Dual Layer Processor

[H2SO4]

so whats up with these "daul layer" processers. Supposedly there two processors in one. Please correct me if im wrong. Intel just recently released them.

 

Would this be any better than Hyperthreading?

[5614]

It's dual core.... it means that there are physically two cores inside one processor, unless you are talking about dual-processing, which is linking two processors on one computer. They are both better than hyperthreading yes.

[Dak]

whats hyperthreading?

 

and are two 1G prosessers better than one 2G prosesser?

 

what about, are two 32bit prosessers better than one 64bit prosesser?

 

and whats the prosessers 'core'?

 

you have piqued my interest!

[Dave]

whats hyperthreading?

 

Processors with hyperthreading have one physical core and one logical core; it speeds up applications that make use of threading, basically.

 

and are two 1G prosessers better than one 2G prosesser?

 

Not usually.

 

what about, are two 32bit prosessers better than one 64bit prosesser?

 

Depends on the processors

 

and whats the prosessers 'core'?

 

It's the bit that does the calculating, basically.

[Dak]

and whats the prosessers 'core'?

 

It's the bit that does the calculating, basically.

 

umm... what else does a prosesser do? I thought that it just calculated?

[Aeternus]

I think the advantage of the dual-core processors (being made by both Intel and AMD) over things like using multiple processors in SMP etc setups (Symmetric Multi-Processing) is that as the cores are much closer together and the communications are handled alot better. When the two cores need to communicate (perhaps for access timings, bus usages etc (ie making sure the two dont try to do the same thing at the same time etc) they can communicate through things like the new Hypertransport (specifically between these two cores) in AMDs case which allows for far far faster communications between the two cores when compared to the communications via the system bus between two seperate processors. Intel still seem to be using the Northbridge FSB interface and so there doesnt seem to be much difference between their dual core implementation and a classic dual processor arrangement but to be honest I don't really know what theyve done. Most of what i'm saying im getting from a PC Format article on the subject (Great magasine, great article).

 

Also from what i can gather, both cores will be able to use the same cache, obviously increasing the speed of communications between them (as communicating via the RAM is several factors slower) and allowing seperate threads to access and change set data more easily (although obviously that can get rather complicated and hard to ensure robustness) and much more quickly (Not sure if both Intel and AMD are doing this exactly or have different ways of doing something very similar, seems Intel may be doing this and AMD may be simply relying on their HyperTransport Link to allow fast communications between the two processors).

 

The major problem with using dual core over single core (normal processors) is that alot of applications still don't make use of threading (splitting the program up into various processing threads which are processed (seemingly - depending on whether its dual or single core) simultaneously rather than one after the other as a block). This means that as the program is only a single thread, it will only be able to use one of the cores to process it. Dual core setups still help in the case of multitasking (Running multiple programs/jobs (seemingly) simultaneously), and with Intel's hyperthreading technology, along with the increasing number of multiprocessor setups in the server industries, I would imagine alot of software houses have already come out with or are in the process of changing alot of their software to a multithreaded approach.

 

 

 

Links -

 

http://www.short-media.com/review.php?r=261 ( Seems decent enough )

http://www.datafuse.net/page.php?news=289 ( Seems ok )

http://www.pcformat.co.uk ( Excellent article, although youll have to actually buy the magasine (although they may open the article online after the end of the month))

http://multicore.amd.com/en/Technology/ (Straight from the horses mouth)

http://www.intel.com/technology/computing/dual-core/index.htm (Straight from the horses mouth)

 

 

In Response -

 

Originally Posted by me

and whats the prosessers 'core'?

Originally Posted by dave

It's the bit that does the calculating' date=' basically.

[/quote']

 

umm... what else does a prosesser do? I thought that it just calculated?

 

There are also other things on the processor such as registers, various cache elements (different levels of cache), some have seperate memory controllers (i think, not sure) etc. Plus you have all the small interconnections between all these components on the processor and probably more components that i can't remember or don't know about (I'm sure dave or someone will list more).

 

 

P.S - I AM NOT A PROCESSOR ARCHITECHTURAL ENGINEER

So obviously I may be talking complete nonsense. Please feel free to correct me, as i want to learn more about this. This is only what i can gather from articles I've read so...

[5614]

There are 3 mains part to the central processing unit (CPU)... these are the CU (control unit) which controls what data goes where, the ALU (arithmetic and logical unit) that will do mathematical and logical operations (so 1+1 (arithmetic) or deciding if 5>2 is true or not (logic)) and finally the IAS (immediate access store) which stores the data used by programs.

 

are two 32bit prosessers better than one 64bit prosesser?

It depends, do you mean a dual-core 32 bit processor, or two physical 32bit processors linked together? Either way normally it'd be the 64bit as when you have 2 cores, or even 2 processors you will never quite get twice the power. Just like with SLi (that's having 2 graphics cards in one computer, you get the power of say 1.5 graphics card,) I believe you get a similar effect with processors, although I don't know about figures, with SLi I know that the 1.5 is supposedly quite accurate. Basically dual core or dual processing is more powerful, but not quite twice as powerful.

[Dak]

so which bit does the dual layer core prosesser have two of? (the ALU?)

[5614]

It's dual core, or just dual processing (when referring to two seperate processors linked together)... But I'm honestly not sure, I'd have assumed it has all 3 sections replicated, however that is a major assumption.

 

The major problem with using dual core over single core (normal processors) is that alot of applications still don't make use of threading (splitting the program up into various processing threads which are processed (seemingly - depending on whether its dual or single core) simultaneously rather than one after the other as a block). This means that as the program is only a single thread, it will only be able to use one of the cores to process it.

I don't think that is correct based on the fact that data will go into the CPU where it'll then be sorted by the CU, the CU decides what data is processed by which core, not the program, but I really don't know for certain.

 

As for what you said about dual core processors using the same cache, that is quite interesting, although I'm sure that could/will be updated over time, although with The Cell with it's 9 core and 234 million transistors coming out, it could be interesting to see where things go!

[Aeternus]

It's dual core' date=' or just dual processing [i'](when referring to two seperate processors linked together)[/i]... But I'm honestly not sure, I'd have assumed it has all 3 sections replicated, however that is a major assumption.

 

I don't think that is correct based on the fact that data will go into the CPU where it'll then be sorted by the CU, the CU decides what data is processed by which core, not the program, but I really don't know for certain.

 

As for what you said about dual core processors using the same cache, that is quite interesting, although I'm sure that could/will be updated over time, although with The Cell with it's 9 core and 234 million transistors coming out, it could be interesting to see where things go!

 

 

Says Here that the programs must be multithreaded. Dual Cores or Dual Processing in general are advantageous when multitasking but when running a single threaded application, you cant really try to process bits of it with each core or processor as some bits/chunks may be dependent on the results of others and may require certain other sections of the code to have been executed beforehand. This may not be obvious from the way in which the "code" (i say code but obviously at this point it is simply binary instrutions being passed to the processor) is laid out or implemented (hence why multiple threads (well i say threads but child processes are forked on linux/unix systems instead of threads but the idea is similar)) and so it is extremely hard to simply split single threaded programs/applications up to be executed simultaneously on multiple processors. Not to say that it isn't possible, analysis of the instructions ahead of the actual processing could be done to try to provide some form of dependency checks etc and I think this might be what is being done with the new Cell architecture (as the PPE (PowerPC Element) divides the job into discrete tasks to hand out to each of the 8 cores on the cell processor). This may be what is being done by the CU (Control Unit) as you said, but I don't think it is being done to a high degree as I would imagine it would require more than a simple Control Unit to do it (perhaps more in line with the power of another core) and I doubt the articles shown would mention the problem if the CU took care of it. If you find I'm wrong, please mention it, because as I said im not expert and Id like to learn more about this (this is only what i can gather from what I've read and know already).

 

Further Evidence

[H2SO4]

Ive been reading along, and it makes sense now. Now im curious, what will my computer performance be if I disable hyperthreading. How much do you think a dual coare processor e cost? Whats the speed they run at?

[5614]

The average home user will not notice the difference with hyperthreading unless they do a lot of renderring, even gamers can't really tell the difference.

 

Dual core processors will cost more than a current processor, although realise that there will be e.g. 3GHz dual processor as well as 3.6GHz, this will make a difference in the price, however dual processor at a certain speed will cost more than a normal processor at the same speed.

 

ummm, well, dual core processors will run at the same speed, although they won't make really slow ones, to get some figures I went to the Intel site and found http://www.intel.com/products/processor/pentium_D/index.htm the Intel Pentium D is a dual core, note it says 2.80 GHz to 3.20 GHz..... the reason I said "ummm" was because I also noticed that it says L2 Cache 2x1MB I'm guessing what I said in post #9 about them making two caches for dual core is already true!

 

Aeternus thanks fot the link, I still believe that a modified CU could be programmed to do the task of the SMP side of the program... but obviously not so at the moment.