Sign in with
Sign up | Sign in
Your question

AMD emergency price cuts?

Tags:
Last response: in CPUs
Share
November 30, 2006 2:01:52 PM

Maybe they got a nice letter from AMD telling them to retract the article.

Or maybe the Inq. jumped the gun again.
Related resources
November 30, 2006 2:29:35 PM

It's what intel is doing, only that they've got so much 'fat' that they can stand those prices a lot better :lol: 
December 2, 2006 9:45:17 PM

Quote:
http://theinquirer.net/default.aspx?article=36076

But now it's gone! Eh, what gives?


They put it back up:
http://www.theinquirer.net/default.aspx?article=36076

Odd that they need price cuts when they are 'selling everything they can make'.

This is a great point, Jack. Maybe it's a sign that not all of AMD's products are flying off the shelves in certain segments. Server CPUs, perhaps, but they'll probably to take a beating in desktop CPUs this Xmas now that the Core2 cat is out of the bag.
December 3, 2006 1:23:13 AM

65nm will give around twice the chips per wafer compared to 90 nm, not 3 to 4 times as many.
65nm ^ 2 = 4225
90nm ^ 2 = 8100

8100 / 4225 = 1.91715
The dies may be relatively smaller or relatively larger than the old ones (transistor count wise), and may use an extra layer or two in construction to get over double the chips per wafer.


They slashed Opteron 200/800, 2000/8000 prices by around 45% on the 10th - 11th October 2006 to remain competitive in the server space.

Getting beyond 2.8 - 3.2 GHz on 65nm will take them 9 months or so once 65 nm production gets into full swing. If it takes them until April 2008 to get there they'll start falling behind pretty badly.

They really need a cost effective part at 3.4 - 3.73 GHz to remain competitive in the desktop / workstation space.

AMD need to keep the price/performance curve in their favour as many gamers are cash limited, and other users just look at price/performance for what they can afford.

As much as they should've been expecting Core 2 Duo, they would've known since slightly before March 2006, AMD haven't done enough to counter it.

Intel just timed its release to maximum their market growth.

To anyone looking at Quad-FX I'd recommend looking at Opteron 2000 series instead.
December 3, 2006 1:30:55 AM

Quote:
65nm will give around twice the chips per wafer compared to 90 nm, not 3 to 4 times as many.
65nm ^ 2 = 4225
90nm ^ 2 = 8100

8100 / 4225 = 1.91715
The dies may be relatively smaller or larger than the old ones, and may use an extra layer or two in construction to get over double the chips per wafer.


They slashed Opteron 200/800, 2000/8000 prices by around 45% on the 10th - 11th October 2006 to remain competitive in the server space.

Getting beyond 2.8 - 3.2 GHz on 65nm will take them 9 months or so once 65 nm production gets into full swing. If it takes them until April 2008 to get there they'll start falling behind pretty badly.

They really need a cost effective part at 3.4 - 3.73 GHz to remain competitive in the desktop / workstation space.

AMD need to keep the price/performance curve in their favour as many gamers are cash limited, and other users just look at price/performance for what they can afford.

As much as they should've been expecting Core 2 Duo, they would've known since slightly before March 2006, AMD haven't done enough to counter it.

Intel just timed its release to maximum their market growth.

To anyone looking at Quad-FX I'd recommend looking at Opteron 2000 series instead.


You're forgetting that Fab36 uses 300mm wafers, which in and of itself means 2x the amount of chips at the same transistor size.
December 3, 2006 8:49:56 AM

Quote:
http://theinquirer.net/default.aspx?article=36076

But now it's gone! Eh, what gives?


65nm = 3-4x more chips from one wafer.

That is only if they get close to 100% yield from each wafer. I bet you dollars to donuts that they wont get even 60% yield from each wafer initially. Yield will most likely follow the standard Bell curve. If their top processors are going to be at 2.6GHz and their low is at 2.0GHz then 60% to 70% of the yielded processors will be at 2.2 and 2.4GHz.

You can also bet that even with Intel's mature 65nm process they are only getting in the mid 90's for yield. But since they have 3 plants producing wafers that is a lot of die that they are getting. Again the yield will be in a Bell Shape curve with most of the die binning out at the 2.4 to 2.6GHz range.

At least this is my educated guess from what I have heard from my Intel friends.
December 3, 2006 1:03:51 PM

Quote:
http://theinquirer.net/default.aspx?article=36076

But now it's gone! Eh, what gives?


65nm = 3-4x more chips from one wafer.

That is only if they get close to 100% yield from each wafer. I bet you dollars to donuts that they wont get even 60% yield from each wafer initially. Yield will most likely follow the standard Bell curve. If their top processors are going to be at 2.6GHz and their low is at 2.0GHz then 60% to 70% of the yielded processors will be at 2.2 and 2.4GHz.

You can also bet that even with Intel's mature 65nm process they are only getting in the mid 90's for yield. But since they have 3 plants producing wafers that is a lot of die that they are getting. Again the yield will be in a Bell Shape curve with most of the die binning out at the 2.4 to 2.6GHz range.

At least this is my educated guess from what I have heard from my Intel friends.


If we assume the same yields for both, it's still 2x, so the statmemnet that they'll get 3-4x from Fab36 still stands.
December 3, 2006 1:43:04 PM

Quote:
http://theinquirer.net/default.aspx?article=36076

But now it's gone! Eh, what gives?


65nm = 3-4x more chips from one wafer.

That is only if they get close to 100% yield from each wafer. I bet you dollars to donuts that they wont get even 60% yield from each wafer initially. Yield will most likely follow the standard Bell curve. If their top processors are going to be at 2.6GHz and their low is at 2.0GHz then 60% to 70% of the yielded processors will be at 2.2 and 2.4GHz.

You can also bet that even with Intel's mature 65nm process they are only getting in the mid 90's for yield. But since they have 3 plants producing wafers that is a lot of die that they are getting. Again the yield will be in a Bell Shape curve with most of the die binning out at the 2.4 to 2.6GHz range.

At least this is my educated guess from what I have heard from my Intel friends.


If we assume the same yields for both, it's still 2x, so the statmemnet that they'll get 3-4x from Fab36 still stands.

AMD will have much higher yeilds from 90nm than they will from 65nm to start with. They have had quite some time to get 90nm worked out properly...
December 3, 2006 2:38:16 PM

Quote:
AMD will have much higher yeilds from 90nm than they will from 65nm to start with. They have had quite some time to get 90nm worked out properly...


Since when?
December 3, 2006 8:09:24 PM

KLA-Tencor, leading semi process control company, recently indicated that almost all semi companies have 65 nano yield well below 50%. As Amd converts single core 90 to dual core 65nano, there will be a decline in unit production. They not only need good yied on 65 nano but a ramp of Fab36 to 20,000wsm from the current 10,000wsm. Fab 30 will actually see declining production in 2007 from the 25,000wsm at 200 millimeter in 2006 as they retrofit for 65nano at 300 millimeter in 2008.
AMD will have a difficult time making products in 2007 at price and/or quanity anyone would want to buy. They are in real trouble.
December 3, 2006 8:17:37 PM

Quote:
http://theinquirer.net/default.aspx?article=36076

But now it's gone! Eh, what gives?


65nm = 3-4x more chips from one wafer.

That is only if they get close to 100% yield from each wafer. I bet you dollars to donuts that they wont get even 60% yield from each wafer initially. Yield will most likely follow the standard Bell curve. If their top processors are going to be at 2.6GHz and their low is at 2.0GHz then 60% to 70% of the yielded processors will be at 2.2 and 2.4GHz.

You can also bet that even with Intel's mature 65nm process they are only getting in the mid 90's for yield. But since they have 3 plants producing wafers that is a lot of die that they are getting. Again the yield will be in a Bell Shape curve with most of the die binning out at the 2.4 to 2.6GHz range.

At least this is my educated guess from what I have heard from my Intel friends.


If we assume the same yields for both, it's still 2x, so the statmemnet that they'll get 3-4x from Fab36 still stands.

Process 90nm Wafer size 300mm ((pi x D2/4) = 3.14 x (150 mm)2 = 70,650 mm2)X2 die size 230mm2 = approximately 307-62.1(edge die) = 245 x2 die per 300mm @ .90nm. My math is below:

-The circumference of a 12-inch wafer is (pi x D) = 3.14 x 300 mm. = 942mm. The diagonal width of a 230mm die is ( SQRT (2) x 230mm. ) = 15.16mm. The approximate number of partial die around the perimeter is (942mm/15.16mm) = 62.1 die.

Process 65nm wafer size 300mm ((pi x D2/4) = 3.14 x (150 mm)2 = 70,650 mm2)X2 die size 133mm2 (1.72x smaller, for straight 90nm 65nm transition) = approximately 531-81.7(edge die) = 449.3 x2 die per 300mm @ .65nm. My math below:

-The circumference of a 12-inch wafer is (pi x D) = 3.14 x 300 mm. = 942mm. The diagonal width of a 133mm die is ( SQRT (2) x 133mm. ) = 11.53mm. The approximate number of partial die around the perimeter is (942mm/11.53mm) = 81.7 die.

1.83x more x2 die at 65nm than 90nm on 300mm wafers. If my math is wrong I am sure someone will point that out.
December 3, 2006 8:45:06 PM

[code:1:412069cbe3]Process 90nm Wafer size 300mm ((pi x D2/4) = 3.14 x (150 mm)2 = 70,650 mm2)X2 die size 230mm2 = approximately 307-62.1(edge die) = 245 x2 die per 300mm @ .90nm. My math is below:

-The circumference of a 12-inch wafer is (pi x D) = 3.14 x 300 mm. = 942mm. The diagonal width of a 230mm die is ( SQRT (2) x 230mm. ) = 15.16mm. The approximate number of partial die around the perimeter is (942mm/15.16mm) = 62.1 die.

Process 65nm wafer size 300mm ((pi x D2/4) = 3.14 x (150 mm)2 = 70,650 mm2)X2 die size 133mm2 (1.72x smaller, for straight 90nm 65nm transition) = approximately 531-81.7(edge die) = 449.3 x2 die per 300mm @ .65nm. My math below:

-The circumference of a 12-inch wafer is (pi x D) = 3.14 x 300 mm. = 942mm. The diagonal width of a 133mm die is ( SQRT (2) x 133mm. ) = 11.53mm. The approximate number of partial die around the perimeter is (942mm/11.53mm) = 81.7 die.

1.83x more x2 die at 65nm than 90nm on 300mm wafers. If my math is wrong I am sure someone will point that out.[/code:1:412069cbe3]


But 80%+ of AMDs 90nm chips are coming out of 200mm Fab30 which at some point early this year was runnign at a reported 125% capacity. I heard Fab36 was shipping revenue 90nm but that may mean as few as 3000WSPM. Chartered is also doing about that.
December 3, 2006 9:44:46 PM

Quote:
KLA-Tencor, leading semi process control company, recently indicated that almost all semi companies have 65 nano yield well below 50%. As Amd converts single core 90 to dual core 65nano, there will be a decline in unit production. They not only need good yied on 65 nano but a ramp of Fab36 to 20,000wsm from the current 10,000wsm. Fab 30 will actually see declining production in 2007 from the 25,000wsm at 200 millimeter in 2006 as they retrofit for 65nano at 300 millimeter in 2008.
AMD will have a difficult time making products in 2007 at price and/or quanity anyone would want to buy. They are in real trouble.



Doom and gloom. Doom and gloom. If they can make 3X more chips and have 50% yields, that's still 1.5x more chips, maybe more since 90nm is probably not at 100%.
December 3, 2006 10:51:44 PM

Quote:
KLA-Tencor, leading semi process control company, recently indicated that almost all semi companies have 65 nano yield well below 50%. As Amd converts single core 90 to dual core 65nano, there will be a decline in unit production. They not only need good yied on 65 nano but a ramp of Fab36 to 20,000wsm from the current 10,000wsm. Fab 30 will actually see declining production in 2007 from the 25,000wsm at 200 millimeter in 2006 as they retrofit for 65nano at 300 millimeter in 2008.
AMD will have a difficult time making products in 2007 at price and/or quanity anyone would want to buy. They are in real trouble.

Oh really??? When is "recently"?? "Almost all" does not mean "all". What is the definition of "well below 50%"? Who at KLA Tencore said it? Was it a press release or is it all a figment of your imagination??
December 3, 2006 11:14:40 PM

Intel has actually disclosed some info on the advantages of two die pakages. Their yield or capacity increases nearly 50%. Even more importantly by matching die, they obtain a 23% increase in average clock frequency per processor. This of course applies in the first 12 months of a process. The advantage declines with the maturity of the process.
You would have to be an idiot not to construct high end multicore processors this way. I would have to assume that AMD does not have the packaging expetise to do the same as Intel.
December 3, 2006 11:58:10 PM

Question:Not knowing the processes involved,doesnt creating a whole new architecture reduce yield as well?As in the c2d?
December 4, 2006 12:47:41 AM

Quote:
KLA-Tencor, leading semi process control company, recently indicated that almost all semi companies have 65 nano yield well below 50%. As Amd converts single core 90 to dual core 65nano, there will be a decline in unit production. They not only need good yied on 65 nano but a ramp of Fab36 to 20,000wsm from the current 10,000wsm. Fab 30 will actually see declining production in 2007 from the 25,000wsm at 200 millimeter in 2006 as they retrofit for 65nano at 300 millimeter in 2008.
AMD will have a difficult time making products in 2007 at price and/or quanity anyone would want to buy. They are in real trouble.



Doom and gloom. Doom and gloom. If they can make 3X more chips and have 50% yields, that's still 1.5x more chips, maybe more since 90nm is probably not at 100%.

Remember that the quad-core chips are physically larger then the dual core 90nm....

Assuming redwood's facts are true (50% is rediculous, I don't think it's that low..), then AMD would probably be making less chips/wafer on the 300mm wafers, and may break even when you compare the 200mm to the 300mm ones... I do not have the necessary intelligence, information (do not know transistor count of the quad-cores), or inclination to to do the math, so I am just estimating...
December 4, 2006 2:21:50 AM

Quote:
KLA-Tencor, leading semi process control company, recently indicated that almost all semi companies have 65 nano yield well below 50%. As Amd converts single core 90 to dual core 65nano, there will be a decline in unit production. They not only need good yied on 65 nano but a ramp of Fab36 to 20,000wsm from the current 10,000wsm. Fab 30 will actually see declining production in 2007 from the 25,000wsm at 200 millimeter in 2006 as they retrofit for 65nano at 300 millimeter in 2008.
AMD will have a difficult time making products in 2007 at price and/or quanity anyone would want to buy. They are in real trouble.



Doom and gloom. Doom and gloom. If they can make 3X more chips and have 50% yields, that's still 1.5x more chips, maybe more since 90nm is probably not at 100%.

Remember that the quad-core chips are physically larger then the dual core 90nm....

Assuming redwood's facts are true (50% is rediculous, I don't think it's that low..), then AMD would probably be making less chips/wafer on the 300mm wafers, and may break even when you compare the 200mm to the 300mm ones... I do not have the necessary intelligence, information (do not know transistor count of the quad-cores), or inclination to to do the math, so I am just estimating...

AMDs best selling retail chips are now low-end Semprons and Turion. Since the server sphere is lower volume they don't need to make as many; especially since there will still be 90nm Opterons around for entry level prices.

Most of the 65nm stock in 07 will be X2, Sempron and Turion. Agena and Agena FX are not scheduled until Sep or so while Brisbane is due before the end of the year while 65nm Semprons are due in time for Vista.

939 is gone and as Fab 36 ramps, Chartered can also ramp 65nm. 90nm X2 will still make a nice HTPC to entry-high -level PC.

Intel still has yet to actually reach EOL on all NetBurst and is due to still be 70% "fire sale" until at least Q207. Ramping 65nm fabs to 45nm just allows removal of PD SKUs(hopefully the Core 2 fabs aren't affected) but not more volume for mainstream Core 2 (resources wise).

I mean right now Intel has 95 SKUs on Newegg, 2/3 of which are under $300, 40 are under $200. Intel needs a lot more growth than the market to really catch up as only 5 of the 95 are COre 2 derivatives.

Talk about taking a bath. I guess it's a good thing they kept their money-making stranglehold for so long.
December 4, 2006 2:43:42 AM

It was a live web cast last week by the CFO. I would take "almost all" to mean everyone but intel. Everyone can make their own interpretation of well below 50%.
December 4, 2006 2:50:03 AM

Quote:
It was a live web cast last week by the CFO. I would take "almost all" to mean everyone but intel. Everyone can make their own interpretation of well below 50%.



That's just a slightly ambiguous statement, especally without the COVETED LINK showing trends or graphs or something.
December 4, 2006 3:23:42 AM

Quote:
Well, again I am not a fan of using the Inq. as a source, but they did put up a revised 'price cut' article. It is interesting --
http://www.theinquirer.net/default.aspx?article=36129

If this data is true, which from the Inq is immediately questionable, but if true then Woodcrest/Clovertown is having a bigger impact on their server sales that what AMD is alluding to in their press reports.

Not only is their fastest DT processor dropping below the 300 mark, but they are cutting prices on their high end server products as well.... this is going to kill ASPs, they are in for a rough ride.... even if they do sell everything they make, the margins are going to be under tremendous pressure.

Jack


Sorry to bring this discussion back to a post on the first page everyone, but I've got a question for Jack(or anyone else who may be able to shed som light). It's been a while since I've seen discussion about AMD taking CapX against their new fab and equip. Were you able to find out if they ever took any for Q2 and/or Q3? If not, what kind of an affect will it have on Q4 assuming they still have the brunt of it to claim? I'm looking at it from the perspective of your last statement above (bolded) and wondering how AMD will manage to turn a profit. ???
December 4, 2006 11:21:37 AM

Quote:
AMDs best selling retail chips are now low-end Semprons and Turion. Since the server sphere is lower volume they don't need to make as many; especially since there will still be 90nm Opterons around for entry level prices.

Most of the 65nm stock in 07 will be X2, Sempron and Turion. Agena and Agena FX are not scheduled until Sep or so while Brisbane is due before the end of the year while 65nm Semprons are due in time for Vista.

939 is gone and as Fab 36 ramps, Chartered can also ramp 65nm. 90nm X2 will still make a nice HTPC to entry-high -level PC.

Intel still has yet to actually reach EOL on all NetBurst and is due to still be 70% "fire sale" until at least Q207. Ramping 65nm fabs to 45nm just allows removal of PD SKUs(hopefully the Core 2 fabs aren't affected) but not more volume for mainstream Core 2 (resources wise).

I mean right now Intel has 95 SKUs on Newegg, 2/3 of which are under $300, 40 are under $200. Intel needs a lot more growth than the market to really catch up as only 5 of the 95 are COre 2 derivatives.

Talk about taking a bath. I guess it's a good thing they kept their money-making stranglehold for so long.


The market takes time to catch up... C2D may be the fastest chip around, but that does not mean that the average joe six-pack really gives a damn wether MS outlook takes 9 seconds to open rather than 10...

By the time 65nm C2Ds are in the average consumer's hands, C3D or P5 will be out.. Just as P4's are pretty much in everyone's computers now. The same goes for AMD, the XP series is probably the most saturated right now.

This is really rudimentary stuff, perhaps the quote below me is correct?

Quote:
You are more on the ball than BaronMatrix is because you are correct.

Baron does not understand anything about computing, CPUs, or the industry in general -- his arrogance prevents him from learning.
December 4, 2006 11:48:00 AM

Quote:
his arrogance prevents him from learning.

Not his arrogance, but his stupidity.
December 4, 2006 2:11:41 PM

Quote:
his arrogance prevents him from learning.

Not his arrogance, but his stupidity.


I'm smart enough to know that the MAJORITY of 65nm chips WILL NOT BE Quad in 07.
December 4, 2006 7:24:37 PM

Quote:
his arrogance prevents him from learning.

Not his arrogance, but his stupidity.


I'm smart enough to know that the MAJORITY of 65nm chips WILL NOT BE Quad in 07.

You're actually right about that. At last check, Intel intends to only have two QC models for almost if not the entire year. As for K8L/Rev H, it will even have single-core variants (theoretically), and most definitely DC variants. QC is only the high end-or is only supposed to be.

Which makes me wonder (almost scared actually) what the (effective) yield rates for a 65nm QC 2.9GHZ AMD processor are going to be...but that's beside the point.
December 4, 2006 7:53:25 PM

Exactly. It would appear that Intel sees a large benefit to building quad cores with two dual die cores even at 45 nano. AMD is nuts to make a quad on one core. In the past they have not even shared the cache between the cores which is one benefit of all on one die.
Getting back to the emergency AMD price cuts. Is this an indication that sales through Dell are less than expected?
December 4, 2006 8:36:36 PM

Quote:
his arrogance prevents him from learning.

Not his arrogance, but his stupidity.


I'm smart enough to know that the MAJORITY of 65nm chips WILL NOT BE Quad in 07.
BRAVO....you should be proud because of that. :roll:
December 4, 2006 10:54:37 PM

Quote:
AMD was 15th in the rankings in silicon before the ati aquisition.They are now at 7th.

I dont see any emergency,i see a bs news site having fun at the enthusiasts expense."here fishy fishy"

"Get the ball fido ,fetch boy"

I hope the inquirer goes offline someday.permanently.


Exactly. It could also be so that they can put a 5-10% premium on 65nm and still be at the current price. The same story about rankings said that including ATi revenues, AMD will see 90% more profit this year.
8O

I think they'll be just fine. Just being No.7 means at least $2B in credit. Maybe $5B.
December 5, 2006 1:45:41 AM

Quote:
they should build a 45nm fab with that credit.


They've already started one in upstate NY. I guess they can pay for it now.
December 5, 2006 4:01:58 AM

Jack, it's been a while since I've seen discussion about AMD taking CapX against their new fab and equip. Were you able to find out if they ever took any for Q2 and/or Q3? If not, what kind of an affect will it have on Q4 assuming they still have the brunt of it to claim? I'm looking at it from the perspective of your last statement above (bolded) and wondering how AMD will manage to turn a profit. ???
December 5, 2006 5:37:02 AM

This is my understanding of how capx works. I'm not an accountant, but had one explain it to me.
The aquisition of ATI has made capx a non issue.
Primarily capx becomes an issue when you have a depreciation greater than the reduction in secured debt. That has never been an issue with AMD.
However, to the unenlightened, capx makes it appear that the net worth of a company is reduced by the depreciation. (fab 30 has been totally depreciated, and so is valued at $0. I wish they'd sell it to me at that price).
At this point, AMD has a large debt on the books for the ATI purchase.
They can write the depreciation of fab 36 off against the debt reduction on ATI.
December 5, 2006 5:43:34 AM

AMD has thier new price list up, if you want to check it out.
Seems to reflect a reduction in production costs.
December 5, 2006 2:15:37 PM

Quote:
they should build a 45nm fab with that credit.


They've already started one in upstate NY. I guess they can pay for it now.

From what I understand they have not yet broken ground.... and AMD has a letter of intent with a withdrawl clause.... they are hell long gone away from build that fab..... they don't have the capital at this point, they just blew their wad on a graphics company.


You're wrong, NY State is giving them more than $1B. And that graphics company purchase made them teh No7 semi manuf and raised their revenue 90%.

You forget that ATi does have a reasonably good revenue. I like to say imagine paying on a loan with what you bought. That's the case here. AMD can forego most of the profit from ATi and pay on the loan.

65nm will enable them to seriously cut costs, raise supply and compete with even lower prices.
!