View Full Version : Is my 920 a Good Chip ?
I havae a 920 with the Asus Premium board. Right now I am running 4.2 @ 1.5V Prime 95 stable. Is that a lot of volts for that kind of clock. Any less on the voltage is not stable. I can run up to 4.5 but at a whopping 1.58V and it is not Prime stable. My temps are astronomical, even on water....42 idle and 62 load.
Does this sound like a good chip or should I sell it and buy another one ?
nikhsub1
01-24-06, 05:30 PM
Seems just below average... what water setup do you have? Temps seem very high. Have you reseated the WB?
I have the Swiftech H20-120 Premium kit. I have resated the waterblock and it is making good contact, because I can feel the heat.
It seems like I have to give my chip so much voltage to get stable at a certain speed.
M Diddy
01-24-06, 09:30 PM
I noticed the same with mine...
I'm rock soild at 4.0 and 1.41 VCore... If I go to 290 FSB, I would fail Prime even with 1.45... I think there might be something else causing this issue...
I also wonder if it is my 955 Premium board. I had the same thing happen with my 820. I wonder if two things aren't off with my P5WD2 Premium.....the temp sensor and the voltage regulator. I seem to get more V droop than normal and my temps readings are pretty high.
Or maybe some of us have a mediocre batch of chips ?
decent chip :)
I know yours is but I am not so sure about mine ! :)
FWIW, I'm getting 286FSB / 4Ghz at 1.425v, which after dual priming for 8 hours or so produces 58c load temps (large volume water system - takes a while to stabilize). 4.2 is probably possible, but its clear that FSB increases above about 286 really start to require way too much Vcore for the gains realized. Mine fails dual primes in about 10 seconds or so at 4.2 / 1.465v. No point in testing higher, as it doesn't make sense to run at 4.2 if it needs 1.5x volts, but 4.0 Ghz only needs 1.425.
I've also noticed that if I ease off of 4Ghz, say to maybe 3.85 / 3.9, stock voltage is just fine.
I've got 3 more (920) chips coming in tomorrow - will post results when available.
- Chris
FWIW, I'm getting 286FSB / 4Ghz at 1.425v, which after dual priming for 8 hours or so produces 58c load temps (large volume water system - takes a while to stabilize). 4.2 is probably possible, but its clear that FSB increases above about 286 really start to require way too much Vcore for the gains realized. Mine fails dual primes in about 10 seconds or so at 4.2 / 1.465v. No point in testing higher, as it doesn't make sense to run at 4.2 if it needs 1.5x volts, but 4.0 Ghz only needs 1.425.
I've also noticed that if I ease off of 4Ghz, say to maybe 3.85 / 3.9, stock voltage is just fine.
I've got 3 more (920) chips coming in tomorrow - will post results when available.
- Chris
Three more 920's coming in ? Are they all for you ? Wow !
I am anxious to hear your results to see if my 920 is average or not. At 4Ghz my chip needs about 1.45 to be stable. At 4.2Ghz it needs 1.5V at 4.5 it needs about 1.58, and is not Prime stable but I can work on it and surf. I have not tried anything above that because of temps and high voltage.
Is this average for these 920's ?
I think there is a lot of variability on this forum when it comes to defining "stable". I define, and have always defined stable as not crashing, freezing, BSOD'ing or doing anything that, say a stock $750 "Dell" PC wouldn't do. And that doesn't mean the occasional offense is OK. If something EVER causes an issue, and I can eliminate it by down-clocking, then it wasn't stable under those conditions. Period.
That said, one should temper people's results here. Not to say people aren't fastidious about reporting things accurately, but people define stable a little differently.
I've built enough Intel P4 OC'd systems to have a pretty good feel for whats causing issues, and in this case (as in most in my experience) the chip is the limiting component (as long as you keep the mem within its limits, of course).
So, when I said I need 1.425v for stability at 4.0 Ghz, that means I can dual prime at LEAST overnight, and run a .mpg2 -> divx encoder at the same time. I find that running "autoGK" really can find weak spots, sometimes that dual priming won't pick up, believe it or not.
That also means that I know 1.400v will NOT do it. 1.4125 MAY, but I know 1.425 is the minimum that I have tried that WILL. I would consider 48hrs dual prime stable and letting autoGK do a list of video conversions thrown in there to be about as stable as a known overclocked machine is going to be (which should / could be "perfect").
So, hopefully we can get some qualifications as to what criteria people are using to declare a rig "100% stable", and get a better answer to your question as to what is average - I'd like to know too.
- Chris
iamjcl,
How can we say an overclock is 100% stable when our software and OS are not 100% stable? :)
Ya, I think we need to say something like "Dual Prime stable" instead of saying an unqualified "100% stable".
iamjcl,
How can we say an overclock is 100% stable when our software and OS are not 100% stable? :)
Good point !
- Chris
I know I am using the standard of Dual Prime Stable. I would like to see if others are too when talking about their chips ?
Here is a screen shot showing 4 different 920s on the same rig (mine).
Its far from conclusive, but it shows some of the variance in these chips (not a lot).
Red indicates failure, and how long it took. Blue is success, and for how long.
- Chris
My stable definition is dual folding 724 without folding errors.
current stable 920 config is on a p5wd2 folding for nearly 2 weeks now at 290 (4.06) at 1.425 vc on air but caseless on a rack with cool ambients.
I should really do some more fine grained testing to see if i can hit a sweeter spot with near stock VC for only a small sacrifice in fsb. According to my initial OC'ing notes I was running stock VC up to 280, but didn't do enough stability testing to declare it stable there. All my testing up to 290 was in 10 fsb increments.
Big VC jumps required past that which are not sustainable on air dual folding.
I also had a very interesting result of some tuning with an 830 temporarily on a p5wd2-e. After letting it fold during the day today, i came home to find one instance had errored. I lowerd the fsb by 5 (my standard way to make it stable). Then restarted the failed work unit from the begining.
The suprising part was than my folding performance actually when UP from 894 to 911 by LOWERING the FSB by 5! :cool: :shrug:
My guess is that this finds a sweeter spot for the NB /mem timings and ratios.
Here is a screen shot showing 4 different 920s on the same rig (mine).
Its far from conclusive, but it shows some of the variance in these chips (not a lot).
Red indicates failure, and how long it took. Blue is success, and for how long.
- Chris
Thanks for posting the results of the testing of your chips. Maybe then I will keep mine and just work with it. It doesn't seem like a big difference between the chips.....like one needing 1.3V at a certain speed and another needing 1.5V.
It doesn't seem like a big difference between the chips.....like one needing 1.3V at a certain speed and another needing 1.5V.
I see the difference between 1.3v and 1.5v as being very large. Huge, in fact. You seem to have a lot of chips to bin, so if I were you, I would go for the lowest voltage for a good OC, not the absolutely top Ghz at a dangerous voltage. (although they could be 1 and the same)
bobad,
I agree.
Chips can have a "personality", as well, like only failing prime, or rebooting, or giving windows errors, and only 1 core failing a certain way.
Some chips have a "wall", also, where no matter what, the cache (probably) just won't run above a certain speed, even if a fairly low voltage is all thats needed at a slightly lower speed.
So, you hope that the one that gives the highest OC will also be the one that delivers a reasonable speed - I say 4.00 Ghz for a 920 - at a "low" Vcore - I say 1.4 or less.
Sure, its fun to push it, but in 3 or 4 months, when the rig isn't cutting edge, whats 100 or 200 mhz (2-5 %) ? It surely isn't worth pumping lots of extra juice into the chip to get it.
Look at it like this - rule of thumb is 10% additional voltage is in the "green" zone (1.3v * 1.10 = 1.43v). I say go by BIOS settings, not what you see in CPUz, MBM, etc...
So, if you look at the additional .13 volts that is considered OK as being 100% of the additional safe voltage that you have available to OC, and you need 1.5v to reach a speed that is 150mhz higher than what the "safe 100%" 1.43v will do, you are pumping almost an additional 50% MORE of what I would call the needed "overage" voltage into the chip to achieve only a 3.6% clock speed improvement.
BTW, still testing on chip 3, and its been stable (dual primes, pcmark, autoGK video conversions) for ~10 hours now, at 1.4875v. Got a windows error after 3 or 4 hours or so at the 1.4750 listed in the chart.
Also, not reflected in the chart, I did a little low voltage testing (not comprehensive) , and it appears chip 3 is also well-suited to this (better than the others). It will do 3.8 Ghz (272 fsb) at BELOW stock voltage...This was only tested for a fairly short period, but was stable nevertheless, when the other chips would reboot.
- Chris
Yeah but considering only Chip 0 and Chip 3 seemed to pass Prime at any given speeds so they are the only ones to look at.
I guess I am looking at this from the perspective that some chips at no amount of voltage will reach a given speed, these chips seem to take the voltage better and reach higher speeds. Some more than others but that is to be expected. For instance.....I see some of members here got their old 800 series Pentium D's to 4Ghz or above, stable. With my 820, I could never go above 3.85Ghz even for a suicide shot at any voltage and stable for me was 3.73Ghz. I think there always will be variances, but not like there have been in the past.
Here is a screen shot showing 4 different 920s on the same rig (mine).
Its far from conclusive, but it shows some of the variance in these chips (not a lot).
Red indicates failure, and how long it took. Blue is success, and for how long.
- Chris
wow..guesss my chip really is "special"
but hey 4.2ghz is very niec :) #3 is the keeper?
wow..guesss my chip really is "special"
but hey 4.2ghz is very niec :) #3 is the keeper?
Yep like I said earlier....your chip is special ! :)
d94 clearly has an above average chip, at least based on my little batch for testing.
Still, at 4.5G -vs- 4.2G, the clock-speed gap between a "special" chip and a more typical one is no more than 7% - while I'd take it, that increase is purely academic in the final analysis.
I would speculate that a "poor" 920 may only hit 3.8 on 1.5v or less, and, using d94's chip as a "special" one that does 4.5 on 1.5v or less, that puts about an 18% variance in achievable clockspeed between the "worst" 920 one is likely to get, and the best, with the likely result at 1.5v or less being 4.0 Ghz.
I think that about sums it up.
- Chris
d94 clearly has an above average chip, at least based on my little batch for testing.
Still, at 4.5G -vs- 4.2G, the clock-speed gap between a "special" chip and a more typical one is no more than 7% - while I'd take it, that increase is purely academic in the final analysis.
I would speculate that a "poor" 920 may only hit 3.8 on 1.5v or less, and, using d94's chip as a "special" one that does 4.5 on 1.5v or less, that puts about an 18% variance in achievable clockspeed between the "worst" 920 one is likely to get, and the best, with the likely result at 1.5v or less being 4.0 Ghz.
I think that about sums it up.
- Chris
Good summary. Since my chip falls at 4.2Ghz @1.5V, I am average or I should I say slightly better than average ?
vBulletin® v3.8.7, Copyright ©2000-2013, vBulletin Solutions, Inc.