These are difficult times for Intel, which seems to have no idea which direction to take in order to bounce back.
First launched by Intel on Xeon processors for servers (in February 2002), Hyper-threading technology was soon rolled out on "client" chips, the Pentium 4 (in November 2002). At the time, very few processors featured multiple cores, and Hyper-threading was designed by Intel to overcome this problem by doubling the number of tasks (threads) managed by each core. In fact, as long as the application in use was able to exploit this extra power, performance could be increased by 50%, 75% or even almost 100%. Logically, the technology met with great success, and continued to do so as more and more cores were integrated into chips. The proof? AMD was quick to jump on the bandwagon, and today continues to use it even on its most powerful server processors, the Threadripper.
When Intel explained the benefits of Hyper-threading © Intel
However, despite this indisputable success, it was Intel itself that created the surprise at the beginning of last year: the American firm considered that it was no longer very useful to offer Hyper-threading, pointing out that the ever-increasing number of cores on the most modern processors made it obsolete. So, with the launch of the Arrow Lake generation in autumn 2024, Intel officially put an end to Hyper-threading, and the Core Ultra 200K should be able to do without it without having to blush at the comparison with the competition. Unfortunately, this was not to be, and today Intel is forced to make amends.
Lip-Bu Tan, Intel's CEO, was quick to point out that " moving away from SMT will have been a competitive disadvantage ". SMT stands for simultaneous multi-threading, and Lip-Bu Tan's words have at least the merit of being clear. Intel therefore plans to bring back the famous technology in its next generations of processors, namely Granite Rapids for servers and Panther Lake for clients. Arrow Lake is unlikely to be remembered.
