E8400/8500 degradation myth possibly busted?

[jcniest5]

Interesting finding...I'm gonna subscribe to this thread.

[tranceaddict]

I am not sure if it is degradation, but my chip was 15h+ Orthos stable @ 4200MHz with 1.344 volts. After just 1 hour of running 3DMark05/06 @ 4500 with 1.43V, I now need 1.352 volts to be stable at 4200MHz. The wierd thing is that now with 1.344V when I run 4200MHz, Orthos fails between 35min-7.5h randomly. All other settings and temperature variables are pretty constant. As long as it does not degrade any further I am content with that small increase in voltage for stability, but I am never going to bench with it above 1.4V.

[Leeghoofd]

I really don't see why peeps wear out their hardware with these tests over and over again... really beats me.... STOP IT !

[mrcape]

None of this is helpful unless we stay on topic and try to get some new info. And I don't think the interesting info is how to test a chip, that's been done to death. It makes me want to puke how often I read arguments about what "stable" is.

There are plenty of short tests that will work to see what a chip can do at a certain clock with a certain voltage. Using those, I can identify just how much my chip has degraded.

An interesting data set could be - chip's vid, what it's baseline clocks and req. voltage were out of the box, what was done to it out of spec (benchmarks used & how much v) and what vcore it takes after degradation to meet the previous clocks and pass the same tests.

Another set could be used to illustrate whether the chips continue to degrade, or can sustain a bit of damage and then act as normal, just needing a tad more vcore. I've been testing mine with the same methods as before and after degradation and it seems to be holding its own at the same clocks/vcore as just after degrading. That tells me it's not still sliding downhill.

The $hitty part about this is that noone wants to believe that these chips degrade, or that they didn't take readings on the way up, and have no way to compare. Everyone who bought one of these chips is an early adopter and has to expect grey areas with overclocking them. Another road block are the armchair quarterbacks who don't even own a 45nm chip and derail the discussion with an attempt at a debate about how long to stress test a chip.

Crazy Nuts or anyone else with a chip that's degraded - how is it holding up weeks down the road? Still sliding down? Holding its ground? This is the stuff that's going to interest me at least.

[CrazyNutz]

With my method, he'll sort it out in few minutes
I'll take a look by the way at WCG too, looks to be interesting

I'm pretty certain that increasing the voltage would likely fix the issue here,
but that does not tell us anything that we don't already know (i.e. increased
voltage will/should stabilize our overclock).
The state that I'm at right now is giving me very reproducible results:

1. I let the machine idle for 18-24 hours.
2. I then run prime95 10k FFT. Error in < 2-5 minutes
3. Continue restarting prime95 10k until stable at > 2 hours.

I think doing this (for days/weeks) until this "warm up theory"
does not work any longer will tell us if any degradation is occurring.
If it does that would suck since this CPU has absolutely never been
passed 1.33vcore.

Consider this scenario:
The first thing I do when I get a new CPU is:

1. Up the FSB a little 10 - 20 mhz at a time
2. run prime95
3. Continue step 1 - 2 until prime starts giving errors

then

4. Up the vcore
5. Run Prime95
6. Continue steps 4 - 5 until prime stops giving errors


At some point I'm at or around my target frequency and a comfortable
vcore, temp, and prime stable for 8 hours.

So in this scenario no one would ever notice the need to have a
"warm up period" since the initial OC procedure involves a ton of stress
testing which as a side effect completes the theoretical "warm up period".

Now at a later date, I (and others) will likly run the stress tests again (unaware of the
"warm up period") only to find the test returning errors, yikes! This is where I think
some start to label this as Degradation.

Here are the real questions:
1. Will the "warm up" procedure eventually fail to make the OC prime stable again like
it was initially? I intend to figure this out. If it does fail then Degradation is certain.

2. Is the overclock considered a success If there is a need for a "warm up" to make it
prime stable on a daily basis? I would say no (depends on what you're doing i.e.
scientific computing or just playing games, or have OCD like me ) .

3. Will just a little more vcore eliminate the needed "warm up" all together? Maybe,
and I think this is what you're saying. I will test this once I'm sure the needed "warm up"
is not a sign of degradation.

Thanks

[Mav451]

Another road block are the armchair quarterbacks who don't even own a 45nm chip and derail the discussion with an attempt at a debate about how long to stress test a chip.

Crazy Nuts or anyone else with a chip that's degraded - how is it holding up weeks down the road? Still sliding down? Holding its ground? This is the stuff that's going to interest me at least.

Lol while the armchair QB's are annoying, I do agree on consistency. Probably update the first post on a widely-agreed "stability" test that EVERYONE will use. Problem is that I doubt people used the very same test before stressing their chip with too much vcore/FSB/etc. Agreeing on a test after the damage is done means we probably can't squeeze more info...unless more degradation is to be had -_-

[mrcape]

Well, thing is you can't go backwards and people aren't like to spend cycles re-testing. And never mind getting people to agree on the same test.

As long as users used some form of test, both before and after degradation at the same exact settings, the info could help.

edit - But then again Mav41, if you're proposing a method going forward for fresh chips I think you're right that the more consistent the better. It would be ideal to have say 20 fresh chips from the same batch and run through a cycle of tests with each chip in the same environment.

[CrazyNutz]

None of this is helpful unless we stay on topic and try to get some new info. And I don't think the interesting info is how to test a chip, that's been done to death. It makes me want to puke how often I read arguments about what "stable" is.

There are plenty of short tests that will work to see what a chip can do at a certain clock with a certain voltage. Using those, I can identify just how much my chip has degraded.

An interesting data set could be - chip's vid, what it's baseline clocks and req. voltage were out of the box, what was done to it out of spec (benchmarks used & how much v) and what vcore it takes after degradation to meet the previous clocks and pass the same tests.

Another set could be used to illustrate whether the chips continue to degrade, or can sustain a bit of damage and then act as normal, just needing a tad more vcore. I've been testing mine with the same methods as before and after degradation and it seems to be holding its own at the same clocks/vcore as just after degrading. That tells me it's not still sliding downhill.

The $hitty part about this is that noone wants to believe that these chips degrade, or that they didn't take readings on the way up, and have no way to compare. Everyone who bought one of these chips is an early adopter and has to expect grey areas with overclocking them. Another road block are the armchair quarterbacks who don't even own a 45nm chip and derail the discussion with an attempt at a debate about how long to stress test a chip.

Crazy Nuts or anyone else with a chip that's degraded - how is it holding up weeks down the road? Still sliding down? Holding its ground? This is the stuff that's going to interest me at least.

All Good points.

My first chip that I initially considered Degraded is not
sliding down hill. But I am not sure If it is actually degraded as it appears
I may have misdiagnosed it because of the "warm up" theory, and I did not
reliably log the frequency @ vcore it would do initially. I just know that at some
point I set the vcore up over 1.4v and frequency over 4ghz and it was prime stable
for over 8 hours. But a few days later It began failing prime in < 5 minutes. So
in fear of degradation I backed off the vcore and frequency, and finally got it prime
stable again, then prime stated failing again. I decided to put this proc in my
HTPC and run it at stock. I then got a new e8400 for my main rig and decided
to not go beyond 1.36v (It still has not had over 1.33v) I got it to 3960mhz 440fsb
@ 1.328 vcore (reported by CPU-Z) prime stable 8 hours(7 hours blend, 1 hour 10k)
, and it happened again a few days later 10k prime was failing (in < 5 mins) on my
new CPU! That's when I realized that it would prime a little longer after every time
I restarted prime95 like this:

1st run error in < 5mins
2nd run error in >5min but <10
3rd run error in > 20 but < 40
4th run 2 or more hours...

So what I thought was degradation, I now think is caused from the CPU
not getting the "warm up" period. Also this may be (and probably is) an indication that
it was not so prime stable like I initially thought.

[CrazyNutz]

I really don't see why peeps wear out their hardware with these tests over and over again... really beats me.... STOP IT !

We all have Obsessive Compulsive Disorder No but seriously I do

[CrazyNutz]

I am not sure if it is degradation, but my chip was 15h+ Orthos stable @ 4200MHz with 1.344 volts. After just 1 hour of running 3DMark05/06 @ 4500 with 1.43V, I now need 1.352 volts to be stable at 4200MHz. The wierd thing is that now with 1.344V when I run 4200MHz, Orthos fails between 35min-7.5h randomly. All other settings and temperature variables are pretty constant. As long as it does not degrade any further I am content with that small increase in voltage for stability, but I am never going to bench with it above 1.4V.

Help us out. Set it back to 4200mhz @ 1.344vcore and keep rerunning orthos after it fails to see if it gradually gets back to the stable state (or a more stable state).

[Leeghoofd]

Your chip will get degraded with all these hours of prime and co as normal software nor games doesn't torture ya rig as much...

I can vouch that my benching sessions have degraded one QX for sure... out of the box it wasn't really up to standards as it never ran at stock VID... but going from 1.16 to 1.19V at stock speeds tells the story doesn't it... and that is about the same value on 3 different mobo's... I'll burn this sucker up ! as my high end Wc setup with triple rad should be able to cool this CPU at 4.5ghz at 1.55 volts for a short duration... other QX in my gaming folding rig has never been over 1.4volts and still runs happily at 3.6ghz 24/7 with 1.14 volts...

[tranceaddict]

Help us out. Set it back to 4200mhz @ 1.344vcore and keep rerunning orthos after it fails to see if it gradually gets back to the stable state (or a more stable state).

It does not, when it initially failed at 4200 with 1.344, I automatically started another session of Orthos, and then another and yet another. They all varied in the duration of stability, but it seems that the first one was the longest. In other words, when I leave the PC off for the night and the next morning I run 4200 @ 1.344 it will be the longest stability (about 4-6h), after it fails if I restart it, the duration of stability will only decrease.

BTW, these are my system specs:

C2D E8400 - 4200MHz @ 1.35V
ASUS P5K-E WIFI
Big Typhoon
2GB Patriot XP PC2-8500 4-4-5-8 @ 2.2V
XFX 8800 GTS 512 @ 785/1900/2160MHz
WD Raptor 74gb
Fortron AX-500 Bluestorm PSU
21" Hp 1130 CRT/NZXT Lexa case
Razer Diamondback 3G mouse

Update:

Something is really screwed up here, today I failed after 2.5h at even 1.35V where as it was perfectly stable yesterday. Maybe it is those goddamn timings, will run overnight with 5-5-5-15 to check.

[CrazyNutz]

It does not, when it initially failed at 4200 with 1.344, I automatically started another session of Orthos, and then another and yet another. They all varied in the duration of stability, but it seems that the first one was the longest. In other words, when I leave the PC off for the night and the next morning I run 4200 @ 1.344 it will be the longest stability (about 4-6h), after it fails if I restart it, the duration of stability will only decrease.

BTW, these are my system specs:

C2D E8400 - 4200MHz @ 1.35V
ASUS P5K-E WIFI
Big Typhoon
2GB Patriot XP PC2-8500 4-4-5-8 @ 2.2V
XFX 8800 GTS 512 @ 785/1900/2160MHz
WD Raptor 74gb
Fortron AX-500 Bluestorm PSU
21" Hp 1130 CRT/NZXT Lexa case
Razer Diamondback 3G mouse

Update:

Something is really screwed up here, today I failed after 2.5h at even 1.35V where as it was perfectly stable yesterday. Maybe it is those goddamn timings, will run overnight with 5-5-5-15 to check.

Hmm interesting. Just the opposite of what is happening with my CPU.

Those timings are wierd.

[Hughyhunter]

I read the original post on this thread and read through a few of the others. I agree that the 45nm has quite a bit more degradation than maybe an older 65nm conroe.
What we have to remember though is that just like automobiles... when you drive them things degrade (kind of like tires and an engine) so it's really not a big deal.
I have found from my own personal experience though that when you run prime95 for hours on end (with massive amounts of heat) that the CPU has no choice but to degrade (once again because of heat) and of course you'll have to bump voltage up in the future because of this.
We have to remember the natural order of things and that is that things (no matter what the matter organic or non-living) degrade, some at greater rates than others and absolutely things will degrade much faster when you add heat to the equation.
So... dont run PRIME95 for hours on end! I only run PRIME for about 30mins max... Why more when everything I do presents itself as stable? I play games and it doesnt crash. I rarley get system errors and if I do it's because of Vista not an overclock.

[aoch88]

I think Prime has been around for more than a decade and with recent technology such as dual core and 45nm, perhaps it is no longer that relevant anymore as a stress tool?

I'm not sure but what do you guys think? I'm sure there are good people around which could code a better stress test program?

[tranceaddict]

I do not agree with the idea that Prime can wear out a processor in a matter of days. If on average Prime is twice as stressful as the most stressing game, you telling me that 100 hours or so of gaming would degrade your CPU to wher it is noticable. On my Opteron 146 I ran about 50 hours of Prime and few 1000 hours worth of games and it was just as stable as when I got it.

[Cronos]

I'm not sure but what do you guys think? I'm sure there are good people around which could code a better stress test program?

There is one - Linpack 64bit. 10-15C over Prime Small FFTs on C2Q is guaranteed. This is for average overclock, not exceeding 3.6GHz (talking about 65nm C2Q). Those with presumably "stable" 4+GHz will likely to see 20C or more.

[aoch88]

Thanks Cronos but is it safe to stress test the Wolfdales on 1.4V++? I personally to not quite agree with that degration idea because burn-in process have always been helpful and I reaffirm that my heavily burned-in E2140 earlier had me overclocked it to 3.48Ghz Orthos stable.

[jonny_ftm]

Here are the real questions:
1. Will the "warm up" procedure eventually fail to make the OC prime stable again...

It should continue to degrade unless you lower the volts/frequency. Usually it is an "infinite loop" pocess

2. Is the overclock considered a success If there is a need for a "warm up" to make it prime stable on a daily basis? I would say no (depends on what you're doing i.e.
scientific computing or just playing games, or have OCD like me ) .

Suerely no, you also say it. How can you sell a chip saying you need a warm up period Of course, it can suit your needs, but this is going to change as you should warm it up longer and longer: annoying task that will fail at a moment

3. Will just a little more vcore eliminate the needed "warm up" all together? Maybe,
and I think this is what you're saying. I will test this once I'm sure the needed "warm up"
is not a sign of degradation.

Thanks

Well, if increasing vcore solves this, it is a sign of degradation. In any case, you know well a normal CPU shouldn't warm up. If you're obliged to go in such a stage, means your CPU is no longer normal, it is degraded. Even Intel won't exchange it for sure (I would try RMA though before it completely degrades).

I think Prime has been around for more than a decade and with recent technology such as dual core and 45nm, perhaps it is no longer that relevant anymore as a stress tool?

I'm not sure but what do you guys think? I'm sure there are good people around which could code a better stress test program?

Well, a normal CPU, not overclocked, ran at specs, will run Prime95/OCCT/Linpack64/TAT without issues for more than a month. CPUs are complexe structures with many functions. You need many tests for each part to be sure no problems are. Also, with overvolting, overclocking, degradation, the shorts causing Prime95 and other sort of errors are kind of electric interferences at the electron level. So, to produce such a short with some electrons loosing their path and conflicting with other paths, it can take time. The less time it takes, the more frail is your setup

Personally, I like to OC and like my PC still stable: in my case, I have to load it, it should work like in real world. Don't pump crazy voltages especially on 45nm chips, don't forget overshooting problem with all of us disabling vdroop, keep temperatures under control, and you should be fine benching every thing you like

[Hughyhunter]

I do not agree with the idea that Prime can wear out a processor in a matter of days. If on average Prime is twice as stressful as the most stressing game, you telling me that 100 hours or so of gaming would degrade your CPU to wher it is noticable. On my Opteron 146 I ran about 50 hours of Prime and few 1000 hours worth of games and it was just as stable as when I got it.

You have to remember that even the most stressful game (like crysis for GPU or FSX for CPU) isnt going to raise the temps on the proc as much as prime. 60 is a lot for any electrical components. I work on aircraft electrical systems and the #1 leading cause of a component failing is excessive temperatures. Also temps that are too low... but prime takes a cpu to the ultimate hottest place it's ever been. So I can understand or believe that when you prime with not stock voltages you will get degradation in a short period of time... It might not be that much but it's there for sure. It's not like an organic material where you stress it and it grows back stronger... this is metal and silicon we are talking about.

[Leeghoofd]

I do not agree with the idea that Prime can wear out a processor in a matter of days. If on average Prime is twice as stressful as the most stressing game, you telling me that 100 hours or so of gaming would degrade your CPU to wher it is noticable. On my Opteron 146 I ran about 50 hours of Prime and few 1000 hours worth of games and it was just as stable as when I got it.

You have to compare apples with apples, these CPU's are very very sensitive to voltage and heat... if you stress them with Prime temps will go way higher like stated in the post above then when you use it daily... 65nm and older CPU's are better resistant to abuse... peeps tend to forget that...

[Cronos]

You have to remember that even the most stressful game (like crysis for GPU or FSX for CPU) isnt going to raise the temps on the proc as much as prime.

There are some who're using they computers for more than gaming. It is fairly easy for number crunching software to exceed the Prime95 level of load, as it is nothing really impressive in the first place.

60 is a lot for any electrical components.

Not really. For silicon technology, 60C is nothing.

but prime takes a cpu to the ultimate hottest place it's ever been.

Nah. It is very easy to exceed Prime95 load level by 5-10C, and there are some apps. which can give 15-20+C higher (for 65nm Quads, depending on speed and voltage)

[mrcape]

I'm also interested to see how stability scales up testing with linpack on these 45nm suckers from the get go. I'm done testing the two I have, but I think I'll give linpack a go as a testing tool for my next chip, which will be in a couple months.

[oublie]

well guys this is interesting so i'm trying a little reverse old school overclocking theory. Many of you may already be aware that it is possible to "train" your chip to run higher speed at lower voltage by starting low.

I've lowered my volts as low as they will go on a p5k prem 1.1v 1.072 loaded and am gradually increasing fsb. then orthos for an hour then fsb up again. Last night my box wouldn't boot to windows at that speed so i had to use set fab to run higher.but i'm now running at 345 fsb and raising it by 5 fsb every 1 hour othos stable. I'm hoping this will allow me to get a better clock on phase change when i switch over by using less volts for the same clock. This is on the stock intel cooler and full load on both cores temps are 43 degrees c.

Don't know if this will help or if it will delay chip degradation once finaly sub zero overclock is set but its worth trying.

[aoch88]

We should start a poll to see how many people has this degration problem and define what "degration" is really about. I think that would probably help.