But before that you could just fill up the tube and bench?Quote:
Originally Posted by xoqolatl
I'm not sure Intel were 100% coldbugless even if they did just do what I just asked about. One would have to run some oldies on LHe to really find out:D
Printable View
But before that you could just fill up the tube and bench?Quote:
Originally Posted by xoqolatl
I'm not sure Intel were 100% coldbugless even if they did just do what I just asked about. One would have to run some oldies on LHe to really find out:D
they were somewhat but not as bad as I7 from some of the old school benchers ive chatted with. I actually predidcted it before I7 came out, was talking to moviman like 6 months before I7 released and said I bet i7 gets a nasty CB with the addition of the MMC.Quote:
Originally Posted by xoqolatl
Sometimes the source was the only persosn that was supposed to use the chip, when it gets posted on a forum by someone else, well someone can lose a good paying job. Thats 9/10 the case.... Covering the entire lower line is really the only safe way to post it.Quote:
Originally Posted by largon
How did that CS chip clock HT under cold ;) or did you just do multi?Quote:
Originally Posted by xoqolatl
I like that batch #, only had 1 lower than that in my entire life.
yea my chip isn't a cv or a cs. it was just produced the other day. i think they made up a special batch for amd employees idk. its most likely fine cause i see now the only difference is that the memory controller revision letter is different and its an ES and that it was produced in 2009. im putting it in this weekend so if i see anything different from the ordinary i won't post anything. but every amd chip i have had before is an ES anyway.Quote:
Originally Posted by xoqolatl
I did not test HT under cold. It does 320 stable on air though.
Here's AMD's explanation of date codes. It's still a mystery to me.:)
http://www.amd.com/us-en/assets/cont...koptions_1.pdf
http://img107.imageshack.us/img107/2...odesxr0.th.jpg
makes sense. i know it has been posted earlier. also does anyone have any info on how the cpu dies are numbered on the wafer? many say that they are numbered from the middle outward so if you get a 0001 chip then you got the middle of the wafer. this doesn't really make sense and i have heard that they just number it normally from the top left corner left to right.
I only read the first post. :rolleyes:
I graphically compiled a small amount of FX chips for you all to investigate.
http://img152.imageshack.us/img152/6...7068tk3.th.jpg
They're mostly review samples so they have the same week, 5 and 4-letter codes so it probably doesn't help much. To be honest, finding a good clocker is probably just a matter of getting lucky.
yea thats what i think too. the batch number isn't affecting anything and same with the steppings.Quote:
Originally Posted by uncle john
THat's because you don't have a database...For example, chartered semiconductor AND AMD make CPU chips that AMD sells. Chartered's Foundry is not as good as AMD's...how do you know what you are going to get?
AMD engineers changed the doping slightly...on a specific set of wafers...and they clocked exceptionally well...or are lower wattage...how do you find them?
AMD used to sell cpus with 1/2 cache disabled with same basic part number as cpus that were not made with the extra cache the first were...how do you tell?
How do you find the same wafer used for Opteron chips, which have higher tolerances, in the retail line?
Batch Numbers and Steppings....
amdgeeks.net is a database and I uploaded both of the 939's I had but it doesn't require the wafer unit number or whatever those last 4 digits represent. I can't remember my Opti's number but the X2 was 12xx something and it had a poor IMC. 242MHz was all it would allow.
These are all xPMW CPUs. The P and the W are always the same because these CPUs made in Germany (Dresden). The M letter is always a good sign. For example the MPMW stepping always signs good OCing modell.Quote:
Originally Posted by uncle john
Well, I'm even less sure the markings matter than before now. The AMD document clearly states that the markings are only based on when the thing is packaged and not directly tied to the wafer in any way. 8001 may have been the center most die or a flaky one on the outside edge. There is no guarantee of anything. And M just stands for mixed, meaning there's even less consistency in the product than the other markings.
Yes, particle, so if you know HOW the person doing the binning is doing thier job, you know which chips they will put into the batch first...
I mean, it's not cut and dry, but it IS definately meaningful.
The point here is that it's all based on guesses at this point. There is absolutely no solid link here. We're just assuming that the same chips end up numbered the same way reliably, and we have no way of knowing if that's true. If it were, the numbers wouldn't represent packaging times but would instead be linked to chips themselves. Especially considering that they have mixed batches, I really doubt the same dies end up with the same numbers all the time.
Just to be clear: It's not that your idea is conceptually difficult to understand or that I don't get it. It's just that you cannot properly draw that conclusion with the information we're given.
Ah, but I can becuase I have info you do not, and also I have garnered results worldwide from AMD cpus, as have others, that led to this conclusion. I HAVE that tidbit that elludes you...you also forget that Macci works @ AMD, and at one point, was very helpful to all of us when it came to searching for cpus...because in the end, AMD does not test every chip to thier max, but are VERY interested in the results.
I'll share a site with you that used to keep really good results, that may help you with the correlation.
http://www.amdgeeks.net/processors
There used to be a time when if I posted an issue with hardware or something, I'd get a pm from a company rep asking what they could do to solve the problem...Don't forget Francis being on here as well...
That customer interaction fell off when companies started to do good, and AMD bought ATI...the tech world changed then...but it's those companies that make a grassroots effort with us in these hard times, that will excel.
The letter for the first five is for L2, and that's it, hence I'm pretty sure the AC is related to L3 cache, and to support this these letter only came up with the K10, my K8 doesn't have those letters (I can post a pic if you don't believe)Quote:
Originally Posted by largon
Since some of you so insistent that the batch # means little or nothing, anyone care to muster a guess what batch numbers over 20 chips I saw at the boston event had. I saw them personally with my own eyes.
I read they weren't cherries but I'll submit a guess; 0001-0020.Quote:
Originally Posted by chew*
I will say they were all less than 0100 and leave it at that, apparently they weren't cherries though as others posting with inside info apparently KNOW FOR A FACT that it means nothing ;)Quote:
Originally Posted by uncle john
Nothing the average user can't duplicate however.......
My own experience hasn't done much to improve my belief that it matters. My 9850 for instance was an 80016 and it was awful. It topped out at 2.7GHz stable with 1.45V. My 940 is 80074 and it's not great either. You'll notice on the OC chart that I'm in dead last needing the most voltage for the slowest stable frequency.
My 9850 was 0816GPMW--80259...and Overclocks like crazy
The only conclusion I can draw from my experience with stellar clocking chips is that they were xPxW coded. Unfortunately I haven't a clue what the batch numbers were.
0310XPMW Thoroughbred (desktop) box
0329XPMW Thoroughbred (desktop) box
0348TPMW Barton (mobile) oem only
0614EPCW Venus (server) box
my 9850 is same letters, diff. week,,, 0820GPMW--80436...and also oc's good