GCN Graphics, ARM-Powered Security Processor
And of course, no APU refresh for AMD would be complete without goosing-up its graphics engine. Not only did AMD kick up clock speeds, but they were able to drop in another two additional GCN cores into Carrizo, offering an immediate 20% kicker in performance, clock for clock. This iteration of the GCN architecture in Carrizo is also DX12 compatible, which is an important checkbox item that helps take advantage of the chips HSA-compatible design as well.
Combined with better tesselator performance, a larger L2 cache and an updated instruction set, these additional two, higher speed 3rd generation AMD GCN cores offer a combined claimed 65% increase in graphics performance versus the previous generation AMD Kaveri chip. We should note, previous generation top-end Kaveri SKUs had 8-core GCN implementations, but only in a 35W TDP chip, not 15 Watts. Regardless, these are huge gains and it will be interesting to see how it pans out in real-world game engines. For now, AMD is referencing 3DMark 11 as its watermark, but we should underscore that this is a measurement taken again at the 15 Watt power envelope. Think thin and light notebook performance here in general, perhaps with a lot more gaming muscle than comparable Intel-powered machines. Again though, we'll remain only cautiously optimistic until we see retail product for hands-on testing.
The final piece to the puzzle that AMD needed to work on for their notebook offering was security processing. Crypto algorithms take a lot of CPU cycles to process and with the platform requirements of current gen cryptography like RSA, Sha, and AES, Intel has held an advantage in their architecture for some time now. So it's fitting that AMD has carried over their PSP (Platform Security Processor) engine from their recent Beema and Mullins designs to Carrizo. This is a dedicated 32-bit ARM core with its own secure ROM and RAM area that makes use of ARM's Trust Zone® technology. It's intentionally an island of silicon in AMD's overall architecture, that operates autonomously handling security processing requests, though it does have access to system memory and other on-chip resources. This was a block that didn't make it into AMD's previous generation Kaveri APU but it's a welcome addition that will increase overall system performance. It's also an important feature especially for enterprise applications.