Help at "bench-OC" on 920 D0

[cadupradoo]

Hi guys.. I had a C0 till two weeks ago and decided to go for the D0.. With the C0 I could bench (3dmarks) at 4.25GHz with HT On or 4.35GHz withou HT..

With my D0 I'm not being able to get good results.. Better results than with the C0, but not as good as I've been seeing on the internet.. I know that each chip is different, but maybe with some of your ideas I can get a better result.. Since I'm running a tri-crossfire setting, any MHz I can get more on the cpu is very welcome. So far I'm at 12th on the RV770's, but if I could run the cpu faster, I could probably get on top ten.

Anyway, I can bench with 210x21 @ 1.48V with vdroop. My temps get HOT, like 88C maximum on the cores.. but I'm already buying a WC..

Here's the settings:

Vcore: 1.48V with vdroop
Vtt : +300mV ( doesn't help setting it to +400mV = 1.5V)
CPU PLL: 1.950V
Vdimm: 1.7V
Qpi PLL and IOH Vcore: 1.475V
IOH/ICH/ i/O : 1.65V
ICH Vcore: 1.2V
Pci-E : 103MHz (104Mhz doesn't help either)

I think I don't have a good D0 chip anyway.. I can get 100% prime95 stable 4.0 with 1.25V but can't do 4.1 with less than 1.3V.. that's too much voltage for a D0.. maybe mine is just under average...

Oh, I forgot to say.. if I try 211x21 I don't get a BSOD or something, the screen just looks like a detuned tv.. Keeps going crazy on horizontal lines on a vertical movement..

Well, appreciate any help!
Thanks a lot!!

Sorry about the english skills btw..

[cadupradoo]

Any ideas?

[YMAA]

There doesn't seem to be a voltage limit on Nehalem - no degradation due to massive amounts of voltage like we were used to seeing with Yorkfield/Wolfdale and to a lesser extend Kentsfield/Conroe...at least there's no evidence of that yet. So the definition of "too much voltage" is really a misnomer. You just have to keep the chips cool.

So pour on the volts if temps aren't a problem - you're benching anyway.

And a lot of people posting high clocks with low voltages are probably full of it...yes, even the ones with D0s. 1.3v for 4.1GHz with HT is pretty decent from the legitimate posts in the "8hrs stability test" thread. So don't worry - your chip is fine. What does that extra 100MHz buy you, anyway? 4GHz i7 is plenty fast.

[Peen]

Check out my thread Worst W3520 or whatever I called it. Mines even worse and its a Xeon. Main problem I have is it runs insanely hot! Much hotter then your chip

[ReverendMaynard]

YMAA, if there ever was a misinformed post on XS, you just took the cake. Not one thing that you've said is true and you highlight your lack of knowledge. You obviously have never overclocked an i7, so why bother giving advise? 775 clocking does not apply...at all.

[Splave]

YMAA, if there ever was a misinformed post on XS, you just took the cake. Not one thing that you've said is true and you highlight your lack of knowledge. You obviously have never overclocked an i7, so why bother giving advise? 775 clocking does not apply...at all.
qft new ballgame completely, if you are stable @ 1.25vcore 4ghz then it takes 1.30vcore for 4.1ghz I would look into my other voltages needing to be tweaked. :yepp:

[ReverendMaynard]

qft new ballgame completely, if you are stable @ 1.25vcore 4ghz then it takes 1.30vcore for 4.1ghz I would look into my other voltages needing to be tweaked. :yepp:

Word. Time to take things back to square one.

the video corruption (detuned TV) the OP described is troubling. Wonder if the board is shot?

[Peen]

YMAA, if there ever was a misinformed post on XS, you just took the cake. Not one thing that you've said is true and you highlight your lack of knowledge. You obviously have never overclocked an i7, so why bother giving advise? 775 clocking does not apply...at all.

Who are you talking to?

[ReverendMaynard]

Who are you talking to?

not you tough guy lol.

Read the post, it starts with the persons handle. ;)

[YMAA]

Alright, if I'm wrong what's the correct information? I'd wager that'd be infinitely more useful to the OP than calling me out as a novice to i7 (which I'll admit I am...parts are arriving today and next week).

I haven't seen too much testing done on voltage scaling with frequency on i7 platforms, but a trait common to most kinds of chips (not just Core 2) is that they reach a point where the voltage required for a given clock starts to scale exponentially rather than linearly, and the point varies from chip to chip. I assumed this was also the case for i7. Also, most other sources of information I've seen have stated that there does not seem to be a maximum safe vCore before chip degradation occurrs. This could be guesswork however, the chips haven't been out long enough to consider one theory or the other as Gospel truth.

If I'm wrong on these points please correct me without getting hostile - there's no reason to say I'm wrong and not explain why. You're not helping the OP...shouldn't that be the goal?

[Splave]

with i7, you cant just crank 1.6vcore like you could on say an e8600. I have lowered my vcore before and its added to stability. After a certain voltage with i7 adding more does nothing. Its more about a harmony of your 3 or 4 voltages that need to be changed. Vcore, vtt, vdimm and in some cases the ioh vcore.

OP other then the voltages I mentioned you dont need to touch the other ones. put them back at stock.

[ReverendMaynard]

I apologize, but anyone who tells someone to blast the volts to a setup that they have no experience with, it pisses me off. Voltage/freq scaling with the D0's are erratic at best. 1.2v might get you 4.2ghz stable, then 1.3v might make the same clocks unstable where as with the later 775's, you could aim high with the volts as long as you cooled it well they'd eat it up and spit it out. The OP is already putting way too much volts into the entire system, so how are you helping him by telling him to pump more into it?


This is also a blind statement.

And a lot of people posting high clocks with low voltages are probably full of it...yes, even the ones with D0s.

Here is the absolute first boot with this system after loading the OS. While these tests definitely do not make it certified stable, which is subject to opinion anyway, it is impossible to deny that some of these chips are just pure awesome.

http://clubnboc.com/forum/index.php?action=dlattach;topic=876.0;attach=3514; image

Those voltages are real, and it's far from BS.

Now lets get back on track with the OP and work to help him achieve his goals from experience, not theory.

[cadupradoo]

There doesn't seem to be a voltage limit on Nehalem - no degradation due to massive amounts of voltage like we were used to seeing with Yorkfield/Wolfdale and to a lesser extend Kentsfield/Conroe...at least there's no evidence of that yet. So the definition of "too much voltage" is really a misnomer. You just have to keep the chips cool.

So pour on the volts if temps aren't a problem - you're benching anyway.

And a lot of people posting high clocks with low voltages are probably full of it...yes, even the ones with D0s. 1.3v for 4.1GHz with HT is pretty decent from the legitimate posts in the "8hrs stability test" thread. So don't worry - your chip is fine. What does that extra 100MHz buy you, anyway? 4GHz i7 is plenty fast.

Unfortunately, the chip isn't cool to add more volttage... I hope to get better results with a WC, but I have to wait for performance-pcs to get the Heat Killer block on stock again. Thing is that with the vgas I have, I get a 2k difference on 3d06 score from 4.2 to 4.4GHz.. It's a lot of bottleneck..
Thank you.


YMAA, if there ever was a misinformed post on XS, you just took the cake. Not one thing that you've said is true and you highlight your lack of knowledge. You obviously have never overclocked an i7, so why bother giving advise? 775 clocking does not apply...at all.

Well, the first I read your post, I thought YMAA meant something and you were talking to me..

Anyway, I could notice that adding more volttage than what's needed isn't a very good think to do on i7's, since the C0 stepping I used to have. But if I set the Vcore a couple of steps lower, I can't bench with the given oc. Anyway, I'll try setting some voltages a little lower, maybe VTT. Thing is, on the evga forum, everybody says that the 0x000124 BSOD is due to lack of VTT, while the 0x000101 BSOD is due to lack of vcore. I don't think that's 100% correct, since when I set the vcore a little lower i get a 0x0000124 BSOD instead of a 101 one.

Thank you.


qft new ballgame completely, if you are stable @ 1.25vcore 4ghz then it takes 1.30vcore for 4.1ghz I would look into my other voltages needing to be tweaked. :yepp:

I tried to do something like that, but it hasn't been easy.. And actually, I don't care a lot about 100% stable clocks, only "bench clocks". I tried to get 4.2GHz 100% stable just in order to have some idea on how far can I get 3d06 stable...

Thanks.


Word. Time to take things back to square one.

the video corruption (detuned TV) the OP described is troubling. Wonder if the board is shot?

I don't think so. This problem doesn't happen when I'm stock or even 4.1GHz prime95 stable... I think it's funny because it's kind of an awkward error to get when a OC isn't stable..

Thank you.


Alright, if I'm wrong what's the correct information? I'd wager that'd be infinitely more useful to the OP than calling me out as a novice to i7 (which I'll admit I am...parts are arriving today and next week).

I haven't seen too much testing done on voltage scaling with frequency on i7 platforms, but a trait common to most kinds of chips (not just Core 2) is that they reach a point where the voltage required for a given clock starts to scale exponentially rather than linearly, and the point varies from chip to chip. I assumed this was also the case for i7. Also, most other sources of information I've seen have stated that there does not seem to be a maximum safe vCore before chip degradation occurrs. This could be guesswork however, the chips haven't been out long enough to consider one theory or the other as Gospel truth.

If I'm wrong on these points please correct me without getting hostile - there's no reason to say I'm wrong and not explain why. You're not helping the OP...shouldn't that be the goal?

Thanks for the concern on helping me. And the exponentially increase needed for a chip when it reachs the point you talked about still happens, I just don't know if that's the best way to do the OC..

Thanks.


with i7, you cant just crank 1.6vcore like you could on say an e8600. I have lowered my vcore before and its added to stability. After a certain voltage with i7 adding more does nothing. Its more about a harmony of your 3 or 4 voltages that need to be changed. Vcore, vtt, vdimm and in some cases the ioh vcore.

OP other then the voltages I mentioned you dont need to touch the other ones. put them back at stock.

I think maybe PCI-E have something to do with it too.. And I've heard a little about qpi pll and ioh vcore (raising both equally). Can that help?

I'll put the other voltages on stock back again.

Thanks.


I apologize, but anyone who tells someone to blast the volts to a setup that they have no experience with, it pisses me off. Voltage/freq scaling with the D0's are erratic at best. 1.2v might get you 4.2ghz stable, then 1.3v might make the same clocks unstable where as with the later 775's, you could aim high with the volts as long as you cooled it well they'd eat it up and spit it out. The OP is already putting way too much volts into the entire system, so how are you helping him by telling him to pump more into it?

Here is the absolute first boot with this system after loading the OS. While these tests definitely do not make it certified stable, which is subject to opinion anyway, it is impossible to deny that some of these chips are just pure awesome.

http://clubnboc.com/forum/index.php?action=dlattach;topic=876.0;attach=3514; image

Those voltages are real, and it's far from BS.

Now lets get back on track with the OP and work to help him achieve his goals from experience, not theory.

Thanks again for the info and help. I used to bench with 1.62V on the C0 stepping, but it ran WAY cooler... that's what it used to take to bench at 4.25GHz with HT or 4.35GHz w/o HT.

Thanks to you all, will try do some more tweaking as some of you have adviced...

[ReverendMaynard]

right on cad. Keep at it and stay positive. It's usually the easily overlooked stuff that throws a wrench into it.

If you could try some things, and share the results we might be able to point something out. I also ran the C0, and got the most I could out of it with my cooling and so far I've taken the same approach with the D0 but adjusted for lower volts.

Try doing all of this with a single gpu card. Just adjust the vtt and vcore but either keep them 1:1 with each other or keep the vtt a touch lower to start. Also, try with a 17x uncore. It seems to be the sweet spot at least with the eVGA board that I have.

[Splave]

no need to play with the pci-e until you want over 220 bclk, could your ram be failing? what are the settings for it

[ReverendMaynard]

no need to play with the pci-e until you want over 220 bclk, could your ram be failing? what are the settings for it

QFT, some cards, no matter what model just can't handle PCIE freq.

Simplify!

[cadupradoo]

right on cad. Keep at it and stay positive. It's usually the easily overlooked stuff that throws a wrench into it.

If you could try some things, and share the results we might be able to point something out. I also ran the C0, and got the most I could out of it with my cooling and so far I've taken the same approach with the D0 but adjusted for lower volts.

Try doing all of this with a single gpu card. Just adjust the vtt and vcore but either keep them 1:1 with each other or keep the vtt a touch lower to start. Also, try with a 17x uncore. It seems to be the sweet spot at least with the eVGA board that I have.

I'll test a lot more, i don't usually give up :D

Anyway, I'm very busy this weekend, some college stuff, but as soon as I have a little free time, will do it.

Do you really have a sweet spot with 17x uncore? I've always tried with 16x and auto settings cause 99% of the comments i've seen says that 16x is required for stability. Will do that too, thanks again!


no need to play with the pci-e until you want over 220 bclk, could your ram be failing? what are the settings for it

I've tried running them with 10-10-10-30 at a 8x multiplier with something around 1.65V~1.7V and around 1660MHz.. but for benchs, i set the timings as 8-8-8-24 and get the same problems when trying to raise the OC. I might run memtest overnight tonight. Is there any difference between the windows and dos version (performance wise?).

Thank you :D

QFT, some cards, no matter what model just can't handle PCIE freq.

Simplify!

I've been reading a lot, and it seems that you need to set pci-e at 110~115MHz after 221 bclk... and I'm pretty far from it, i was just wondering..


Thanks to all of you. Will do some more tests tonight (here in Rio, it's 3pm right now) and will be back with some results..

[ReverendMaynard]

man I'd be on the beach right now, testing in the early morning lol.

Good luck and keep us updated on what you've tried and what has worked/not worked.

[cadupradoo]

man I'd be on the beach right now, testing in the early morning lol.

Good luck and keep us updated on what you've tried and what has worked/not worked.

As incredible as it sounds, it's raining a lot today here :D

I'll be back with results, thanks a lot :up:

[Peen]

I thought YMMA was a new code word for something lol...

Anyways, whats your uncore? Mine runs really hot too but I dropped the uncore temps went down a bit. Going to play with this more, only 2nd day with i7

[cadupradoo]

I thought YMMA was a new code word for something lol...

Anyways, whats your uncore? Mine runs really hot too but I dropped the uncore temps went down a bit. Going to play with this more, only 2nd day with i7

Which uncore? Frequency or voltage?

The frequency is at 16x bclk.. the least I can get, since it's twice RAM clock..

The voltage is on the first post...

I've been even more busy than I though I'd be.. But for sure will do the tests suggested above tomorrow.. (sunday)

[Peen]

Well I'm new to i7 so discovering some things that help me OC the CPU more and run lower temps. I was running 4ghz 1.29ish volts now (95c!) now down to 1.248v and (69c)

I found running less mem frequency with tighter timings, lower uncore along with low QPI voltage (1.16-1.2v) and DRAM (1.5v) helps it run cooler and OC more, without any real loss in speed.

Try it out see what happens

[cadupradoo]

Ok, after some tests I got exactly the same results... Here's the changes I've done:

PCI-E 100MHz, All voltages but vcore, vtt, dram, ioh vcore and qpi pll back to stock. Uncore x17.

I'm starting all over again, from 4.1.

Actually, I forgot to say... After setting some of the voltages at stock, I could run at 211x21, but 3dmark06 always stops responding on cpu test...