In the next five years, Intel plans to launch no less than eight CPU architectures, keeping its annual cadence intact. Out of those eight CPU architectures, five belong to the high performance segment, and three belong to a completely revamped mobile lineup which is consisted of System-on-Chip designs.
In this article, we will be focusing on the upcoming high performance CPU architectures, i.e. the ones that will grace our notebooks, desktops and servers. The traditional line in the notebook segment will become blurrier with the Haswell (2013) and especially Broadwell (2014), as these parts will be directly attacking the thin-and-light segment held by the "brotherly" Silvermont architecture (i.e. the new Atom).
Sandy Bridge Gets Refreshed: Meet the 22nm Ivy Bridge
Given the amount of changes between Sandy and Ivy Bridge, as Ivy Bridge is more than just a 22nm die-shrink of the existing Sandy Bridge processor. Intel is bringing in several changes to the mix, such as official support for DDR3-1600 (is there anyone who is not running Sandy Bridge with DDR3-1600 already?) even with ECC switched on (for Xeon), and extending the vPro functionality with the eighth generation of AMT (Virtualization, Trusted Computing Platform).
Furthermore, the graphics part is completely new, up from current Gen6 to new Gen7. That’s right, no more partial DirectX 10 support, Ivy Bridge carries a DirectX 11 implementation. The performance will be significantly increased, but if you ask us "is it enough" – we’re not quite certain. One of biggest issues Intel face are the drivers, as the default apology "our graphics are not for hard-core gamers" fails badly when the CPU cannot even start mainstream games such as The Sims or Civilization V without a mass display of glitches.
Gen7 brings OpenCL 1.1 support as well, while the popular Decode/Encode/Transcode feature that debuted with Sandy Bridge is being improved as well. Overall, Ivy Bridge is an evolutionary step, and we welcome the integration of USB 3.0 in the chipset.
Haswell Architecture: 22nm Haswell Tock and 14nm Broadwell Tick
No doubt about it, Haswell codename had its fair share of shifts and changes. Originally, the Haswell codename was used for one of the successors to Tejas, a canceled representative of the NetBurst architecture. Haswell was also supposed to be the second, then the first part with Larrabee graphics. Today, Haswell is actually the fourth different product carrying the same codename.
Haswell is a new CPU architecture which brings further performance optimizations across the board. This is not a completely new CPU architecture, though. The last big reset Intel did with the Conroe architecture back in 2006. Majority of effort Intel did with Nehalem, Sandy Bridge and now Haswell is to optimize the package and increase the performance inside that package as much as possible.
The graphics part is being upgraded to DirectX 11.1 and this is going to be the first native-Windows 8 design. Three displays and support for 4K resolution are being inherited from Ivy Bridge and you should expect decent graphics performance, even though Intel will focus on lower-power system memory.
Haswell is Intel’s first take on dividing the product strategy into discrete and System-on-a-Chip parts. As we reported earlier, Haswell is going to split into an Ultrabook part which is going to be pitched as an SoC. In reality, Haswell will be a MCM (Multi-Chip Module), single substrate playing home to a CPU and PCH (Chipset) silicon dies.
That is just the beginning though. Haswell will utilize new mobile (947 pin) and desktop sockets (1150 pin), which will stay the same through Broadwell "tick".
Debuting in 2014, Broadwell will be the first part manufactured in brand new 14nm technology and represents a further optimization of the Haswell design. According to our sources, Broadwell brings the first true System-on-Chip design, integrating features such as Ethernet, Thunderbolt or USB 3.0 – all inside one single package. This part is already being widely discussed by Far Eastern motherboard vendors as a margin killer, since unless something changes, there won’t be room to earn money off building parts for that platform.
Skylake Architecture: 14nm Skylake Tock and 11nm Skymont Tick
There isn’t a lot of information available on the 14nm Skylake CPU architecture (2015), sans the completely new CPU core and GPU core, which will work as one. What Larrabee architecture failed to do, Skylake intends to fix, bringing DirectX 12(?) support straight through the CPU pipeline. There is no doubt that Intel will be first to 14nm process, just like the company was with practically every manufacturing process out there. Skymont will be a 11nm "Tock", a die-shrink and a component refresh coming out in mid-2016.