It’s got roughly one billion 32nm transistors, fabbed at Globalfoundries. Four CPU cores and a single graphics core. It’s what AMD calls an Accelerated Processing Unit (APU). And we’ll see it in 2011.

Unfortunately that’s a bit late. The APU, codenamed Llano, was originally scheduled for 2010 but got pushed back. In 2009/2010 Intel will be the first to deliver on-chip graphics with Clarkdale/Arrandale, and in late 2010 Sandy Bridge will have on-die graphics.

The first die shot of AMD's 32nm Llano APU based on 32nm Phenom II cores

Above is what I believe to be a die shot of AMD’s first APU. The CPU doesn’t use AMD’s next-generation microarchitecture, that’s only for the server and high end in 2011. The first APU will use the existing Phenom II architecture on the same die as DX11 graphics, but at 32nm. Sandy Bridge will use a brand new microprocessor architecture on 32nm but with updated Intel integrated graphics. It looks like Sandy Bridge will have the CPU advantage while Llano might have the GPU advantage, assuming Intel can't get their GPU act together by then. Llano is on schedule to debut in 2011 with OEM sampling happening before the end of the year.

Also on schedule is AMD’s next-generation microarchitecture, codenamed Bulldozer. AMD listed its client PC goals for 2010 at this year’s Financial Analyst Day, one of them is to start sampling its next-generation microprocessor next year - in 2010. If the chip is ready for OEMs by the end of 2010, that means it’ll go on sale as early as 1H 2011.

Unfortunately AMD isn’t talking much about Bulldozer architecture, I suspect we won’t see that disclosure until mid to late 2010. It’s not to keep things secret, we already have many estimates of what Bulldozer’s architecture is going to look like. And if the public already knows, then Intel is also well aware of what AMD has coming in 2011. Updated: AMD has given a high level overview of its Bulldozer and Bobcat architectures here

A major focus is going to be improving on one of AMD’s biggest weaknesses today: heavily threaded performance. Intel addresses it with Hyper Threading, AMD is throwing a bit more hardware at the problem. The dual integer clusters you may have heard of are the route AMD is taking...

AMD's 2010 - 2011 Desktop Roadmap

Today’s high end desktop platform is called Dragon. It’s what you get when you combine a Phenom II, AMD 790FX/GX chipset and a Radeon HD 4000 series graphics card. Nice, but that’s old news now.

Next year the high end platform will be called Leo. It’s made up of a Thuban CPU, which is an updated Phenom II rev sporting as many as six cores. It’s still 45nm so don’t expect much in the way of new architectural features. Graphics comes courtesy of the Radeon HD 5000 series, which we all know and love. The chipset is going to be AMD’s new 8-series, complete with a new SB850 south bridge.

In 2011 we get Bulldozer and it comes in the form of the Zambezi CPU (AMD’s codenames are such fun). You’ll see four and eight core versions of Zambezi. Both will support DDR3 and both will work in Socket-AM3. Obviously guaranteeing motherboard support this early in the game is difficult, but AMD is usually good about maintaining socket compatibility. You may be able to slip a Zambezi into your current day Socket-AM3 motherboards.

All this plus a chipset and AMD’s 6000-series graphics makes up what’s going to be called the Scorpius platform.

Moving down one notch to mainstream, today we have the Pisces/Kodiak platforms (yep, never heard them called that either). That’s made up of Athlon II CPUs and the 785G chipset with integrated graphics.

Next year we’ll have Dorado, which consists of Athlon II CPUs and next-year’s integrated graphics chipset. Unfortunately it won’t be a DX11 chipset, AMD is only listing 10.1 support in the 2010 platform. Which means that the graphics we’ll see in 2011 integrated on-die will be a derivative of the DX11 Radeon HD 5000 series we have today. Sweet.

The 2011 mainstream desktop platform is called Lynx, purr. It comes with the Llano APU, which as I mentioned before, doesn’t use Bulldozer. Instead Llano is made up of as many as four 32nm Phenom II-like cores. Llano also features an integrated DX11 GPU. Llano will require a new socket as the pinout will have to support video out just like Intel’s Clarkdale.

Cheesy Marketing Names for Cool Tech, AMD Velocity Ensures New Designs Every 12 Months

AMD’s first APUs drop in 2011, but what happens in 2012? Intel is committed to new microprocessor architectures every 2 years as a part of its tick-tock strategy. AMD’s GPU-inspired equivalent is called Velocity.

About every year we get a new GPU architecture, whether it’s a strict doubling of execution resources or something more significant, it happens like clockwork assuming TSMC isn’t fabbing the chips. AMD Velocity just states that, in turn, every year we’ll get a brand new chip that integrates this new GPU architecture. The CPU side may or may not change, but with yearly design cycles we could see regular improvements on that end as well.

Velocity also means that even if it’s difficult getting more performance out of a CPU architecture, AMD can always rely on a beefed up GPU core to give users a reason to upgrade.

This is all going to get real interesting once we have some good GPU compute applications to run on these things. For GPU compute apps, every year could be another Conroe, with ~20% performance gains just from the GPU improvements.

We just need the apps to support it. And no NVIDIA, what we have today isn’t enough.

The Notebook Roadmap

For mainstream notebooks today AMD doesn’t really offer anything sexy. We have the Tigris platform based on the Caspian CPU (45nm Athlon II X2 derivative) and RS880M chipset with integrated Radeon HD 4300 series graphics (DX10.1).

Next year we get the Danube platform, complete with Champlain CPU (Athlon II X2 or X4 derivative) and relatively similar graphics to what we have now.

Like the desktop roadmap, things don’t get interesting until 2011; that’s when we meet Sabine.

Sabine comes with a Llano APU, just like the desktop, and four 32nm Phenom II-like cores. There isn’t a dual-core mainstream offering for mobile on the roadmap. Llano of course comes with a DX11 GPU on die, which AMD is calling GigaFLOPS-class. I guess we won’t have 4870 level performance on die at 32nm.

AMD doesn’t have a high performance CPU for ultrathin notebooks today. The only options are Consesus and Huron, both first generation K8 derivatives. They’re also fabbed at 65nm so they’re not exactly as small or power efficient as they could be.

Next year this starts to change with the Nile platform. It comes with a Geneva CPU, a lower power Athlon II derivative. That should help fix the performance issues. At 45nm we’ll hopefully see good battery life from the platform.

Come 2011, AMD has another CPU core for us; it’s called Bobcat. I first talked about Bobcat over two years ago, and we won’t see it for nearly another two years, incredible. It’s an ultra low power microprocessor architecture specifically designed for mobile. Assuming AMD didn’t exaggerate its claims, this should be the first real mobile competitor Intel has seen.

Bobcat will appear in the Ontario APU as a part of the Brazos platform. Hooray for more codenames.

If you look at AMD’s roadmap there are a couple of subtle hints to pay attention to. Note that Brazos extends top to bottom from Ultrathin down to Netbook, whereas no other AMD platform has ever completely gone down to the bottom of the Netbook segment. Could this be an indication of AMD’s ability to hit ultra low, Atom-like, pricepoints with Bobcat?

Secondly - the Brazos platform isn’t colored yellow, green or red for 65nm, 45nm or 32nm. I'm guessing that means Ontario will be built on a bulk 28nm process to save cost.

When I spoke to Globalfoundries earlier this year they indicated that fabbing on half nodes could be one option exploited by CPU companies looking to compete with Intel.

Server Roadmap

AMD’s previous CMO, Henri Richard, was a huge Ferrari fan. In turn, AMD became a sponsor of Ferrari’s Formula 1 team and some of AMD’s roadmap developed Italian-inspired codenames.

The Maranello platform is AMD’s high end server platform, due out in 2010. It supports the 8 and 12 core Magny Cours processors. These are multi-chip-modules with two quad or hexa core die on a single package.

With twice the die, you get twice the memory controllers. Magny Cours has four DDR3 memory channels. With more memory channels, Magny Cours needs a new, higher pincount socket which AMD is calling Socket G34. The chips will be branded as Opteron 6000 series.

San Marino is the second major server platform of 2010 and it’s for more normal servers. Four and six core Lisbon processors find their homes in San Marino. There’s also a low power Adelaide platform that’ll be available.

These Lisbon processors will be branded as Opteron 4000 series chips and will work in Socket C32.

In 2011 we get Interlagos and Valencia platforms, both based on AMD’s next-generation Bulldozer cores. Interlagos is a Socket-G34 platform supporting 12 and 16 core processors, while Valencia is Socket-C32 and can accommodate 6/8 core chips.

Final Words

Unfortunately for AMD, 2010 isn’t really interesting. The company will have to rely on aggressive pricing and the continued success of its graphics teams to carry it for the next 12 - 18 months.

Bulldozer, from what I know, appears to be a bold enough architecture to really challenge Intel if AMD can get it done properly. Bulldozer should arrive between Sandy Bridge and Intel's first 22nm CPUs. It's too early to tell how well Bulldozer's execution is going; AMD absolutely must sample in 2010.

It's disappointing that Llano won't use Bulldozer. With 32nm Phenom II cores, Llano will be roughly one to two architecture generations behind Sandy Bridge. The GPU side should be strong though, it is ATI after all.

AMD’s graphics strategy is much stronger. Bringing an already industry leading GPU architecture on die and then revving it every year is going to completely change the way we look at annual CPU releases. The big question here is what apps are we going to be running on these integrated GPU cores? The market has roughly two years to start finding out.