• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD: Anti-Hyperthreading function due

mrk

mrk

mrk

Man of Honour
Joined
18 Oct 2002
Posts
103,759
Location
South Coast
AMD are reportedly working on a sort of ‘Anti-Hyperthreading’, which could allow multiple physical CPU cores to emulate one physical core. The problem with dual (and soon quad) core processors like the AMD X2 range is that most applications are still written for single core processors, so do not make effective use of multiple cores.

According to Bit-Tech.net AMD’s new platform, AM2 won’t be able to compete with Intel’s Conroe offering in terms of raw native performance, but this feature would give them a huge edge in single core applications. The idea of taking two pieces of hardware and making them function as one isn’t new. Hard drives do it in Raid 0 configuration, and graphics cards also emulate one card with two via crossfire or SLi. The performance gains with these can equal over 80% in added performance, so the potential of an Anti-Hyperthreading CPU could be huge.
Click to expand...

thinking of the "now" is a good idea =)

http://www.neowin.net/forum/index.php?showtopic=453169&st=0
 
sounds like itl be a nice little stop gap but as soon as more programs are multi threaded it will become a useless feature
 
blitz2163 said:
sounds like itl be a nice little stop gap but as soon as more programs are multi threaded it will become a useless feature
Click to expand...


There's a long way til that happens, it's like the AGP situation now where it's proven that AGP can still deliver and there are lots and lots of people using it.
 
Theres also the problem that if software developers prefer the AMD one then thats what will take off. Still, id say that multiple threading is the "proper" method. Just like like the Itanic (Itamium :P) was the proper method. . . . . .
 
see i agree with both sides, the fact that right now we dont have many if any multi threaded applications allows this ''anti-hyperthreading'' method to come into action and provide users power in single threaded applications, which will prove usefull. On the other hand it will eventually become obselete due to multi threaded applications, but no where did it say that these cores could not support multi threaded applications therefore this core must beable to do both and would eventually even the odds of this Intel Conroe and AMD AM2 battle.
 
The rumours are exaggerated. Basically AMD are working on instruction level parallelism (ILP) improvements for the K10 architecture. Similar to the ones that Conroe has. Nothing new there. Every architecture improves its ILP in some way over its predecessor.

It's impossible to split a single thread's execution across multiple cores. It just cannot be done. It would have to defy physics. The only way it can be done is by identifying certain instructions and running them in parallel in the pipeline (or the pipeline of the other core), then collating their results at the end. This is nothing new, it is called superscalar architecture and has been used on X86 chips since the Pentium Pro.

It sounds to me like the AMD spokesman was trying to describe how superscalar works and blurted out the phrase "it's like the inverse of Hyperthreading" as a running shot at Intel. This phrase then probably got exchanged between several reporters before finally being published on some website as "Stop the press! AMD preparing whole new architecture with anti-Hyperthreading technology for the K10". Funny :D
 
Last edited:
But isn't there theoretically a way for multi cores to be linked as one if there is no multithreading in an app?

I mean if there is an actual physical connection between the 2, register to register, would this be able to set-up?

Or am I talking rubbish?
 
Soul Rider said:
But isn't there theoretically a way for multi cores to be linked as one if there is no multithreading in an app?

I mean if there is an actual physical connection between the 2, register to register, would this be able to set-up?

Or am I talking rubbish?
Click to expand...
Register to register, definately no.

You can however have a "multi-core aware" pipeline. So instead of a pipeline job only being limited to the core it was started on, it can also wait for a pipeline slot on the other core(s). This is what AMD is talking about when they say "inverse of Hyperthreading".

The Hyperthreading present on P4's simply exposes a virtual core to the operating system. The operating then schedules threads to this virtual core and the P4 can then better keep its large pipeline filled with the work of two threads.

It certainly is a promising technology. But it is being way overexaggerated by news sites and discussion forums, for the simple reason that only a limited selection of instructions can be parallelised. I would also not be surprised at all if Conroe has this inside it already. Intel let it slip a while ago that NGMA does actually have a "form of Hyperthreading" but that it was disabled for the time being. Hyperthreading technologies (especially those that span multiple cores such as AMD's) are notoriously difficult to test and validate which is why Intel disables the functionality in the first revisions of its chips.

I really hate the wording "anti-Hyperthreading", it makes no sense. If anything this is simply Hyperthreading Version 2.0, for a multi-core world.
 
Last edited:
Sounds like marketing BS, any `driver` that made this possible, and it would be a driver, would probably add so much overhead, it would cancle out any realworld gains in performance.
 
Lanz said:
Sounds like marketing BS, any `driver` that made this possible, and it would be a driver, would probably add so much overhead, it would cancle out any realworld gains in performance.
Click to expand...
Why would it be a driver? As I said the only way this can be done is by adding a multi-core aware pipeline to the chip :)

It's definately possible, some of the IBM Power chips do it, but the gains are nowhere near what people are expecting.
 
I didnt read the article (took more than 5 seconds to load when i clicked;), i was presuming the `function` was for X2's. Hmm surely by the time this gets developed and added to the core, most new stuff will be multithreaded anyway.
 
It doesn't matter if everything is multithreaded by the time the technology arrives, there is still a performance gain to be had (albeit fairly small) even with multithreaded software. Don't read too much into the hype surrounding this, i.e. "it turns single threaded software into multi threaded software!!!!1111oneoneone" As it is largely false and like I said, exaggerated :)
 
I think Vista will be the daddy for Duel Core / Quad core / 64 Bit and SATA2 Hard Drives technology. I am hoping it is going to be they have the chance, AM2 is just a DDR2 memory controller on die and the Conroe will eat it alive. Even though I am an AMD fan boy it is going to happen unless AMD get to work on that 0.65nm (Afaik just a die shrink). Am I allowed to post a link to another forum that has live Conroe results?
 
Sam666 said:
Am I allowed to post a link to another forum that has live Conroe results?
Click to expand...
Go for it. As long as it complies with the forum rules and isn't a link to that "Sharikou, Ph. D" blog that is doing the rounds.
 
You must log in or register to reply here.
Back
Top Bottom