PDA

View Full Version : Intel's Nehalem 8 cores per processor = 16-core Mac Pro in the near future.


Vincent Bergbahn
07-06-2008, 06:35 PM
Intel's coming Nehalem processor will introduce up to 8 cores per processor. If combined into a dual-processor configuration, we could see a 16-core Mac Pro in the near future.


http://www.macrumors.com/2008/07/06/intel-recommends-developers-plan-for-massive-multi-core-processing/


WOW!

GtrJazz
07-06-2008, 06:46 PM
Does that mean by christmas the mac pro 8 will cost half as much?

A.Leung
07-06-2008, 06:46 PM
We have enough cores as it is. We need Snow Leopard to make USE and propagate efficiency of those cores..

Muziksculp
07-06-2008, 07:27 PM
"Near Future" is a very relative term, another year, six months, three months ? How near is near ?

I have been holding back on ordering a second Mac Pro 8 core, but I'm not sure I can wait another six months, three months would be the max. I'm willing to wait for the new "Nehalem" based multi-processor Mac Pros.

Obviously, having more cores + a very efficient OS such as Snow Leopard, and SW applications/Plug-Ins that are designed to take advantage of these new resources would be a great development.

The "Near Future" looks very bright and promising :)

Memphis
07-07-2008, 12:01 PM
But how many DAW´S and regular programs can use so many cores? I wonder why they dont try to make them stronger instead of more. Make like a 5-6GHz core? maybe its harder? o_0

/M

gstitt
07-07-2008, 12:47 PM
I wonder why they dont try to make them stronger instead of more. Make like a 5-6GHz core? maybe its harder? o_0


The shift towards multicore was caused by the inability to further increase clock speeds. There is still definite room for improvement, but power is now the limiting factor.

Regarding improved usage of multicores, there are huge research efforts going on trying to make it easier to develop applications for multicore systems, but I wouldn't expect any significant results for at least 5 years. Things will get even crazier when multicores shift away from a shared cache architecture and start using networks on a chip. Very few people can effectively program these types of systems.

Even with 16 cores, I doubt you would see much of an improvement, except for certain specialized applications (DAWs luckily may be one of them). Sharing cache between 16 processors has to have a large arbitration overhead. Also, at some point, memory bandwidth becomes the limiting factor. If 16 processors are constantly fetching data from memory, you basically need 16x the memory bandwidth.

Spinning poo machine
07-07-2008, 02:49 PM
We have enough cores as it is. We need Snow Leopard to make USE and propagate efficiency of those cores..

Couldn't agree more.

GtrJazz
07-07-2008, 03:35 PM
anyone know when the nehalem cpu might be coming out?

Djoule
07-08-2008, 04:07 AM
to me the interesting thing in creating 16 cores is... that 4 and 8 cores will be at a lower cost !
I run my CCCollection on a Toshiba laptop with 1.7 GHz single core, 2 GB RAM and I kind of get afraid when I read posts like "Do you think 8 cores with 16GB of RAM is enough to run xxx ?".
I plan to buy a new computer next year when prices will decrease so I may get a laptop (mac I think) with multicores so the more they create, the more I may get at a lower price next year ! :D

I wonder why they dont try to make them stronger instead of more. Make like a 5-6GHz core? maybe its harder?

In fact there are physical limitations because of physical laws (I'm not going to give you a quantum mechanics lesson but it deals with it) so they can actually increase a bit the clock speed but it's not interesting to do this now because once they will reach the physical limitations the computer market will sort of "stop" (by 2020 we will actually reach these physical limitations and there may not be improvements until scientists manage to develop quantum computers, which is technically extremely difficult and will take decades).
So they try to make improvements by adding cores rather than GHz ;)
(I hope I haven't scared you ! :p )

A.Leung
07-08-2008, 12:17 PM
I think we have ALREADY nearly reached the physical limitations already with speed and core technology. I think by 2010 not 2020...

A.Leung
07-08-2008, 12:22 PM
personally, I for one will be glad when we DO hit a limitation and I'll tell you why. It will force software developers to 'trim the fat' and instead focus on efficiency.

Steve Jobs did it with OSX. The first Leopard release was in my opinion and many others, part 'bloatware'. With every subsequent release it continues to get trimmed and ive seen MUCH faster performance and efficiency as a result.

I'd LOVE to see other software developing follow suit.

gstitt
07-08-2008, 02:06 PM
I think we have ALREADY nearly reached the physical limitations already with speed and core technology. I think by 2010 not 2020...


There won't be any physical limitations hit by 2010, but there are plenty of practical limitations. Power being the biggest one.

personally, I for one will be glad when we DO hit a limitation and I'll tell you why. It will force software developers to 'trim the fat' and instead focus on efficiency.

I couldn't agree more. The quality of software engineers graduating from even the top 10 universities is ridiculously bad. It is no wonder why most commercial software is a big pile of crap.