What Took So Long?

AMD announced the acquisition of ATI in 2006. By 2007 AMD had a plan for CPU/GPU integration and it looked like this. The red blocks in the diagram below were GPUs, the green blocks were CPUs. Stage 1 was supposed to be dumb integration of the two (putting a CPU and GPU on the same die). The original plan called for AMD to release the first Fusion APU to come out sometime in 2008—2009. Of course that didn't happen.

Brazos, AMD's very first Fusion platform, came out in Q4 of last year. At best AMD was two years behind schedule, at worst three. So what happened?

AMD and ATI both knew that designing CPUs and GPUs were incredibly different. CPUs, at least for AMD back then, were built on a five year architecture cadence. Designers used tons of custom logic and hand layout in order to optimize for clock speed. In a general purpose microprocessor instruction latency is everything, so optimizing to lower latency wherever possible was top priority.

GPUs on the other hand come from a very different world. Drastically new architectures ship every two years, with major introductions made yearly. Very little custom logic is employed in GPU design by comparison; the architectures are highly synthesizable. Clock speed is important but it's not the end all be all. GPUs get their performance from being massively parallel, and you can always hide latency with a wide enough machine (and a parallel workload to take advantage of it).

The manufacturing strategy is also very different. Remember that at the time of the ATI acquisition, only ATI was a fabless semiconductor—AMD still owned its own fabs. ATI was used to building chips at TSMC, while AMD was fabbing everything in Dresden at what would eventually become GlobalFoundries. While the folks at GlobalFoundries have done their best to make their libraries portable for existing TSMC customers, it's not as simple as showing up with a chip design and having it work on the first go.

As much sense as AMD made when it talked about the acquisition, the two companies that came together in 2006 couldn't have been more different. The past five years have really been spent trying to make the two work together both as organizations as well as architectures.

The result really holds a lot of potential and hope for the new, unified AMD. The CPU folks learn from the GPU folks and vice versa. Let's start with APU refresh cycles. AMD CPU architectures were updated once every four or five years (K7 1999, K8 2003, K10 2007) while ATI GPUs received substantial updates yearly. The GPU folks won this battle as all AMD APUs are now built on a yearly cadence.

Chip design is also now more GPU inspired. With a yearly design cadence there's a greater focus on building easily synthesizable chips. Time to design and manufacture goes down, but so do maximum clock speeds. Given how important clock speed can be to the x86 side of the business, AMD is going to be taking more of a hybrid approach where some elements of APU designs are built the old GPU way while others use custom logic and more CPU-like layout flows.

The past few years have been very difficult for AMD but we're at the beginning of what may be a brand new company. Without the burden of expensive fabs and with the combined knowledge of two great chip companies, the new AMD has a chance but it also has a very long road ahead. Brazos was the first hint of success along that road and today we have the second. Her name is Llano.

The Llano A-Series APU
POST A COMMENT

177 Comments

View All Comments

  • JarredWalton - Tuesday, June 14, 2011 - link

    The only way to make sure that Intel's current processors aren't at the top of most charts is to leave them out, particularly on notebooks. If we only look at IGP/fGPU, AMD comes out on top of graphics charts, but is that fair to NVIDIA's Optimus technology that allows dynamic switching between IGP and dGPU in a fraction of a second? The overall tone of this article (apart from the CrossFire section) is positive, but still people look at the charts and freak out because we didn't manipulate data to make Llano look even better. It's not bad, but it's certainly not without flaws. Reply
  • kevith - Tuesday, June 14, 2011 - link

    Oh too bad.

    I would like to use a laptop for music production with Nuendo and Win 7.

    It actually reqires a little more graphics-musclle than you might think to run an app like Nuendo.

    And,up to now, it has not been possible to get both a powerful CPU and GPU in the same machine for the money I have.

    So now the fGPU is powerful enough, that's great. But it seems, that the CPU-part of these APU's are too weak.

    Øv...
    Reply
  • krumme - Tuesday, June 14, 2011 - link

    Øhhh

    Just make sure your computer have 1Gb ram and win xp sp2, Nuendo even runs on single core 2Ghz whatever old shit.
    I would save the money and buy a e350.
    Heck you could even buy an Atom 510, acording to Anandtech, its just as fast as e350 for the cpu side.

    When i think about it. Just do that.
    Reply
  • ET - Tuesday, June 14, 2011 - link

    As madseven7 commented correctly, this isn't the fastest Llano CPU. There are 45W parts which perform better. They will have less battery life, but a significant increase in core speed. If you're interested in Llano you might want to wait until they get reviewed. Reply
  • JarredWalton - Tuesday, June 14, 2011 - link

    I suspect the 45W Llano parts will only have less battery life if you're specifically doing CPU/GPU intensive tasks. At idle, SNB and Llano should both bottom out at similar levels. For example, if you have a 2630QM and a 2820QM doing nothing, they both run at a very low clock and voltage. We'll test any other Llano chips we can get and report our findings, but other factors (BIOS and firmware optimizations) will generally be more important than whether the TDP is 35W or 45W, at least for our particular battery life tests. Reply
  • Shadowmaster625 - Tuesday, June 14, 2011 - link

    I dont get the Cinebench single threaded results. An N660 is about the same as a desktop X2 250/255 on that benchmark. Yet this A83500M scores only 61% of what an X2 250 does. That would seem to indicate that it is only running at 1.8GHz during that single threaded test. Why so low with 3 idle cores? It should be running at 2.5GHz and scoring 2500, or just neck and neck with a P520. Turbo is clearly not working anywhere near as well as it should be. Reply
  • krumme - Tuesday, June 14, 2011 - link

    Well this is AMD business at work. They are in a constant learning process and have been for the last 40 years.

    Next time they might consider the following:

    1. Dont send some half baked prototypes to the reviewers. Wait fx. 3 more weeks. This is just old Jerry Sanders style.

    2. Consider not sending stuff to Anandtech. As anandtech lives from backlinking also, the site needs the new product. And AMD, - and for the sake of the consumers right decisions, can live without 3 similar i7 plus high end discrete gfx, at 1.200 usd at the top of each chart. If AMD dont understand they have other interest than Anandtech - its business for all - they cannot serve their own interest. And its about time they start to earn their own money. They are competing against Otellini not some stupid schoolboy.
    Reply
  • JarredWalton - Tuesday, June 14, 2011 - link

    Thanks, krumme; always a helpful response. Lenovo has taken this to heart, I'm sure you'll be happy to know, and is not sending any review samples our way. Amazingly, we're still able to survive. And FWIW, if AMD hadn't sent us anything, we'd have had more content earlier through other sources. The only way they can get us to abide by NDAs is by actually working with us. Reply
  • krumme - Tuesday, June 14, 2011 - link

    Well thank you Jarred. That was an helpfull answer! that explains a lot.

    I hope AMD gives you attention and work with you in the future, its in all your readers interest.

    That means AMD giving you priority, broad access to the right people and more time to do the reviews.
    Reply
  • JarredWalton - Tuesday, June 14, 2011 - link

    This is something I discussed with AMD numerous times, and it's one of the reasons we want a utility that will show us true CPU clock speeds in real time. Unfortunately, they don't have anything they're willing to share with us right now. They said they have test units where they can monitor this stuff, but it requires special BIOS hooks and those are not present in our preview samples. In theory, Turbo Core should allow the single-threaded Cinebench result to run up to 60% faster than non-Turbo. Of course, we can't even disable Turbo Core either, so we don't know how much TC is actually helping.

    P920 is clocked 6.7% higher than A8-3500M, but 3500M has twice the L2 cache and some other enhancements. With 3500M coming in 17% faster than P920, that would suggest that 3500M averages around 1900MHz, but that could mean it runs at 2.4GHz for a bit and then 1.5GHz for a bit, or somewhere in between.

    Given the way AMD does Turbo Core (monitoring instruction workloads and their relative power requirements), I think that at least right now, it's not being as aggressive as Intel's Turbo Boost. It feels more like Bloomfield and Arrandale turbo, where you got an extra 2-4 bins, rather than Sandy Bridge where you can get an extra 5-10 bins. Hopefully we'll see refinements with Turbo Core over the coming months and years.
    Reply

Log in

Don't have an account? Sign up now