Ivy Bridge Architecture
The graphics engine has been overhauled and now is DirectX 11 compatible. The GPU core is armed (depending on mode) with sixteen EUs -- Execution Units (another synonym for unified processors like CUDA cores / Execution Units / Shader Cores / Stream processors). Clarkdale had eight of these, Sandy Bridge has twelve with Ivy Bridge containing four more EUs.
Each Execution Unit can manage more work per clock opposed to the older ones as well. Now there are two different GPU configurations as Intel applied a dual-graphics core to segment their products, available with 1 core (Intel HD Graphics 2500) and 2 cores (Intel HD Graphics 4000). A single GPU core holds 8 EUs and some of Ivy Bridge Core i5 and higher products will have two of these cores, so we'll see configurations with 8 (Core i3) or 16 (Core i5/i7) EUs.
So one last time, as you can understand from the specifications, don't expect heaps of gaming and fragging fun (though very simple games should be playable) but see this more as a desktop integration and implementation for very good Windows usage and, importantly, high-definition 1080P transcoding, decoding and acceleration.
The AVX (Advanced Vector Extension) Instruction Set
Certain improvements that will boost its performance in general applications by around 20% compared to Core i7 "Sandy Bridge" chips can be found in enhanced AVX acceleration. AVX stands for Advanced Vector Extensions. AVX is very well suited for floating-point-intensive applications and has become really popular.
AVX provides features and instructions. The easier way to grasp AVX is to look at it as a new sophisticated form of SSE. Intel opened up the data path which has been expanded from 128 bits to 256 bits, the two-operand instruction limit is increased to three operands, and advanced data re-arrangement functions are included. Now, AVX doesn't just add new instructions, Intel states it allows the CPU to execute older instructions more quickly over the AVX extensions as well.
AMD also has it planned for all their next-generation processors. So let's put it really simply; SB and IB processors will support running x87 FPU code which is capable of up to 2GB/s of double-precision throughput per core. Now here is the trick, the very same amount of work once written to be supported over the AVX extensions boils down to a 4x performance increase (theoretical) so that's 8GB/s per core (double precision), a fourfold increase in performance.
Software that caters for and supports AVX will show huge performance incensement, transcoding for example. In our review we'll show you a trick or two with AVX and Media Espresso video transcoding. Nothing to date can beat the phenomenal speed as we'll show you.
