CHIP DESIGNER AMD is planning to refresh its Brazos platform in time for Microsoft's Windows 8.
AMD launched its Brazos platform last year, though actual products tipped up earlier this year featuring a dual core processor and a DirectX 11 GPU. However with Windows 8 coming out in 2012, it is a little surprising that leaked slides point to AMD planning a refresh of Brazos, codenamed Hondo. According to the slides, Hondo is designed to operate with passive cooling, and have 2W "app power" usage, about half that of the current Brazos chips.
AMD outlines Hondo as a "re-architected" chip from its Ontario accelerated processing unit (APU) core found in Brazos. AMD claims that the 2W "app power" rating applies when the chip has to cope with 720p video playback, with "wireless web browsing" drawing 1.6W. The thermal design power (TDP), or maximum power the chip will draw, will be less then 4.5W.
AMD says the 1GHz dual core chip will be fabbed at 40nm and have a DirectX 11 capable GPU that will be clocked at 276MHz. To support it there will be an updated Hudson Fusion Controller Hub (FCH), which has a TDP of 1W.
The cut in TDP is impressive and should put a little more pressure on Intel to ramp up its Oak Trail Atom production in order to make it a viable chip for tablets, come the arrival of Windows 8. The slides indicate that AMD will start sampling Hondo in December 2011 with production ramping up in the second quarter of 2012.
Although Intel has been chided for its absence in the tablet market, Hondo is the first AMD chip that looks like it might be able to challenge Intel for design wins. The question is if Intel can get its Oak Trail chips out the door and into the hands of tablet makers fast enough, because even though it doesn't have the graphics capabilities to take on Hondo and win, its TDP is 30 per cent lower and Chipzilla has said it expects to transition to a 22nm process node in 2012, further lowering power consumption.
We asked AMD to comment on the leaked slides but it said that it would not comment on rumour or speculation. µ
Huh huh huh, Dairy Queen "Brassiere"
CAL NAUGHTON, JR.: THERE IS SOMETHING I WANT TO GET OFF MY CHEST. IT'S ABOUT THAT SUMMER, WHEN YOU WENT AWAY TO COMMUNITY COLLEGE. I GOT AN OFFER TO DO PLAYGIRL MAGAZINE, AND I DID IT. I DID A FULL SPREAD FOR PLAYGIRL MAGAZINE. I MEAN SPREAD MAN, I PULLED MY BUTT APART AND STUFF. I WAS TOTALLY NUDE. IT WAS WEIRD, I... I MEAN YOU PROBABLY DIDN'T HEAR ABOUT IT BECAUSE I WENT UNDER THE NAME OF MIKE HONCHO. BUT I JUST WANTED YOU TO KNOW THAT. IF YOU CAN HEAR ME, IF IT GOT INTO YOUR BRAIN SOMEHOW. THAT I SPREAD MY BUTTCHEEKS AS MIKE HONCHO.
Nice move but a little too late and not bold enough!
This has always been the problem with AMD; always one or two steps behind.
They ought to fab this sucker at 22nm or at worst 32nm.
I know it may make sense for them to use the oldest therefore cheapest process to fab this chip
as they are likely to look at it as a low margin and high risk product.
This strategy is wrong in several ways:
-They can use this chip to iron out their 22nm process in a low profile way. The benefit is that they will down the road have sufficient performance head room to play the market. This is also likely to be a simpler design than trying the process on more complex design.
-They are overlooking the embedded and equipment market. if Fabbed at 22 or 32nm AMD can position quality products of this caliber in the embedded market. They can even make SOC versions of it for premiums easily by packaging the separate dies (CPU/GPU, ram, flash, Ethernet, small Fpga ...) into a socket
-AMD should invest some small amount into a design house. Some place slightly separate from AMD that is tasked with designing reference products using AMD parts and technology that they can license or sale to third parties or even end users in some cases. A 10 man outfit can easily be funded by AMD and the payback can be great. Areas they can look into is entertainment, medical visualization, medical equipment, telecommunications, industrial control/automation, automotive, portable computing, education ...
A good example of this is to see where the atom is and where it is going. A lot of atom CPUs are finding their way into medical, industrial equipment. Eventually they are even vying for the server. It was not quite design for that, but we all know a good thing when we see it and will try to leverage it.
Wally