Sign in with
Sign up | Sign in
Your question

thoroughbred

Last response: in CPUs
Share
May 1, 2002 12:10:31 AM

does anyone have any idea how much the new processors will cost when they come out...and when exactly will they be in stores?

More about : thoroughbred

May 1, 2002 12:45:57 AM

0.13m usually WOULD cost less. But they want profit, I expect the prices of 0.13m versions of i.e. an AXP 1800+, to be a tad less than its 0.18m. But no real changes, prices will always remain below Intel's.

--
Thunderbirds in wintertime, Northwoods in summertime! :lol: 
May 1, 2002 12:52:54 AM

Yes usually they would cost less, (since smaller die=more yeild=less cost per chip) but knowing AMD and their current money-losing position, I'd say they'll sell it for the same price, because a average user won't know the difference between a .13um XP1800+ and a .18um XP1800+ and since there's really not much difference in the cores, why should the pricing be different right?
Related resources
May 1, 2002 12:56:48 AM

Well since nothing else was added than just a die shrink (unless the rumors are true), I would expect a 1-3% less cost. P4 NWs had this at some point until the prices were revised and went above. But even with more cache, it proved that the price didn't increase more than 5% in most cases.

--
Thunderbirds in wintertime, Northwoods in summertime! :lol: 
May 1, 2002 4:46:57 AM

AMDs cost per chip will be significantly decreased by moving to .13 micron. This is because not only is the process now at .13 micron, but they are now using larger platters to make the chips as well. I think before they used 8" platters, now they are switching to 10" or 12" platters. at 8" the are is about 32,000 sq. inches, and at 10" the area 50,000 sq. mm. At 120 sq. mm for Palomino, they could get about 260 dies per platter, and at 80 sq. mm for TBred, they can get about 625 dies per platter. Thats over twice as many dies per platter for not much more of a cost. But, being that AMD wants to make money, you will probably see them introduced at a similar price as Palominos, with a price drop shortly after, or they'll introduce the TBreds in conjuction with a price drop. My 2 cents anyways.

"Trying is the first step towards failure."
May 1, 2002 6:41:26 AM

Challenges:
1. When and where did you hear that AMD is using 12" wafer?
2. The cost of a die is not only determined by the size. Yield is equally important. Why do you assume AMD will get same yield on .13 as on .18?
May 1, 2002 8:25:58 AM

Quote:
AMDs cost per chip will be significantly decreased by moving to .13 micron. This is because not only is the process now at .13 micron, but they are now using larger platters to make the chips as well. I think before they used 8" platters, now they are switching to 10" or 12" platters. at 8" the are is about 32,000 sq. inches, and at 10" the area 50,000 sq. mm.


No they arent, they are still using 8 inch wafers.



:wink: The Cash Left In My Pocket,The BEST Benchmark :wink:
May 1, 2002 10:05:56 AM

mmm. niiiiilla waaaaaaafers. aaaathlon waaaaaafers.

<b>Studies have shown that most people prefer the taste of AMD to the taste of Intel</b>
May 1, 2002 12:55:52 PM

AMD may fit more die on per wafer, but their yields are lower, Intel's are the best in the Industry, hence the fact that they managed to maintain their margins in the last qtr at a better than industry level. AMD won't make the 12 inch jump for a while yet. Also, the rumours I have heard in relation to the ongoing Thoroughbred roadmap saga is that it's not actually yield which is the problem for AMD, it's getting the higher speed bins whih is troubling them, hence the lower PR rated models already being on the market. Availability of the higher rated models will be v. tight indeed.
May 1, 2002 1:07:19 PM

Quote:
AMD may fit more die on per wafer, but their yields are lower, Intel's are the best in the Industry, hence the fact that they managed to maintain their margins in the last qtr at a better than industry level.


Actually, Dresden, AMD's main CPU FAB purportedly has the highest yields in the industry. Sanders has said that he believes Dresden has higher yields than ANY production Intel FAB.

In January AMD claimed yields for TBred were similar to AXP. Since then I expect AMD has significantly improved TBred yields. AMD expects around 50 million CPUs from Dresden at the 80mm die size (.13 micron) when at full capacity. Of course, it will be lower because there is always at least one line being revamped I suspect. All the same, if total CPU shipments (PC/X86) were to remain similar to 2001, Dresden ALONE would be able to produce nearly 1/3 of all PC processors. I doubt very much that would be the case if yields out of Dresden were lower than Intel's yields in any of their FABs.

Mark-

<font color=blue>When all else fails, throw your computer out the window!!!</font color=blue>
May 1, 2002 3:40:15 PM

Dammit, Nilla Wafers sound really good right now.

But back on topic, the Northwood settled into pricing below that of the Willamette. I wouldn't expect the same thing from AMD however, due to the same size wafer (mixed with assumedly lower yields, at least at the beginning) and the fact that AMD doesn't have as high a profit margin as Intel does for the majority of CPUs.

<font color=blue>Hi mom!</font color=blue>
May 1, 2002 6:31:26 PM

As always, I wouldnot believe Sanders' words. He has been saying everything then taking back later. Yeah, yeah, he said AMD is gonna turn profitable in Q2... AMD gonna release Hammer in 2003. Then someone jumped out covering his ass in the 2nd day.
May 1, 2002 8:28:06 PM

I'd like to hear from Matisaro on this one. He was with Fugistsu, and IIRC they had quite some affiliations back then with AMD, so he might know of their yeilds at Dresden. Even if it doesn't yeild as much, it still has the best yeilds IMO, look at how good the AXP silicon is compared to a P4 chip heat and power-wise as well as size.

--
Thunderbirds in wintertime, Northwoods in summertime! :lol: 
May 1, 2002 8:28:50 PM

Quote:
AMD may fit more die on per wafer, but their yields are lower, Intel's are the best in the Industry, hence the fact that they managed to maintain their margins in the last qtr at a better than industry level. AMD won't make the 12 inch jump for a while yet. Also, the rumours I have heard in relation to the ongoing Thoroughbred roadmap saga is that it's not actually yield which is the problem for AMD, it's getting the higher speed bins whih is troubling them, hence the lower PR rated models already being on the market. Availability of the higher rated models will be v. tight indeed.



Dresden is the best fab in the industry, and like the other poster mentioned, the tbred yields are about the same as axps, which are much higher than the industry average.

As for the binning issue, they did have a problem with the first stepping of the shrink, and they repaired that issue, the chips made with the first stepping should be sold as mobiles and low end tbreds, the supply should not last long.



:wink: The Cash Left In My Pocket,The BEST Benchmark :wink:
May 1, 2002 8:31:21 PM

^^^
Well that cleared it, thank you!

But what about the first bad OCed samples that the vrzone website told us about? These aren't the real OC yeilds we'd get if we wait a bit for newer chips, right?

--
Thunderbirds in wintertime, Northwoods in summertime! :lol: 
May 1, 2002 8:33:54 PM

We at fujitsu didnt have ties with amds cpu division, but the people I talked to from amd told me about dresden and the fact they have VERY high yields, especially for logic, they blew our yields at GMD AWAY, and we were only making flash and dram, much less complex than logic.


I dont think sanders is lying, he may brag, and he certainly antagonizes intel, but he dosent lie IIRC.

As for yields/die shrink, when you shrink a die you will lose yield, for the simple fact that smaller defects can(and do) cause fatal issues to the die. Also your specs are tighter and more prone to killer variation.

BUT, you always (unless your fab sucks ass) make up for the lost yield (and then some) with the fact you now have more die/wafer, if this wasnt the case there would be little reason to shrink non speed limited product(like dram/flash/slow logic etc).



:wink: The Cash Left In My Pocket,The BEST Benchmark :wink:
May 1, 2002 8:36:57 PM

Quote:
But what about the first bad OCed samples that the vrzone website told us about? These aren't the real OC yeilds we'd get if we wait a bit for newer chips, right?


I dont know for sure, but here is my guess.


Given the info I have read/gathered, amd had a timing issue on their first shrink stepping, they realized this when samples were not binning high(but much higher than axp mind you, just not the 45% they should have gotten).

They went back and fixxed the issue and the new chips should bin where the shrink should put them.


HOWEVER, there were chips produced on the bad stepping, chips which have to be sold, they should be sold as mobiles and low rated tbreds(avoid the 1800+ for a few weeks!).

Once the supply of these chips is exhausted, the tbred will scale quite nicely.

:wink: The Cash Left In My Pocket,The BEST Benchmark :wink:
May 2, 2002 12:45:25 AM

so you think the problem chips will only be limited to the 1800's...i have to get a processor soon and my choices are the 18 19 or maybe the 2000...think its worth it to chance it with a 1900+ or play it safe and get the t-bred 2000+?
May 2, 2002 12:50:46 AM

It isn't limited to them, it's limited to the early samples of them, which aren't the best or indicative of their actual overclockability AT ALL. Mat could explain it again, but we all agree, wait a month, see the Tbred 1800+ again and you'll probably see a better view then.

--
Thunderbirds in wintertime, Northwoods in summertime! :lol: 
May 2, 2002 12:56:50 AM

Well, any yield number is top secret in the whole industry. I highly suspect the claim that "Dresden is the best", let alone "much better". Also, yield is nothing to do with the die size or power consumption difference between AXP and P4.

<P ID="edit"><FONT SIZE=-1><EM>Edited by castle on 05/01/02 09:01 PM.</EM></FONT></P>
May 2, 2002 1:46:02 AM

Yes but the succesful yeilds, are the best IMO.

--
Thunderbirds in wintertime, Northwoods in summertime! :lol: 
May 2, 2002 5:25:08 AM

Quote:
so you think the problem chips will only be limited to the 1800's...i have to get a processor soon and my choices are the 18 19 or maybe the 2000...think its worth it to chance it with a 1900+ or play it safe and get the t-bred 2000+?


I think if amd had chips (which arent problem chips btw they can clock more than enough to be sold as 2200+'s) but they topped out lower, they would be more likey to sell said chips as mobile and the lowest bin tbred untill said stock was eliminated.

:wink: The Cash Left In My Pocket,The BEST Benchmark :wink:
!