• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Heard a rumor, want to know to know if it's true

B

brett1

Member
This one guy recently told me he read about some upcoming Intel technology that would trick older programs that only support one core into supporting all cores of the processor. I asked him to give me a link to what he read or an Intel codename or some such but conveniently he could not find it.

Does anyone know if this is true? Is such a technology coming out anytime soon? I play several old games that are very CPU dependent and do not take advantage of multicore processors.

thanks
 
I don't see how this would be possible, as you can't just arbitrarily take single-threaded code and magically make it multi-threaded. You have to specifically code for these things in order to keep the order/flow of logic the same, otherwise you'll have some parts of the code complete before others. How do you pick which parts of the code to run on one core vs the other from a completely arbitrary standpoint?
 
Originally posted by: brett1
This one guy recently told me he read about some upcoming Intel technology that would
Click to expand...

your friend is mixing technologies.

AMD was suposidely to have a feature on the new cpu's that would allow this. But it failed.

And if Neha does indeed have this feature, its under NDA so there not allowed to discuss this even if it did have the feature.

So only time can answer this question.
 
Originally posted by: brett1
This one guy recently told me he read about some upcoming Intel technology that would trick older programs that only support one core into supporting all cores of the processor. I asked him to give me a link to what he read or an Intel codename or some such but conveniently he could not find it.

Does anyone know if this is true? Is such a technology coming out anytime soon? I play several old games that are very CPU dependent and do not take advantage of multicore processors.

thanks
Click to expand...

The concept is real and it is called Mitosis. Also known as "speculative processing".

Anandtech had a nice write-up about it back in 2005: http://www.anandtech.com/trade...howdoc.aspx?i=2507&p=7

Turning Single into Multi-Threaded with Speculative Threading

Intel had a particularly interesting research project being demonstrated called Mitosis, a hardware and compiler solution to implementing speculative threading.

On modern day Out of Order microprocessors, the CPUs themselves will speculatively execute code based on what it thinks will need to be executed in the future - thus improving processor utilization and overall performance. This research project proposes that the same sort of speculative execution be applied on a thread level, meaning that threads are created on the fly and speculatively executed by idle cores in a system in order to improve performance on future multi-core CPUs.

The Mitosis project relies on both hardware and software (compiler) support to work. First, on the software side, blocks of code that have very few inputs and outputs are detected and considered for use as a separate thread.
Click to expand...

Now Mitosis was supposedly dropped by Intel once the power consumption issues of 90nm and newer processors came to the forefront of marketing's eye.

To my knowledge Mitosis is a dead project at Intel, no CPU's are intended to actually put the concept it to work. This is where your friend's information is very much incorrect.
 
Guys, Mitosis is a research project. There currently are no plans at this time to release a product.

For one thing, we don't have enough cpu cores yet.
 
Originally posted by: Nemesis 1
I would very much like to see a link were Intel has dropped Mitosis.
Click to expand...

It's the opposite way around, show me a link that actually shows Intel talking much about it since that "debut" in 2005 IDF.

Why would Intel bother to expend resources to come right out and state "we killed the mitosis program"? But the near total lack of discussion on it basically leaves one to infer that it is dead. (my take on it)

I can't think of a single product which Intel has brought to market where once the rumors and speculation on the product started to build that the "hits" on the topic didn't continue to increase and increase until the eventual products release. Nehalem and Larrabee for example, no shortage of ever increasing amounts of Intel driven talking points for these. Same with tick-tock, EUV, etc.

Where's the consistent "drip, drip" of info on Mitosis since its "debut" in the public knowledge space at IDF 2005? It's a never mentioned topic whenever Intel takes the stage now, which historical means only one thing...
 
Really If intel doesn't talk about it . It doesn't exist. So when I was saying befor any info was released about it that. Intel would do High K and metal gates on 45nm. I got flamed alot over that statement but Intel did say 45nm High K and I believed. I also believe 3D silicon will be on 32Nm . But there are people who all ready know this as intel has shown the 32nm wafer. .

So if intel doesn't shout about it it doesn't happen . Now we have known about Larrabee for some time . Tell me everthing Intel has said about larrabee. NOT alot at all . Hopefully in 30 days we know much more . Mitosis when it comes won't be announced befor hand.

If you notice in my graph. linked. Their is alot more info than the AT article you linked. Than it also speaks of terra scale . So that kinda shoots the P4 theory down . If you read the link with info . You will plainly see why Mitosis is a better fit with terra scale than it is for single or dual processors.

You already said mitosis was killed because of what ? C2D . Not likely . If you would read the second link . all of it . You will see that ist a later paper than 05. Mitosis will be better served on terra scale cpus. Now how many cores are required to = terra scale is unknown to myself. Does anyone know the intel defination of what terra scale is ? Core wise.
 
Originally posted by: brett1
This one guy recently told me he read about some upcoming Intel technology that would trick older programs that only support one core into supporting all cores of the processor. I asked him to give me a link to what he read or an Intel codename or some such but conveniently he could not find it.

Does anyone know if this is true? Is such a technology coming out anytime soon? I play several old games that are very CPU dependent and do not take advantage of multicore processors.

thanks
Click to expand...

That would be reverse hyperthreading. The idea is to take several N-wide issue cores, and parallelize them to appear to be a single MxN wide core. Essentially it takes a single execution thread, and spreads it out over the execution resources of multiple cores on a CPU.

Rumored to exist in the C2D design, according to early Intel BIOS option called "Core Multiplexing", that was mysteriously taken out and never fully explained.
 
If you would . Iwant to lay out what we know and match them up to when and were we might see mitosis.

1). Mitosis works well in a terra scale system .

2) Simple x86 cores small and fast .

3)1 Speciality core.


Intel at last IDF said Larrrabee was the first terrra scale project.

larrabbee has 32 simple inorder x86cpus x 4 threads=128threads . I think this would be enough for mitious to work.

Larrabee has 32x86 cores 1 speciality core A large vertex engine a new kind of large shared cache.


Could someone please explain again what it is that the new Hydra chip does. It works off of software and hardware which means a compiler is involved. So this is like a what say a Speciality chip that is worth much more as such than something say you would addto the General cpu. Now go look at the graph In my link and re read what mitosis chip is .
 
I had to add this you got to love the guy who came up with these names.

Mitosis = Xcell division

Hydra= Dragon Jason fought and every time ya cut of one of the seven heads more would replace it. Than of course the hydras teeth.


.



 
Originally posted by: Nemesis 1
Now go look at the graph In my link and re read what mitosis chip is .
Click to expand...

Nemesis all that the Intel article you linked to is talking about is the basics of existing and long known trade-offs in parallelization when dealing with fine-grained versus course-grained computing in Beowulf-class supercomputers.

Sure Intel is hypothesizing in the article about the prospects of putting "Beowulf on a chip". But they are not talking about subverting any of the known limitations (Amdahl's Law, Alamasi and Gottlieb's Law, etc) of parallelized computing. Hence the analysis of trade-offs for core-size versus throughput per core.

Only at the absolute very end of the article do they toss a few words towards Mitosis.

The "special purpose hardware" is basically an attempt to improve upon the known latency and bandwidth issues of core-to-core communications, even within the same die. In beowulf-class systems this is generically referred to as the network fabric. The "graininess" of your program determines how critical the network fabric is to the total system's performance. Special hardware is already in use today in all high-dollar Beowulf-class supercomputer installations. (quadrics is but one example)

I spent nearly a decade building and optimizing Beowulf-class systems for computational chemistry calculations. Trust me when I say there isn't really nothing knew under the sun here in your links to Intel, save for the hardly unexpected possibility of have an advanced enough process node (the Intel paper assumes a 22nm node is available) so as to enable the creation of a single-die beowulf-class system on a chip.

Nothing about this validates or invalidates the concept of Mitosis. Mitosis can exist in tomorrow's CPU's as sure as branch speculation exists today. You don't need >2 cores to make it happen, but the more cores you throw at the speculation (the more hands a single player seats at a single blackjack table) the more likely one of the cores will be to have the correct answer.

The problem with this, as always, is that it comes back to power-consumption. Turn on 1 core and get your answer in 5min at 1W or turn on 100 cores and get your answer in 1min at 100W...how many customers need that answer 4min sooner at 100x the power consumption? The answer is >1, but is it enough to create an entire mass consumer volume product?

To me the end conclusion here to Mitosis is obvious, as obvious as the end conclusion to Intel's foray into the netburst architecture.
 
I do remember reading an article describing architecture that would use idle cores to add unused cycles to the active core...can't remember if it was Intel or AMD, or for which generation it was supposed to be implemented.

-z
 
i'd be happy if they just tricked older programs into handling multiple cores without crashing. specifically, thief 1, 2, and 3. i have to set affinity for all 3 or else i cant even play them.
 
Originally posted by: fishjie
i'd be happy if they just tricked older programs into handling multiple cores without crashing. specifically, thief 1, 2, and 3. i have to set affinity for all 3 or else i cant even play them.
Click to expand...

That's easily done (assuming they know how) by the game developer, with a patch. I don't think it's possible at the hardware level, and if it's possible, it would likely be expensive to implement, and most likely would sap performance, I'd guess.
 
Correct me if I'm wrong, but I remember reading in another thread either here or @ XS (though for the life of me I can't remember it) that mitosis was done completely at a software level, not hardware?

Anyone have any more info on this

Thanks.
 
Originally posted by: VirtualLarry
Originally posted by: brett1
This one guy recently told me he read about some upcoming Intel technology that would trick older programs that only support one core into supporting all cores of the processor. I asked him to give me a link to what he read or an Intel codename or some such but conveniently he could not find it.

Does anyone know if this is true? Is such a technology coming out anytime soon? I play several old games that are very CPU dependent and do not take advantage of multicore processors.

thanks
Click to expand...

That would be reverse hyperthreading. The idea is to take several N-wide issue cores, and parallelize them to appear to be a single MxN wide core. Essentially it takes a single execution thread, and spreads it out over the execution resources of multiple cores on a CPU.

Rumored to exist in the C2D design, according to early Intel BIOS option called "Core Multiplexing", that was mysteriously taken out and never fully explained.
Click to expand...

Reverse Hyperthreading? No such thing.

Stop reading the Inquirer, you will be better off for it.

 
1 possible way to solve it would be :

I would think that since all Intel chips overclock so well.

And in order to stay within the thermal envelope, the processor would shut down it's unused cores completely and clock 1 core higher automatically. Also all the shared resources between the cores would be dedicated to that single core. But if you have a modern os that already uses the advantages of multiple cores this would not work. Unless maybe there can be some coöperation between the OS and processor through a driver to let the processor know how the program is threaded.
The OS would have easy knowledge about it and force the processor in the "single core" state for as long as the thread is running.

Would that be viable ?

 
You must log in or register to reply here.

TRENDING THREADS

Back
Top