AMD's Cayman - HD 6900 series info delayed till week of Dec 13th

[bryanW1995]

No doubt that 6870/6850 is significantly more important to amd's bottom line, and you are correct that high end buyers are less likely to jump on nonexistent BF deals on 6970 (or 580 for that matter). however, many people plan their builds around BF and if 580 is the only high end card out there then we know what they'll buy. UNLESS amd has flooded the internet with rumors/speculation/marketing slides about how much better 6970 is than gtx 580. I can't believe we're all still talking about this. AMD has stated for years that they will target the high end with dual-gpu cards. the only difference this round is that nvidia forgot to execute so some non-amd people started thinking that they might go back after the single gpu crown. that isn't necessary as long as they get antilles out in a timely manner, there's just more pressure on cayman now b/c nvidia got their act together and executed this round.

 

[ModestGamer]

No doubt that 6870/6850 is significantly more important to amd's bottom line, and you are correct that high end buyers are less likely to jump on nonexistent BF deals on 6970 (or 580 for that matter). however, many people plan their builds around BF and if 580 is the only high end card out there then we know what they'll buy. UNLESS amd has flooded the internet with rumors/speculation/marketing slides about how much better 6970 is than gtx 580. I can't believe we're all still talking about this. AMD has stated for years that they will target the high end with dual-gpu cards. the only difference this round is that nvidia forgot to execute so some non-amd people started thinking that they might go back after the single gpu crown. that isn't necessary as long as they get antilles out in a timely manner, there's just more pressure on cayman now b/c nvidia got their act together and executed this round.

Well then we should all just wait for the announcement, the single GPU crown doesn;t mean squat and looking at how good xfire is on new amd 68xx cards I thknk the dual gpu proposition makes alot of sense.

even if the 6970 is = or a few % better or worse if its $399 or so. It'll sell well while AMD gets to matures its new architecture.

 

[bryanW1995]

Yes becuase black friday is when historically they sell alot of high dollar low volume graphics parts.

I'll give you an example.

I'm a car dealer so I'll use car sales, but many of you have probably seen this dynamic in action over the years. The week between xmas and new years is always our best week of the year. We plan around it, we market for it, etc etc. kind of like BF for many retailers. We get tons of bargain shoppers, we cater to them, we advertise towards them, and they buy lots of cars at discount prices. however, our average profits are typically the same or nearly so as they are the rest of the year. Why? because high end buyers also come out of the woodwork, just because the perception of a deal is there. If you're buying a $25,000 truck for $15,000 then yes, you could literally get the best deal of the year. But if you're looking at that $45,000 grand cherokee overland that is severely supply-constrained? Same deal that you'd get any other time of year, and on top of that you'd probably wait longer b/c so many other people are there buying cars at the same time. But it still happens. We might not get AS MUCH of a spike in our high end sales, but it's typically enough to offset our losses on the deep disounts for particular models.

I recognize that this isn't completely appicable to this situation as amd is clearly suffering from undersupply/overdemand already in their new 68x0 series, but you see my point that it's better to release on nov 22 instead of dec 22, right?

 

[ModestGamer]

Yes becuase to the dealer a high dollar premium model is a good profitizer, however from a manufacturers standpoint a premium model is also a high per vehicle profit. The issue is that you need to sell alot of chevy cavaliers to make the division pfoitable becuase cashflow makes it possiable to have a profit on the other division with the halo models.

Cashflow.

Cashflow


AMD did the right thing. they got a great performing batch of hardware out the door in time for the busy sales season and if romours I hear from various retailers are to be belived they are outselling nvidia hardware buy a large margin.

I'll give you an example.

I'm a car dealer so I'll use car sales, but many of you have probably seen this dynamic in action over the years. The week between xmas and new years is always our best week of the year. We plan around it, we market for it, etc etc. kind of like BF for many retailers. We get tons of bargain shoppers, we cater to them, we advertise towards them, and they buy lots of cars at discount prices. however, our average profits are typically the same or nearly so as they are the rest of the year. Why? because high end buyers also come out of the woodwork, just because the perception of a deal is there. If you're buying a $25,000 truck for $15,000 then yes, you could literally get the best deal of the year. But if you're looking at that $45,000 grand cherokee overland that is severely supply-constrained? Same deal that you'd get any other time of year, and on top of that you'd probably wait longer b/c so many other people are there buying cars at the same time. But it still happens. We might not get AS MUCH of a spike in our high end sales, but it's typically enough to offset our losses on the deep disounts for particular models.

I recognize that this isn't completely appicable to this situation as amd is clearly suffering from undersupply/overdemand already in their new 68x0 series, but you see my point that it's better to release on nov 22 instead of dec 22, right?

 

[Cheex]

If I was at the helm of the GPU division looking at the christmas season. I would be looking to make sure my midrange more affordable higher volume products had significant QOH becuase these high end parts while badass and good margin. aren't big movers.

My prediction.

Big prices cuts on 58xx cards starting next week.

Expect hard launch by dec13th.

A hard launch then would be awesome for AMD an would be great for the consumers during the holiday season. The price points would cause some good deals to become available and also fuel the excitement. However, a paper launch might give Nvidia a clear advantage in the holidays. We'll just have to wait and see.

 

[RussianSensation]

There is no need for HD5850/70 price cuts. With HD5870 at $270, HD6870 at $260-270 has no purpose when the 5870 handily beats it. At $210, the HD5850 is also better than the $200 6850. What AMD needs is to meet the demand for the 6850/70 series in the very important $180-250 price bracket.

 

[maddie]

They wouldn't necessarily issue a press-release statement about it to the WSJ but they would most likely issue a statement to their institutional holders and the major analysts...which would eventually find its way into our domain in a matter of hours.

But since that hasn't happened...hmmmm.

Was not online and on catching up wondered why everyone until you ignored this observation.

What about this?

Lead times have reached 20 weeks for power MOSFETs and small signal transistors, and 18 weeks for bipolar power devices and rectifiers, up from a more typical 10-12 weeks.

“When lead times enter the 20 week range, they indicate a major schism between component supply and demand, iSuppli analyst Rick Pierson said in a statement.
http://blogs.barrons.com/techtraderd...ppli-contends/

If true, AMD is getting screwed again for something outside their control. The Gods love Nvidia.

 

[RussianSensation]

Thanks for the link maddie. Looks like there is a definite shortage of components/ie. un-forecasted demand issue.

 

[exar333]

Was not online and on catching up wondered why everyone until you ignored this observation.

What about this?

Lead times have reached 20 weeks for power MOSFETs and small signal transistors, and 18 weeks for bipolar power devices and rectifiers, up from a more typical 10-12 weeks.

“When lead times enter the 20 week range, they indicate a major schism between component supply and demand, iSuppli analyst Rick Pierson said in a statement.
http://blogs.barrons.com/techtraderd...ppli-contends/

If true, AMD is getting screwed again for something outside their control. The Gods love Nvidia.[/QUOTE]

Yeah, 40nm went so smoothly for NV and was such an issue for AMD. *eyeroll*

 

[maddie]

Was not online and on catching up wondered why everyone until you ignored this observation.

What about this?

Lead times have reached 20 weeks for power MOSFETs and small signal transistors, and 18 weeks for bipolar power devices and rectifiers, up from a more typical 10-12 weeks.

“When lead times enter the 20 week range, they indicate a major schism between component supply and demand, iSuppli analyst Rick Pierson said in a statement.
http://blogs.barrons.com/techtraderd...ppli-contends/

If true, AMD is getting screwed again for something outside their control. The Gods love Nvidia.[/QUOTE]

Yeah, 40nm went so smoothly for NV and was such an issue for AMD. *eyeroll*

Are you serious?

Nvidia screwed themselves at 40nm. No one made the design mistakes but they themselves. Even their CEO said as much.

AMD had manufacturable designs, but TSMC could not produce enough.

Do you really see the situations as the same?

 

[GaiaHunter]

Yeah, 40nm went so smoothly for NV and was such an issue for AMD. *eyeroll*

AMD was still supply constrained and the 4770 was a disaster.

And that is without speculating how many more sps Cypress has for yields purpose that forced AMD to cut somewhere else.

 

[AnandThenMan]

AMD was still supply constrained and the 4770 was a disaster.

Why was the 4770 a disaster? I was hard to fine, but no one cared because other models were readily available. And it was a very important proving ground chip for AMD that taught them how to get excellent yields out of TSMC's craptastic process. So from that perspective, it was wildly successful.

 

[digitaldurandal]

I think one issue that is worth noting is that if the issue really is MOSFET component related they could have used the MOSFETs available to create 69x0 but chose not to, instead building 68x0 and 5xxx

 

[AnandThenMan]

I think one issue that is worth noting is that if the issue really is MOSFET component related they could have used the MOSFETs available to create 69x0 but chose not to, instead building 68x0 and 5xxx

It's possible. But don't you think AMD would reallocate say 1000 batches to make some 69xx cards so they could launch as scheduled? Then have shelf availability a couple of weeks later? A soft launch, paper launch, whatever.

Maybe AMD didn't want a paper-ish launch I don't know.

 

[Silverforce11]

They have enough for a launch, AIBs have stock, rooms full of cards actually. Why they aren't launching earlier, all speculation.

Edit: What i've heard from various AIBs is they have been sitting on stocks for awhile now. 22nd is the AIB launch briefing. Other than that, they still dont know what AMDs plans are.

 

[bryanW1995]

how big would you say the room is? is it a large room? small room? medium sized? where are these hypothetical "room(s)"? Maybe you could burnish your credentials just a smidgen if you could actually supply us with a picture.

Any comment on your "magic driver" for 68x0 claim? since amd clearly isn't devoting their resources to manufacturing high end gpus in a timely manner, they must all be busy writing drivers for the last cards released before the new high end. Right?

 

[AnandThenMan]

how big would you say the room is? is it a large room? small room? medium sized? where are these hypothetical "room(s)"? Maybe you could burnish your credentials just a smidgen if you could actually supply us with a picture.

Any comment on your "magic driver" for 68x0 claim? since amd clearly isn't devoting their resources to manufacturing high end gpus in a timely manner, they must all be busy writing drivers for the last cards released before the new high end. Right?

Huh??? :\

 

[ginfest]

Huh??? :\

He's looking for proof/sources to some of the posters wild assertions!!
Is it really that hard to comprehend

 

[Douglar]

Was not online and on catching up wondered why everyone until you ignored this observation.

What about this?

Lead times have reached 20 weeks for power MOSFETs and small signal transistors, and 18 weeks for bipolar power devices and rectifiers, up from a more typical 10-12 weeks.

Isn't that article kind of old? It's like 20 weeks old. Oh wait. Never mind. Parts that were ordered on 7-2 would be arriving on 11-19 if the 20 week lead time is right. But I though DrMosfet's had a shorter lead time for some reason.

 

[ModestGamer]

Isn't that article kind of old? It's like 20 weeks old. Oh wait. Never mind. Parts that were ordered on 7-2 would be arriving on 11-19 if the 20 week lead time is right. But I though DrMosfet's had a shorter lead time for some reason.

figure in time to run the boards, box and ship them dec 13th sounds ver reasonable.

 

[GaiaHunter]

Why was the 4770 a disaster? I was hard to fine, but no one cared because other models were readily available. And it was a very important proving ground chip for AMD that taught them how to get excellent yields out of TSMC's craptastic process. So from that perspective, it was wildly successful.

AMD cared. They could have sold near 4850 performance with a smaller chip.

And of course it was a disaster because the yields sucked.

While you can call it a success, AMD would have much preferred for the 4770 yields to be high to begin with. It is like saying GF100 was a success because of whatever successors of it in 40nm have higher yields.

 

[AnandThenMan]

Comparing the 4770 to GF100, seriously? AMD lost basically no revenue due to low supply of the 4770, and it was designed to be a testbed card anyway. Horrible comparison.

 

[GaiaHunter]

Comparing the 4770 to GF100, seriously? AMD lost basically no revenue due to low supply of the 4770, and it was designed to be a testbed card anyway. Horrible comparison.

Just because the 4770 wasn't critical for AMD doesn't make the card a success (as a consumer product, as a testbed it was a success as it detected problems of the 40nm process) and doesn't mean that 40nm was trouble free for AMD.

Both GF100 and RV740 had yields and leakage problems.

Sure, AMD gained critical knowledge, but that doesn't mean Cypress didn't suffer design changes due to the 40nm problems.

If you want to compare company strategies, that is another thing and I wasn't focusing into that.

 

[Dark_Archonis]

So where are all of you AMD supporters that claimed the 69xx cards would definitely not be delayed? Time for you to eat some crow and admit you were wrong.

 

[Nemesis 1]

I right here. But never said anything about the 6900 series being late . As I never knew AMD had set an official launch date . A link from amd sighting an official launch date would be helpful . Than I could appauled your wading around is a sest pool of NV outraged fanboyism over a delay . Oh gosh the 6900 series won't make it for Black Friday . Its Armageddon . Make sure your around at launch day . I want to here your comments on that day .