[Spliff]

http://arstechnica.com/news.ars/post/20050207-4591.html

I hope Apple surprises us this year.

[turtle777]

Originally posted by Spliff:
http://arstechnica.com/news.ars/post/20050207-4591.html

I hope Apple surprises us this year.

I demand a dual-core G5 with 3.0 GHz in a $999 iBook !

-t

[goMac]

Originally posted by Spliff:
http://arstechnica.com/news.ars/post/20050207-4591.html

I hope Apple surprises us this year.

Intel is known for running late on stuff..

Then again... so is everyone else in the industry.

[olePigeon]

Originally posted by Spliff:
http://arstechnica.com/news.ars/post/20050207-4591.html

I hope Apple surprises us this year.

Cell Processor > Dual Core anything.

I think Apple already has the chip they need. They just need to stick it in something (or everything. )

Don't think I need to list the advantages the Cell has over traditional processors and even dual core CPUs.

[jcadam]

I'll take my two, one-core 64 bit G5s over a single dual-core, 32-bit P4 any day of the week.

[willed]

Originally posted by olePigeon:
Cell Processor > Dual Core anything.

I think Apple already has the chip they need. They just need to stick it in something (or everything. )

Don't think I need to list the advantages the Cell has over traditional processors and even dual core CPUs.

Apple hasn't officially said anything about using Cell yet have they? Even if they did, it's still a long way off, so don't get too excited yet!

[Eug Wanker]

Apple is not going to be using Cell any time soon.

However, I would not be surprised to see a dual-core G5 at WWDC, hopefully at 3.0 GHz.

[olePigeon]

http://news.zdnet.com/2100-1040_22-5566340.html

Another cool article on the Cell with some specifics. Here're some nifty excerpts:

Test production of Cell chips is set to begin later this year, and the processors will appear later in workstation PCs optimized for animation and other graphics chores.

Dual 32-bit cores, I say HAH!

Cell will have a 64-bit IBM Power processor and eight "synergistic processing units" capable of handling separate computing tasks, said Jim Kahle, an IBM Fellow. The Power processor will act as the brain of the chip, running the main operating system for an application and divvying up chores for the other processors.

The first version of the chip will run at speeds faster than 4GHz. Engineers were vague about how much faster, but reports from design partners say 4.6GHz is likely.

Cell can process 256 billion calculations per second (256 gigaflops)

The chip will have 2.5MB of on-chip memory...


I wonder who had a bigger crowd? Cell kicks Intel's chip in the nuts.

[Kernel Angus]

A dual core pentium is as useful as a third armpit. Until Windows & its applications are multithreaded, it seems like a kludge.

[goMac]

I'm wondering if with a dual core design Apple will move back to single processor and change the cooling system to something less... complex...

Maybe keep Dual processors on highest end Macs and XServes.

[Eug Wanker]

Originally posted by olePigeon:
Test production of Cell chips is set to begin later this year, and the processors will appear later in workstation PCs optimized for animation and other graphics chores.

This was mentioned many months ago. And the general consensus was that it'd likely be for dedicated high-end graphics workstations, not for OS X Macs, if that's what you were suggesting.

[olePigeon]

Originally posted by Eug Wanker:
This was mentioned many months ago. And the general consensus was that it'd likely be for dedicated high-end graphics workstations, not for OS X Macs, if that's what you were suggesting.

It doesn't rule out future possibilities for Apple. Plus, since it's PowerPC based, Mac-on-Linux would be a cool way to test it out.

[effgee]

no speeds announced.
no pricing announced.
won't run on existing chipset - needs new mainboard.
there'll be one "consumer" model with HT turned off (on purpose).

and that is great news ... how and why??

[olePigeon]

Originally posted by goMac:
Maybe keep Dual processors on highest end Macs and XServes.

Nothin' like 8GHz and 18 CPUs.

Must multitask like mad.

[Eug Wanker]

Originally posted by olePigeon:
It doesn't rule out future possibilities for Apple. Plus, since it's PowerPC based, Mac-on-Linux would be a cool way to test it out.

It seems that Cell is based off the POWER5, and supports VMX. What I'm thinking is that any 970FX replacements that show up would be cousins of POWER5 and Cell, but new Macs won't actually run Cell per se.

[Thorin]

There's an interesting pair of articles over at El Reg:

http://www.theregister.com/2005/02/0...ysis_part_one/
http://www.theregister.co.uk/2005/02...ysis_part_two/

Here's an excerpt from the first one:

"It isn't just a single microprocessor or even a family of processors," writes Tom. "It's a top-to-bottom architecture for a broad range of computing systems, from servers and workstations at the high-end to game consoles, PDAs, digital TVs, and other consumer electronics at the low end".

So it sounds like we could end up with something from the Cell range in a Mac .

[Big Mac]

Originally posted by Thorin:
There's an interesting pair of articles over at El Reg:

http://www.theregister.com/2005/02/0...ysis_part_one/
http://www.theregister.co.uk/2005/02...ysis_part_two/

Here's an excerpt from the first one:

So it sounds like we could end up with something from the Cell range in a Mac .

Thank you for the articles. The first part was quite informative, while part two was sensationalist, speculative and just plain uninformed in many respects. And neither part explained why the PlayStation 3's use of Cell will be so significant. I am not a computer scientist, but here is my take. Cell is an interesting concept, but I am not at all convinced that it can change the world in the manner some are envisioning. Cell seems like it will really thrive in supercomputing clusters, where the data set lends itself inherently to parallel processing and there are speedy network connections between each of the machines. It should also revolutionize SETI processing, as if that's going to rock anyone's world. But how much could it possibly do for regular desktop computing? The type of software we use on the desktop needs a lot of fast, low latency processing. It doesn't matter if one can tap into the resources of one hundred machines over the Internet, since the task you want to perform faster will already be done by your local machine by the time the distributed data packets come back and are assemble. Moreover, desktop software can only be multithreaded to such an extent, and it takes a skilled programmer to get optimal threading. This is why you don't get much of a boost after a certain point from adding additional functional units to processors, or by adding processor after processor to a machine. So, unless I am missing something, I don't see Cell taking over the world.

[Scotttheking]

Originally posted by jcadam:
I'll take my two, one-core 64 bit G5s over a single dual-core, 32-bit P4 any day of the week.

What about a single dual-core, 64-bit P4? Or, dual dual-core, 64bit Xeons?

[demograph68]

Too much... speculation in here... must... escape...

[itistoday]

There is now a story on /. about this.

[Brass]

In reference to the original article and the title of this thread...

"Beat by Sun again?" might be more appropriate - only time will tell.

Sun announced 2 mulit-core processor lines about a year ago. "Rock" for the high end servers (no expected date of delivery yet), and "Niagra" for workstations, due this year some time.

I think Sun are more likely to deliver on time than Intel.

None of which is terribly significant for Macs, I suppose.

[itai195]

Originally posted by Brass:
In reference to the original article and the title of this thread...

"Beat by Sun again?" might be more appropriate - only time will tell.

Intel announced their multi-core ambitions a while ago too, they're just clarifying the product lines now. Something Sun also has to do. Anyway, rumors have been swirling for a while that there's a dual-core G5 on the way. Makes sense, since the G5 was based on a dual-core server CPU (the first such CPU, IIRC).

I wouldn't be too worried, Apple is in good company with IBM. The really amazing thing about the Intel CPUs is how much power they'll need -- 130 watts!

[goMac]

Originally posted by effgee:
there'll be one "consumer" model with HT turned off (on purpose).

This begs the question, why would you need HT on a dual core chip?

[CD Hanks]

Originally posted by Kernel Angus:
Until Windows & its applications are multithreaded, it seems like a kludge.

HUH?

[goMac]

Windows isn't threaded very well but a lot of applications for it certainly are.

[Eug Wanker]

Originally posted by goMac:
why would you need HT on a dual core chip?

To make each individual core function more efficiently of course.

[MilkmanDan]

At least until we get a G6 processor, or a new 'potato' based chip set.

[Scotttheking]

Originally posted by itai195:
Intel announced their multi-core ambitions a while ago too, they're just clarifying the product lines now. Something Sun also has to do. Anyway, rumors have been swirling for a while that there's a dual-core G5 on the way. Makes sense, since the G5 was based on a dual-core server CPU (the first such CPU, IIRC).

I wouldn't be too worried, Apple is in good company with IBM. The really amazing thing about the Intel CPUs is how much power they'll need -- 130 watts!

Sun already ships dual core systems, IBM already ships dual core systems...

[jcadam]

Originally posted by Scotttheking:
What about a single dual-core, 64-bit P4? Or, dual dual-core, 64bit Xeons?

If it can run OS X I guess I'll take it.

[itai195]

Originally posted by Scotttheking:
Sun already ships dual core systems, IBM already ships dual core systems...

We're talking about workstations.

[Eug Wanker]

New round of info and speculation:

1) Cell's PowerPC core is NOT directly related to the G5, nor it is a stripped down POWER5. Because of the reported SMT support, everyone assumed it would be based off the POWER5, but it is not.

2) Cell's PowerPC core is similar to Xbox 2's Xenon, but Xenon has three of them. Xenon doesn't have any SPE cores though, while Cell has up to 8 SPE cores.

3) Cell's PowerPC core does in fact support Altivec. (The SPE cores do not however.)

4) Because of its PowerPC + Altivec heritage, OS X could likely be easily modified to support this core, although real world performance in apps might suck in certain instances. In other situations it might do extremely well, however, because of the very high clock speed.

[Spliff]

Eug,

What are the chances that Apple would dump the G5, which is a new processor, for a Cell-based processor? Or simply put, is there any chance that Apple will use this processor in its computers?

[Eug Wanker]

Originally posted by Spliff:
Eug,

What are the chances that Apple would dump the G5, which is a new processor, for a Cell-based processor? Or simply put, is there any chance that Apple will use this processor in its computers?

Hell if I know.

I suspect though in the short term, Apple is simply waiting for upgraded and low power G5 9xx series chips, for the Power Macs and PowerBooks respectively (and for the iMac).

[olePigeon]

http://www.blachford.info/computer/Cells/Cell1.html

Here's a link explaining more how the Cell works and what it's made of. I think it's one of the coolest things to come out in a long time.

[olePigeon]

The article mentions that the Cell V processor could be used as a coprocessor to the G5. With a little modification to the standard G5, it could act as the primary processor with multiple Cells working with it.

[Superchicken]

I was just thinking that, imagine a G5 with a bunch of cell coprocessors. Or if these things are really that scalable, even a G4 with a bunch of cell coprocessors depending on the market.

[goMac]

Originally posted by Eug Wanker:
To make each individual core function more efficiently of course.

HT only makes about a %6 difference, its not a huge deal.

[Eug Wanker]

Originally posted by goMac:
HT only makes about a %6 difference, its not a huge deal.

You cannot make such a declaration.

1) Earlier implementations of HT are not as robust as later implementations of HT.
2) Dunno about Cell, but IBM claims its SMT implementation in the POWER5 is more robust than Intel's on Prescott.
3) Some code will be slower with SMT, and some code will only be marginally faster, but some code will be significantly faster. For instance, Cinebench 2003 on the P4 gets a 20% speedup with HT turned on.

[Apple Pro Underwear]

whatev


speed doesn't mean the same thing it meant even 2 years ago.

[goMac]

Originally posted by Eug Wanker:
You cannot make such a declaration.

1) Earlier implementations of HT are not as robust as later implementations of HT.
2) Dunno about Cell, but IBM claims its SMT implementation in the POWER5 is more robust than Intel's on Prescott.
3) Some code will be slower with SMT, and some code will only be marginally faster, but some code will be significantly faster. For instance, Cinebench 2003 on the P4 gets a 20% speedup with HT turned on.

However, HT only makes a difference with programs that work with dual processors. Quite a few Windows programs don't care about dual processors. In order for HT to be effective on a Windows machine with two cores, you'd need programs coded for four processors.

That said, I'm sitting at my HT P4 system at my desk right now. HT does make a difference, however, its not huge.

[olePigeon]

Originally posted by Apple Pro Underwear:
speed doesn't mean the same thing it meant even 2 years ago.

The difference between an Apple ][ and a IIgs. Those were the days.