Slashdot Mirror
Apple To Discuss HyperTransport For Future Macs
macrealist writes "CNET is reporting that Apple will discuss the use of HyperTransport in Macs at the Developer's conference. The interesting thing is that the article claims that Apple is not likely to use hypertransport to link the CPU to the memory, but instead to link chipsets together because IBM would have to 'to adapt it to the Power architecture.' But according to arstechnica, the 970 does have a frontside bus that operates at similar speeds to Hypertransport."
As many people (keep) saying, Apple kit isn't necessarily the fastest out there in terms of raw speed. However, from a day-to-day point of view, is raw speed what you want on a minute-by-minute basis? Probably not. If you do, then you've probably got a dual or quad processor x86 box churning away with your favourite SMP kernel-based OS. For everyday use (productivity apps, Internet, media manipulation) Apple kit does a really good job. Firewire is fast and convenient. More importantly, Apple kit (and software) is very stable in my experience. Apple looks like it is selective in its choice of cool new tech (tm) to incorporate into its products. This is a Good Thing.
Contribute to the online videogame encyclopedia: GamerWiki
I have been thinking of switching over to Apple, and now that many designers are coming up with cool products with OSX support, I am paying much more attention to Mac. I can remember back in the day when I first saw an Apple 2e, and I thought that it was so much better than my TI 99/4A, because of the games mostly. Oh and it had it's own monitor, and at the time I needed a TV for my TI. :)
I like the idea that Mac develops the hardware and software together under one roof. I think following the process from all angles like that would make for a better product. It's a better philosophy than the Windows/PC mish-mash way of thinking, primarily because no person sees all ends of the production for PC, and you can bet that there are quality issues with computability under PC that just aren't there with Apple (or at least that is what one would expect). So looking at Hyper Transport, at this stage, I'm a tad leery of it because it didn't come from Apple. I'm worried that it might have some kind of negative impact on the technology.
The necessary question is; is this going to be the next evolutionary step for Apple, or is it just an added hardware feature that is relatively minor?
<pedantically> I think that Apple has already developed a tried and true solution for external, non-ethernet-based, high-speed data transfer. It is called FireWire800.
Of course, an IP substack can be built on top of the FW, to have additional networking options. (Check out)</pedantically>
0.02â
nah, smeagol was picked cos of the general attitude of mac users to our hardware...
"We loves it we do! my precioussss"
(twelve powermacs and counting)
It's true - Macs are harder to upgrade the CPU than PCs. Everything else is fair game, but the CPU is expensive to upgrade. On the plus side, Macs retain their resale value for much longer. If you buy a top of the line Mac in year X, by the time you ebay it in year X+2, it'll probably pay for half of your new top-of-the-line mac. I'm trying to sell my laptop right now to buy a mac... because my Laptop is windows based, I can't get crap for it - even though it's only 8 months old. -- Funky.
Apple seriously needs to buy SGI.
Seriously.
Hear me out.
The Origin family of computers is basically the coolest thing out there. It's getting a little long in the tooth--the node-to-node interconnect is only 3.6 GB/s which isn't awe-inspiring any more--but it's got serious coolness.
IRIX is a great operating system, but it's also long in the tooth, and it's tied to MIPS processors.
The Origin design transcends the CPU, however; behold the Itanium Origin, which SGI calls the Altix. (Yes, it's hideously ugly. I'm getting to that.)
Basically Apple could buy SGI for a song right now. In the transaction, they'd get all of SGI's IP and sales contracts. They could continue operating SGI as an independent company to serve existing customers and customer bases--NASA, NOA, that sort of thing--but shut down everything that's not profitable. Like the nationwide sales channel, for example. What a cluster-fuck that is. Get existing big MIPS customers migrated over to Darwin on PowerPC or Darwin (Linux compatible) on Itanium (I see no good reason to keep going with Linux development) over the next five years or so.
The Origin 350 gets PowerPC processors and Mac OS X Server and becomes the next-generation XServe. Four processors (or maybe even eight; PowerPC's are small and cool) in two rack units, scalable with external 3.6 GB/s high-speed interconnect up to 32 (or maybe 64) processors. No clustering required; that's a single system image for those who need or want it. Ideal price point for an entry-level four-processor system: ~$5,000. Maybe get really cool and do a two-processor system for ~$3,000. Get in bed with Oracle and Sybase (moreso, I mean) and get their stuff running better on PowerPC than it does on any other platform. (Are databases candiates for vector optimizations?) If it makes sense, get big applications like Oracle and Sybase and Web Objects and BLAST and so on running on Itanium 2 versions of the Origin 350-based Xserve. If it makes sense. Keep the PowerPC available throughout the product line, however, for compatibility with existing software.
The Power Mac-based Xserve would remain as the lower-cost and non-scalable option.
If the demand ever exists for it, which it might someday, Apple can put PowerPC's in the Origin 3900 family (the SN-2 family) and sell supercomputers that scale from 4 to 1,024 processors and up. If the demand exists.
As for SGI's visualization systems, IR4 and IP, if it makes business sense, keep 'em around. If it doesn't, kill 'em.
Fuel? Kill it unless it can pay for itself. Same with Octane2. Same with Chimera, the upcoming Origin 300-inspired workstation product.
But the good news is that Apple would be able to take the truly cutting-edge work SGI has been basically pissing away, like GSN with ST for multi-gigabyte-per-second communication over TCP/IP and XFS/CXFS for high-performance direct-attach and fabric-attach storage and FailSafe for high availability and (1) continue developing them in a sort of skunkworks R&D environment, and (2) move them down to the servers and desktops as it becomes practical to do so.
SGI has seriously got it goin on, but they've lost all momentum in the marketplace and need a hero. Apple has momentum in the marketplace and would benefit from a leg up on some cool advanced technologies. Also, Apple has assloads of cache and valuation to trade with.
Please please please please please...
To do it right, you'd have to get a new bus as the chips are being strangled by bandwidth bottlenecks on current bus designs.
I don't know where these rumors get started.
The combination of fat caches, low latency, and predictive fetching basically negates the memory bus bottleneck in the current-generation (MaxBus-based) Power Macs. Even in SIMD instances, the processor generally doesn't have to wait on data that much. (This is especially true in SIMD instances, because these are almost always sequential-read applications, which makes those fat caches and predictive fetching work up a sweat.) Consider Apple's AltiVec-optimized BLAST, for instance. It's 10X faster than BLAST on a Pentium 4. It's not memory-bound. It's compute-bound.
What's that famous Seymour Cray quote? "A supercomputer is a device for turning compute-bound tasks into I/O-bound tasks."
So if anybody produces PowerPC 970 upgrades with MaxBus interfaces, they're almost certain to be good buys. Unless they cost thousands of dollars, of course. A dual-1.8 GHz (pulled that number out of my ass, guys) Power Mac G5 (pulled that out of my ass, too) will be faster than a dual-1.8 GHz upgrade in a MaxBus G4, but it'll still be considerably faster than the G4 was originally.