Inverse Hyper Threading: Explained
Two weeks ago, Intel launched their latest architecture, Skylake. At first nobody knew anything about the architecture as Intel did not released any information, but some documents leaked 2 days ago shows something that the rumor mill calls Inverse Hyper Threading. An interesting technology none the least.
Hyper Threading is Intel`s take at simultaneous multithreading or SMT. This allows the OS to allocate two threads to every physical core. This is a really good technology as it improves the performance in multithreading tasks like video rendering, image rendering and a lot of other scenarios.
The problem with SMT is that it only helps multithreaded workloads, but most of the applications we use on a daily basis are not optimized for multithreaded processor, and most of the time are not using more than one thread anyway as they don`t know how. This is where Inverse Hyper Threading comes into play.
At its core, this technology lets the application see only one big thread made out of all the threads that the CPU has. This improves the performance in day to day tasks by a significant margin. This idea has multiple prototype implementations, most notably VISC and MorphCore.
Some benchmarks done by heise.de show that unlike previous generation CPUs, on Skylake the Performance hoes down as you add more cores to the game. The first 4 points on the x-axis are the physical cores and the next 4 are logical ones. As you will see in the diagram, the single-threaded performance got significantly beefed up.
Basically this is all you need to know, but you may have one more question: “Why they do not say anything about this in their presentation?”. The answer is: “Global Shortage”. If you did not know, Intel suffers now from a shortage of Skylake CPUs, and releasing this to the consumers will increase the demand exponentially.