AMD Trinity review
AMD’s Fusion technology was mooted when AMD bought ATI Technologies in 2006, but it took until last year for the plans to fully come to fruition. By teaming a CPU with capable integrated graphics, AMD finally had something that it could say it did better than Intel.
Trinity, the successor to last year’s Llano chips, doesn’t stray too far from this winning formula. This year, AMD has spent time improving both the performance and efficiency of its processor and graphics cores.
The processing cores used in the Trinity chips are based on the 32nm Piledriver architecture, which is an upgraded version of the Bulldozer technology that underpinned first-generation FX processors. It’s a big improvement on Llano – those parts were based on the creaking architecture that was used in Phenom II-based processors, among others.
There are six desktop Trinity chips in this year’s range, each of which include two or four cores clocked between 3.4GHz and 3.8GHz. We were sent the high-end A10-5800K and mid-range A8-5600K for testing.
While many of last year’s chips didn’t use Turbo Core at all, the entire range can now dynamically increase the clock speed between 200MHz and 600MHz. Despite the speed increases, efficiency improvements mean that the TDPs remain at either 65W or 100W.
The HD 7000 Series GPUs that accompany these CPUs are based on the 40nm Cayman architecture, which formed last year’s Radeon HD 6930, HD 6950 and HD 6970 desktop graphics cards.
The decision to use the Cayman architecture rather than 28nm HD 7000 Series technology means that Trinity misses out on the more efficient VLIW5 instruction set that was introduced with the Radeon HD 7970. However, AMD counters this by updating Trinity with DisplayPort 1.2 support and four-screen Eyefinity compatibility, plus support for 7.1 channel audio over HDMI.
Top of the Trinity pile is the A10-5800K. It’s an unlocked quad-core part clocked at 3.8GHz with a Turbo Core limit of 4.2GHz, and it includes Radeon HD 7660D graphics. Despite the stock speed bump of 800MHz over last year’s top-end A8-3870K, our benchmarks show it isn’t that much faster, with an overall score of 0.76 compared to 0.7 in our application tests.
|Cores (number of)||4|
|L2 cache size (total)||4.0MB|
|Thermal design power||100W|