Too many cores: four cores should be limit

[azrealhk]

The number of cores things seems to be getting out of hand.
An average PC user can only do so many things interactively at one time and not every piece of code can be be run in parallel, e.g. a sum of 8 numbers with 8 cores still needs about 3 cycles (assuming one cycle for an add) because there are things that still need to be serialized.
OK so you can benefit from multitasking, video encoding and games. But...
Even with HD movies, and 7.1 surround sound, the need for many cores for encoding is still limited.
For games there still needs to be synchronization with the main loop, e.g. the main draw loops needs to wait for physics to be calculated before it can draw the object at the new position etc.
For multitasking, well as I said, there is only so many things one person can do. So unless the next gen PC will have an O/S which runs some stuff at 100% CPU, or our PCs are so cluttered with stupid applications running in the background using 100%, multi-cores are not really going to improve performance that much.
What I am saying is that why are Intel and AMD going to put so much effort into multi-cores (32 cores in 2010 ????). OK the server market might benefit, but not us PC users. I think 4 core will be the limit and then they will have to go back to improving performance of the single core. Intels new line up seems the right direction, but AMD seems to be slightly behind, with only smaller die size and increase clock speed for the near future. It may be AMD has something up its sleeves, I hope so because this multi-core thing is getting a bit stupid.

 

[MagicPants]

While it's true that today's software doesn't take advantage of multi-core, I think large numbers or cores 5+ is a good idea.

When 3d hardware first came about there were few applications that could take advantage of it, so it's use was questionable. After it was imbraced 3d hardware became one of the most exciting things to happen to computers ever.

The same will probably be true of multi-core. Programmers just need to learn to take advantage of it. Also learning to jump from programming one core to two is the biggest leap. After that it's not as big a deal, and things may scale quite easily.

 

[elpresidente2075]

I don't know what anyone could ever do with more than 128K of ram

 

[Heyyou27]

Sure, let's stop innovation right now.

 

[TheMaster]

It's not exactly innovation.. they're just adding more cores based on the same design. Laziness.

 

[azrealhk]

Sure, let's stop innovation right now.

Exactly, Conroe is a totally different architecture (OK, based on PIII), but adding cores is NOT innovation. Thats what AMD and Intel should be working on.

Do a single taks faster is better than more cores, even if there are advancements in parallel programming.

There was parallel programming over 20 years ago with languages to get the most from these servers. Even now most multi-core/CPU servers still do not use these parallel language compilers and at most uses threads. Parallel programming is not working.

I wouldn't like to see a 3.2GHz 8 core Conroe processor in 4 years time, I want something totally new, (or a 32 core Woodcrest for that matter)

 

[buckiller]

fLAME SUITS AT tEH rEADY?!

Lol...cmon man of course we will need more than 4 cores...in fact i would benefit already from a quad core...

 

[runt23]

It's not exactly innovation.. they're just adding more cores based on the same design. Laziness.

Im not trying to troll but...If I am, say so and I'll delete this post.

Thats quote is like saying "Hey, why use a 6 cyl engine, you know for more horse power so you can haul that big load of dirt when you could use this 2 ton single piston engine. I mean cmon, you're just adding pistons for more HP, nothing good will come of that"

Just because it is simple doesnt mean its not innovative. With price and cooling reality being a large factor here, if companies decide to use the dual core feature, you will see a large increase in many aspect of games and functionality of software.

It isnt nessesarily all about speed.

Google for example. They dont use many "super computers" to do their searching job. They use MANY MANY small pc's to do their work. I'll search for a recent article about it. I think it related to cost of keeping them running with electricity. It may not be "OMG that is SOOO amazing that they do that" but it is still innovative to do something simple rather than completely complex and new.

 

[elpresidente2075]

One Word:

Virtualization.

 

[azrealhk]

Dont tell me you dont think Conroe is amazing.

OK, so 4x4 is interesting but it doesn't have the punch even if it can out match a dual core conroe.

Like I said, I would prefer not to see a 8 core Conroe in 4 years time.

 

[elpresidente2075]

First of all, noone has even SEEN 4x4, much less done any performance numbers. Conroe still has yet to be released (I know, just a couple days :roll: ). Bad comparison

Second, multicore-optimised apps scale very well, and with 8 cores and a multicore app, you can expect to see approximately 4 times the performance. Yes, it will be less, due to overhead and whatnot, but you get my point.

If all the chipmakers go mulitcore (Hmm, has this already happened? Yes.), the software will eventually follow. 4 years is a lifetime in software time.

"If you build it, they will come."

 

[npilier]

Well, I guess we'll have to see and wait what matures from all of these multicore "innovations," for a lack of a worst word. However, I believe that we'll be able to benefit somehow from the not-so-distant technologies like from 4, 8, 16 or maybe 32 cores at once. At least the military will, that's for sure!!!

 

[ktasley]

I agree multi-core solutions is where the industry is going right now and justly so, but I understand where azrealhk is comming from in a point of view that continued focus and development of the individual core is essential. If a core's efficency is raised 10% in a quad-core platform say for example, then thats a 40% gain. I also know that conroe for example is a overhaul on architecture so I am not worried. Intel is just happy about its new venue of multiple cores and broadcasting its sucess and that does not mean it will not continue its work with individual architecture.

 

[xsandman]

I agree multi-core solutions is where the industry is going right now and justly so, but I understand where azrealhk is comming from in a point of view that continued focus and development of the individual core is essential. If a core's efficency is raised 10% in a quad-core platform say for example, then thats a 40% gain. I also know that conroe for example is a overhaul on architecture so I am not worried. Intel is just happy about its new venue of multiple cores and broadcasting its sucess and that does not mean it will not continue its work with individual architecture.

You also can not ONLY add cores and not make any improvements to the individual cores. If this was the case, then Intel with its process technology advantage (Being first to 65nm and expected to hit 45nm first) would keep adding cores and AMD with its disadvantage would find a way to make the cores efficient enough that they can beat Intel with fewer cores. This is similar to what AMD did when they had the frequency disadvantage. They made the cores more efficient to beat Intel with its freq advantage. So in order for Intel to stay competitive they will need to keep their process advantage and keep improving the cores. For AMD to remain competitive, they will need to make even more efficient cores while still trying to catch up with their process. If either company slacks off, they will lose market share and profits.

 

[Atolsammeek]

You must be blind there azrealhk. Here why amd and intel started with new dual core and comming up with quad core. It due to the fact we are Ghz cant go any higher. If it could we be to 10ghz by now. But looking at intel trying to reach 5ghz. They seem to hit a brick wall. That why they aim for dual core. For it better then hypertreading. basicly tricking the computer thinking there 2 cpus.

 

[INeedCache]

Four cores and seven years ago...