Graphics Overview, Turbo Core 3.0
Adding The 7000 Series To The APU
The inclusion of a powerful integrated graphics component on nearly all modern CPUs means that the updates to the graphics component are at least as important as the updates to the processor cores. AMD realizes this better than anyone – after all, graphics is one area where they’ve retained a nice lead over Intel in the past.
Llano’s Radeon 6000 series IGPs were able to hand Intel HD 3000 its hat in most games, but Intel HD 4000 has raised the bar. AMD’s response is to update its own APUs to the 7000 series. This time around the company is going to offer not three but five different IGPs. All of them are based on the same architecture but offer a different core count at different clock speeds.
Core counts are actually down slightly when compared to Llano, but this is made up for by significantly higher clock speeds. Llano IGPs were restricted to 400MHz or 444 MHz, but the mainstream Trinity APUs max out at 655MHz or 686MHz.
The expansion of the lineup from three to five IGPs is due to the new 25W and 17W APUs. AMD apparently ran in to power limitations which prevented these low-voltage parts from performing like their 35W cousins. Intel has done the same in the past with its own low-voltage processors.
This means the 25W APU offers an IGP that has 384 cores clocked at 497 MHz, while the 17W APU has an IGP with 256 cores clocked at 424Mhz. This will significantly hamper the graphics performance of those parts, though they should still be a fair shake quicker than Llano’s best IGP.
AMD has included a number of feature enhancements in Trinity’s IGP. AMD Eyefinity is now natively supported by the IGP via DVI, HDMI and also DisplayPort 1.2 (on up to four displays). A new media acceleration feature will allow hardware assisted video conversion in supported applications. Tessellation is said to be improved, anti-aliasing is said to be improved and the IGPs are now compliant with DirectCompute 11 and OpenCL 1.1.
None of this is ground-breaking, though the inclusion of Eyefinity on up to four displays is a nice one-up on Intel, which just last month was boasting about support for up to three displays with the IGP on Ivy Bridge. Ultimately, as with the CPU, the real story is going to be told by our performance benchmarks.
Turbo Core 3.0
Intel Turbo Boost is one of the best features introduced by the company in the last decade. AMD quickly introduced its own solution to try and catch up, but at first seemed a bit blind-sided by Intel’s success with Turbo Boost. AMD Turbo Core worked but was much less aggressive than Turbo Boost.
Turbo Core 3.0 is, of course, meant to close the gap. AMD has designed Trinity so that power can be shifted across both the CPU and GPU as a unit. The advantage of this is a better ability to place power where it’s needed most. It’s actually rather rare for any application to place a significant and roughly equal load on both components – some games do this, as well as some video or photo editing suites, and that’s about it. Most games and most applications only fully tap one or the other, and many applications don’t require the use of the GPU at all.
The results are obvious in AMD’s specifications. The A10-4600M, which is the fastest Trinity processor at launch and the one found in our reference platform, has a base clock of 2.3 GHz but can top out at 3.2 GHz when a single CPU thread is tasked – an increase of 900 MHz. That’s a 200 MHz improvement in relative Turbo Core clock speed and a 500 MHz improvement in maximum clock speed over the A8-3550MX, which is the quickest Llano quad-core. That part has a base clock of 2 GHz and a maximum clock of 2.7 GHz.
Even the low-voltage components aren’t left out of the party. The A10-4665M, which has a 25W TDP, can increase its clock speed by up to 800 MHz. The A6-4455M, which has a 17W TDP, can increase its clock speed by up to 500 MHz. Intel’s Turbo Boost is still generally more aggressive, but AMD is on the right track.