Bulldozer frequency and thermal output

When developing Bulldozer, AMD pursued the goal of high clock speed, while at the same time “hold-the-ground” on the IPC front. So, has AMD reached its high-frequency goal? Probably not.

Consider the FX-8150 processor: this new, bulldozer-based 32nm chip has a base clock of “only” 3.6 Ghz. I'm stating “only” because the old K10-based, 45nm Phenom X6 1100T has a base frequency of 3.3 Ghz, only 300 Mhz less than the new AMD flagship. Sure, the FX-8150 can turbo-boost to 4.2 Ghz, but this rather high clock can be used only in some very light workloads. The reality is that even when a single thread is running, the frequency often raise to 4.2 Ghz only for some short bursts, quickly returning to the intermediate 3.9 Ghz turbo clock.

However, some overclocking tests show that the Bulldozer architecture is absolutely capable of attain very high clock speed. So, where is the problem? To tell in two word: thermal output. The new FX-8150 processor show quite high thermal output at default clock speed already, but when overclocked, it become a real furnace.

So, total thermal (heat) output is going to be the first, major problem of the new FX-8150 processor.

You can state that this is more a GlobalFoundries's 32nm process problem rather than a Bulldozer's one, and this is true; on the other hand, until recently GlobalFoundries was part of AMD itself. Anyway, AMD must now live with GlobalFoundries fabrication processes for high-end chips, whether they are good or bad.