E8400/8500 degradation myth possibly busted?

[prime81]

Of all the things I mentioned, I firmly believe VTT did me in with that chip in particular.

Prime81, did you do a Vdroop mod or something to have your Vcore constant when both idle and under load in windows? It's very odd to see such especially with all the Vdrop values being reported - average of about 0.03V from supply voltage. You may have a gem of a motherboard there.

Well I'm running a Maximus Formula SE with the latest Rampage Formula BIOS. Loadline enabled creates this setting, the only catch with loadline is to turn off speedstep.

[mrcape]

I can add that air benching sessions with same e8400 going to >1.5vcore, 600fsb, 1.75pll etc have not caused any further signs of degradation. I am thoroughly convinced these chips are pretty tough. Next tests will be putting these under cold and really giving some juice.

[tmatrix]

Prime81, I had a feeling it was along the lines of Loadline calibration. I didn't want to assume since we have different motherboards. Still, even without that turned on, I wouldn't feel right about leaving speedstep on in the BIOS since I'm overclocking in the first place

I can add that air benching sessions with same e8400 going to >1.5vcore, 600fsb, 1.75pll etc have not caused any further signs of degradation. I am thoroughly convinced these chips are pretty tough. Next tests will be putting these under cold and really giving some juice.

Yowza Mrcape!!! I'm impressed. 600FSB?!!! What multiplier did you use, to get what speed? Cool! All this benching on air, no less. To support your findings, so to speak, I've been up all night feeding my "degraded" E8400 some high voltage: I managed to just post 4.5GHz with 1.8V on my WC kit. I'm figuring that was more than necessary. Anyway, I brought it back down to 1.472V CPUZ idle, running at 3.8GHz - it's stress voltage is 1.456. I was even able to run orthos blend stress tests for about 34 minutes before I stopped it; I was actually thinking 5 minutes would be enough considering how at 4GHz, no matter what fsb/multiplier combination I tried, not up to 10 seconds into the start of the test, my PC would hang and reboot.

That brings me to another observation I suspected before. With my new E8500, I was able to pull off 4GHz with ease, but I couldn't get anything higher than 4GHz stable. The E8500 would post and boot into windows, but it failed orthos small fft tests after a couple of seconds - it was a relief it didn't hang and reboot on me though as the E8400 was fond of doing. As I mentioned before, I managed to get the E8400 to get to 4GHz with about 1.6V, but it hangs and reboots the moment I start a stress test, be it small ffts or blend. However, now I'm using this E8400 chip stable as can be at 422x9, just as I was stable with my E8500 at 422x9.5. It would appear my motherboard has seen better days. It can't be the memory since I'm running it at the moment at 1125MHz, as opposed to its 1066 default...and I'm still using the extreme profile with a TRD of 7. I guess I can be glad I can use this motherboard at all. I would have seriously considered an RMA for this mobo, but I just had to go and remove that plate glued to the NB portion of the Crazycool heatsink assembly. I might look into the X48 offerings from Gigabyte later to test my theory of a faulty motherboard. For now, I'll leave the E8400 in here to avoid messing up my E8500 by accident.

[tiborrr]

Question related - yes, i believe in cpu degradation, but it depends on the operating temperature of the chip. E.g.: I never run more than 1.4-1.42V into a chip on aircooling, but under SS cold i feel safe to go to about 1.65-1.7V and under cascade up to 1.85-1.9V for extended benching period.

[mrcape]

What multiplier did you use, to get what speed?

Well, mobo has a lot to do with the high fsb. This DFI/bios is just made for it. About you getting stable over 4ghz, you should be able to do that with way less vcore. You gotta set gtls manually.

For example first try to get stable with manual gtls at 500x8 keeping ram ratio 1:1 to keep memory out of the equation. Try to set NB & fsb term relativeley low and use gtls between 58-62. For me a real sweet spot is 8x500, NB 1.4 with gtls 59/59/59. Toy around and you'll get it. Doing that, it should be stable and require less vcore too! So that buys you more headroom to try 8.5x500 with the same settings but maybe just a few core bumps. Over 4.3ghz and the wolfdales need for more juice scales up pretty quick.

[tmatrix]

Well, mobo has a lot to do with the high fsb. This DFI/bios is just made for it. About you getting stable over 4ghz, you should be able to do that with way less vcore. You gotta set gtls manually.

You're so right about the kind of mobo playing a big part the success of overclocking. Yeah! I know I should be able to pull of 4GHz with a lot less than what I'm working with at the moment; heck, I recall pulling off a stable 4.25GHz OC at 500x8.5 on that E8400 for about 2 weeks before it started struggling. Damn "high VTT" penryn bully!!! I have noticed though about getting stable at lower NB overvoltage settings. I'm using about 1.525V for the NB right now for that 3.8GHz OC and "normal" for the FSB term, which should be about 1.2V for this Gigabyte board - I wonder whether anyone knows how to bring down that VTT some more in the BIOS.

After getting myself accustomed to how the different manufacturing components this board uses applies my settings, I have come to terms with how that 1.525V NB value is the minimum I've been able to stick with, without it suffering some kinda temperature spike instability penalty when I start running games and the like that stress the PC to a good enough degree. In the middle of a game, it would get to a point where my PC would hang without warning and reboot; I tend to blame the NB (aka motherboard) temp for getting an anxiety attack of sorts when it's doing its thing at up to 54C.

Did you forget manual gtl settings for a Gigabyte user don't translate the same for us as for other users of mobos by other manufacturers. The only thing(s) I've got to work with is the single-digit percentage variations setting: Auto, +3%, normal, -3%, -6%, and -9%. Is there some kinda conversion formula I could use for the manual gtl settings you mentioned. Ah! "Brings back memories...!" 1.4V NB was what I used to use before I bumped into that penalty I referred to earlier. Now not being able to safely rely on 1.4V NB is part of why I'm suspecting my motherboard might be crippled now that I can't stabilize a 34% or more OC. It was pretty wierd when I started taking 43C core temp readings from Realtemp as idle at what should be a relatively easy 4GHz OC for an E8500, my new chip. It did shock me though that I could pull it off using just 1.264V CPU-Z idle. As much as I'd like to believe Newegg sent me a faulty chip, the motherboard is still in question. Oh well!

Nevertheless, I'll try to apply the settings you've suggested on that wacky E8400 I have in there now - save for the gtl ones - and report back here. I'll see how it goes and share what happens. Oh by the way, what speed did you pull off with 600MHz FSB on air?

[tmatrix]

So I tried what I could with my E8400 to reach 4GHz stable with up to 1.58V for 500x8, but still no cigar. It would stay on for a while and hang and reboot either after a long stall while there was disk activity - that was some bad latency there - or right after I started an orthos stress test. Just goes to show either my motherboard and/or that chip can't handle it. I do remember one thing that may be a clue to when and how this started: after adding the first spot cooler for the NB, my formerly stable 4GHz OC was useless; this was back when my E8400 wasn't so beaten up as it is now. I'm now wondering whether that fan messed something up on my motherboard. I'll do more troubleshooting to get to the bottom of this.

By the way, I'm still awaiting your reply about your 600MHz benchable OC on air.

[mrcape]

Hi, I was testing the max fsb capability of the board and chip. Just wanted to see how hi I could get it and still run 1m or validate with cpu-z. The multi, overall clock and temps were kept low, so I didn't need to beat on the chip too much during the process.

Didn't realize your bios had that type of gtl setting. Good luck and have fun!

[Leeghoofd]

Jeez Tmatrix and you wonder why your CPU's degraded ? 3.8GHz with over 1.47Volts... nothing can survice that punishment... buy another board now , seems ya giga is wasted...

[ScottFern]

Guys, I think I finally broke my e8400. I had my VTT at 1.37v for a long time and recently I have lost any OC ability. I am running my e8400 at 8X333=2.6Ghz because if I leave everything at default my PC will shutdown and I lose video during the Windows splash screen. Its quite bizarre. I am actually not sure if its my DFI BloodIron, or my e8400?

Also, what is the difference between VTT and vcore? On this mobo all I can find is VTT and CPU VID Special ADD?

[CrazyNutz]

Guys, I think I finally broke my e8400. I had my VTT at 1.37v for a long time and recently I have lost any OC ability. I am running my e8400 at 8X333=2.6Ghz because if I leave everything at default my PC will shutdown and I lose video during the Windows splash screen. Its quite bizarre. I am actually not sure if its my DFI BloodIron, or my e8400?

Also, what is the difference between VTT and vcore? On this mobo all I can find is VTT and CPU VID Special ADD?

VTT = FSB termination voltage. Default VTT for 45nm is 1.10v

[ScottFern]

Well, its safe to say my last E8400 was toast. I am not laying blame to anyone specifically, but there was one post in a thread over at overclocker.net giving advice on certain settings, and I noticed it stated to have the VTT=1.45v when Intel states the max voltage for VTT to be 1.10v. I have done some research and it looks like anything over 1.15v is going to degrade your 45nm Intel cpu.

DO NOT, I REPEAT DO NOT set VTT to anything over 1.15v or else you CPU will degrade over time very rapidly.

[Leeghoofd]

Well I can state that some CPU 's prefer a very low setting in FSB Term or CPU VTT and even CPU PLL... most peeps think that more voltage is required to get a CPU stable, well with 45nm it can be totally the other way around...

My E8400 was before at 1.6PLL and 1.26 VTT and needed 1.26 Vcore to be stable at 4ghz, ow I used 1.52 PLL and 1.2 CPU VTT and now it only needs 1.22Vcore to be prime stable... all these voltages seem to be connected some way or the other... I have had other wolfdales and yorks where I needed to pump way more volts to get stable...

Till now I didn't need to go over 1.35 CPU VTT to get any CPU stable at 500FSB and I only had my first Qx degrade in a few weeks due to the 780i with it's underpowered circuitry.... there's a burn in period like also found out in this thread.. but I take it slow and don't push the CPU's direct up to 4Ghz and beyond anymore... first a fw weeks at moderte clocks and then see if she can hold it...

Think about it MORE isn't ALWAYS better...

[RealTelstar]

Well I can state that some CPU 's prefer a very low setting in FSB Term or CPU VTT and even CPU PLL... most peeps think that more voltage is required to get a CPU stable, well with 45nm it can be totally the other way around...

My E8400 was before at 1.6PLL and 1.26 VTT and needed 1.26 Vcore to be stable at 4ghz, ow I used 1.52 PLL and 1.2 CPU VTT and now it only needs 1.22Vcore to be prime stable... all these voltages seem to be connected some way or the other... I have had other wolfdales and yorks where I needed to pump way more volts to get stable...

Till now I didn't need to go over 1.35 CPU VTT to get any CPU stable at 500FSB and I only had my first Qx degrade in a few weeks due to the 780i with it's underpowered circuitry.... there's a burn in period like also found out in this thread.. but I take it slow and don't push the CPU's direct up to 4Ghz and beyond anymore... first a fw weeks at moderte clocks and then see if she can hold it...

Think about it MORE isn't ALWAYS better...

Good post!

[CrazyNutz]

Well, its safe to say my last E8400 was toast. I am not laying blame to anyone specifically, but there was one post in a thread over at overclocker.net giving advice on certain settings, and I noticed it stated to have the VTT=1.45v when Intel states the max voltage for VTT to be 1.10v. I have done some research and it looks like anything over 1.15v is going to degrade your 45nm Intel cpu.

DO NOT, I REPEAT DO NOT set VTT to anything over 1.15v or else you CPU will degrade over time very rapidly.

Actually as stated in the intel docs VTT in spec. is: min=1.045v typ=1.10v max=1.155v
however the absolute maximum is 1.45v same as VCC (vcore). But yes
Intel also says even within the absolute max the CPU will degrade.

[CryptiK]

My e8400 has definitely degraded over a period of 4 weeks. I have never gone over 1.360v, and that was idle, 1.32v load was the most vcore it was ever given.

At first it was stable at 4GHz at 1.29v. It passed 8 hours of small & also large FFT orthos, and 30 passes of linpack.

After 2 weeks, I noticed it would only pass 1 out of 5 linpack tests. I had not changed any settings at all, it just went from stable to unstable. I tried altering my vdimm, underclocking my ram, increasing vNB, loosening the PL, nothing worked. Then I remembered I had heard about this degradation, so I upped the vcore from 1.29v load to 1.304v load, and set everything else the way it had been originally. My system was now rock stable.

Another week passed, and I again noticed that I could only pass 1/5 linpack tests. I again changed everything but the vcore, to no avail. I changed eveything to it's original settings, and begrudgingly bumped up the vcore to 1.32v load. Once again it became rock stable.

After 5 days, it again became unstable. Again after trying everything else, I set the original settings for all components, and increased the vcore to 1.33v load. Yes, once again it became rock stable.

I am absolutely certain my e8400 has degraded from when I purchased it 4 weeks ago. I have had to raise the vcore 0.04v (1.29v load up to 1.33v load) in three steps over a 4 week period to remain stable. I have never gone over intel's maximum of 1.3625v, in fact I never went over 1.360v. It has also been cooled with a T.R.U.E since day one, and has never gone over 52*C at any time.

I am irritated at this situation. CPU's degrading in 4 weeks when kept under their rated maximum voltage and kept cool, is very poor, and has never to my knowledge occurred in the past. I'm not sure about basis for this, but I would guess electromigration and physical degradation of the transistors/gates/silicon wafer in the core to be occurring - perhaps intel rushed it's die shrink, and this rapid degradation is the result of poor engineering?

[aztec]

Looks like I might be keeping my E6600 after all, just in case.

[ahmad]

My e8400 has definitely degraded over a period of 4 weeks. I have never gone over 1.360v, and that was idle, 1.32v load was the most vcore it was ever given.

At first it was stable at 4GHz at 1.29v. It passed 8 hours of small & also large FFT orthos, and 30 passes of linpack.

After 2 weeks, I noticed it would only pass 1 out of 5 linpack tests. I had not changed any settings at all, it just went from stable to unstable. I tried altering my vdimm, underclocking my ram, increasing vNB, loosening the PL, nothing worked. Then I remembered I had heard about this degradation, so I upped the vcore from 1.29v load to 1.304v load, and set everything else the way it had been originally. My system was now rock stable.

Another week passed, and I again noticed that I could only pass 1/5 linpack tests. I again changed everything but the vcore, to no avail. I changed eveything to it's original settings, and begrudgingly bumped up the vcore to 1.32v load. Once again it became rock stable.

After 5 days, it again became unstable. Again after trying everything else, I set the original settings for all components, and increased the vcore to 1.33v load. Yes, once again it became rock stable.

I am absolutely certain my e8400 has degraded from when I purchased it 4 weeks ago. I have had to raise the vcore 0.04v (1.29v load up to 1.33v load) in three steps over a 4 week period to remain stable. I have never gone over intel's maximum of 1.3625v, in fact I never went over 1.360v. It has also been cooled with a T.R.U.E since day one, and has never gone over 52*C at any time.

I am irritated at this situation. CPU's degrading in 4 weeks when kept under their rated maximum voltage and kept cool, is very poor, and has never to my knowledge occurred in the past. I'm not sure about basis for this, but I would guess electromigration and physical degradation of the transistors/gates/silicon wafer in the core to be occurring - perhaps intel rushed it's die shrink, and this rapid degradation is the result of poor engineering?

You have a P5Q. If you used "AUTO" for any of the voltages, values can and WILL change with every boot. Never use AUTO for any of the voltages if you want a guaranteed OC on the asus boards.

[CryptiK]

I set absolutely everything manually, and I have measured the voltages with a DMM and I am sure that none of them change. This is straight up CPU degradation, I'm certain of it.

[mskvorc1]

@CryptiK

I don't get it. How come your CPU has degrade and you didn't even pass 1.3625V, while Intel states that as max safe voltage on their site, under processor specific finder? I found that pretty strange, especially while looking random setups of people with higher voltage than you for 24/7 usage... Maybe you just got defective processor?

All of this is crazy to me. While Intel states 1.3625V, other say 1.4V is max for 24/7, here are also test's of max voltages and Intel processors running under unbelivable voltages, than CPU degradations under 1.3V...

To who to belive anymore?

Sorry for my bed english...

[ahmad]

I think I might sacrifice my e8400 just for this cause...

[003]

Yeah something else has to be going on. I keep my Q9450 at 1.425v 24/7 with LLC enabled, and it has been rock solid since day one. I have gone up to 1.55v for benching sessions as well.

[G4h4o8s6T]

I originally voted no to the poll, but it seems I jumped the gun, as my cpu has degraded twice.

First needed 1.36v for 4ghz, after 4 months needed 1.39v, and now after another 3-4months need 1.41v to be prime stable.

Just figured Id share my "results", and hope the E0s dont degrade like these chips.

[aztec]

I'm still stable with 1.36v @ 4GHz. Been OK for several months now.

[radaja]

@candymancan, ive been seeing the same thing.friday igot my best clock to date (ran prime95 for three hours stable)wake up saturday ran prime95 fails in less than 10 minutes. wth didnt change anything in bois whatsoever. so tried some test. one run pass/one run fail.so now im at a loss as to whether im really ever stable