...have so much L3 cache (32-64MB)? Is it because the things it's used for deal with humungous files? It can't be because it's bottlenecked by it's FSB. The Itanium 2 seems to do just fine with 6MB of L3 cache, and that much high speed DRAM is pretty expensive already.
Comments
Would the damn things cost so bloody much if the cache were in the desktop range? In other words, would the POWER4 be utiziable in a Powermac at a reasonable price without the ginormous cache?
ting5
Possibly. But why would IBM bother? Apple doesn't sell enough to justify it.
R&D costs alone would drive the price up out of desktop range even with the cache whacked, not to mention it would require new motherboards, and probably would take a lot of work to get OS X running on them. Power4 is not just a PPC processor, it implements a superset of the PPC instruction set.
[ 07-12-2002: Message edited by: jasonf ]</p>
<strong>
Possibly. But why would IBM bother? Apple doesn't sell enough to justify it.</strong><hr></blockquote>
I betcha Apple sells more G4's than IBM sells POWER4's by at least 10-fold.
Jet
<strong>Because it's intended for use on systems running massive databases and other operations that require work to be done on massive data sets?
R&D costs alone would drive the price up out of desktop range even with the cache whacked, not to mention it would require new motherboards, and probably would take a lot of work to get OS X running on them. Power4 is not just a PPC processor, it implements a superset of the PPC instruction set.
[ 07-12-2002: Message edited by: jasonf ]</strong><hr></blockquote>
If it's a new mobo that is required, that's Apple's issue, not IBM's.
Jet
Even if it is dealing with huge files like that I would've thought that IBM would build in a fast enough FSB to make that much L3 cache mostly unnecessary. And the Itanium 2 which is built for the same type of processes does without that much L3 cache.
It's relatively simple to change the size of the caches, I don't think that would be holding IBM back from developing a Power4 derivative for Apple. They'd probably welcome it because it would be at least doubling their sales of the Power4, plus giving it a smaller cache would keep a PowerMac equipped with one competing directly against IBM workstations.
That said, I don't think the G5 will be a Power4, it'll probably be a RIO equipped G4 on an 0.13 micron process, maybe that BookE thing that's been going around. I would figure the next gen chip from Apple would be a derivative of the IBM Power5.
<strong>I bet Apple would prefer being IBM's bitch over being Motorola's bitch.</strong><hr></blockquote>
And who's bitch would you prefer for Apple to be?
ting5
Why not use a faster FSB? I don't know, but there must have been a good reason for it.
If you got rid of the L3, would Power4 be cheaper? Not really; the chip itself is very expensive.
Why doesn't Apple use it? Because it would require a new, very expensive chipset, pushing the cost of the machine over $10K (just check out the price on a 1-processor IBM p630).
It would be interesting to see someone port Darwin to the pSeries, though.
Well, we don't really know how expensive the chip is by itself because all the caches are factored into the overall cost of the chip. I forget how much the high-speed DRAM that is used on chip caches costs, but I know it's really expensive on a MB/$ basis.
The reason the IBM Power4 workstations are so expensive isn't just because of the chip it uses, though that's part of it. These are specialized, powerful products here, using high quality components and lots of advanced technology. It would be possible to design a Power4 chip with say 2MB of L3 cache. This would reduce the cost of the chip substantially as well as reducing the size and heat produced, making it possibly feasible for a PowerMac or ProMac. Whether this will happen nobody can say, it's not likely.