We were delighted to report about AMD’s Trinity performance in the new and long anticipated computer game, Blizzard’s Diablo III. We’ve mentioned the imbalance between AMD’s chip and Intel’ processor regarding the power consumption and especially the price, but it seems there’s much more to that.
AMD’s top performing Fusion mobile part is called A10-4600M and works at a base frequency of 2300 MHz. The processor can Turbo up to 3200 MHz and is packaged in a 722-pin micro-PGA socket called FS1 (FS1r2). The processing cores are kept fed with data by 4 MB of level 2 cache. That’s 512 KB of level 2 cache for each core. Unlike AMD’s FX processor line, there is no level 3 cache, but being a Fusion processor, AMD’s A10-4600M comes with an integrated graphics processing unit (iGPU) that runs at a default of 497 MHz with the option to Turbo up to 686 MHz when all the 384 Shaders are subjected to a heavy 3D task. All of this is done with a maximum heat dissipation of 35 watts and this is exactly what Intel is lacking.
Most of AMD Trinity reviews have compared AMD’s mobile flagship with Intel’s top performing mobile solution right now and that’s the Core i7-3920XM. It all good that AMD’s Trinity gets around 50% better results in Blizzard’s Diablo III, but the most important aspect is that AMD’s APU does that at a huge frequency disadvantage when compared with Intel’s CPU, on every side. The Ivy Bridge mobile part works at a base frequency of 2900 MHz with a turbo option to 3600 MHz. This is a clear frequency advantage for Ivy Bridge of 26% on base frequency and 12.5% when in Turbo mode.
While AMD’s iGPU has a default frequency of 497 MHz, Intel’s HD4000 starts from 650 MHz and that’s a 31% frequency advantage, but it can turbo up to 1300 MHz, showing a whooping 89% frequency advantage versus AMD’s Turbo mode of only 686 MHz. Therefore it’s surprising to see AMD’s Trinity win real life benchmarks with considerable better results despite Intel’s frequency advantage of 31% to 89%.
It’s clear that Intel’s finer manufacturing process of 22 nm has paid off and that their CPUs are able to run at much higher frequencies while consuming acceptable amounts of power. We’re sure AMD would have raised the frequency of Trinity and would have shown us an Ivy Bridge “killer” if it could have, but they are still building Trinity in the same 32 nm manufacturing process they’ve used for Llano, albeit a more mature 32 nm. Considering the manufacturing process disadvantage along with the resulting working frequency disadvantage, we can only draw the conclusion that AMD’s Trinity is, in fact, a much more efficient architecture than Intel’s Ivy Bridge.
AMD’s APUs are able to do more while using less power despite the less advanced manufacturing process. It is obvious that Dirk Meyer’s bet on more INT and less FPU while hoping for a future GPU based FPU amelioration, was a good idea. The 17 watts AMD Trinity model that will soon surface will cement AMD’s position as a true mobile competitor for Intel and this is practically the first time in AMD’s history that they’ve managed to beat Intel at the mobile game. There was some friction between the companies during the Pentium 4 vs Athlon K8 period, but those were simple cherry picked desktop CPUs and were not a specifically mobile architecture.
During those times, because of Intel’s hot and bad performing Pentium 4 architecture, AMD’s K8 would easily beat the competition at any level, mobile, desktop or server. Once Intel’s Centrino mobile architecture stepped up in the ring, AMD never had a true mobile competitor. Now they do have and, surprisingly Intel has no match. Their i7-3920XM mobile Ivy Bridge consumes a maximum of 55 watts while being likely 400% more expensive. That’s 57% higher power consumption for likely 400% more money as, we’d like to remind you again that Intel’s Core i7-3920XM sells for more than 1000 USD (783 EUR).
There is no quad core Ivy Bridge processor in Intel’s mobile line-up that consumes 35 watts so practically Intel has no real competitor for AMD’s Trinity.
0 comments:
Post a Comment