llano supply issues ?

[Soulkeeper]

it seems like the A8-3850 has been out of stock for 2 weeks at newegg ..
A8-3800/3600 seem like they've never really been in stock
and still no A8-3870/3670

Something tells he it's not because these are that great and selling out fast ...
Has AMD/GF really scaled back production this much ?

thanks

 

[Phynaz]

Yes, AMD/GF is having major supply issues. I think they even mentioned it in their last earnings release and SEC filings.

 

[exar333]

Isn't this why the AMD/Apple deal fell through? AMD has had supply issues for some time.

 

[superccs]

What a shame, I am sure people would love to put together machines based on that sillicon... GF get your sh!t together.

 

[jhu]

There was a recent extremetech article about the yields being very low, but AMD only has to pay for chips that work, so GF is losing a lot of money as a result.

 

[Idontcare]

There was a recent extremetech article about the yields being very low, but AMD only has to pay for chips that work, so GF is losing a lot of money as a result.

IIRC the skuttlebutt right now is that those specific contract terms (pay per yielding chip instead of paying per wafer) expire at the end of this year and GloFo was not willing to extend that favorable contract into 2012.

So 2012 could be a very rocky situation for AMD when the new foundry contracts come active (if the rumors are true).

 

[jacktesterson]

I bought my A8-3850 from newegg (like 2 weeks ago too )

 

[Vesku]

Yields are low enough that I would not be surprised to see some major performance and feature compromises made in their next APU, Trinity, in order to improve yields.

 

[Idontcare]

Yields are low enough that I would not be surprised to see some major performance and feature compromises made in their next APU, Trinity, in order to improve yields.

That would be tough IMO only because it seems like the yield issues we are seeing are the functional yield kind, not the parametric yield kind.

Parametric yield loss can be addressed by respins (if desired), binning differently (selling lower clocked parts).

But functional yield loss basically requires die harvesting or good old fab improvements if you want to increase yields.

I remember when we (TI) were having serious functional yield issues on our 90nm process for SUN's ultrasparc chips at the time. We were yielding maybe 2-3 chips per wafer at the time. SUN only paid us per chip, same as the AMD situation, and it wasn't a matter of SUN accepting slower binned CPU's because the other 80 chips on the wafer were simply dead, truly dead from functional yield loss and not a parametric yield issue.

 

[blckgrffn]

That would be tough IMO only because it seems like the yield issues we are seeing are the functional yield kind, not the parametric yield kind.

Parametric yield loss can be addressed by respins (if desired), binning differently (selling lower clocked parts).

But functional yield loss basically requires die harvesting or good old fab improvements if you want to increase yields.

I remember when we (TI) were having serious functional yield issues on our 90nm process for SUN's ultrasparc chips at the time. We were yielding maybe 2-3 chips per wafer at the time. SUN only paid us per chip, same as the AMD situation, and it wasn't a matter of SUN accepting slower binned CPU's because the other 80 chips on the wafer were simply dead, truly dead from functional yield loss and not a parametric yield issue.

I really appreciate your insight and first hand experience with these matters Thank you for being here and sharing!

 

[Soulkeeper]

excellent comments, glad to see I wasn't the only one thinking these things.

 

[LoneNinja]

They lowered expected revenue months ago for the rest of the year due to poor yields on 32nm and a tight supply. Bulldozer, being the under performing failure it is, is also in tight supply for this reason. Sadly even if AMD had the performance crown right now, they wouldn't be making much money due to supply issues.

 

[Arachnotronic]

AMD needs to move to TSMC. GlobalFoundaries sucks.

 

[exar333]

IIRC the skuttlebutt right now is that those specific contract terms (pay per yielding chip instead of paying per wafer) expire at the end of this year and GloFo was not willing to extend that favorable contract into 2012.

So 2012 could be a very rocky situation for AMD when the new foundry contracts come active (if the rumors are true).

At some point, is there even a point to selling a chip with yields like that? If you only get (for arguments sake) 10% yields, and the 'bad' 9 out of 10 chips costs you more than you can make in profit on the one functional CPU, why bother?

 

[sm625]

They should have made llano a native 5 core design so yields on quads would be good enough.

 

[Rvenger]

I had to RMA my Llano twice. One was crashing with core errors, the other had video artifacts. Funny thing is, they worked fine at first, then they degraded shortly after use. I never even overclocked it! I also notice when you start to have everything integrated, it gets very hard to diagnose and pinpoint issues. I was so frustrated, I RMA'd the mobo and CPU.



I am going to call our AMD rep here at work and I am going to see if this is a common occurence.

 

[antisocialmunky]

I would think that the yields aren't that completely terrible and its more demand for laptop being filled first and the rest getting sold as desktop.

 

[Rast]

I had to RMA my Llano twice. One was crashing with core errors, the other had video artifacts. Funny thing is, they worked fine at first, then they degraded shortly after use. I never even overclocked it! I also notice when you start to have everything integrated, it gets very hard to diagnose and pinpoint issues. I was so frustrated, I RMA'd the mobo and CPU.



I am going to call our AMD rep here at work and I am going to see if this is a common occurence.

what is that video artifacts problem? i have also somehow problems with my a8-3850 rig. my video animation is someone very delayed. i will make soon a own topic about my problems with video etc.

 

[sm625]

My own experience with the A4 can best be described as disgusted. This could be the end of AMD. I think the professional reviews were incomplete at best...

 

[Phynaz]

At some point, is there even a point to selling a chip with yields like that? If you only get (for arguments sake) 10% yields, and the 'bad' 9 out of 10 chips costs you more than you can make in profit on the one functional CPU, why bother?

For one reason, contractual obligations.

 

[Arkaign]

AMD selling their fabs has been full of fail. It probably resulted in less than ideal products, given the additional hassles of trying to do re-spins without the efficiency of a top-down 100% control of the facility, and now it's f'ing them in the arse as far as getting product that people want to buy into the customer's hands.