Venice 3200+ at 2.5 wants more volts a year later.. ideas?

[jahutch]

Hey all,

I've got a bit of a weird situation going. My computer is as follows:

- Venice 3200+ (2GHz) OCed to 2.5GHz
- FSB set to 250
- HT multiplier = 4x
- RAM Divider = 166 (effective speed 209MHz or DDR418)
- Default memory timings for the memory (eg, "by SPD" or "auto")
- Geforce 6800 AGP
- 2GB Ram (4x512MB, Crucial Value)


When I built this rig a little over a year ago, I went through the standard overclocking procedure. My result was a rock-solid Prime95 stable system. Now, just lately, I've been having problems with crashing. On a hunch I ran Prime95 last night and sure enough, it errored after 1hr 9min. The voltage was at +.05 (or 1.45v for a Venice). I upped that to +.075 and am in the process of running Prime again. Its already gone further than it did before, and I suspect this will fix the problem.

Thus my question: Why would a system that was stable for 8-12 months at 1.45v suddenly decide "I want more voltage?" The only thing I changed in my system was I added 2 more RAM sticks. I eliminated memory as an issue, as I ran memtest86 overnight with no errors with the overclock in place. I fail to see why more memory would make the CPU need more voltage. Is this just something that can happen as a CPU "ages" or is there somethign I should be concerned about going on here?

Thanks for your expertise as always!

Edit:
Assuming there is nothing to be concerned about, is 1.475v ok for a Venice on stock cooling? Heat doesn't really seem to be an issue, as its well withing acceptable ranges under full load (say 105-110F or 40.5-43.3C), but I just want to make sure I'm not hurting anything by upping the voltage. Thanks!

 

[n7]

More memory = more stress on teh on-die memory controller.

That'd be my theory.

 

[jahutch]

Thanks n7, makes sense. Because the problems did seem to start when I upped the RAM. So the CPU is working a tiny bit harder, and thus needs a little more voltage.. makes sense to me. So if its Prime stable at +.075 I should be good to go. Any other thoughts on this?

Thanks!

 

[n7]

It may also be that your new RAM isn't stable @ 209.

I had some Mushkin DDR400 that wouldn't get past 204 @ stock timings lol, so you never know.

 

[jahutch]

n7:

I was thinking the same originally, but wouldn't an overnight run of memtest86 return some kind of error if that was the case? I ran it overnight 2 nights ago with no errors, yet Prime95 set on the CPU intense but less memory intense max heat setting errored in under 1.5 hours... also, dropping the FSB to 240, which resulted in stock 200MHz on the RAM still resulted in crashes in games. These 3 things combined led me to believe its likely the CPU is the culprit.

 

[TheRyuu]

n7 is probably right. Memtest just stresses the RAM and maby not the on die memory controller??? Who knows.

Use a multiplyer or somehow lower your HTT (FSB) so that it's stable. Also make sure your on 2T command rate.
And instead of using auto/default timings try loosing some of them up.

 

[n7]

Originally posted by: jahutch
n7:

I was thinking the same originally, but wouldn't an overnight run of memtest86 return some kind of error if that was the case? I ran it overnight 2 nights ago with no errors, yet Prime95 set on the CPU intense but less memory intense max heat setting errored in under 1.5 hours... also, dropping the FSB to 240, which resulted in stock 200MHz on the RAM still resulted in crashes in games. These 3 things combined led me to believe its likely the CPU is the culprit.

Yeah, that would seem to point toward the CPU, but who knows.

As wizboy mentioned, make sure it's set as 2T (though i am sure you never could booted with 4 dimms @ 1T).

I have seen too many issues with 4 dimms on A64s, including an issue i had myself.

I had 4x512 MB Corsair value on my old Neo2 Platinum, & i couldn't get any kind of OC stable till i dropped down to 2 dimms again.

IMHO = A64 mem controller doesn't really like 4 dimms, period.

Sure, it works, but you can see how it used to only work @ 166, & then they improved it to 200 2T, but it's still far from ideal IMHO.

 

[jahutch]

Well good news is Prime95 is still running steady (4 hours or so total now) with the voltage set to +.075. This gives me hope that it will be okay with the 4 DIMMs at the new voltage. I'm going to stop it in about 30 minutes and try some games, then tonight I'll let Prime run overnight. Hopefully all will be well! If not, I'll look into losing 2 of the DIMMs. As for command rate, I'm almost certain its been 2T all along - this is value ram and I've never used anything other than stock settings.

 

[TheRyuu]

Originally posted by: jahutch
Well good news is Prime95 is still running steady (4 hours or so total now) with the voltage set to +.075. This gives me hope that it will be okay with the 4 DIMMs at the new voltage. I'm going to stop it in about 30 minutes and try some games, then tonight I'll let Prime run overnight. Hopefully all will be well! If not, I'll look into losing 2 of the DIMMs. As for command rate, I'm almost certain its been 2T all along - this is value ram and I've never used anything other than stock settings.

So whats the total voltage on the DIMM's? (since +.075 doesn't tell me anything)
Just wanna make sure it's not too much

 

[Avalon]

Besides all the good advice in this thread, also be sure to check your HSF. Dust accumulation will catch up to you quickly and can cause what you are seeing. 1.475V should be fine for a Venice. I usually feel that voltage within 10% of stock is alright for long time use.

 

[jahutch]

DIMMs are still at stock voltage. I was only talking about CPU volts. Because of the RAM divider, the DIMMs are only at 209MHz, and seem fine with stock voltage. As for dust, I just took canned air to the system about a week ago, so all should be well in that department.

I went out with friends tonight so decided to leave Prime95 running.. its now about 9.5 hrs, still with no errors. I'm going to try some games, and hopefully all will be well.

Thanks for all the help!

 

[yacoub]

could also your power supply be fading a little bit over time? fading is the wrong word but you get my question.

 

[akshayt]

Isn't 2T much slower. How would a little higher oc with 2T be better than a little lower oc with 1T?

 

[jahutch]

Well guys, for those of you saying Athlon 64s don't like 4 DIMMs, looks like you may be right. Despite being prime stable for 9+ hours, when I started up a game, sure enough, I got random blue screens.. often enough to be more than irritating. I loosened the memory timings, and that seemed to help a little, but I still got the blue screens. So I finally pulled out the extra two dimms, and sure enough, I was able to play without any BSOD. Now, I can't say for sure I didn't just get lucky, but over the next day or two I'll find out I suppose. Quite annoying my mobo chokes on having 4 RAM sticks.

Now, does anyone think the RAM itself might have been bad? Error free memtest results, etc, seem to suggest that's not the case, but I figured I should ask.

 

[DrMrLordX]

Best way to know if the new RAM is bad is to pull the two old sticks and isntall the two new ones and see how it works

 

[jahutch]

One other question - is there any chance *really* loosening memory timings might help this? Something like 3-4-4-8? I'm not sure the extra RAM is worth it (1gig vs. 2gig doesn't matter in most gaming situations), but is it even an option? Or do these memory controllers choke on 4 dimms when OCed regardless? Thanks!

 

[CaiNaM]

likely nothing to do with ram at all, but rather electromigration. heat/voltage will break down a cpu over time. overclocking/overvolting can weaken the "paths" over time, requiring more voltage to strenghten internal signals..

 

[n7]

Originally posted by: jahutch
Well guys, for those of you saying Athlon 64s don't like 4 DIMMs, looks like you may be right. Despite being prime stable for 9+ hours, when I started up a game, sure enough, I got random blue screens.. often enough to be more than irritating. I loosened the memory timings, and that seemed to help a little, but I still got the blue screens. So I finally pulled out the extra two dimms, and sure enough, I was able to play without any BSOD. Now, I can't say for sure I didn't just get lucky, but over the next day or two I'll find out I suppose. Quite annoying my mobo chokes on having 4 RAM sticks.

Now, does anyone think the RAM itself might have been bad? Error free memtest results, etc, seem to suggest that's not the case, but I figured I should ask.

Perfect example of why P95 is a joke for stability testing.
I've said it many times, yet people still like to think it's such a great indicator.
It's not.

Anyway, that's exactly what happened to me.
I could run P95 (or SP2004), etc, for hours, but it'd crash in games right away.

I would recommend you try lowering your OC to 240 (2400 MHz) so the RAM is sitting @ 200 (DDR400).

Also, try using the 133 divider instead of the 166 divider.

What i'm trying to say is take the RAM out of the equation.
Right now you are still OCing the RAM, which makes it hard to know for sure that's what's causing the issue.

Either at least lower your OC so the RAM isn't running past 200, or lower the divider, so it's below 200.

And yes, loosening timings might help a bit, & won't hurt performance too too much.

 

[jahutch]

Ok, I think my plan of attack is going to be as follows. For now, I've removed the 2 new RAM sticks. I've left the slight RAM OC in place. In short, this computer is back to the way it was the day I built it (when it was super stable). If it is stable in this configuration, im going to swap out the current RAM for the newer 2 sticks and test the same way. Finally, if all is well there, I'll know the issue is in having 4 sticks, and I'll put all 4 in and start lowering the divider / lowering the OC / loosening the timings to see if I can get it stable.

Two side questions however:

(1) When I lower the RAM divider from 166 to 133, Windows won't boot. First reboot, it blue screens. Second reboot, I get a black screen with a mouse pointer I can move and nothing else. Putting it back at 166 fixes the problem. Note this is a fresh install. Any ideas what might be causing this?

(2) As I mentioned above, I had the computer at +.05v (1.45v) when I built it and it was very stable. Now, even back to its original configuration, it needs +.075 to be stable. I think I found the culprit though - even thought the voltage is *set* to +.075 (1.475), the actual voltage the CPU is getting as shown by the PC Health section of the BIOS is only 1.44-1.45. Is there any reason the voltage received would be so much lower than it should be? PSU weakening maybe?

Thanks again, and sorry for all the questions!