E8400/8500 degradation myth possibly busted?

[MosIncredible]

Yeah, I was disappointed in the performance of my chip compared to others but more than satisfied because of the jump I made (P4 2.0GHz). I also only paid $200. As far as the actual thread subject, I had a similar experience of a warm-up period. I ran the same settings 4 different times and it was more stable each time. From 3 minutes to 30, to 50, to 2 hrs.

[CrazyNutz]

Yeah, I was disappointed in the performance of my chip compared to others but more than satisfied because of the jump I made (P4 2.0GHz). I also only paid $200. As far as the actual thread subject, I had a similar experience of a warm-up period. I ran the same settings 4 different times and it was more stable each time. From 3 minutes to 30, to 50, to 2 hrs.

Yeah, I know what you mean. However I am happy with it, the low cost and the performance of the wolfdales makes it a damn sweet deal

As for the warm-up period, it's is definitely strange. Also my stability changes
day to day, however I must say it does not effect any of my real world applications, (i.e. games, video encoding, programming/compiling, etc..)
just the torture test are affected, so I'm not concerned as much.

[Technophobe]

What a huge thread! I have quickly flicked through so sorry if this has been raised before (I just didn't see it). I wanted to go back to basics as to what my recommended max CPU voltage should be (as voltage seems to be the main cause of degradation).

The consensus seems to be 1.3625v. But after reading the datasheet (http://www.intel.com/design/processo...hts/318732.htm) it seems more likely that this wasn't my CPU's individual max at all but rather my Vid was my max (in my case 1.1125v). If you have a look at the datasheet at Table 5 (pg 19) and Figure 1 (pg20), it certainly indicates this.

I personally can't get to 4GHz without going over the magical 1.3625 mark. I'm on 3.8GHz @ 1.344v at my CPU (idle), but if I'm reading the datasheet correctly, I'm still 0.2351v (20%) over recommended. I don't think I would get any overclocking if I used 1.1125v. Interested in your comments.

[RunawayPrisoner]

Here's something interesting that I read:

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

According to that page and the page after it, without vdroop and voffset (or vdrop), something unpleasant may happen, and we may "accidentally" put too much voltage into our cpus when it's not really necessary. Maybe that's why we've heard a lot of degradation stories from people although they were well below 1.36v specified by Intel. If the article proves true, then even at 1.300v without vdroop or vdrop, we are still NOT safe. Mmm... this has just made me feel relieved that I have my Yorkfield in when there are vdroop and vdrop on my motherboard.

[CrazyNutz]

Here's something interesting that I read:

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

According to that page and the page after it, without vdroop and voffset (or vdrop), something unpleasant may happen, and we may "accidentally" put too much voltage into our cpus when it's not really necessary. Maybe that's why we've heard a lot of degradation stories from people although they were well below 1.36v specified by Intel. If the article proves true, then even at 1.300v without vdroop or vdrop, we are still NOT safe. Mmm... this has just made me feel relieved that I have my Yorkfield in when there are vdroop and vdrop on my motherboard.

Finally something is making since. Very few people take these transient voltage spikes into account. And these spikes will get exponentially higher
with not only increased vcore but also increased frequency since the power/
current requirements grow exponentially with higher frequency, and high
processor load. So when were priming @ high frequency (i.e. 4ghz) the
cpu is sucking some serious current, and the vreg's have to open the floodgates to keep up with the demand, and as soon as you stop priming
the cpu current demand instantly drops while the vregs have not yet
sensed the lower demand, so the flood gates are still open for a period of time, which equates to a substantial voltage spike, and depending on our
vcore & vdroop this spike could be well beyond our so-called safe range.


What we need are tourture tests that bring the CPU load down at a slower
controlled rate like from 100% to 10% load over 3 seconds (even when an error is detected) this will give the vreg's time to sense, and compensate for
the current demand drop.

[NJDevilsFan21]

If voltage spikes are the cause then why do 65nm chips seem to not suffer from it too much? People put 1.5V+ into 65nm chips so the voltage spikes should be even higher.

You people are all raising the voltage to 1.4V from 1.1V, which is 27% increase of voltage. On 65nm, take an average VID of 1.275V, and 27% increase in voltage translates to 1.62V!

[mrcape]

Both interesting points.

Voltage spikes are scary with this more fragile 45nm architecture.

Since the beginning of this the distance from vid to Max voltage, or headroom, has concerned me. My better e8400 has low 1.0875 vid. If I compare the size of the headroom I used on 65nm and apply similar percentage to the 45nm the max vcore is pretty damn low, like 1.3vcore max.

[RunawayPrisoner]

If voltage spikes are the cause then why do 65nm chips seem to not suffer from it too much? People put 1.5V+ into 65nm chips so the voltage spikes should be even higher.

You people are all raising the voltage to 1.4V from 1.1V, which is 27% increase of voltage. On 65nm, take an average VID of 1.275V, and 27% increase in voltage translates to 1.62V!

Imagine yourself trying to blast your way through that tiny mouse hole, and yourself trying to blast your way through your front door. Which would cause less damage? And the article was analyzing the structure of 45nm processors only (namely QX9650) thus the same thing may not happen to our 65nm friends. The article did say that this is a new direction in overclocking, and maybe that means... it's time we play by as many rules that were laid down by the manufacturer as possible... which includes intentional vdroop and vdrop.

[dnottis]

Straight from Tankguys.com today. My E8400 did 4.1 @ 1.41v orthos, check out this little lady. Not even lapped, just jumped up slight volts and the bus and away we go!! BTW Orthos is over a half hour now.

This is water cooled, specs in sig. Xeon I think I love you....

[mrcape]

Those 805 are supposed to be good.

Just curious, do you ever try 8.5x500 or 8.5x500? I bet you can get stable with lower vcore and manual gtl settings.

Oh dang, never mind, I thought you meant 1.4 on the new chip.

Still, I have better luck and stability at lower vcore with 8.5x than with 9x. Lot's of differences between our rigs I'm sure.

[Eldonko]

Know what = degradation? This:



That was a P5K I owned. Scary huh?

Chip is fine, was 65nm. :p

[dnottis]

Those 805 are supposed to be good.

Just curious, do you ever try 8.5x500 or 8.5x500? I bet you can get stable with lower vcore and manual gtl settings.

Oh dang, never mind, I thought you meant 1.4 on the new chip.

Still, I have better luck and stability at lower vcore with 8.5x than with 9x. Lot's of differences between our rigs I'm sure.

I just got it today, I'm at work now, it's home priming lol.

Didn't play with the multiplier at all, VTT and PLL are stock NB, SB too all stock volts manually configged in the BIOS. I just popped it to 1.33v and started orthos havent even played with vcore yet.

I couldnt find anything on Q805 before I bought this, which is why I was on the fence for about a week. So far I'm very happy with it.

[CrazyNutz]

Know what = degradation? This:



That was a P5K I owned. Scary huh?

Chip is fine, was 65nm. :p

Holy Baby Jesus!!! That looks like a 45nm killer for sure!!!! I need to run this
on my p5k-e, I hope it's not that bad. If it is, I getting on the first flight to
taiwan to personally kick Asus's ass.

[mrcape]

At high clocks I've seen spikes in Smart Guardian, but I thought it was just that program being unstable.

Scary!

[RunawayPrisoner]

Use OCCT to determine if your current OC has a spike. Sincerely, if vCore is stable, it should go:



^= Very uniformly.

Same goes for the rest:







When an OC goes wrong, you'll see the voltage... drawing weird shapes all the time. When this happens, it's actually advisable to... reduce the amount of voltage you are cranking in instead of putting in more.
And OCCT IS the hardest of all stress tests to me thus far. It failed my 6h 10K prime (8h 8K prime) in... 30 seconds.

Edit: And no, your eyes don't deceive you. 1.12v is plenty for stable 3.2GHz...

[mrcape]

That graph is pretty cool i never looked at those in occt. Sometimes I have an easier time passing occt auto test than prime, depends on the oc settings.

[icecpu]

my E8400 is happenning exactly as CrazyNutz described on the first page "warm-up period"

first run failed, then run again it passed

this is anoying, I want to get rid of this cpu and get somesome else

Or anyone think it's a unsteady PLL voltage ?? Currently I set PLL to auto, maybe set to v1.5 will help ??

[CrazyNutz]

my E8400 is happenning exactly as CrazyNutz described on the first page "warm-up period"

first run failed, then run again it passed

this is anoying, I want to get rid of this cpu and get somesome else

Or anyone think it's a unsteady PLL voltage ?? Currently I set PLL to auto, maybe set to v1.5 will help ??

Good Luck because I tried tweaking everything, and anything to make it stop,
and the worse thing is, it finally got to the point to where the "warm-up period" did not even make it stable anymore. This cpu has never been over 1.36vcore (bios) so I'm accrediting this degradation to transient voltage
spikes, obviously made worse be enabling "Load-Line Calibration".
I have since turned off LLC, drop my vcore, and my cpu freq (about 100mhz)I'll now have to wait to see how it goes.

[icecpu]

Does Intel know about this ??
I never went beyond absolute max vcore recommend by Intel. These kind of crap cpu is not acceptable.

At least my cpu still stable , passes prime95 if I run it again

maybe call Intel request for an RMA

[CrazyNutz]

Does Intel know about this ??
I never went beyond absolute max vcore recommend by Intel. These kind of crap cpu is not acceptable.

At least my cpu still stable , passes prime95 if I run it again

maybe call Intel request for an RMA

It does not matter if you never went beyond some vcore, the voltage spikes
are capable of going way beyond that for a small period of time (i.e. a few uSeconds).

[sfdmalex]

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.

[RunawayPrisoner]

Oh my god... that's a close-to-1.6v spike! Holy smokie... if that doesn't fry or kill the processor, you oughta be the luckiest guy in the world. It's like you are stabbing the processor really hard whenever something like that happens!

[icecpu]

So what is the cause of voltage spikes ??

[cadaveca]

either power consumption the board or PSU can't keep up with, or software reading error.

[sfdmalex]

So what you guys figure is the best PSU?

I dont need any megawatts....

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