Talk:Hyper-threading

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Hyperthreading is pure marketing term, which this article does not say clearly[edit]

Hyperthreading is pure marketing term for three totally different things: Either SMT(Core series processors and Pentium 4/Xeon), coarse-grained multi-threading(Itanium) or fine-grained multi-threading(old Atoms). In the core-series "hyperthreading" there is nothing proprietary in the implementation. It's just SMT, which intel calls with it's marketing name. --Hkultala (talk) 22:29, 9 December 2019 (UTC)[reply]

All the technical things about SMT in this article should be moved to the SMT article, and this article should be made into page which explains the high-level idea of multi-threading, and links to these three different multi-threading technologies, and only explains, that which of these three techonlogies are meant be term "hyperthreading" on which processors; Currently this page contains a LOT of misinformation because it talks technical details about SMT which do not hold true for "Hyperthreading" in Itaniumn and Atom processors, whith are NOT SMT. --Hkultala (talk) 22:34, 9 December 2019 (UTC)[reply]

Each logical core can run an independent process (vs can run a thread)[edit]

In the second paragraph, it states that each logical core can run its own independent process. I was under the impression that there can only be one independent process per physical core, but that each process could spawn multiple threads to run concurrently on the different logical cores. Descriptions from Intel specifically stick to stating that only multiple threads can be run, and they don't mention processes, and other sources I've seen have stated that you can only run one process per physical core, but probably someone here knows more than me. The sentence in question: "With HTT, one physical core appears as two processors to the operating system, allowing concurrent scheduling of two processes per core." - Captain-Apone (talk) 09:31, 16 June 2021 (UTC)[reply]

On Linux, there's no scheduling difference between process and thread. In fact, a Linux process is just a container for at least one thread. Thus, discussing threads running on cores makes even more sense in that context. 141.162.101.52 (talk) 20:16, 7 February 2024 (UTC)[reply]

Performance vs. Efficient cores (P-cores vs. E-cores)[edit]

As of this writing, Intel current 12th generation processors that include efficient cores (E-cores) are single threaded--not hyper-threaded. I'm thinking this needs to be at least mentioned in this article, and with links to relative articles including:

Intel Core#12th generation

References/External links:

https://www.intel.com/content/www/us/en/gaming/resources/how-hybrid-design-works.html Macgyver24x7 (talk) 15:35, 9 August 2022 (UTC)[reply]

Processor vs. Core[edit]

This article seems to flip-flop between the term processor and the term core. The terms should be clearly defined and used consistently. I am no expert, but my understanding is that each element that can do work for a thread of execution is a logical processor and that a core can have one or more logical processors, though physical processor is used in lieu of that meaning of core. (To my knowledge, there can be only one or two logical processors per core today, but in the future? Who knows?) Then there's the question of what is plugged into a socket and, today, has multiple cores. Is that a processor? Anyway, clear definitions and consistency of usage will be helpful. 141.162.101.52 (talk) 20:19, 7 February 2024 (UTC)[reply]