All Barcelona Shipments Halted

[Viditor]

Originally posted by: Accord99

Originally posted by: Viditor
Firstly, I said 10% of the total market, not last quarter's sales.

The most optimistic estimates indicate AMD will be selling in the hundreds of thousands range at best. Nor is the video card market for >$200 cards that large to support millions of card sales per quarter.

Whose estimates? Were they talking US or worldwide?
The market for add-in graphics cards (not on-board chips) last quarter was 25.74 million cards sold (8.58 million cards/month). Revenue for those cards was $6.617 Billion.
That means that the average cost of the cards was $257...

My estimate is that HD38xx should give AMD an additional $300 million in revenue while cutting costs on GPUs by at least 40%...

$300 million is more than the entire graphics divisions revenue in the previous quarter. And far more importantly, AMD's problem is not revenue, it's profit.

You are correct that AMD's revenue for all graphics in Q3 was only $252 Million...
However, the VAST majority of that was for on-board graphics chips (where revenue is only a third of add-in cards). Remember that AMD hasn't had a competitive card for a year now. By comparison, in Sept of 2006, ATI (pre-merger) had revenues at near $520 Million for their Q4 (which was AMD's Q3)...

 

[eRacer]

Originally posted by: Viditor
My estimate is that HD38xx should give AMD an additional $300 million in revenue while cutting costs on GPUs by at least 40%...

I'd say the $300 million figure is off by at least a factor of ten for Q4. If we assume AMD ships 300,000 38xx GPUs this quarter (which is a large number for a launch), and we assume a healthy ASP of $100 then HD 38xx revenue will only be $30 million.

 

[smthmlk]

As more pieces of the Hindenburg catch fire and explode and the crew jumps ship, it's refreshing to see someone explain away each incident as 'ordinary.'

 

[myocardia]

Originally posted by: NXIL

Umm, take a business class. Manufacturers don't sell products for less than it costs the company to make them.

Dear Myo,

well, not always: inkjet printers are sold at below cost...they make money on the ink, of course.....right now DDR2 is being sold below cost in some instances, oversupply; cars: lots of small cars are sold at a loss, even factoring in reduced fines for failing to meet fuel efficiency standards. (Of course, GMs numbers look beyond atrocious, so here's another company which may not survive hemorrhaging red ink).

Dear NXIL,

When you take one sentence out of a two sentence quote, it's the epitome of taking something out of context. Go back and read the second sentence, and you'll understand. BTW, CAFE is the only reason some American car companies will sell their best MPG cars below cost-- so they can continue to sell their worst gas mileage, but highest profit SUV's.

 

[bryanW1995]

Originally posted by: Stoneburner
I think they meant to say the shipments were halted because nobody was buying.

LOL

 

[SickBeast]

It's really sad what's happening to AMD.

Not many users even need 4 cores at this point, and with this 'bug', the X2 chips are pretty much performing at parity with the Phenom clock-for-clock, plus they're cheaper, plus they can clock higher, plus they're stable and proven.

If the chips have indeed been recalled, hopefully the problems will be rectified. After the delays the chip went though, this type of problem is completely unacceptable IMO.

I can see why AMD's marketing director quit. He should not have slated a release date for Barcelona until he had a better idea of what they were dealing with. In the end, it looks like they rushed it out to meet some sort of hard deadline that they set to co-incide with the chipset/graphics launch.

If AMD were smart, they would release only one type of product per quarter, and make it a good one. By releasing all 3 at once, they probably stretched their already-limited engineering resources too thin.

 

[nonameo]

Originally posted by: SickBeast
It's really sad what's happening to AMD.

Not many users even need 4 cores at this point, and with this 'bug', the X2 chips are pretty much performing at parity with the Phenom clock-for-clock, plus they're cheaper, plus they can clock higher, plus they're stable and proven.

If the chips have indeed been recalled, hopefully the problems will be rectified. After the delays the chip went though, this type of problem is completely unacceptable IMO.

I can see why AMD's marketing director quit. He should not have slated a release date for Barcelona until he had a better idea of what they were dealing with. In the end, it looks like they rushed it out to meet some sort of hard deadline that they set to co-incide with the chipset/graphics launch.

If AMD were smart, they would release only one type of product per quarter, and make it a good one. By releasing all 3 at once, they probably stretched their already-limited engineering resources too thin.

But they had to drop a massive hardware bomb for great media coverage!

Just messin'

 

[NXIL]

Hi Myo,

well, the second sentence didn't make much sense to me:

That means that if they're selling all that they can possibly make of a certain product, that product is generating profits.

Not true....there are many other factors other than % of production capacity.....and, not to be an annoying logicator, that would mean that if their yields were low, and they could only crank out 1,000 Phenoms a month, they would sell them all at a profit.....there are economies of scale, marginal return, etc.....

But, I did not intend to quote you out of context/pester/annoy--it's just that a lot of big companies can take losses for a while, and recover.

As for the high profit SUVs: GM/Ford not doing so great....Honda/Toyota are....of course this is due to high gas prices--but it doesn't look like that is goign to turn around any time soon. I think Honda/Toyota had some leaner times when gas prices were lower--I remember reading something about Honda not making bigger cars back a few years ago when gas was relatively cheaper--Honda designer said they were doing the right thing, and that time was on their side.

Cheers--and by the way I think your posts are always excellent--well written, helpful...thx.

NXIL

 

[Viditor]

Originally posted by: eRacer

Originally posted by: Viditor
My estimate is that HD38xx should give AMD an additional $300 million in revenue while cutting costs on GPUs by at least 40%...

I'd say the $300 million figure is off by at least a factor of ten for Q4. If we assume AMD ships 300,000 38xx GPUs this quarter (which is a large number for a launch), and we assume a healthy ASP of $100 then HD 38xx revenue will only be $30 million.

Actually, 300,000 is a very small number...
Remember that purchases from foundaries are made by the wafer, not the chip. Even at say 2000 wafers/week (to give you an idea of proportions, TSMC is currently ramping their 300mm to 100,000 wafers/month), that's 2.8 Million candidate dice per month.
If it's as low as a 50% yield (doubtful), then that's still 1.4 Million chips a month or 4.2 Million for the quarter.

 

[bradley]

Originally posted by: SickBeast
It's really sad what's happening to AMD.

Not many users even need 4 cores at this point, and with this 'bug', the X2 chips are pretty much performing at parity with the Phenom clock-for-clock, plus they're cheaper, plus they can clock higher, plus they're stable and proven.

Can you show me the sites that are benching the K8 even with the K10? If anything I'm seeing it trend towards an approximate 10-15% speed increase, especially running on an AMD 790 chipset.

 

[taltamir]

when they say they are loosing money it doesn't mean that they are selling below cost, that is RETARDED and would never be done by a real company.

What it really means is that they are selling above cost, but not by enough to cover operational expenses and or development costs... Not selling would mean that the operation costs and cost of developments are still there, but now they aren't making any money to cover them... selling at too low a margin means they are loosing LESS money they they would from not selling, but the company overall is loosing money... selling with healthy margins means that they are making enough profit on the chips to cover operational expenses and RnD.

 

[bfdd]

Originally posted by: taltamir
when they say they are loosing money it doesn't mean that they are selling below cost, that is RETARDED and would never be done by a real company.

What it really means is that they are selling above cost, but not by enough to cover operational expenses and or development costs... Not selling would mean that the operation costs and cost of developments are still there, but now they aren't making any money to cover them... selling at too low a margin means they are loosing LESS money they they would from not selling, but the company overall is loosing money... selling with healthy margins means that they are making enough profit on the chips to cover operational expenses and RnD.

Original Xbox's were sold below cost because MS wanted market share. Even the 360 and PS3 were sold below cost. Although no company in AMDs position where they're bleeding money with a lackluster product would be selling below cost. They need marketshare and they need money to stay alive. Hopefully they can work out their problems with the phenom.

 

[Keysplayr]

Originally posted by: Viditor

Originally posted by: eRacer

Originally posted by: Viditor
My estimate is that HD38xx should give AMD an additional $300 million in revenue while cutting costs on GPUs by at least 40%...

I'd say the $300 million figure is off by at least a factor of ten for Q4. If we assume AMD ships 300,000 38xx GPUs this quarter (which is a large number for a launch), and we assume a healthy ASP of $100 then HD 38xx revenue will only be $30 million.

Actually, 300,000 is a very small number...
Remember that purchases from foundaries are made by the wafer, not the chip. Even at say 2000 wafers/week (to give you an idea of proportions, TSMC is currently ramping their 300mm to 100,000 wafers/month), that's 2.8 Million candidate dice per month.
If it's as low as a 50% yield (doubtful), then that's still 1.4 Million chips a month or 4.2 Million for the quarter.

Then why are we seeing things like this?

300,000 is a small number? Ok, then where are they all?

 

[Viditor]

Originally posted by: keysplayr2003

Originally posted by: Viditor

Originally posted by: eRacer

Originally posted by: Viditor
My estimate is that HD38xx should give AMD an additional $300 million in revenue while cutting costs on GPUs by at least 40%...

I'd say the $300 million figure is off by at least a factor of ten for Q4. If we assume AMD ships 300,000 38xx GPUs this quarter (which is a large number for a launch), and we assume a healthy ASP of $100 then HD 38xx revenue will only be $30 million.

Actually, 300,000 is a very small number...
Remember that purchases from foundaries are made by the wafer, not the chip. Even at say 2000 wafers/week (to give you an idea of proportions, TSMC is currently ramping their 300mm to 100,000 wafers/month), that's 2.8 Million candidate dice per month.
If it's as low as a 50% yield (doubtful), then that's still 1.4 Million chips a month or 4.2 Million for the quarter.

Then why are we seeing things like this?

300,000 is a small number? Ok, then where are they all?

You mean where besides Europe, Asia, Australia, etc...? I have to say that there is better supply of the HD38xx cards here in Oz at launch than there ever was of any of the nVidia launches...
Then there's the OEMs...even though it has to go through design and system testing, do you really think Dell is going to stick with the XT series when these are available?
How many cards do you imagine Dell alone has ordered?

Also note that there are a ton of the 3850 cards in stock and at MSRP of $179...
Trying to guage the channel's supply worldwide by a single store's pricing of a single model is an effort in futility...

 

[Accord99]

Originally posted by: Viditor
You mean where besides Europe, Asia, Australia, etc...? I have to say that there is better supply of the HD38xx cards here in Oz at launch than there ever was of any of the nVidia launches...

The $200 video card market is small, a couple of hundred thousand cards would easily saturate that portion of the market for months. And what little profit they would make would be eaten up by AMD's losses in a matter of a week.

Then there's the OEMs...even though it has to go through design and system testing, do you really think Dell is going to stick with the XT series when these are available?

Dell never sold 2900XTs in their systems. The AMD cards used were 2400/2600s on a few entry level models mainly.

How many cards do you imagine Dell alone has ordered?

Probably none, it's likely Dell has already standardized on the 8800 GT as the performance mainstream option. HP likewise uses only Nvidia video cards on their performance desktops.

 

[eRacer]

Actually, 300,000 is a very small number...
Remember that purchases from foundaries are made by the wafer, not the chip. Even at say 2000 wafers/week (to give you an idea of proportions, TSMC is currently ramping their 300mm to 100,000 wafers/month), that's 2.8 Million candidate dice per month.
If it's as low as a 50% yield (doubtful), then that's still 1.4 Million chips a month or 4.2 Million for the quarter.

It isn't a small number for a new GPU on a new process targeting the high-end of the mainstream. To assume that practically overnight TSMC has dedicated 8% of their overall 300mm wafer production to the production of ATI HD 38xx GPUs on a new 55-nm process is quite a stretch to say the least.

 

[trajan2050]

Umm, take a business class. Manufacturers don't sell products for less than it costs the company to make them. That means that if they're selling all that they can possibly make of a certain product, that product is generating profits.




For your information, businesses sell products at a loss all the time, for a variety of reasons. AMD didn't generate huge losses because their sales were wonderfully profitable.Loss leaders, dumping to gain market share, liquidations of unpopular products are just some of the reasons companies sell products for less than the manufacturing cost.

 

[KingstonU]

Originally posted by: wwswimming

looks like this came out after the stock markets close.

http://quote.yahoo.com/q?s=AMD

be interesting to see what happens tomorrow.

they're still shipping Barcelona Phenom's.

OMG, And I can't believe how I was certain, certain that the stock price would never do anything like this.

But then again, I was also certain that Phenom would be The Return of the Jedi (to use a great analogy from Anand ) but it was instead more of The Empire Strikes Back, and Intel just cut off AMD's hand.

And now we wait, again, for Bulldozer to bring back some hope

 

[Viditor]

Originally posted by: Accord99

Originally posted by: Viditor
You mean where besides Europe, Asia, Australia, etc...? I have to say that there is better supply of the HD38xx cards here in Oz at launch than there ever was of any of the nVidia launches...

The $200 video card market is small, a couple of hundred thousand cards would easily saturate that portion of the market for months. And what little profit they would make would be eaten up by AMD's losses in a matter of a week.

Well, I have to ask if you have anything to back that up...
I did provide a link with numbers showing that the average price for add-in cards is $257 and that over 25 million were sold last quarter...but if you have something that disputes that I would certainly be interested.

Then there's the OEMs...even though it has to go through design and system testing, do you really think Dell is going to stick with the XT series when these are available?

Dell never sold 2900XTs in their systems. The AMD cards used were 2400/2600s on a few entry level models mainly.

And the 2600 was priced at launch for more than the 3850 is...$199 vs $179

 

[Viditor]

Originally posted by: eRacer

Actually, 300,000 is a very small number...
Remember that purchases from foundaries are made by the wafer, not the chip. Even at say 2000 wafers/week (to give you an idea of proportions, TSMC is currently ramping their 300mm to 100,000 wafers/month), that's 2.8 Million candidate dice per month.
If it's as low as a 50% yield (doubtful), then that's still 1.4 Million chips a month or 4.2 Million for the quarter.

It isn't a small number for a new GPU on a new process targeting the high-end of the mainstream. To assume that practically overnight TSMC has dedicated 8% of their overall 300mm wafer production to the production of ATI HD 38xx GPUs on a new 55-nm process is quite a stretch to say the least.

The 3850 actually targets the middle of the mainstream.
As to ordering a significant portion of TSMC capacity,

"AMD IS PREPARING an interesting offering. After ordering massive amounts of 55nm 300mm wafers from TSMC containing Number of The Beast (RV670 comes with 666M trannies), it turns out that RV670 was just the beginning.

There are several new models coming to market based on these new 55nm chips"

Inq article from Nov

 

[SexyK]

Originally posted by: myocardia

Originally posted by: ELopes580

Originally posted by: myocardia

Originally posted by: Skott
I dont see ATI keeping AMD afloat.

You really only have to click this link to find out why that statement couldn't be correct. It's definitely a good thing that they released the 3800 series when they did.

Your statement doesn't prove anything at all either...... It doesn't prove what profits are being made.

Umm, take a business class. Manufacturers don't sell products for less than it costs the company to make them. That means that if they're selling all that they can possibly make of a certain product, that product is generating profits.

As others have pointed out, it doesn't matter if they are making a profit on each card sold. All that matters is whether or not the company is profitable. It's all well and good if they make $1, $10, or even $100 per card, but it's going to take many, many, many cards sold at that level to come close to breaking even on the R&D, manufacturing, etc... costs.

 

[trajan2050]

AMD hits yet another multi year low.

ADV MICRO DEVICES (NYSE:AMD)

Last Trade: 9.37
Trade Time: 11:57AM ET
Change: 0.29 (3.00%

Never try and catch a falling dagger.

 

[Viditor]

Originally posted by: trajan2050
AMD hits yet another multi year low.

ADV MICRO DEVICES (NYSE:AMD)

Last Trade: 9.37
Trade Time: 11:57AM ET
Change: 0.29 (3.00%

The stock price quoting troll hits again...:roll:

When are you going to learn that it just doesn't matter...

 

[Phynaz]

Originally posted by: Viditor

Originally posted by: eRacer

Actually, 300,000 is a very small number...
Remember that purchases from foundaries are made by the wafer, not the chip. Even at say 2000 wafers/week (to give you an idea of proportions, TSMC is currently ramping their 300mm to 100,000 wafers/month), that's 2.8 Million candidate dice per month.
If it's as low as a 50% yield (doubtful), then that's still 1.4 Million chips a month or 4.2 Million for the quarter.

It isn't a small number for a new GPU on a new process targeting the high-end of the mainstream. To assume that practically overnight TSMC has dedicated 8% of their overall 300mm wafer production to the production of ATI HD 38xx GPUs on a new 55-nm process is quite a stretch to say the least.

The 3850 actually targets the middle of the mainstream.
As to ordering a significant portion of TSMC capacity,

"AMD IS PREPARING an interesting offering. After ordering massive amounts of 55nm 300mm wafers from TSMC containing Number of The Beast (RV670 comes with 666M trannies), it turns out that RV670 was just the beginning.

There are several new models coming to market based on these new 55nm chips"

Inq article from Nov

Funny you would quote somebody who you agreed was a clueless clown.

Anything to support an argument I guess.

 

[Viditor]

Originally posted by: Phynaz

Funny you would quote somebody who you agreed was a clueless clown.

Anything to support an argument I guess.

I agree, he is a clown...but that doesn't mean his data is wrong (in fact it usually isn't). What it means is that both his understanding of the data and his conclusions are wrong.