E8400/8500 degradation myth possibly busted?

[synergy]

I haven't followed the details of this thread enough to know what kinds of spikes are happening when, but if Vcore spiking is being suspected I'd highly suggest you look at the following link and the linked page and the couple pages thereafter:

http://www.anandtech.com/cpuchipsets...spx?i=3184&p=5

Specifically it's about how the tendency to minimize Vdroop by various options in the BIOS and overcompensating for Vdroop by boosting Vcore directly leads to predictable overvoltage spikes to the CPU whenever the load of the CPU changes momentarily from heavy load to something much less.

They suggest that the "Vdroop" is actually proper and necessary and beneficial in many cases, and it is trying to disable / compensate for it that is a problem in many cases.

Now of course some motherboards / PSUs may just have poor regulation or load handling ability, but those issues aside, even the best PSU / Motherboard has to have some momentary variability of output voltage versus load change since the response of the power unit can't be as instantaneous as the load change of the CPU. Given that, the MB/PSU should put out a voltage that is within specification, but not so close to the limit that the predictable spike or brownout when the load decreases or increases heavily causes the voltage to exceed the lower or upper limit.

[synergy]

Best brand / model? That's subjective. I'd say you'd be in good shape with a solid 550W unit with PFC and 80% efficiency though.

I'd look for an Antec Earthwatts 550 on sale just because they're inexpensive when on sale at places like Fry's and they've been independently reviewed / tested to indicate decent performance / quality.

I'm sure you could go with a Corsair or any number of other reputable models too.

So what you guys figure is the best PSU?

I dont need any megawatts....

Single 8800GT, X-fi and 2 HDD + optical.

[RunawayPrisoner]

Any decent power supply will do. Actually, I think any will do just fine. It's just that... you should not try to eliminate vdroop or vdrop on your boards, lest you want to end up with a degrading chip. These 45nm processors require much less voltage to operate nicely at high clock speeds, anyway.

[kiwi]

I can confirm now again that whoever wrote about cpus recovering slowly from degradation has a point. I am still testing one of my cpus that seemed to require more volts after DI/LN2 benching but now after a few weeks it is back to normal. 1.33V 4.1ghz prime blend 12 hours I've tested this on different setups ddr2 and dd3.

The thing is that high cpu pll kills/degrades 45nm chips rather severely but they can recover from just high voltage while benching

Why would RAM play a role in CPU degradation?

Because it might be RAM causing errors and not cpu Some micron D9 chips are known to degrade over time even at manufacturers rated voltages

[RunawayPrisoner]

PLL wasn't intended to be so high. In fact, it might be ideal to have low PLL for... maximum stability. Trust me, my PLL is only at 1.58v for a quad 45nm... and stock 1.50v for a dual 45nm. The only thing that's dying here is my NB chip, which is running at 1.55v versus 1.25v stock.

[Infa]

Kiwi, i have the samep problem, first of all , this is my third board i put this cpu in, and in the first...it was 2h+ stable at 1.45 4,3 , And later on, i did put it in a Gigabyte board... now it suddenly required little more..from nothing to more vcore just like that.. needed then 1.47 approx..same clocks...

And now in my brand new premium board... it needs like 1.50 , dont get this Either i have to swallow the apple and accept that it HAS degraded..

i have benched in several times with hiigh vcore, not for days though, some hours, from 1.60-1.75 for a while.

maybe ill just keep it now at 1.45 and it will get back to normal?

[dnottis]

I can confirm now again that whoever wrote about cpus recovering slowly from degradation has a point. I am still testing one of my cpus that seemed to require more volts after DI/LN2 benching but now after a few weeks it is back to normal. 1.33V 4.1ghz prime blend 12 hours I've tested this on different setups ddr2 and dd3.

The thing is that high cpu pll kills/degrades 45nm chips rather severely but they can recover from just high voltage while benching


Because it might be RAM causing errors and not cpu Some micron D9 chips are known to degrade over time even at manufacturers rated voltages

I'm finding the same thing. Took my Xeon to 1.45v, then it got flaky for a bit. Same thing with my E8400. After a few days things seem to be ok again.

[RunawayPrisoner]

So now... the big question is... have they REALLY degraded, or they were just unstable for a while?

[mrcape]

Mine definitely can't do what they did in the first weeks, before taking them to high fsb/vcore. Still, it seems they have "broken in" and haven't slid any further in a month. I can run all different stress test for hours even 10kfft for hours at 1.33vcore/4.25ghz. On the way up the first time I could do that with 1.28vcore.

So, no signs of recovery for my chips, but no further degradation either.

[RunawayPrisoner]

Or maybe your chip was not stable with 1.28vcore to begin with.

[mrcape]

We can maybe all day long. I don't want to have an argument about the definition of stable.

The chip passed stress tests at certain settings for blocks of hours. Now it doesn't.

[RunawayPrisoner]

Yeah, I've had that with my chip when the motherboard was set to auto-adjust voltage settings. I could only eliminate it when I manually adjust the voltage values to fixed values. I'm not saying that you're lying or anything. I'm just saying that... maybe it's part of your motherboard's fault as well. If your chip doesn't stay stable with some certain voltage settings that you set before, then maybe it's the chip...
But hey, that could lead to something actually. Were (or are) your motherboard settings at Auto?

[mrcape]

I've never had any voltages at auto and I've already discussed this in this thread. You're kinda preaching to the choir here : ). I've troubleshot this pretty carefully with new RAM etc.

[RunawayPrisoner]

It's not that I want to discredit you. Please don't misunderstand. And okay, read up on your posts. Seems like you did press your chip really hard at high voltage.
Just out of curiosity, did you try to stress your CPU at voltage higher than 1.36v?

[mrcape]

No worries man, I just didn't want to detail it all. I had it up to around 1.42-1.44. Not that high really you would think, but that got me to 4.7ghz and sub 10sec superpi.

I'm thinking that the spikes could have done it. I'd like to think that those were software errors, but I think the spikes are reallu causing slight damage.

[RunawayPrisoner]

I might just try to see what would happen with my E2140 tomorrow... when I mod my mainboard again with the pencil method. If it happens, then I can say... that's probably the cause of voltage spikes.

As for how the spikes damage hardwares, I haven't seen much on my X3350 so I can't say much. But it seems to happen.
Oh yeah, and different voltage settings do make differences in the graphs. From my own experiences, linear or pattern graphs tend to be more stable than non-linear graphs.

[mrcape]

It would make sense for a standard methodology to test these chips. That would require a set of testing from the ground up for each brand new chip. I know it's idealistic, because it would also require pushing them to high volts to get the degradation.

Out of curiosity, what would people suggest as a method to confirm degradation? What is acceptable proof? I'm assuming there's a range of opinion here.

[RunawayPrisoner]

I guess... if your chip suddenly BSOD's continuously at a frequency at which it worked stably 24/7 prior?

[MosIncredible]

Im worried about this spike I got at the near end....The rest of my voltages are 100% linear cept for 2 or 3 dips but 5 volts and 3 is dippy

However if its of any interest, my 5VOLT graph has a spike in the exact same spot as the Vcore spike.

Wow, that spike is crazy. OCCT doesn't reveal any real spikes with my voltage. Starts @ 1.31v drops to 1.28-1.29 through the whole test and back up to 1.31.

[mrcape]

I guess... if your chip suddenly BSOD's continuously at a frequency at which it worked stably 24/7 prior?

That would indicate instability for sure, but I'm talking about a comprehensive test method that could identify degradation.

For example, if someone bought a 45nm chip and took it to 1.5v without first recording how high it could go at low voltages, there's no way they'd even know it degraded.

[RunawayPrisoner]

That's because you have safe vdroop and vdrop. On boards with vdroop and vdrop mods, it's easier to stabilize the CPU with low voltage, but it might be easier for spikes to appear as well. Double-edged blade, I guess...
But I'm going to test the vdroop/vdrop mod = spikes theory tomorrow.

[mrcape]

I think we're talking about testing different things, but I'm interested in seeing your results.

I've done the pencil mods on asus boards but no need on my DFI as I can disable in the bios and get almost exact volts steady.

Let us know if you see the spikes. It would be nice to have a stack of fresh chips to test.

[KTE]

There's no myth just a lapse of understanding by users hence they assume too much because they can't explain what actually is happening.
We just need to work together to figure it out and discuss idea's and possibilities without biting each other, since none of us are Intel employees here and we all want to know. It's not aomething only limited to 45nm Wolfdales.

Ok now at some point ( Next day/Next week) you'll run that torture test again and then OMG!!!
WTF??? Prime95 FATAL ERROR!! But it used to be prime stable for hours and now it's failing in
less than 5-10 minutes? Instantly that horrifying thought comes to mind "Degradation".

This happened with my 65nm Phenom and Q6600, Phenom more than Q6600. Around 8 other users I know have had this with Phenom and I did with 6x B2 Phenoms, 1x E6750, 1x E6850 and 2x Q6600 B3/G0. I've not pushed my E8400 yet.

And here is what I found:

If my system is either off, or has been at Idle for a long period of time (i.e 18 - 24 hours) and I run prime
(especially the 10k FFT's) It will fail in < 5 minutes sometimes < 1 min. If I then run prime again it will take
prime a little longer to fail but maybe still < 5 mins. If I run It again it will take a lot longer to fail > 20 mins.
And then I run it once more Bam!!! Smooth sailing no more failing, It's back to the stable state it was
initially. 10k priming for hours on end.

With my Q6600, if I leave it off for say more than 3 days, it doesn't boot any MHz at any volts above or below stock. That's until I bootup stock, use it for about an hour and then reboot to change settings. Next shot I try 4G is a piece of cake at 1.38V boot, stock HSF but stability is still not there until I prime it bits on and off. Then if I reboot and set my previous stable settings of 450x8 1.36V, it'll test fully Linpack 32b/TAT/P95 stable. I then get no issues leter.

But yep, it did degrade from 3.85G 1.36V stable to 3.6G 1.36V stable max, not in a night, but after 2 weeks.
Overnight stability loss is quite impossible unless your real Vdd pin volts were at or above ~1.36V 45nm. It would usually mean your stability test never covered the chip area's which were unstable the first time round, Prime 95 is not the be it all end. If it's severely unstable later even in a few hours after perfect stability i.e. bench SPi at X MHz Y Volts, then you can tell quite clearly if degradation took place, no need to go further. This is why max MHz you may never get after once on any CPU.

[alpha0ne]

Are these Vspikes only happening to Anus boards ????

[CrazyNutz]

Are these Vspikes only happening to Anus boards ????

Anus?