AMD admits to 4770 supply problems

[thilanliyan]

http://www.hexus.net/content/item.php?item=18876

Not really a surprise given what we've heard about TSMC's 40nm issues.

 

[JEDIYoda]

I Love my 4770!!!

 

[Idontcare]

I am told they have a line-down situation, the hotfixes they were pinning their hopes on didn't deliver the needed results so the line is down while they expedite second round hotfixes thru the line to confirm a valid solution before restarting the line again. This may sound drastic, and to be sure it is, but this happens semi-frequently when critical/fatal flaws are uncovered in fab situations so it isn't as dire or unusual as it might sound.

 

[nitromullet]

AMD didn't actually 'admit' to anything, they just said that they "couldn't foresee the tremendous enthusiasm from the marketplace for the ATI Radeon HD 4770".

 

[Keysplayr]

Originally posted by: nitromullet
AMD didn't actually 'admit' to anything, they just said that they "couldn't foresee the tremendous enthusiasm from the marketplace for the ATI Radeon HD 4770".

Yah. Positive spin. Even if they could foresee it, they couldn't have done anything about it. It is what it is. TSMC will sort things out hopefully soon. This is going to delay ALL successive 40nm parts. Unless UMC can deliver.

 

[Phynaz]

Originally posted by: Idontcare
I am told they have a line-down situation, the hotfixes they were pinning their hopes on didn't deliver the needed results so the line is down while they expedite second round hotfixes thru the line to confirm a valid solution before restarting the line again. This may sound drastic, and to be sure it is, but this happens semi-frequently when critical/fatal flaws are uncovered in fab situations so it isn't as dire or unusual as it might sound.

Don't line restarts take on the order of weeks?

 

[OCGuy]

I know it is Apples and Oranges, but this seems strange seeing how Intel and AMD didnt appear to have any major issues with this node (45nm)......

This is bad news for all of us....

 

[Wreckage]

Originally posted by: OCguy

This is bad news for all of us....

Well maybe not all of us....

http://www.techreport.com/discussions.x/17067

 

[Keysplayr]

Originally posted by: Wreckage

Originally posted by: OCguy

This is bad news for all of us....

Well maybe not all of us....

http://www.techreport.com/discussions.x/17067

"The company goes on to say we should see the new GPUs in "over 100" notebook designs by the end of the year."

Yes, maybe for all of us. You missed this little chestnut.

 

[Idontcare]

Originally posted by: Phynaz

Originally posted by: Idontcare
I am told they have a line-down situation, the hotfixes they were pinning their hopes on didn't deliver the needed results so the line is down while they expedite second round hotfixes thru the line to confirm a valid solution before restarting the line again. This may sound drastic, and to be sure it is, but this happens semi-frequently when critical/fatal flaws are uncovered in fab situations so it isn't as dire or unusual as it might sound.

Don't line restarts take on the order of weeks?

It really depends on how many mask levels the engineering lots needed to cycle thru before getting to a point in the line where they can be inline probed to generate the data needed to make a decision regarding baseline change before resuming the line again. I've seen linestops last a brief as 6 hrs, and last as long as 2+ weeks. They rarely last longer than 2 weeks though as by that point you really have no choice (financially) but to run wafers even if the yields are just going to be 10% for the next month.

 

[OCGuy]

Originally posted by: Idontcare

It really depends on how many mask levels the engineering lots needed to cycle thru before getting to a point in the line where they can be inline probed.....

I was inline probed by a guy in a mask onc.......wait, this isnt Off Topic? Sorry.

 

[Phynaz]

Originally posted by: Idontcare

Originally posted by: Phynaz

Originally posted by: Idontcare
I am told they have a line-down situation, the hotfixes they were pinning their hopes on didn't deliver the needed results so the line is down while they expedite second round hotfixes thru the line to confirm a valid solution before restarting the line again. This may sound drastic, and to be sure it is, but this happens semi-frequently when critical/fatal flaws are uncovered in fab situations so it isn't as dire or unusual as it might sound.

Don't line restarts take on the order of weeks?

It really depends on how many mask levels the engineering lots needed to cycle thru before getting to a point in the line where they can be inline probed to generate the data needed to make a decision regarding baseline change before resuming the line again. I've seen linestops last a brief as 6 hrs, and last as long as 2+ weeks. They rarely last longer than 2 weeks though as by that point you really have no choice (financially) but to run wafers even if the yields are just going to be 10% for the next month.

Okay, I was under the impression that when a line is stopped the wafers that are WIP have to get tossed.

 

[Idontcare]

Originally posted by: Phynaz

Originally posted by: Idontcare

Originally posted by: Phynaz

Originally posted by: Idontcare
I am told they have a line-down situation, the hotfixes they were pinning their hopes on didn't deliver the needed results so the line is down while they expedite second round hotfixes thru the line to confirm a valid solution before restarting the line again. This may sound drastic, and to be sure it is, but this happens semi-frequently when critical/fatal flaws are uncovered in fab situations so it isn't as dire or unusual as it might sound.

Don't line restarts take on the order of weeks?

It really depends on how many mask levels the engineering lots needed to cycle thru before getting to a point in the line where they can be inline probed to generate the data needed to make a decision regarding baseline change before resuming the line again. I've seen linestops last a brief as 6 hrs, and last as long as 2+ weeks. They rarely last longer than 2 weeks though as by that point you really have no choice (financially) but to run wafers even if the yields are just going to be 10% for the next month.

Okay, I was under the impression that when a line is stopped the wafers that are WIP have to get tossed.

Yeah the WIP which has already progressed past the point in the flow at which the fatal flaw is believed to be introduced (the process responsible for causing leakage, or failing to prevent it) would be scrapped.

And all the WIP prior to that point will continue to process up to a "safe" point in the flow prior to the process that is expected to be altered pending results from the engineering splits.

But what I gather from your post is that you are referring to the time lag at wafer outs for the fab that stems from a line-stop, not just the line-stop time itself. (which is what I was talking about)

So yeah, if you are asking "how long will it take for wafers to flow out of the fab again" after a critical linestop (no reworks, etc) then it will be 2-3 weeks minimum as the point in the flow for this issue is in the FEOL. So once that issue is fixed and wafers are allowed to start processing thru it again you have a good 2-3 weeks (more likely 4-5 weeks but I'm sure TSMC will expedite the lots to minimize damage to AMD) before they start hitting the end of the BEOL.

 

[Phynaz]

Yep, you got it that's exactly what I meant.

Thanks for the insight.

 

[bgeh]

Speaking of which:



Yields of Taiwan Semiconductor Manufacturing Company's (TSMC's) 40nm process will not see any significant improvement until August 2009, according to sources at graphics card makers.

Since TSMC's yield rates at 40nm are currently less than 25%, they have seriously impacted AMD's shipment schedule for its ATI Radeon HD 4770 graphics chips, the sources noted, adding that TSMC is offering preferential quotes to AMD as compensation.

TSMC and AMD both declined to comment.

Nvidia has said previously that its 40nm GPU production at TSMC will be limited to OEM parts initially. 40nm GPUs destined for the channel market will be introduced when the process is more mature, likely around the end of the year.

http://www.digitimes.com/news/a20090619PD210.html