AMD's new "Richland" APUs are a refined version of the Trinity parts the company introduced last year. Nothing about the underlying architecture has changed – the CPU is still based on AMD's second-generation Bulldozer architecture, codenamed Piledriver, while the GPU is built on the Cayman architecture AMD first launched in late 2010. The only difference between the new AMD A10-6800K we’re reviewing here and its predecessor, the A10-5800K, is their clock speeds and power consumption.
The A10-6800K is a 4.1GHz part with a 4.4GHz Turbo Mode and a GPU clocked at 844MHz. That's a modest tap upwards compared to the A10-5800K, which offered a 3.8GHz base clock, 4.2GHz Turbo, and an 800MHz GPU. The A10-6800K also adds formal support for DDR3-2133 (A10-5800K topped out at 1866MHz) and a new power management unit that's designed to give the CPU better visibility.
The A10-6800K can raise clocks automatically for the CPU, GPU, or both. This type of incremental year-on-year improvement is in line with previous manufacturing trends. Typically, AMD and Intel are able to extract an additional speed grade or two at a given TDP as a process node matures and chip designs are tweaked for optimal results.
AMD is launching multiple 6000-series APUs, with each dropping in as a replacement for an older 5000 series part. The new 6000 parts use the same motherboards and chipsets as the older 5000 hardware and they keep the same TDP ratings. The A10-6800K and 5800K are both 100 Watt parts, the A10-5700 and A10-6700 are both 65 Watt, and so on. The difference is that, in each case, AMD has managed to increase the CPU and GPU clocks by 6 to 8 per cent while holding the same TDP.
Our performance tests were performed using 8GB of DDR3-2133 on an MSI FM2-A85XA-G65 motherboard. The A10-6800K is capable of using DDR3-2133, while the older A10-5800K tops out at 1866MHz. While others may have different results, our test results confirmed that limitation – the AMD A10-5800K couldn't run our Mushkin Redline memory at full speed and was forced to drop back to 1866MHz, while the A10-6800K had no problems with 2133MHz.
We've compared the AMD solutions to the Intel Core i5-3470 which we previously benchmarked nearly a year ago. That chip is an Ivy Bridge-based quad-core CPU (no HyperThreading) at 3.2GHz with a 3.6GHz Turbo Boost clock speed and Intel HD Graphics 2500. Graphics performance is going to be an automatic win for AMD – the Cayman-derived GPU at the heart of the A10 family is far more powerful than Intel's HD 2500 – but what about CPU performance?
In Cinebench R11.5, the A10-6800K turned in 42.47 frames per second (fps) on the OpenGL rendering scene, and scored 1.14 fps in the single-threaded CPU benchmark test, and 3.58 fps for the multi-threaded rendering test. The AMD A10-5800K hit 39.24 fps in OpenGL, 1.08 fps in single-threaded rendering, and 3.29 fps in multi-threaded rendering. The Intel Core i5-3470, in marked contrast, turned in scores of 1.48 fps (single-threaded), 5.67 fps (multi-threaded) and 12 fps in the OpenGL test.
Cinebench, in this case, neatly demonstrates the respective strengths and weaknesses of AMD versus Intel. Intel's cores are far superior to AMD's when it comes to single-threaded code – the Intel Core i5-3470 was 30 per cent faster than the A10-6800K in single-threaded performance, despite being clocked 20 per cent slower. The A10-6800K improved on the AMD A10-5800K's POV-Ray 3.6 benchmark rendering time (289 seconds vs. 298 seconds), but again, the Intel Core i5-3470 is faster, at 258 seconds total.
In TrueCrypt 7.1a, the A10-6800K turned in a Serpent-TwoFish-AES encrypt speed of 136MBps, compared with 125MBps for the AMD A10-5800K and 162MBps for the i5-3470. Performance in PCMark 7 showed a similar gap – the AMD A10-5800K turned in a score of 4,590, the A10-6800K scored 4,740, and the Intel Core i5-3470 scored 5,511.
The quad-core AMD processors tend to sit a ways back from the quad-core Intel chip, and while that distinction is captured in their pricing, it's not a great place for AMD to be. In our Photoshop CS6 12-filter test, the A10-5800K took 289 seconds to perform all 12 actions. The A10-6800K took 277.5 seconds (an improvement of about 4 per cent), and the Intel Core i5-3470 clocked in at 182 seconds, or 35 per cent faster.
Gaming benchmark scores continue to be the exception to this rule. In 3DMark 2011's Performance preset, the Core i5-3470 scored a bare 430 points, compared with 1591 for the AMD A10-5800K and 1,685 for the A10-6800K. Gaming – albeit low-end gaming – is far more palatable in general on an AMD APU compared to an Intel chip with HD 2500 or HD 4000 graphics. We should point out, however, that the AMD platforms often struggle to deliver modern titles at an acceptable frame rate as well. "Better" at the low-end doesn't always mean "good."
Richland's power consumption figures are only slightly different from Trinity's. Both the AMD A10-5800K and A10-6800K idle at 48 Watts, which is in line with even the best chips from Intel. Under load in Cinebench 11.5, the AMD A10-5800K drew 116 Watts, while the A10-6800K drew 121 Watts. Richland drew 4 per cent more power, but it was 8 per cent faster – an overall win for AMD. Under what we call peak load – maximum power consumption generated by running Prime95 in its Torture Test mode – the A10-6800K hit 132 Watts, compared with 128 Watts for the AMD A10-5800K. Overall, that's a small win for AMD as far as Richland's total power efficiency is concerned.
In an odd way, Richland mirrors Intel's Haswell, at least as far as what it offers in the short-term as compared to its immediate predecessor. The A10-6800K is a modest improvement over the AMD A10-5800K, with slightly higher performance and power efficiency. It's an iterative design, not a major leap, and if you already own a higher-end Trinity-based system, Richland isn't going to be an attractive upgrade path for you.
There's also the matter of the price increase. The A10-6800K has debuted at around 15 per cent more than the A10-5800K cost at launch, offering a performance increase of
8 to 10 per cent over the A10-5800K. Objectively, AMD needs stronger average selling prices and higher revenue if the company is to have any hope of regaining a competitive position against Intel, but the just-over-a-hundred-pound mark is a tough price bracket.
If you care about peak CPU performance, a fast Intel dual-core or higher-end quad-core will deliver better overall performance than the AMD A10-6800K, even if the total cost will be somewhat more.
If, on the other hand, you want a combined CPU+GPU with better graphics capabilities than you can get from an Intel chip right now in this low-end price range, the A10-6800K is a solid deal. The subjective difference between using an Intel and AMD solution is much smaller than you might think, even in cases where benchmark performance points to significant gaps.