Intel CEO Confirms SMT To Return to Future CPUs

[Cpt.Jank]

Intel today announced its Q2 results, and it was a bit of a mixed bag, with the earnings largely down and projections showing little overall growth for the foreseeable future. Ahead of this announcement, though, Intel's CEO, Lip Bu Tan, sent an internal memorandum to employees, which has since been made public, detailing his plan to "step in the right direction." While much of this revolved around AI, its foundry business, and job cuts—a 15% cut in overall head count and a 50% cut in management layers—one of the more relevant takeaways for the PC enthusiast community is that Intel will be reintroducing SMT or Hyper-Threading to its processors in the future.

It's unclear whether the change to reintroduce SMT, aka as "Hyper-Threading", will only apply to the data center, since it was only called out when talking about that segment, but generally the desktop and data center are not too far apart in terms of architecture, so it stands to reason that if SMT returns to Intel's data center CPUs, it will likely make its way back to the desktop, too. Intel first started moving away from Hyper-Threading around the 12th-Gen CPUs, around the same time as the company moved towards asymmetrical CPU designs, first enabling SMT on only performance cores and later ditching Hyper-Threading entirely in the 15th-Gen CPUs. It's already clear that reintroducing SMT won't happen with Nova Lake, or likely even its successor, and as we've recently shown in our testing, it isn't always a necessary feature for high performance—especially in gaming—but most PC enthusiasts would probably rather have it and be able to disable it than outright not have the option.


The direct quote from Lip Bu Tan's letter to Intel employees reads:

Revitalize the Intel x86 Ecosystem

We will focus on growing share in our core client and server segments. To that end, I am working closely with our product and engineering teams to strengthen our roadmap.

In client, Panther Lake is our top priority as it will reinforce our strength in notebooks across consumer and enterprise. We also must drive continued progress on Nova Lake to close gaps in the high-end desktop space.

In data center, we are focused on regaining share as we ramp Granite Rapids while also improving our capabilities for hyperscale workloads. To support this, we are reintroducing simultaneous multi-threading (SMT). Moving away from SMT put us at a competitive disadvantage. Bringing it back will help us close performance gaps. We are also making good progress in our search for a permanent leader of our data center business, and I plan to share more on that this quarter.

Across client and data center, I've directed our teams to define next-generation product families with clean and simple architectures, better cost structures and simplified SKU stacks. In addition, I have instituted a policy where every major chip design is reviewed and approved by me before tape-out. This discipline will improve our execution and reduce development costs.

View at TechPowerUp Main Site | Source

 

[Claudio]

I'm afraid you've got it a bit confused... HT is present on Intel processors up to the 14th generation. It's not present on the 15th (ULTRA)...

 

[john_]

but most PC enthusiasts would probably rather have it and be able to disable it than outright not have the option

If I am not mistaken, it makes cores bigger. I believe the idea behind removing HT, was to make P cores smaller, living room for more P cores or even more E cores. And because consumers can somewhat understand cores, but not threads, from a marketing perspective advertising 40 cores instead of 40 threads was preferable.

 

[JohH]

CEO noted this while talking about data center. Diamond Rapids will be a one off without it.
But HT is not returning to client designs. No product on their latest client roadmap has it. If it is returning it's 2029+.

 

[JustBenching]

Unless they got some breakthrough with quad way SMT I don't see the point. E cores already cover more than adequately the MT performance part of SMT, what else do we need that thing for?

 

[RejZoR]

Thing is, stuffing more E cores into desktop CPU's doesn't really help gaming since they are still too crap to run games primarily on them. Where using HT/SMT on larger fast cores, they are still large fast cores where you want to run the game.

I don't think Intel can get around the fact that games prefer large fast cores with fuck ton of L3 cache to make framerate go up. I don't know the numbers, but I'd say SMT is helping 9800X3D and not nerfing it like it was years ago where SMT often wasn't beneficial. I think those days are long gone.

 

[iameatingjam]

I'm afraid you've got it a bit confused... HT is present on Intel processors up to the 14th generation. It's not present on the 15th (ULTRA)...

And this is being shown on the front page. Yikes. Perhaps he was thinking like... meteor lake? Or maybe he was considering RL and RL refresh the same gen, which is kinda fair I guess, since they are the same silicon, but if we are to do that, then we might have to consider removing some other intel gens too and that would get messy and really confusing.

But yeah its probably just a mistake.

 

[JustBenching]

Thing is, stuffing more E cores into desktop CPU's doesn't really help gaming since they are still too crap to run games primarily on them. Where using HT/SMT on larger fast cores, they are still large fast cores where you want to run the game.

I don't think Intel can get around the fact that games prefer large fast cores with fuck ton of L3 cache to make framerate go up. I don't know the numbers, but I'd say SMT is helping 9800X3D and not nerfing it like it was years ago where SMT often wasn't beneficial. I think those days are long gone.

Smt does indeed help the 9800x 3d (in heavy core usage games) but that's because it has a core deficit. If zen 6 comes with 12 core CCDs i guarantee SMT will be entirely pointless, the 10800x 3d with 12 cores in a single ccd will be faster with SMT off in every single game.

Also its not the case that ecores dont help in games. They definitely do, not so much in averages but 1% lows go way up with ecores on.

 

[_roman_]

Thing is, stuffing more E cores into desktop CPU's doesn't really help gaming since they are still too crap to run games primarily on them.

Some Users, maybe some Intel cpu owners, will disagree on that point.

I'm not fond of E-Cores.

--

I do not want to see SMT. Too many security issues. More cores with same instructions with less cache may be the better approach.

 

[LittleBro]

SMT is beneficial, on AMD platform you get 15% more perf. at no increase in power draw.
Intel was stupid to abandon SMT (HT). Yep, there were multiple security issues present but
instead of abandoning whole SMT technology, they should have make a proper redesign.
Which will they do now. Better late than never, right?

IMHO, SMT is better than e-cores/LP cores, as it:
- does not require additional physical space;
- does not differentiate between instruction sets;
- scheduler issues are much lesser than with multiple types of cores.

By the way, WCCFtech is spreading a rumor that Intel will abandon e-cores and LP cores with generation after Nova Lake.
Having 16-24 good old fashioned P-cores with SMT might be a future. Intel needs another Sandy Bridge effectivity miracle.
i7-2600K was such an amazing processor.

 

[FoulOnWhite]

8 or 16 real cores, no SMT please.

 

[iameatingjam]

- does not require additional physical space;

It does require some space.

 

[LittleBro]

It does require some space.

It's negligible compared to space required by physical cores.

 

[iameatingjam]

It's negligible compared to space required by physical cores.

True. It doesn't require much space, as compared to how much benefit you get in some multi threaded tasks, hence why its been used for so long. Just wanted to clear up that it isn't completely free space wise.

Whether its worth it or not... depends on chip, how many cores, and the job.

 

[BoggledBeagle]

SMT is beneficial, on AMD platform you get 15% more perf. at no increase in power draw.
Intel was stupid to abandon SMT (HT).

So untrue...

You get the SMT benefit ONLY when you have an application that can fully load all the threads, because at least on Intel CPUs the SMT threads are emloyed as the last ones as the intensity of the task increases, because they are the weakest. For normal people running normal tasks the SMT has no benefit at all.

It DOES draw more power, requires more die area and of course brings more complexity with all the connected drawbacks.

 

[LittleBro]

So untrue...

You get the SMT benefit ONLY when you have an application that can fully load all the threads

False.

, because at least on Intel CPUs the SMT threads are emloyed as the last ones as the intensity of the task increases, because they are the weakest. For normal people running normal tasks the SMT has no benefit at all.

With SMT the job can be done quicker and thus reduces overall power required to do the task (cumulated power draw over time).
Also, don't judge AMD's SMT implementation by Intel's one.

It DOES draw more power, requires more die area and of course brings more complexity with all the connected drawbacks.

I strongly suggest you to read this article: https://www.phoronix.com/review/amd-ryzen-zen5-smt/8

 

[Vya Domus]

I'm afraid you've got it a bit confused...

Everyone is confused, including Intel it seems, they remove and add features without much thought.

 

[Tyler-98-W68]

Thing is, stuffing more E cores into desktop CPU's doesn't really help gaming since they are still too crap to run games primarily on them. Where using HT/SMT on larger fast cores, they are still large fast cores where you want to run the game.

I don't think Intel can get around the fact that games prefer large fast cores with fuck ton of L3 cache to make framerate go up. I don't know the numbers, but I'd say SMT is helping 9800X3D and not nerfing it like it was years ago where SMT often wasn't beneficial. I think those days are long gone.

Ever heard of BeamNG Drive? I disabled all but 1 p cores and used the 16 ecores in that game and it ran extremely well. E cores are not capable like most think and the arrow lake e cores have some serious power to them

 

[N/A]

If SMT is only a 15% area saving for 20% less performance, just add real cores.

 

[RejZoR]

Ever heard of BeamNG Drive? I disabled all but 1 p cores and used the 16 ecores in that game and it ran extremely well. E cores are not capable like most think and the arrow lake e cores have some serious power to them

Because the game isn't graphically demanding, it's computational because of physics applied to rigid bodies and there all it maters is moar cores. Where with actual games you need fast cores and just enough of them to drive everything just right. You can't compare that.

Some Users, maybe some Intel cpu owners, will disagree on that point.

I'm not fond of E-Cores.

--

I do not want to see SMT. Too many security issues. More cores with same instructions with less cache may be the better approach.

It's odd how Intel hasn't made GAMING processor with just big cores and crap load of cache. Surely their new approach can just glue whatever together with tiles...

 

[JustBenching]

SMT is beneficial, on AMD platform you get 15% more perf. at no increase in power draw.
Intel was stupid to abandon SMT (HT).

Well of course a cpu that was designed with smt in mind will do worse when you disable it, lol. The question is, is it worth the 5-10% extra die space that it requires or would you prefer something else for that die space

 

[BoggledBeagle]

False.

With SMT the job can be done quicker and thus reduces overall power required to do the task (cumulated power draw over time).
Also, don't judge AMD's SMT implementation by Intel's one.

I strongly suggest you to read this article: https://www.phoronix.com/review/amd-ryzen-zen5-smt/8

You seem to be strongly confused.

Saying that something is false without telling why is a waste of time.

What "job" are you talking about? Be specific. If you look on the broad spectrum of normal PC users, how many people could benefit from it and in what extent?

Why are you even mentioning AMD, this thread is about Intel. I own/ed Intel CPUs and I carefully observed how thread utilisation works and what is the benefit of HT and it is exactly how I explained. May I ask you what experience do you have with Intel CPUs?

 

[JustBenching]

So untrue...

You get the SMT benefit ONLY when you have an application that can fully load all the threads, because at least on Intel CPUs the SMT threads are emloyed as the last ones as the intensity of the task increases, because they are the weakest. For normal people running normal tasks the SMT has no benefit at all.

It DOES draw more power, requires more die area and of course brings more complexity with all the connected drawbacks.

It doesnt really draw more power, for the same amount of power (say 150w cap) the smt on cpu will be faster than the smt off. It will run on lower clocks of course, but it will still be faster since smt will outweight the lowered clockspeeds.

 

[rattlehead99]

SMT is beneficial, on AMD platform you get 15% more perf. at no increase in power draw.
Intel was stupid to abandon SMT (HT). Yep, there were multiple security issues present but
instead of abandoning whole SMT technology, they should have make a proper redesign.
Which will they do now. Better late than never, right?

IMHO, SMT is better than e-cores/LP cores, as it:
- does not require additional physical space;
- does not differentiate between instruction sets;
- scheduler issues are much lesser than with multiple types of cores.

By the way, WCCFtech is spreading a rumor that Intel will abandon e-cores and LP cores with generation after Nova Lake.
Having 16-24 good old fashioned P-cores with SMT might be a future. Intel needs another Sandy Bridge effectivity miracle.
i7-2600K was such an amazing processor.

The thing is that Hyper-Threading gives as much performance as the core is bottlenecked. Meteor Lake got 7% in integer and 13-14% in floating point tasks, while Arrow-lake/Lunar-lake are even less bottlenecked.
Hyper-threading is a crutch and is usually there to maximize profits by allowing for simpler and smaller cores, but of course requires software optimization and creates vulnerabilities with all the exploits we've had since 2017.

Intel won't get another Sandy Bridge as it was entirely carried by 32nm being a massive improvement over 45nm and AVX. Realistically outside of AVX Sandy had only 8-11% higher IPC, but it had 25-30% higher clock speeds due to the 32nm process node. And of course having native AVX sped up everything that needed it by a ton.

Well of course a cpu that was designed with smt in mind will do worse when you disable it, lol. The question is, is it worth the 5-10% extra die space that it requires or would you prefer something else for that die space

Meteor-Lake got up to 7% integer performance and up to 13-14% floating point performance from Hyper-Threading, that's because the core was not bottlenecked much, Lunar Lake and Arrow Lake(Lion Cove and Skymont) are even less bottlenecked so the benefit would be even smaller.

Everyone is confused, including Intel it seems, they remove and add features without much thought.

There is thought behind it. Patt wanted to have the best architecture, the new CEO wants to maximize profits.

Hyper-Threading only yields performance to bottlenecked cores, Meteor Lake was already a very well done, almost non-bottlenecked core. Thus hyper-threading gave only up to 7% integer performance and up to 13-14% floating point performance for the cost of 15% die size.

Arrow Lake(Lion Cove and Skymont) is even less bottlenecked, so that means that it would gain even less performance from Hyper-Threading.

On the other hand Zen cores are very bottlenecked and they gain 15-30% performance.

Also Hyper-Threading is a major vulnerability.

If SMT is only a 15% area saving for 20% less performance, just add real cores.

On Meteor Lake it was 7-14% more performance for 15% more area, because the core was not bottlenecked much. Arrow Lake cores are even less bottlenecked so it would give even less performance.

 

[N3utro]

This guy got it all wrong. When you are behind competition you need to double your efforts to be able to catch up, not cut jobs.

Who's gonna make the innovations intel needs to get back in the game? The ghosts of fired people?