First A10-5800K "Trinity" CPU benchmarks surface
Next generation of "Trinity" Accelerated Processing Units (APUs), coming later this year, will be based on Piledriver (enhanced Bulldozer) core. Like current generation of APUs, Trinity processors will have from 2 to 4 cores, integrate DDR3 memory controller and discrete-class Graphics Processing Unit (GPU) on a chip, and lack L3 cache. AMD claims that Piledriver core will improve Bulldozer microarchitecture performance by 10% on average, and we should also see significant performance boost in 3D applications due to GCN-based integrated graphics unit, as well as big boost in AES/SSE4/AVX-enabled applications. Earlier this month a user in pceva.com.cn forum posted several benchmarks of the flagship Trinity model, A10-5800K, which is helpful in estimating processor actual performance in comparison to existing Llano APUs. Be warned, though, that the results are not verified, and may or may not represent processor's actual performance.
When released, the AMD A10-5800K is going to be a flagship model with 4 CPU cores, 4MB L2 cache and HD 7660D graphics. The part will be clocked at 3.8 GHz, or up to 4.2 GHz in Turbo Core mode, and can be easily unlocked to achieve higher frequencies. The forum user tested A10-5800K processor in default mode, i.e. when the chip was not overclocked. The CPU scored 23.775 in Super Pi 32M single-threaded benchmark. Reported 3DMark06 results for the APU were 4304 CPU score, and 9396 combined 3DMark score. Pceva.com.cn compared these results with AMD A8-3850, and their conclusion was that the A10-5800K is 9% faster in SuperPi, 12% faster in 3DMark06 CPU score, and has about 50% better overall 3DMark06 score. The results look like a decent improvement over existing Llano processors, however we should question their validity. First, it would be more fair to compare A10-5800K to unlocked A8-3870K SKU, than to A8-3850. More importantly, reported benchmark results for A8-3850 look too low to us. 3DMark06 scores for A8-3850, when coupled with DDR3-1866 memory, should be somewhere in the region of 7000 3DMark points, and 4000 - 4100 CPU points, and SuperPi 32M test should complete in 25 seconds. A8-3870K model should be slightly faster owing to 3.5% higher clock speed. Based on this we estimate that the 5800K has only a few percent better SuperPi 32M and 3DMark06 CPU scores than the 3870K, despite the fact that it has up to 40% higher clock speed. A10-5800K on-chip GPU performs up to 30% better than the 3870K. Saying all that, we think that the results should be taken with a grain of salt, and we should wait and see verified independent benchmarks of production Trinity CPUs on production socket FM2 motherboard, using officially released drivers.
Related News (newer articles):
Oct 02, 2012: AMD introduces retail "Trinity" APUs for desktops
Aug 13, 2012: AMD "Trinity" APUs pictured; rumored to launch on October 1
Jun 11, 2012: AMD reveals details of desktop "Trinity" APUs
Related News (older articles):
Feb 15, 2012: AMD Trinity Desktop APU Specs Revealed
Related CPU families