AMD start sampling of ontario

[nyker96]

http://www.xbitlabs.com/news/cpu/di...ors_Targets_Launch_in_First_Half_of_2011.html

not sure what significance this will have on their time table for this fusion chip but I guess it must be a good thing to 'start sampling'. and what does bulk 40nm mean?

 

[GaiaHunter]

http://www.xbitlabs.com/news/cpu/di...ors_Targets_Launch_in_First_Half_of_2011.html

not sure what significance this will have on their time table for this fusion chip but I guess it must be a good thing to 'start sampling'. and what does bulk 40nm mean?

Bulk 40nm is the fab process (can't really explain you the details/differences between this and other processes as I don't really know them, although it is related to adding "impurities" to the silicon subtract) and it means it is going to be made by TSMC using the same process as current GPUs instead of the process AMD uses for its CPUs. I believe bulk is a cheaper process than SOI (which adds an insulator layer to the silicon subtract).

“Ontario is a part of Brazos platform. Just like Llano, we are sampling it to customers today. We are demoing it at Computex [behind the closed doors]. Expect this product to launch in the first half of 2011,” said Rick Bergman, senior vice president and general manager of AMD products group, during AMD’s press conference at the trade-show.

I guess this mean both Llano and Ontario should be on schedule for H1 2011.

 

[Soleron]

Or it could be Globalfoundries' 40nm bulk process too.

In addition to cost, they may be using bulk because it's easier to add third-party IP to the die and spin new products, if they want to compete in the low-power widgets segment that ARM does and Intel is trying to.

 

[Idontcare]

http://www.xbitlabs.com/news/cpu/di...ors_Targets_Launch_in_First_Half_of_2011.html

not sure what significance this will have on their time table for this fusion chip but I guess it must be a good thing to 'start sampling'. and what does bulk 40nm mean?

SOI versus bulk

Both use a silicon wafer, the physical difference is that SOI has a sub-surface oxide layer (the insulator) which is buried under a silicon top-layer (the silicon that is on the insulator - S O I).

It provides electrical benefits which are in part the reason that AMD's processors are power-efficient when idle. (Intel's use of HKMG and powergating gives them an edge too, but that didn't come until 45nm)

Or it could be Globalfoundries' 40nm bulk process too.

Unless GloFo intentionally designed the spice models for their bulk 40nm to be nearly identical to that of TSMC's it is extremely unlikely that AMD taped out Ontario for GloFo 40nm since GloFo has only been developing a bulk 40nm process for less than 2yrs. Ontario was likely locked into a spice model >3yrs ago, the design pipeline is just that lengthy.

 

[lifeblood]

http://www.xbitlabs.com/news/cpu/di...ors_Targets_Launch_in_First_Half_of_2011.html

not sure what significance this will have on their time table for this fusion chip but I guess it must be a good thing to 'start sampling'. and what does bulk 40nm mean?

I assume your comment about Fusion and your question about bulk 40nm are two completly different thoughts. Ontario is being produced by Global Foundries on their 32nm process. Bulk 40nm is used by TSMC to make GPU's. I do not know if GF has a bulk 40nm. They use 40nm SOI to make current AMD CPU's.

 

[Idontcare]

Ontario is being produced by Global Foundries on their 32nm process.

Ontario is Bobcat+integrated GPU and is 40nm bulk Si produced by TSMC.

I do not know if GF has a bulk 40nm.

They do but it is still in development and will be used by Chartered's customers of record prior to their having merged with GloFo.

They use 40nm SOI to make current AMD CPU's.

GloFo produces AMD's CPU's on 45nm SOI.

 

[lifeblood]

Ontario is Bobcat+integrated GPU and is 40nm bulk Si produced by TSMC.

I went to check my facts and I've found articles claiming both, although the majority (of those I read) say 32nm at GF. I've found nothing from AMD officially which say one way or another. If you have an "official" source can you link it? (I don't disagree, I'm just curious.)

GloFo produces AMD's CPU's on 45nm SOI.

I stand corrected, sir.

 

[Martimus]

http://www.behardware.com/medias/photos_news/00/27/IMG0027365.gif
http://www.behardware.com/news/10543/globalfoundries-roadmap.html

Global Foundries Roadmap as of November 11th, 2009.

EDIT: For some reason the image wouldn't embed, so I just posted the link along with the article that goes with it.

 

[lifeblood]

Global Foundries Roadmap as of November 11th, 2009.

Hmm, by that chart it can be anything from the current 45nm process all the way down to the 28nm SLP process. Personally I'll put my money on 32nm. Intel will be releasing 32nm Atoms in 2011 so if AMD wants to compete in the Netbook market they got to step forward with their best. Of course, that's just me making a SWAG.

On the other side of the coin, we have to consider GF's manufacturing capacity. I doubt they can make both Llano and Ontario at the same time on the same brand new process. That adds weight to the Ontario on 45nm, either TSMC or GF.

 

[alyarb]

http://www.xbitlabs.com/news/cpu/di..._System_on_Chip_40nm_Fabrication_Process.html

http://vr-zone.com/articles/-rumour-amd-to-use-40nm-tsmc-bulk-for-fusion-/9092.html

 

[lifeblood]

http://www.xbitlabs.com/news/cpu/di..._System_on_Chip_40nm_Fabrication_Process.html

http://vr-zone.com/articles/-rumour-amd-to-use-40nm-tsmc-bulk-for-fusion-/9092.html

Again, unconfirmed. AMD still refuses to say.

If I was AMD I would be very careful before using TSMC for an APU. TSMC has never done x86 (afaik) and AMD has never done a CPU on TSMC's 40nm bulk process. Apparently AMD is banking on lot on Ontario so for the CPU group to use TSMC will create a pretty steep learning curve for both parties. I'm not sure AMD will be willing to add additional risk for such a critical product.

But as always I could wrong. It does make me wonder why their keeping the process and manufacturer secret.

 

[alyarb]

A couple of months ago there was a battery of news articles confirming that GF and TSMC will not pursue any 32nm bulk production at all. Following that it became clear that Ontario is a 40 nm part. I don't know why you want to "put your money on 32." The uncertainty about Ontario is who will build it.

I would prefer GF to handle the production but I don't care.

 

[lifeblood]

no, it has been clear that Ontario is a 40 nm part. I don't know why you want to "put your money on 32." The uncertainty around Ontario is who will build it.

I would prefer GF to handle the production but I don't care.

It's been made clear by everyone except AMD.

However, as I've been reading other articles the past little while I think I will concede the point. The general feeling among the experts seems to be 40nm at TSMC for Ontario, and I do not presume to know more then them.

In a way it does make good business sense. It doesn't put all the eggs in one basket, and it lets GF focus on the new 32nm process for Lliano while TSMC uses it's mature 40nm process to work on Ontario.

 

[BD231]

Again, unconfirmed. AMD still refuses to say.

If I was AMD I would be very careful before using TSMC for an APU. TSMC has never done x86 (afaik) and AMD has never done a CPU on TSMC's 40nm bulk process. Apparently AMD is banking on lot on Ontario so for the CPU group to use TSMC will create a pretty steep learning curve for both parties. I'm not sure AMD will be willing to add additional risk for such a critical product.

But as always I could wrong. It does make me wonder why their keeping the process and manufacturer secret.

AMD has been planning to use TSMC for years now on the CPU front, three to four years after the announcement of that very fact and you're just now telling them to be careful??

Sorry, these two companies have done well together for some time now and AMD has plenty of influence on the process.

 

[taltamir]

I went to check my facts and I've found articles claiming both, although the majority (of those I read) say 32nm at GF. I've found nothing from AMD officially which say one way or another. If you have an "official" source can you link it? (I don't disagree, I'm just curious.)

Both TSMC and GloFo have canceled their 32nm process to "better focus" on their 28nm process. This makes intel the only company that will ever produce 32nm CPUs and GPUs (intel also produces 32nm MLC and SLC chips, but those are different, and intel is not the only one who does that, Samsung for example also has a 32nm process IIRC)

 

[Martimus]

Both TSMC and GloFo have canceled their 32nm process to "better focus" on their 28nm process. This makes intel the only company that will ever produce 32nm CPUs and GPUs (intel also produces 32nm MLC and SLC chips, but those are different, and intel is not the only one who does that, Samsung for example also has a 32nm process IIRC)

Global Foundries still has the 32nm SOI process. Llano will be a CPU and GPU on 32nm process that isn't made by Intel. According to their roadmap, risk production on 32nm SOI starts about now too.

 

[taltamir]

Global Foundries still has the 32nm SOI process. Llano will be a CPU and GPU on 32nm process that isn't made by Intel. According to their roadmap, risk production on 32nm SOI starts about now too.

when is this map from?
http://www.xbitlabs.com/news/other/...ies_Scraps_32nm_Bulk_Fabrication_Process.html

the map clearly states that all 32nm and 28nm processes listed are HKMG, not SOI. And the article I linked specifies that the 32nm bulk process with HKMG is scrapped.
And the announcement of cancellation was 04/01/2010 02:46 PM

 

[Idontcare]

when is this map from?
http://www.xbitlabs.com/news/other/...ies_Scraps_32nm_Bulk_Fabrication_Process.html

the map clearly states that all 32nm and 28nm processes listed are HKMG, not SOI. And the article I linked specifies that the 32nm bulk process with HKMG is scrapped.
And the announcement of cancellation was 04/01/2010 02:46 PM

Taltamir it is kinda universally known that GloFo's super high-performance CMOS process tech for 32nm is HKMG w/SOI (which is why it is not explicitly stated, the pertinent details are only mentioned and the pertinent details for 32nm is HKMG...you don't see them detailing the use of copper either, for the same reason, it is a given).

 

[taltamir]

thank you for clarifying.
It is good if they kept the high end 32nm process and scrapped the low end one. This lends credence to their claim that it is due to lack of demand (vs the theory that they ran into technical difficulties)

 

[grimpr]

Hans De Vries on AMDs Ontario.

http://www.xtremesystems.org/forums/showpost.php?p=4421074&postcount=125