Slashdot Mirror
Apple Demonstrates A Dual-G4 Power Mac
caligula writes: "Just saw this [macnn.com report]: 'Tuesday, May 17 updated 2:30 pm, top stories. During the hardware keynote of WWDC, which ended just minutes ago, Apple demonstrated a dual-processor G4 Power Mac running Mac OS X. Of note to developers is that Cocoa/Carbon applications do not need to be changed in any way to take advantage of multi-processors. Benchmark demonstrations ran roughly twice as fast on the dual-G4 system compared to the single-G4 Power Mac that was on stage. No mention was given as to when these multiprocessor G4s would ship, although it was stressed that it would not be happening any time soon but that they would definitely be out by next year's WWDC.'" JonahLee pointed out a related link on macosrumors.com, and migooch noted this slightly more informative ZDne t story. Mortals still must wait at least 'til January.
nearware?
tcd004
Here's my Microsoft Parody, where's yours?
Intel chips work on a shared bus architecture - real easy, like any other bus. Request and grant. Each processor requests bus mastering, and they share. Nice, but one (or more) can get held off.
/. article today). Each proc gets it's own dedicated path to the chipset, where there's a switching fabric. This is essentially the same difference between a LAN hub and a switch. One Athlon could be bursting to memory at peak bandwidth while another talks to the AGP bus... good stuff. You only have to wait if the actual resource you are waiting for is being used. This is considerably more dificult (and more costly) to implement than a shared bus.
The Athlon is based on the same SMP style as the Alphas (see earlier
The G4 can use a shared bus (as the earlier PowerPCs could. I'm not privy to the specifics of this new chipset, but it stands to reason that anything priced in the consumer marketspace (ooh, buzzword!) would need to be a shared bus to stay cost-efficient.
The K6 was SMP capable, but not with Intel's chipsets and SMP structure. Therefore, it is essentially non-SMP capable, since nobody supported it. The Athlon has yet to be proven in this arena, but several boards are in the works (or so the rumor mill would have us believe)...
"It's tough to be bilingual when you get hit in the head."
AltiVec vector processing is absolutly killer. Anything dealing with graphics manipulation or signal processesing (amungst others) runs insanely well on a G4 when it uses AltiVec.
In a word, it boils down to coordination. The CPU's need to coordinate their activities. With two CPU's, you no longer have a single "master" that always knows what's going on.
For example, cache coherency is a problem. Each CPU has a write-back cache, which means that the data is not writting to memory right away, but held until it can be written in a burst. If one CPU has unwritten data in its internal cache, the other CPU doesn't normally know about it. So the two CPU's need to talk to each other.
Another issue is interrupts. Some SMP systems let interrupts go to any CPU. Some force all interrupts to go to one CPU. Obviously, the former is more complex.
On recent Intel CPU's, there is something called a system management interrupt (SMI) which is like a super-interrupt that the OS doesn't know anything about. It's meant for the BIOS only. SMI's run in "SMI mode", where all interrupts are turned off, and all CPU's jump the SMI handler simultaneously. Getting all CPU's to exit the SMI handler simultaneously is difficult.
Another issue is the chipset. The system bus needs to be able to handle multiple CPU's vying for PCI and memory devices.
Anyway, that's just a brief list. Maybe someone who's worked on the SMP support for an operating system could chime in.
And the men who hold high places must be the ones who start
To mold a new reality... closer to the heart
Hmmm... well, remember (like the Alpha), GCC isn't optimized as well for the PowerPC line as it is for the Intel Architecture (more people, more time). Of course, that isn't the whole story (nothing ever is).
That aside, I still can't find any specInt or specFP #s... which are a fairly good representation...
Anybody else know of them (and have a source)?
"It's tough to be bilingual when you get hit in the head."
I believe the problem is that AMD have yet to release a memory controller with 2 CPU ports so that we can do 2-way SMP - K7 is supposed to be already to work it's just waiting for the chipset.
Sadly, because this memory controller chip will only be used for SMP systems and most systems are single CPU, this will mean that volumes on this chip will be low and it will likely cost more meaning dual motherboards will be more expensive than their Intel cousins :-(
Back in the day ('95-96 or so) there were dual-604e Macs and quad-604e Mac clones that ran Photoshop like nobody's bidness. The G4 was designed from the beginning (unlike the G3) to accept up to 4-way cache-coherency, IIRC. Four G4 cores on a single die were considered quite feasible when Motorola rolled out the design. Mmmmmm...four-way processing....[hrragglhh]
Hmmm. Could Apple be the first company to introduce a MP laptop?
(Or has somebody already done so?)
I use Macs for work, Linux for education, and Windows for cardplaying.
PowerLogix Announces First Dual G4 Upgrade Card
The advantages of the Apple machine are that it will run with a faster system bus and faster memory than an 8500 that has been upgraded with one of these cards, which has to access main memory at the same speed as the conventional 604 does in the original machine.
The advantage of the upgrade card is that you can run it with older system software - note that when Apple releases new machines, usually they require the latest system to run them so they're not a lot of help to developers wanting to maintain application compatibility with old systems.
And what's really cool is that it's possible that the BeOS will run on the cards, and BeOS applications and the whole BeOS system are pervasively multithreaded and so should take great advantage of these cards.
Note that the reason Be can't make the BeOS run on PowerMac G3 and G4 machines is not because of their processors but because Apple won't give Be the specs for the proprietary chips on the motherboards of the new machines. This is not a problem for an upgraded 8500 and in fact many people are currently using 8500's with G3 upgrade cards.
You can be sure Linux will work on the cards because PowerLogix includes a copy of LinuxPPC with each card they ship.
Sadly, a Be employee who bought a PowerLogix card for his Power Computing home machine found that it didn't work to run the BeOS - he's very happy with it on the MacOS. Many other G3 cards do work with the BeOS though.
I have very enthusiastically urged PowerLogix to support the BeOS in their cards and offered to beta test for them on my 8500.
They also have a USB/Firewire card that allows older mac owners to take advantage of all those spiffy peripherals and video editing software that's available for the new machines. I'm waiting until the dual G4 ships to purchase an upgrade but I'll be getting the rapidfire card so I can use an HP Deskjet USB printer on my mac.
-- Could you use my software consulting serv
That aside, I still can't find any specInt or specFP #s... which are a fairly good representation...
That's because SPEC doesn't run on MacOS. Thus the only PowerPC SPEC scores available are for high end IBM type systems, which have far superior i/o systems to Macs, and thus ought to perform much better on memory-intensive benchmarks like SPEC.
Unfortunately, the latest high-speed PIIIs kick the crap out of them, as do the actually-purchasable GHz Athlons. AltiVec is a great SIMD design, but even if there was enough software to take advantage of it, it'd be starved for data by the Mac's pitiful system bus in most real world conditions.
Basically, the only reasons to buy a Mac for your desktop is the OS, the color, or the fact that the new iMacs don't need fans. The only good points of the PowerPC architecture are low power consumption (which makes it a great laptop chip, and allows for the lack of a fan) and AltiVec (which, again, is starved by the poor i/o system).
For a much more detailed explanation of what I just said (including PPC vs. x86 SPEC comparisons), see Paul DeMone's excellent article.
As for the best SPEC/$ platform, right now it's far and away the GHz PIII. In the near future, it will be the Thunderbird on the low end and the new high-clocked Alphas (finally) for FP stuff. In the medium term, Willamette just might post some amazing SPEC results, although it's too early to tell. In any case, it's clear that the PPC architecture will lose badly for at least the next 12 months or so.
Besides, SPEC has a lot of things wrong with it as a benchmark, not least of which is the fact that even the SPEC "base" tests can be compiled based on optimizations from a previous test run (this is why Itanium will have good SPEC scores but terrible performance), and that it has no graphical component, and that it doesn't model typical code very well, unless you spend your time running scientific simulations all day.
On the other hand, if there were a better cross-platform benchmark, it would also show that the PPC sucks. As a CPU for a personal computer, that is. As well it should: it's designed for embedded/signal-processing applications.
The best reason to get a Mac is if you prefer the look and feel of MacOS to Windows. Although MacOS X is going to be a radical change, I don't think the essential aesthetic rightness of the Mac is going to change as a result.
Certainly I find it more pleasing to work in Photoshop or Painter or any other graphics program on the Mac than to work in Windows, even though the software is basically the same. It has to do with design quality, just like a Mercedes-Benz feels better than a Chevy. The Mercedes might have the exact same statistics - speed to 60mph, quarter mile, etc - but it still feels better because it's better designed and built.
Curiously enough, my Mac has been a pretty cost-effective purchase, too. It's about three years and two generations old, and it still works well with current software. During the time I've owned it, Windows has gone from 95 to 98 to 2000. In the PC universe, I've owned three different systems, which in total cost more than the Mac.
I'll probably replace this machine with one of the multiprocessor G4s (assuming the price isn't too ghastly, which admittedly might be a forlorn hope), and then it will go to serve another owner, or I'll use it to try Mac Linux or something. Either way, its useful lifetime is significantly longer than a PC.
I'll replace it because I enjoy the MacOS, and because I think MacOS X is going to be far superior to any version of Windows. It will be a quality multiprocessing system, unburdened by the tiresome flaws of the Windows world, but it will still run mainstream software like Photoshop and (dare I say it?) Word.
Sure, it's expensive - of course as I grow richer over the years, that's less and less important for me. And of course there's less software than for Windows, but who cares? You couldn't buy every piece of available software for either platform, anyway.
Of course the reality of it all is that I will always own multiple computers and multiple platforms. Right now, I have an old SGI workstation (currently awaiting repairs), a Windows NT 4 system, a Linux notebook, and the Mac. I use the Linux notebook for work-related stuff, and don't take it home every night. My Mac waits patiently for me, smoothly and reliably telling me that there is mainstream life in the computing world after Microsoft.
To me, that's something beyond price.
D
----