- Outstanding IGP
- Native USB 3.0 support
- 6Gbps SATA support
- New FM1 socket
- Doesn’t beat discrete GPUs
The gaming-class graphics CPU is finally here.
Ever since AMD acquired ATI several years ago, it was only a matter of time until the CPU maker that could brought that knowledge to bear on their mainstream product lines. Intel incorporated integrated graphics on-die a while ago – can AMD’s new Fusion APUs compete in the same ballpark? Let’s find out.
Is the Future Still Fusion?
So let’s dive into this whole issue of CPUs and APUs. In terms of how and what they compute, the CPU might be considered a jack of all trades, master of none. It excels at running traditional programs, in-order stuff that GPUs don’t work very well – stuff that requires the processor to compute the output of one part of the program in order to move on to the next part.
GPUs, on the other hand, take a different tact when processing. These guys run best when processing something that the industry terms “massively parallelizable.” What that means is that GPUs work well when whatever it is that they’re processing, be it a video game scene or some kind of video rendering or stabilization, that can be broken up into little pieces and computed all at the same time.
Fusion, AMD’s term for blending aspects of the CPU and GPU onto a single die, brings the best of both worlds together. It adds Radeon stream processors (the simplified GPU cores) to AMD’s bang-for-the-buck x86 cores, removes some of the microarchecture redundancy and streamlines the whole process.
“AMD’s APUs are designed to be as fast as some discrete GPUs but not as fast as the higher end discrete GPUs.”
Those GPU cores are one of the ways AMD will differentiate between the different APU SKUs that it offers; the current high-end A8-3850 will offer what AMD calls the “discrete-class GPU” AMD Radeon HD 6550D while the A6 line will give you the AMD Radeon HD 6530D – that’s a difference of 80 “Radeon Cores” as well as a clock speed difference of 157MHz.
AMD likes to say that they’ve taken a different approach to Intel in making a combination CPU-GPU processing engine. Intel has put out a very capable product in the Sandy Bridge Core series lineup. It’s processors are half CPU, with some GPU added on, where as AMD’s product gives the GPU much more of a first-class citizen status.
AMD’s A8-3850 compares favorably with Intel in a lot of ways on paper. Both chipmakers are now finally using a 32nm manufacturing process (as the process gets smaller, the number of transistors the processor can pack into a space goes up while the power requirements go down). It’s a natively quad-core chip running at 2.9GHz – AMD’s dynamic clock switching Turbo Core functionality isn’t present on this guy.
There’s 128kB of L1 cache and 1MB of L2 for each core. Memory support goes up to 1866MHz and, for once, you’ll really notice the difference, even in a midrange system. The GPUs in an APU-powered machine are so well integrated with the CPU and memory controller that when you’re playing games, upping the RAM from 1066 to 1866MHz can mean a difference of 10% or more, according to our tests.
AMD also updated their chipsets with two new models – the A55 and the A75, which pack in a number of improvements. The A75 offers up to four USB 3.0 ports – natively, finally, thank god – as well as six SATA III (6Gbps) ports. The A55 loses the USB 3.0 support in favor of four more USB 2.0 ports and drops the SATA back a revision.