E8400/8500 degradation myth possibly busted?

[dnottis]

Sounds to me like it's the mobo's fault now... or maybe a bug in premature 45nm BIOSes. Which reminds me, my mainboard also gives different results with the same BIOS settings... after each CMOS reset. I've found some consistent or at least near perfect settings that wouldn't give much of a difference whatsoever, but +-.01v away from any of the voltage options, and I'd see the system acting funky.

The board has been rock solid. Using for awhile now with my Quad and had no issues. If anything I'd be more suspect of the 45nm bios. In any case I've had DFI boards in the past that would flake out and require a CMOS clearing so it's not a big deal to me. After all its electronics.

[mrcape]

I end up clearing my cmos about 20 times a week, even more if I'm doing some testing.. I do make a lot of changes though and I think it's just good practice to clear it.

[KTE]

Back on topic - degradation.

Alrigh I've now played with an E8400 and Xeon E3110 and have a few things to add about degredation.

- After running higher volts then coming back to low volts it will fail prime at a previously stable OC. After a few days this seems to work itself out however and I can prime the lower volts I was before. (Im not even talking high volts, I'm talking 1.4v). I actually found that in a few cases resetting the CMOS and starting over fixed this.

I get the exact same behaviour but with E8400 and Q6600. I'm talking 4 MBs so far and around 16 BIOSes between them. It's the chips in my case not anything else.

I don't use Orthos, established buggy code on stabe platforms, just mucks everything up.

[RunawayPrisoner]

Confirmed: Vdroop and Vdrop modding will NOT cause spikes. Seems like... it's the PSU now.

[cadaveca]

or bad regulation on the board itself, keeping in mind..where is this reading taken from? A sensor that gets it's readings from where in the current pipeline?

[sfdmalex]

Well my conclusions are teh following:

Ive had supposed degradation...

Which was narrowed down to my board being droopier then the first day I installed the chip. At first the droop was quite big, but it got real bad the next few days.

Vdroop mod done, no more vdroop.

System still not stable. I run OCCT and discover that my voltages are all over the map, some bad spikes too. Vcore not stable even though I have the vdroop mod.

I install a new PSU, voltages dead stable, perfect line over a one hour period. Vcore dead stable too.

Now runs like a champ at the old voltages.

--------

From this I can narrow down:

First make sure you have a good PSU that can handel the load, logic says, get more then you need so you never load the PSU hard enough for it to actually work to its full potential.

Then makesure your Vcore is stable, if you have a good PSU which you are SURE is working fine then any vcore instability is the cause of the board, hence perform the appropriate mods to make sure your board is stable.

Here is my vcore after droop mod and psu change
http://www.is-mad.com/upload/userfil...0h05-VCore.png

My voltage graphs are the same, cept for the 5v, which is running substatially lower then 5v (whatever the reason).

My previous Vcore was this

http://www.is-mad.com/upload/userfil...2h15-VCore.png No Vdroop Mod, old PSU

http://www.is-mad.com/upload/userfil...0h25-VCore.jpg With Vdroop mod, old PSU.

And the volts were even more all over the place.

http://www.is-mad.com/upload/userfil...0h05-VCore.png With Vdroop mod, NEW PSU.


So, my system was unstable as the Vcore was a mess, jumping around like crazy, requiring me to bump up the voltage more and more to a point were the voltage is so high that any drop in voltage cant be low enough for the system to crash. The moment I stabilized my Vcore via psu and motherboard mods the chip is working fine at the expected voltage.

------------------


*Old PSU = Enermax 495W
New PSU = Corsair HX Series 620W

[RunawayPrisoner]

Well my conclusions are teh following:

Ive had supposed degradation...

Which was narrowed down to my board being droopier then the first day I installed the chip.

Vdroop mod done, no more vdroop.

System still not stable. I run OCCT and discover that my voltages are all over the map, some bad spikes too. Vcore not stable even though I have the vdroop mod.

I install a new PSU, voltages dead stable, perfect line over a one hour period. Vcore dead stable too.

Now runs like a champ.

You just nailed the coffin. Seems like... case closed?

[sfdmalex]

You just nailed the coffin. Seems like... case closed?

Just updated the post, added a lot more info.

[RunawayPrisoner]

Read up. And what can I say? That's like more coffin nail there. That pretty much kills the "lower than specs but still degrading" theory. Actually, I'd think that with a rock solid system, you can run any kind of voltage you want... as long as you have the cooling for it.

[sfdmalex]

Read up. And what can I say? That's like more coffin nail there. That pretty much kills the "lower than specs but still degrading" theory. Actually, I'd think that with a rock solid system, you can run any kind of voltage you want... as long as you have the cooling for it.

Yep, you just need a rock solid system. This chip just doesnt like instabilities.

The case might be that your Vcore graph is stable, but you check your voltages and thats spiking or dipping, even though that might not reflect in your Vcore graph it still might or will cause a crash.

My suggestion to anyone suffering from "degradation" inspect your voltage and Vcore graphs. I bet they are nothing like this

http://www.is-mad.com/upload/userfil...h05-Volt12.png
http://www.is-mad.com/upload/userfil...0h05-VCore.png

[RunawayPrisoner]

Well, my vCore and 3V are the only thing that stay linear. My 12V and 5V jitter... but they don't jitter that much. Here's a perfectly stable run:

http://i235.photobucket.com/albums/e...0h42-VCore.png
http://i235.photobucket.com/albums/e...0h42-Volt3.png
http://i235.photobucket.com/albums/e...0h42-Volt5.png
http://i235.photobucket.com/albums/e...h42-Volt12.png

And please pay attention to the "Ripple:" note in the picture. It'd tell you if there's any discrepancies. IMO, if vCore goes less than 1% ripple, the test may pass stable.. provided you have enough vCore for the test.

[kiwi]

I have seen a few 2V spikes as well (dont remember what mb and PSU, had too many of them ) but you can't be 100% if that happens for real unless you have an oscilloscope

[dinos22]

On the subject of spikes and dips.

I changed my power supply, an older noisetaker enermax 495W to the Corsair HX Series 620W and guess what, not one spike or dip on OCCT. And Vcore 100% stable.

was going to suggest a better PSU

i don't trust any PSUs after this HX620 i have....awesome psu

those spikes are scary

[dinos22]

I have seen a few 2V spikes as well (dont remember what mb and PSU, had too many of them ) but you can't be 100% if that happens for real unless you have an oscilloscope

this is true as well but bloody hell what else do we need to buy these days lol heheh

[sfdmalex]

I have an osciliscope but its not like IM gonna sit and stare at it for an hour looking for spikes...

UNless someone has one of those fancy LCD scopes that write everything in a file for you.

[dnottis]

So, my system was unstable as the Vcore was a mess, jumping around like crazy, requiring me to bump up the voltage more and more to a point were the voltage is so high that any drop in voltage cant be low enough for the system to crash. The moment I stabilized my Vcore via psu and motherboard mods the chip is working fine at the expected voltage.



I really think that right there is 99% of the problem. I have mine vdroop modded and using 1000 watt TT ToughPower I have no spikes. Occaisionally after testing higher volts and getting Orthos failures I may have to clear the CMOS and reload the profile.

As for buggy code, the only buggy code I've used recently was Vista. Orthos is a pretty good indicator of overclock stability and I'll continue to use it. Vista as an OS, well thats another story. I run orthos in XP exclusively.

[RunawayPrisoner]

But if you can get your system stable in Vista, you know it's more than just stable.

[dnottis]

But if you can get your system stable in Vista, you know it's more than just stable.

If you like Vista you're already challenged

[RunawayPrisoner]

Yeah... and nothing beats a 3.90GHz quad stable in Vista IMHO.



Sometimes my OCs would boot into XP and stay stable for days... but they would not even boot into Vista. Now they do, and if they stay stable, that means the OC cannot be anymore successful. After I've run OCCT for an hour on Vista, it's usually days of Prime95 on XP.

Edit: Or maybe over 54000 hours (roughly 6 years) with Prime95 in Vista would do the trick? (please excuse C1E and SpeedStep. Real clock is 3800MHz)

[mrcape]

Yep, you just need a rock solid system. This chip just doesnt like instabilities.

The case might be that your Vcore graph is stable, but you check your voltages and thats spiking or dipping, even though that might not reflect in your Vcore graph it still might or will cause a crash.

My suggestion to anyone suffering from "degradation" inspect your voltage and Vcore graphs. I bet they are nothing like this

http://www.is-mad.com/upload/userfil...h05-Volt12.png
http://www.is-mad.com/upload/userfil...0h05-VCore.png

Here are my OCCT charts from today with the best stable clock 4.25ghz at 1.325vcore set in bios -

12v

5v

3v

Vcore


Here is what I could get in Feb - same clock 4.25ghz 1.28vcore, now requires 1.325vcore with exact same settings -

[RunawayPrisoner]

That's actually 1.34v... drooping down to 1.33v?

[mrcape]

Yeah that's right. Just stating what was set, not assuming it's real. Real is what's on the chart..

[RunawayPrisoner]

I know. But that means... your motherboard over-volt it?

[mrcape]

It seems to overvolt .01v at idle and pretty much right on at load.

[RunawayPrisoner]

Kinda dangerous... if you ask me. Did you do a vdroop mod?