Jump to content
Sign in to follow this  
supertux1

Opty 165 + Prime95 woes

Recommended Posts

I'm working on priming my first Opteron 165 CCBBE 0615EPMW (IHS removed)

 

Settings are 334x9 @ 1.55v x %110 (1.705v) Water Coooled.

 

I can get both cores to run for over 8 hours using dual prime 95 tests with the first prime 95 torture test option 'small fft'. Temperatures are about 45 c at load. It runs World of Warcraft max settings for 6 hours absolutely fine, it will also run all the 3DMark benchmarks and I could theoretically make a 3ghz OCDB entry.... BUT....

 

'inplace large fft', 'blend' and 'custom' settings will fail after a few minutes on the first core. (!)

 

Does this mean that although the CPU cores might be stable, the on-chip memory controller might not be when stressed?

 

Is there any way to 'fix' that without losing CPU speed?

 

I'm currently running the ram in my signature at 150/200 divider (about DDR500) and it doesn't seem to make a difference whether I run it 2.5-3-2-5 or 3-3-3-8. I've got all the other settings at 'optimized defaults' for the bios listed -- is there one specific setting in there that will affect the CPU's memory controller that might give it more stability so I can

pass Prime95's CPU/RAM tests?

Share this post


Link to post
Share on other sites

Using A64Info find what your settings really are. That's your stating point. Run dual Prime95,Custom Blend with 800MB per instance.The timings that will have the most influence will be Tref, DDS/DDDS and MAL/RP. If you get the timings off of auto and zero in on the optimal values you should be able to get things sorted out.

Share this post


Link to post
Share on other sites

I reset all bios settings to default/auto and reworked the CPU voltage going up in the smallest steps possible until I got Prime95 (small fft) stable on both cores at 334x9.

(I didn't care about running benchmarks or anything this time.)

 

Suprise! I ended up at about 1.58v (like you) so I'm going to stay put there and declare my memory timings as being the culprit. I put it to 1.61 just for giggles since 1.58v was the very edge of stability. (One notch lower than 1.58v = BSOD / Prime errors.) I'm not worried about frying the CPU, I'll just buy another one. I've already hacked it the IHS off it and have awesome water cooling, it NEVER gets above 45 c even when I crank the voltage multipliers up all the way.

 

RAM is running at 250Mhz (DDR500) which was also the limit with my X2, considering it's only rated for DDR400.

 

The crazy thing is it will run absolutely fine at SPD (2.5 - 3 - 2 - 5) / AUTO settings with no errors at 330x9.

 

Anyway, I set all of the DRAM stuff to AUTO save for the divider (150/200), I got memtest #5 errors at 2.5 CAS but none at 3 CAS, so right now I'm priming custom 800mb with 3 CAS and we'll see what happened when I get home. :)

 

I found the A64Info tool and it looks impressive. Do any of you actually fiddle with the settings while Windows is running instead of constantly rebooting etc?

 

I can see where this could cut down on a lot of time. Found a post on extremesystems about running super pi inbetween setting changes. I'm going to try this.

Share this post


Link to post
Share on other sites
I found the A64Info tool and it looks impressive. Do any of you actually fiddle with the settings while Windows is running instead of constantly rebooting etc?

 

I can see where this could cut down on a lot of time. Found a post on extremesystems about running super pi inbetween setting changes. I'm going to try this.

Yes but I always answer no. The trick is knowing which settings to change and by what amount. You only want to make changes that have an impact on performance not stability.

Share this post


Link to post
Share on other sites
Yes but I always answer no. The trick is knowing which settings to change and by what amount. You only want to make changes that have an impact on performance not stability.

 

Right now I'm just trying to find a stable memory setting for my OC.

(Stable being defined as worthy of an OCDB entry, eg: P95 Custom 800mb 8 hours.)

 

Aren't we supposed to adjust for stability first, then performance?

Share this post


Link to post
Share on other sites
Right now I'm just trying to find a stable memory setting for my OC.

(Stable being defined as worthy of an OCDB entry, eg: P95 Custom 800mb 8 hours.)

 

Aren't we supposed to adjust for stability first, then performance?

Yes. The reason for using A64Info right now is so you can get a handle on what those Auto settings really are.

Share this post


Link to post
Share on other sites
Yes. The reason for using A64Info right now is so you can get a handle on what those Auto settings really are.

 

Ah, gotcha.

 

Well, 3-3-2-5 didn't work - was only stable for about 1/2 hour, but that is a remarkable improvement since the best I could do before was 5 minutes. I'm at 3-3-3-8 now and that lasted for about an hour.

 

For some reason, I cannot get the 140/200 divider to work.

 

Anyway, here are the A64Info settings:

 

 

A64Info.jpg

Share this post


Link to post
Share on other sites
For some reason, I cannot get the 140/200 divider to work.

that's not unusual for a 165. Some won't even work on the 150 divider.

Share this post


Link to post
Share on other sites

Okay an update:

 

Yeah, it will boot @ 334x9 @ 1.7v maybe make it through a 3DMark benchmark if I'm lucky, but any stress testing at that speed that involves the memory is going to crash it.

 

To rule things out, I set memory timings as loose as possible for stability according to the guide, divider 100/200 etc.

We're talking super loose (3-4-4-10) and 100/200 for 167Mhz... and guess what: It still crashes at 334x9 @ 1.7v during memory intensive operations.

 

Am I right in thinking that although the cores and DRAM can handle 334, as evidenced by it booting Windows, the embedded A64 memory controller just can't be stressed at that point? I use the memtest86 in the bios, but I have a feeling that's not really the best thing because it's only single core and does not simultaneously load the processing parts of the core.

 

It will boot 330x9 @ 1.65v, but exhibits some of the same problems as the 334x9 settings, running Prime95 for slightly longer and 'only crashing' just a little bit further into the game benchmarks.

 

Now I'm at 325x9 @ 1.55v which is proving to be remarkably stable no matter what. All game benchmarks worked flawlessly, so I felt brave enough to start banging on the RAM.

I managed to tighten that up tremendously, showing about 3089MB/s in a three hour memtest and benching about 6300MB/s Everest. Even kept the rated 2.5-3-2-5 timings. Just about every setting at it's guide recommended maximum performance while running at 243Mhz. Prime95 has been running for about 5 hours with no errors.

 

So there it is, no 3ghz club for me, but so far has beaten the maximum clock and the benchmarks of my X2 4800+ OCDB!

 

So I'm happy, I guess. :) Maybe avoid this particular stepping for 3ghz.

 

As a last ditch effort, I'm going to try to follow the instructions you gave in this post:

 

http://www.dfi-street.com/forum/showthread...highlight=325x9

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Sign in to follow this  

×
×
  • Create New...