SIVE]

Mine allways reports a little low. But other programs report correctly. Try using Everest Home Edition.