AMD A8-7600 Kaveri APU Review - HSA Arrives
System Setup and Review Goals
With this release, AMD is attempting to do things very differently compared to other launches and I want to point this out before we go any further. Over the past 14+ years of doing PC hardware reviews, new processors and graphics cards were always sent to us in a "top down" fashion. If a high end CPU option was available (whatever the flagship part was) it was the one sent to reviewers for the initial release. For Kaveri, the first true HSA-ready part, things were set up a bit differently.
When I arrived home from CES last week, I found a box from AMD that I assumed would have all the necessary hardware to allow us to put together the same comprehensive review for our readers that we are used to. Instead, inside the box was a completely system built around the AMD A8-7600 APU which is the lowest end Kaveri part being announced today for the desktop. This APU includes 10 compute units (4 CPU and 6 GPU) with a maximum clock rate of 3.3 GHz (Boost to 3.8) in its 65 watt option and 3.1 GHz (Boost to 3.3) for its 45 watt setting.
The A8-7600 differs from the A10-7850K flagship Kaveri APU in several key ways. First, it has fewer GPU cores (384 SIMD units down from its bigger brother's 512 SIMD units). That is a drop of 25%. While GPU clock speeds are the same, the A8-7600 has a 400 MHz disadvantage in base clock so you are likely to see some noticeable performance deltas. Of course, all of this comes at the price of power. The A8-7600 will be available in both 65 watt and 45 watt TDP options, configurable in the BIOS if the motherboard vendor implements configurable TDPs.
Let's make it clear: AMD is making a play into the low power space rather than attempting to pit the A10-7850K against the likes of Intel's Core i5 Haswell processor lineup and this is likely because it falters in a lot of areas. It was spelled out to me that, with the move to 28nm on GlobalFoundries, there were some production differences that made the higher power and higher voltage parts less flexible (and less scalable) than expected. But, in contrast, the lower end parts like the A8-7600 shown here today can use less power and hit higher performance levels than was previously possible.
We will test the A10-7850K as soon as we can buy one. For today, however, the comparisons are going to be quite straight forward as we dive into the world of 45 watt TDPs. The Kaveri A8-7600 is going to have company in the form of two Richland APUs, the A10-6700T and the A8-6500T, as well as the Intel Core i3-4330. Intel's Haswell part is a dual-core HyperThreaded unit that has a rated TDP of 54 watts. Both of the 6700T and 6500T APUs from AMD are rated at 45 watts. Thanks the configurable TDP options in the system provided, I have results from both the A8-7600 running at a 65 watt TDP and at a 45 watt TDP.
Keep in the back of your mind that, though the 6700T is an A10 part, the base clock speed of that APU was only 2.5 GHz indicating that the performance of the A8-7600 should be much better (thanks to a 600 MHz edge there). This demonstrates how Kaveri is able to scale, both because of architecture adjustments and because of process technology changes, when compared to Richland.
The system that AMD provided was built inside a nice looking Mini ITX case from Xigmatek called the Nebula which is not yet officially released. The motherboard was the Asrock A88X-ITX+ and the memory was AMD's own DDR3 rated at 2133 MHz. The system OS was Windows 8.1 and it was installed on a Samsung 840 Pro 256GB SSD. The power supply included was a bit overkill; an Antec High Current 750 watt. The complete system was used for all APU testing and I must say that really enjoyed my time with the Xigmatek case. It is well built and fairly easy to work inside for a Mini ITX design. I'll have a separate story and video on that case in the near future.
For the review of the Intel Core i7-4770K we updated our entire suite of CPU benchmarks and thus we have a new set of results to present. The review today of Ivy Bridge-E will continue with that same set of tests.
|Test System Setup|
|CPU||AMD A8-7600 (Kaveri)
AMD A10-6700T (Richland)
AMD A8-6500T (Richland)
Intel Core i3-4330 (Haswell)
AMD A10-5800K (Trinity)
|Motherboard||Asrock A88X-ITX+ (Kaveri, Richland)
MSI Z87I Gaming AC (Haswell)
MSI A85 (Trinity)
|Storage||Samsung 840 Pro 256GB|
|Graphics Drivers||Catalyst 13.30 (Beta)
|Power Supply||Antec 750w High Current|
|Operating System||Windows 8.1 x64|
Never mind the odd processor name; AMD was being very security conscious this time around...
So, while I am disappointed in AMD's decision to withhold the A10-7850K from our testing (although I am sure someone will have results today), there is still a lot to be learn about Kaveri and its position as the first HSA-enabled APU from AMD. From a lower power standpoint, looking at the 45 watt performance levels of Kaveri will be interesting for SFF builders and those of us looking to predict the performance per watt of upcoming Kaveri notebook parts. Is the new architecture, on the new process, more or less likely to find success in the thin and light notebook markets that are growing without a solid AMD option today?
We will also be able to see the architectural efficiency of the entire APU by comparing it to A8 and A10 parts from the previous generation. Does the upgraded graphics and CPU core make that much difference in performance when power is limited?
Let's start with the most interesting capabilities of the APU: gaming and OpenCL.