AMD's Tablet Architectures: Hondo at 4.5W, Future Sub-2W SoCby Anand Lal Shimpi on February 2, 2012 2:06 PM EST
- Posted in
- Trade Shows
- AMD FAD 2012
In its client roadmap AMD revealed Hondo, a 4.5W APU with 1 - 2 low voltage Bobcat cores and an on-die DX11 GPU built on a 40nm process. Hondo will fit into Windows 8 tablets starting later this year. Going forward, AMD wants to get into the sub-2W market although we don't have a codename to associate with that power target. Mobile is very important to AMD going forward both in tablets and ultra thin notebooks and it looks like AMD is planning on building the architectures it needs to be successful there.
Post Your CommentPlease log in or sign up to comment.
View All Comments
Arnulf - Thursday, February 2, 2012 - linkOfftopic: Flood of AMD news for supposedly "pro-Intel" website ;)
Ontopic: why 40 nm in timeframe when everybody and their dog is coming out with 28 nm product for the mobile market ?
SilentSin - Thursday, February 2, 2012 - link"why 40 nm in timeframe when everybody and their dog is coming out with 28 nm product for the mobile market ?"
You answered yourself :) Lots of room at the foundry for 40nm wafer starts and they should be cheaper than 28nm which gives AMD the potential for better margins. Whatever grip Intel held in the netbook/nettop market is going to shrink even further in 2012.
fic2 - Thursday, February 2, 2012 - linkI agree with the OP. 40nm is a bigger, more expensive, more power hungry cpu. Really doesn't make much sense.
SilentSin - Thursday, February 2, 2012 - linkBigger sure. More power hungry maybe. More expensive? Debatable. There's also another variable and that is volume, which AMD wants way more of. OEMs are eating Brazos up like candy and the impending Win8 tablet flood is within sight. They want as many as they can get without having to muscle for room at 28nm.
You also have to include design costs into the 28nm transition. They have a 40nm design that works well enough and generally nodes only get cheaper over time whereas the smaller and newest process commands a price premium. You'd have to think the 40nm LP process is pretty mature at this point otherwise they wouldn't be able to take the same design down from 9W to 4.5W.
duploxxx - Friday, February 3, 2012 - linkanother reason: maturing. a 40nm mature platform can produce lower power silicon then initial 28nm and they are producing right now, just like trinity. There is no time to wait more for the large 28nm mature production.
Ammaross - Friday, February 3, 2012 - linkThe big reason: pennies. Sure, 40nm means more sq mm for the die, but it's etched on sand (read: cheap stuff). With everyone moving to 28nm, no one wants 40nm, which means discount rates before they have to junk the 40nm fab equipment. It's a very smart move, and really the only move AMD can make at this point to remain price-competative with Intel.
JarredWalton - Thursday, February 2, 2012 - linkAs to the other part, it's AMD's Financial Analyst Day, so Anand is there and covering all the news. Anyway, this is actually one of the more interesting pieces today IMO -- a lot of the stuff just doesn't matter much to me, but then I'm not interested in AMD except as a provider of useful tech. ARM in the datacenter? Sure fine whatever... tell me about the next CPU and GPU instead. :-)
Beenthere - Thursday, February 2, 2012 - linkWhile I am not a fan of toy tablets, many sheep buy these so if AMD can sell to these to OEMs at a profit, great. I am however very interested in a Trinity powered laptop as are many other folks so I think AMD will do very well with Trinity.
The slides show exactly what I have been saying for months - AMD will cover all the market segments with APUs except highend discrete CPU/GPU and some server apps. This is a Win-Win for AMD and consumers as APUs offer a better performance/value/power consumption proposition than anything but the highest discrete CPU/GPU combo which offer performance but at a price.
As far as the question on 40nm vs. 28nm it really does not matter if the power is low and the performance acceptable. TSMC and GloFo both need to ramp 28nm processes.
fic2 - Thursday, February 2, 2012 - link40nm vs 28nm matters in both power and price. Price since a lot more 28nm die can be cut from the same size wafer.
Ammaross - Friday, February 3, 2012 - link"Price since a lot more 28nm die can be cut from the same size wafer."
You may get higher yield, but at a much higher fab price. The price of chip wafers pale in comparison to the cost of running it through a 28nm process. Remember, AMD has stat monkies with real numbers at their fingertips. They've already crunched the costs into more charts than you can shake a stick at.