Nearly Half of PCs Will Utilize Multi-GPU Tech in 2012 ? JPR

[MODEL3]

http://www.xbitlabs.com/news/v...n_Peddie_Research.html

Based on the image, there will be nearly half of PCs powered by multi-GPU technology, such as ATI CrossFire or Nvidia SLI, in 2012

Did Xbitlabs misunderstood and JPR meant multi-core GPU technology instead of multi-GPU technology such as ATI CrossFire or Nvidia SLI?

 

[shabby]

Micro stuttering for everyone!

 

[cm123]

Originally posted by: shabby
Micro stuttering for everyone!

yep - and few random lock-ups tossed in there also...

....guessing though mult-core yes, SLi/Cross-Fire no - however micro stutter not as bad today in few games right now, many yes still though

 

[Fox5]

If everyone has multi-gpu setups, it'll be 1 high powered one for gaming, and 1 low power one for desktop usage.

 

[apoppin]

Originally posted by: cm123

Originally posted by: shabby
Micro stuttering for everyone!

yep - and few random lock-ups tossed in there also...

....guessing though mult-core yes, SLi/Cross-Fire no - however micro stutter not as bad today in few games right now, many yes still though

not so strangely, the only ones b!tching the most about micro stutter and other imagined issues have never tried SLi or CrossFire .. at least not for years

 

[Forumpanda]

Multi-GPU? probably, it will likely just be a matter of how you define multi.

Multi PCB? Unlikely.

 

[BassBomb]

Aren't they already multi-core technically?

 

[cm123]

Originally posted by: apoppin

Originally posted by: cm123

Originally posted by: shabby
Micro stuttering for everyone!

yep - and few random lock-ups tossed in there also...

....guessing though mult-core yes, SLi/Cross-Fire no - however micro stutter not as bad today in few games right now, many yes still though

not so strangely, the only ones b!tching the most about micro stutter and other imagined issues have never tried SLi or CrossFire .. at least not for years

wow, assuming is not a very wise thing to do - for example one of my systems has cross-fire, the other sli.

defect from from the start of your statement - however doing os reload this week on my main rig and planning to remove the 2nd hd4890 and go back to non-cf in this system.

 

[Pantlegz]

Originally posted by: BassBomb
Aren't they already multi-core technically?

yes they are, from my understanding each stream processor is a different 'core'.

 

[MODEL3]

Originally posted by: BassBomb
Aren't they already multi-core technically?

If you mean HD4XXX & NV GXXX then No.

There are no independent cores in those models.


If you mean something like SGX543MP, then yes.


http://www.imgtec.com/News/Release/index.asp?NewsID=449

 

[apoppin]

Originally posted by: cm123

Originally posted by: apoppin

Originally posted by: cm123

Originally posted by: shabby
Micro stuttering for everyone!

yep - and few random lock-ups tossed in there also...

....guessing though mult-core yes, SLi/Cross-Fire no - however micro stutter not as bad today in few games right now, many yes still though

not so strangely, the only ones b!tching the most about micro stutter and other imagined issues have never tried SLi or CrossFire .. at least not for years

wow, assuming is not a very wise thing to do - for example one of my systems has cross-fire, the other sli.

defect from from the start of your statement - however doing os reload this week on my main rig and planning to remove the 2nd hd4890 and go back to non-cf in this system.

'assuming' is what most of us do most of the time
- Life would be much more difficult otherwise

i am not wrong in my statement - it was 'general' and partly in answer to shabby's comment about MicroStutter - which has been greatly minimized over the past year or so

.. and if you are getting 'random lockups' with a pair of 4890s in CrossFire, i would tend to believe the problem is with your system; not with CrossFire

 

[Pantlegz]

Originally posted by: MODEL3

Originally posted by: BassBomb
Aren't they already multi-core technically?

If you mean HD4XXX & NV GXXX then No.

There are no independent cores in those models.


If you mean something like SGX543MP, then yes.


http://www.imgtec.com/News/Release/index.asp?NewsID=449

are you including EVGA's 295 thats on one PCB? And while stream processors aren't entirely independent don't they function in clusters and in theory, if you had a small enough process, all be used on different processes?

couldn't one also argue that we don't have independent multi-core cpus because of shared cache and such?

 

[MODEL3]

Originally posted by: Pantlegz1
are you including EVGA's 295 thats on one PCB?

What has 295 to do with, if today's GPUs are multi-core?
And afterall, in what way the EVGA incorporates one multi-core GPU?
It just incorporates 2 seperate GPUs.

But I guess you are joking.

Originally posted by: Pantlegz1
And while stream processors aren't entirely independent don't they function in clusters and in theory, if you had a small enough process, all be used on different processes?

So you mean that everything that functions in clusters in the Graphics tech. world consists a "GPU core"?

No

The SPs consists "GPU cores" becauce they can be used on different processes?

No

Originally posted by: Pantlegz1
couldn't one also argue that we don't have independent multi-core cpus because of shared cache and such?

So you mean that if we exclude the shared cache the cores will be not functional?

No


Shader Processors are just parts of a GPU design.

 

[cm123]

Originally posted by: apoppin

Originally posted by: cm123

Originally posted by: apoppin

Originally posted by: cm123

Originally posted by: shabby
Micro stuttering for everyone!

yep - and few random lock-ups tossed in there also...

....guessing though mult-core yes, SLi/Cross-Fire no - however micro stutter not as bad today in few games right now, many yes still though

not so strangely, the only ones b!tching the most about micro stutter and other imagined issues have never tried SLi or CrossFire .. at least not for years

wow, assuming is not a very wise thing to do - for example one of my systems has cross-fire, the other sli.

defect from from the start of your statement - however doing os reload this week on my main rig and planning to remove the 2nd hd4890 and go back to non-cf in this system.

'assuming' is what most of us do most of the time
- Life would be much more difficult otherwise

i am not wrong in my statement - it was 'general' and partly in answer to shabby's comment about MicroStutter - which has been greatly minimized over the past year or so

.. and if you are getting 'random lockups' with a pair of 4890s in CrossFire, i would tend to believe the problem is with your system; not with CrossFire

you said those bitching the most have never tried sli or cross-fire while quote my post - that would be wrong as one not bitching, two, using both - thats all, nothing big, just not in this case.

 

[dflynchimp]

Originally posted by: shabby
Micro stuttering for everyone!

LOL!!! This post is so made of awesome if truth were measured in awesome XD

On the other hand, the hardware is only one reason for microstuttering. While they are the enablers of the symptom, if drivers were theoretically perfect and everything ran as intended then microstutter shouldn't happen. Alas you can't have your cake and eat it too...

 

[apoppin]

Originally posted by: cm123

you said those bitching the most have never tried sli or cross-fire while quote my post - that would be wrong as one not bitching, two, using both - thats all, nothing big, just not in this case.

i am not at all sure about most of what you just posted (above), so i won't assume [this time]

Yes, i did say: "the only ones b!tching the most about micro stutter and other imagined issues have never tried SLi or CrossFire" and i stand by my post's general comment
- this is not the first time posters with limited to no experience with CF or SLi complained about microstutter; i have been in many (many) threads where the complaint is mostly "2nd hand"

.. and quoting you with "random lockups" .. it is not at all usual to have those with CF; something is probably wrong with your system if you have them regularly
- there is no assumption there

 

[lopri]

The chart and commentary are very vague. What does "enabled system CAGR" mean? (blue bar) And what is "Multi AIB system" (red bar)? Text shows "enabled system: 27%", and "multi AIB: 83%" -> I don't know what that means, either. On the graph, blue bar is about 60+ TAM(? M units?) and red is almost at 30.

I mean, since upcoming CPUs from AMD and Intel will have on-package GPUs, a discrete GPU added to such a system can technically be called 'multi-GPU' system? I don't know.

 

[MODEL3]

Originally posted by: lopri
The chart and commentary are very vague. What does "enabled system CAGR" mean? (blue bar) And what is "Multi AIB system" (red bar)? Text shows "enabled system: 27%", and "multi AIB: 83%" -> I don't know what that means, either. On the graph, blue bar is about 60+ TAM(? M units?) and red is almost at 30.

I mean, since upcoming CPUs from AMD and Intel will have on-package GPUs, a discrete GPU added to such a system can technically be called 'multi-GPU' system? I don't know.

Compound Annual Growth Rate

add-in boards

total available market

 

[BenSkywalker]

If you mean HD4XXX & NV GXXX then No.

There are no independent cores in those models.

Using that same wrong logic Cell would be a single core CPU. Ignoring that though, the GTX260 parts were both GTX280 based parts with varrying 'cores' disabled(clusters of cores more accurately).

Graphics chips were multi core before they were GPUs. Link that lists nV multicore parts- in a CPU list of multicore processors.

 

[MODEL3]

Originally posted by: BenSkywalker
Using that same wrong logic Cell would be a single core CPU. Ignoring that though, the GTX260 parts were both GTX280 based parts with varrying 'cores' disabled(clusters of cores more accurately).

Graphics chips were multi core before they were GPUs. Link that lists nV multicore parts- in a CPU list of multicore processors.

Yes Cell is one-core CPU.

Cell is one-core POWER4 based derivative with 8 (physical, or 6 enabled) SPEs
We are talking about CPUs not processing units (SPEs).
In what way is a SPE a CENTRAL PROCESSING UNIT (CPU)

There is a difference between a CPU that has two or more independent CPU cores (fully functional cores) inside

and CPU with parts inside that consist processing cores (but these cores are not independent from the general architecture nor functional as CPUs if you extract them from the overall design)

Don't let marketing confuse terms

the GTX260 GPU was a GTX280 GPU with a certain percent of some of their "processing element parts" disabled (that "parts" can be labeled "processing cores" but in no way "GPU cores"

There is a diference between GPUs with multi core processing Elements & Multi-core GPUs

 

[BenSkywalker]

Yes Cell is one-core CPU.

Cell is one-core POWER4 based derivative with 8 (physical, or 6 enabled) SPEs
We are talking about CPUs not processing units (SPEs).
In what way is a SPE a CENTRAL PROCESSING UNIT (CPU)

There is a difference between a CPU that has two or more independent CPU cores (fully functional cores) inside

and CPU with parts inside that consist processing cores (but these cores are not independent from the general architecture nor functional as CPUs if you extract them from the overall design)

Don't let marketing confuse terms

the GTX260 GPU was a GTX280 GPU with a certain percent of some of their "processing element parts" disabled (that "parts" can be labeled "processing cores" but in no way "GPU cores"

There is a diference between GPUs with multi core processing Elements & Multi-core GPUs

Where to start-

Multicore doesn't mean multiple independent CPUs, never has, never will. I have no idea where you got that idea from, but it is wrong. Multicore can be SMP style which is what you seem to be saying it must be(along with additonal requirements to match your definition), it can be AMP too, with a singular CPU and additional functional units such as vector processing units or even DSPs. There are many different ways a chip can be configured as multicore without being close to your very narrow definition.

You seem to be under the false impression that because a certain required element is not built in to each core then none of them warrant consideration as dual core. If this is truly your stance, Core2Duo chips are no core CPUs, along with x2s and a slew of the numerous multicore chips we had prior to their gaining mainstream popularity in the x86 market. If you take a close look at almost every given multi core architecture all of them use common hardware that is required to make any of them functional. This is no way whatsoever makes them non multi core.

The Cell processor consists of a general-purpose POWERPC processor core connected to eight special-purpose DSP cores.

I could throw links out all day long. The end effect of which wouldn't do much good if you don't understand what the word 'multicore' means to start with.

 

[dflynchimp]

The idea of CPU's is a general processing unit that can be pretty much used to do anything written in code it understands.

A very dumbed down version of the difference betwee current CPU technologies and GPU technologies is:

CPU: good at processing long strings of code often with lots of branching variations, but is inherently monolithic so one individual cpu can't handle multiple threads in parallel with any sort of efficiency. Even with hyper threading on Intel cpu's the threads aren't technically handled in parallel. Rather, one thread can be preloaded onto the core while another is still in operation, so theoretical computing time is cut in half but CPU's aren't designed with 100 cores due to the complexity of each core (space/power limitations)

GPU: the many cores on the GPU are simple instruction units operating in unison to handle many shorter strings (AKA graphics) in massively parallel format, but the simplicity of each core/SIMD/stream processor means it cannot do the same thing a CPU can do since it doesn't have the ability to handle long branching threads.

Obviously these are the extremes of computing, with single lengthy branching threads on one end and massively parallel but relatively simple math operations on the other. Ideally every "core"in either a CPU or a GPU will be able to handle a thread with heavy branching, but the demands on the respective hardware are inherently different. This is what Intel's Larrabee is trying to do if I'm not wrong, to unify the architecture with something in between the two.

 

[MODEL3]

Originally posted by: BenSkywalker
Where to start-

Multicore doesn't mean multiple independent CPUs, never has, never will. I have no idea where you got that idea from, but it is wrong. Multicore can be SMP style which is what you seem to be saying it must be(along with additonal requirements to match your definition), it can be AMP too, with a singular CPU and additional functional units such as vector processing units or even DSPs. There are many different ways a chip can be configured as multicore without being close to your very narrow definition.

You seem to be under the false impression that because a certain required element is not built in to each core then none of them warrant consideration as dual core. If this is truly your stance, Core2Duo chips are no core CPUs, along with x2s and a slew of the numerous multicore chips we had prior to their gaining mainstream popularity in the x86 market. If you take a close look at almost every given multi core architecture all of them use common hardware that is required to make any of them functional. This is no way whatsoever makes them non multi core.

You know, maybe my definition is narrow but also yours is wide, using your logic.
I don't have a technological background so I don't know how this particular period in time the "technology status" market the notion "multi core".

My definition comes from analyzing the greek word equivalent of "multi" & "core".
Of cource in my definition of the word "multi core", the cores are "of the same quality" or "co-equal".

This is not exactly necessary, they may be uneven, in that case IF we stress the terms then yes you are correct.

You see the intellect meaning of the words exist way before and will exist way after some guy in 20th century market the notion the way he likes or understand.

So I guess if you mean that the technology world right now defines the word "multi core" the way you define it, then of cource you are right.

Oh wait, i just wanted to verify that the technology world right now defines the word "multi core" the way you define it and I Googled the word "multi core" and the first result was from "wikipedia".

In the first line it says:

A multi-core processor is a processing system composed of two or more independent cores.

And later it says:

All cores are identical in homogeneous multi-core systems and they are not identical in heterogeneous multi-core systems

So I guess your definition clicks with the term "heterogeneous multi-core systems" and mine clicks with the term "homogeneous multi-core systems"

Originally posted by: BenSkywalker
The end effect of which wouldn't do much good if you don't understand what the word 'multicore' means to start with.

But I guess I am to stupid to understand what the word 'multicore' means!

 

[aka1nas]

Your interpretation doesn't even really match the definition of homogeneous multi-core. I think you're equating core with die. By your definition, only a multiple die chip such as a Pentium D would be multi-core.

 

[OCGuy]

Originally posted by: shabby
Micro stuttering for everyone!

I have yet to see this. I think it is more than likely an over-blown buzz word.