what is a "ramp issue"?

[taltamir]

From this article: http://www.xbitlabs.com/news/cpu/di..._APUs_Shipments_Delays_of_Bulldozer_CPUs.html

The less-than-forecasted preliminary third quarter 2011 revenue results are primarily due to 32nm yield, ramp and manufacturing issues at Globalfoundries in its Dresden, Germany factory

I know what yield issues mean, what are ramp issues? manufacturing issues?

 

[MagnusTheBrewer]

Increasing production, as in "ramping up" 32nm chip production.

 

[taltamir]

Increasing production, as in "ramping up" 32nm chip production.

Thank you.

what does that involve? what kind of issues could they have with it?

 

[jiffylube1024]

Low yields, limited amount of 32nm-capable tools/equipment compared to preexisting 40nm/65nm capacity... You name it.

There's all kinds of issues that can limit ramping up capacity of a new node.

In general, from what I've seen, Intel often builds new plants or completely converts an old manufacturing plant to a new node, whereas TSMC usually converts part of their old plant(s) to a new node.

TSMC has had issues with their 40nm and 32nm nodes in the past -- AMD's 68xx series of GPU's were supposed to be on 32nm but instead at the last minute had to be done on 40nm.

 

[MagnusTheBrewer]

In addition, the new designs may very well be incompatible with the manufacturing process used in the previous chip design. When I worked at Motorola, the engineering team was famous for creating designs that couldn't be built with the existing tools/process.

 

[taltamir]

thank you for clarifying.

If low yields is a ramp issue then their statement was rather redundant. Then again, low yields would be a manufacturing issue... so yea, it seems their statement was triply redundant.

 

[notty22]

The repercussions of the slow ramp up is partly effecting why their revenue forecasts are off. There is more of this to come, imo. Because they were probably figuring on some 32nm cpu server revenue at some points, points that they are going to continue to miss. The stock took a beating today.
http://www.google.com/finance?q=NYSE:AMD&client=news

 

[TuxDave]

thank you for clarifying.

If low yields is a ramp issue then their statement was rather redundant. Then again, low yields would be a manufacturing issue... so yea, it seems their statement was triply redundant.

You never know. What if high wafer volume @ low yield was the plan. Then you can have low yields but be fully ramped up. 🙂

 

[Tanclearas]

There are various issues that pop up with a new process. "Yield" is an issue related to each wafer. "Ramping" is related to the entire production.

You can have very high yield (99.9&#37😉, but only be able to produce 1000 wafers per month for several months in a row due to other production issues. No ramping.

Conversely, you can have very low yield (5-15%) but go from 1,000 wafers per month to 100,000 wafers per month within 6 months. Good ramp, but still poor yield.

Yield and ramping can definitely go hand-in-hand. If your yield improves (from 10% to 50% to 80%), then you are ramping, even if you're still only producing the same number of wafers per month.

In most cases yield gets better, and other production issues get ironed out, and you get good ramping.

 

[Idontcare]

Hi talta, lots of good responses in this thread already!

FWIW the term "ramp" is usually reserved vernacular in external communications to refer specifically to capacity ramps, unless otherwise expressly stated.

Yield issues = low yields...this could be parametric yield (clockspeed, power, etc) or functional yield (1+1=3, doesn't turn on, etc).

Ramp issues = wafer capacity...how many wafers can the fab start per day (wspd), also related to work in progress (wip) and wafer exits per day (wepd).

Manufacturing issues = can represent all of the above, including ex-fab issues (stuff that happens outside the fab while making the final sellable product) such as test/validate and packaging and so on.

Now you can have "yield ramp" issues which are entirely orthogonal to "capacity ramp" issues, and sometimes the issues can be intimately connected (as was TSMC's 40nm yield/capacity ramp issue).

 

[BD231]

This should explain everything

 

[taltamir]

thanks for the informative replies everyone

 

[Barfo]

 

[carnage10]

Hehe that ramp could also be referred to as a sandy bridge 😛 Notice it's complete, solid, has been around for a while yet is still perfectly useful.

Here is a ramp that better symbolizes AMD

 

[taltamir]

Hehe that ramp could also be referred to as a sandy bridge 😛 Notice it's complete, solid, has been around for a while yet is still perfectly useful.

what? it is a whole foot off the ground.

 

[Mopetar]

thank you for clarifying.

If low yields is a ramp issue then their statement was rather redundant. Then again, low yields would be a manufacturing issue... so yea, it seems their statement was triply redundant.

It's not necessarily redundant. Assume that they mean that the existing yields were lower than expected, TSMC was unable to bring new production capacity online, and there were other manufacturing related issues, which could refer to anything but potentially the working hardware not performing to expectations.

So essentially:

1) The existing equipment produces too many defective chips.
2) They can't get more equipment running, so we can't get more chips.
3) The good chips we do have don't perform as well as we had expected, forcing us to keep prices down.

 

[Mopetar]

what? it is a whole foot off the ground.

It's an early stepping Sandy Bridge 😉