The Lineup and Technology
We can see from the positioning chart that the base clock of the top end FX-9830 is a fairly impressive 3.0 GHz. and a boost to 3.7 GHz. This is a significant jump up from last year’s FX-8800 which featured a 2.1 GHz base clock and a max boost to 3.4 GHz. The rest of the lineup follows suite by increasing base clocks and boost clocks from the products that we saw introduced last year. While Intel still has an overall CPU performance lead, AMD has closed that gap with these products. Where AMD still has a significant lead is in graphics and video playback performance. Games such as CS:GO, League of Legends, and DOTA 2 will play significantly faster and smoother on the AMD mobile parts than the comparable Skylake parts featuring the HD520 and HD530 units.
AMD is hoping that this series of APUs can help further their marketshare increases in the mobile market. HP has already announced that they will be using this latest gen of APUs from AMD in their HP ENVY x360 convertable laptops. These will be premium models with full HD screens with the option to go to 4K. For AMD’s sake we hope other OEMs follow HP’s lead and provide higher end laptops featuring these APUs. While Carrizo broke new ground for AMD, there is a ways to go before we see them in such premium products like the Dell XPS 13/15 models.
With Bristol Ridge AMD has increased performance while not sacrificing power consumption and battery life. They have been able to do this without utilizing a big process node jump. So how exactly have they been able to achieve that? A whole lot of engineering…
Engineering Innovation for Energy Efficiency and Performance
We have a few pieces of housekeeping to cover before we go into the methods that AMD has employed to achieve the advances they have in performance at the same TDP levels. As I had mentioned before, it is hard to design an APU that can adequately cover the 15 watt to 95 watt TDP that AMD has previously covered with single generation parts such as Kaveri. Chips designed for higher TDPs will have a harder time performing well at low TDPs. Parts designed for low TDPs will have a hard time scaling in performance after a point and the amount of clockspeed improvements quickly becomes overshadowed by the extreme amount of power applied to run at those speeds.
Carrizo and now Bristol Ridge were designed at the outset to be low TDP parts that would not scale well above 65 watts. Carrizo was a significant jump from Kaveri for the low TDP markets and provided all the extra parts to make it a true SoC. Bristol Ridge improves upon CPU performance by around 15% over Carrizo, around 30% in graphics performance, and cuts down power consumption in multimedia playback and productivity apps by around 10%.
The first piece of the puzzle is an improvement in process technology. GLOBALFOUNDRIES has continued to improve their 28nm process, which is in line with what others like TSMC and UMC have been doing with their lines. The latest “BR Process” is able to increase the transistor switching speed as compared to the “CZ Process” at any given power/voltage point. This gives Bristol Ridge a clockspeed advantage right off the bat without increasing power consumption.
AVFS was introduced a couple of products ago and essentially looks at individual portions of a die and determines the optimal power to deliver to achieve a certain clockspeed. There are a bunch of AVFS sensors placed around the die that can take advantage of the strengths and weaknesses of individual die rather than applying a “one size fits all” voltage and speed regime that causes a part to run inefficiently.