Linked by Thom Holwerda on Thu 2nd Mar 2006 22:59 UTC
Apple Two articles about the Intel Mac Mini: "As a fan of faster computers from Apple, I'm heartened by this week's release of an Intel-based Mac mini. In fact, with the SPECint_rate_base2000 benchmark indicating speeds of up to four times faster than the Mac mini G4, this new crop of minis sounds just about perfect. I say 'just about', because of one particular issue." And secondly, "Our first Intel-based Mac minis have arrived, straight from the Apple Store, and what was the first thing the cold, cruel alien intellects at Macworld did with one of these innocents? That's right. We got out our putty knife, popped it open, and spilled its guts out." Here are some XBench tests between an Intel and a G4 Mac Mini.
Thread beginning with comment 101240
To view parent comment, click here.
To read all comments associated with this story, please click here.
Disruptor
Member since:
2005-11-06

"Also, what does computer science have to do with chip design? Computer science people tend to have very idealized models of processors, and are just the sort of people who'd tend to think that an elegant external interface (PPC machine code), imples an elegant internal design."

I think that the truth is always somewhere in the middle. Computer scientistists have contributed a lot to modern computers -- and yes their idealism hurts sometimes, but hey, the extreme-practicallity of non-idealists hurts sometimes too doesn't it? As for the elegant design you have to admit that there is nothing wrong with being careful in desinging. If you mean to have a fine result you have to be careful while desinging the thing from the ground up. Even some small flaws in design may cost too much in the end product. It's a `seed winds - harvest tornadoes' situation.

"It's true that benchmarks can be fudged to prefer one architecture over another. Fortunately, SPEC is not one of those benchmarks. Certainly, SPEC does not inherently favor x86 over PowerPC. If it did, how do you explain the fact that POWER5 gets much better per-clock performance in SPEC than the Opteron, while the PowerPC 970 does substantially worse? They're the same ISA after all!"

I've been persuaded by your last posts that Intel (like AMD) `did it's homework' with the Duos and increased their performance. But still the point our friend is trying to make here is valid. G5 processors (like you said rayiner) kinda failed (for Apple) because of lack in R&D and availability. Had there IBM been a little nicer with them and Apple ---I think too that--- it would have been a different story that (unfortunately) we'll never get to see.

Reply Parent Score: 1

rayiner Member since:
2005-07-06

I think that the truth is always somewhere in the middle. Computer scientistists have contributed a lot to modern computers -- and yes their idealism hurts sometimes, but hey, the extreme-practicallity of non-idealists hurts sometimes too doesn't it?

My point is not that idealism is not important. I'm a big supporter of idealism in computer science --- I think there is a distinct lack of it. My point is that the elegance of the G5 is mostly superficial. PowerPC is elegant, x86 is not, but since neither chip is internally PowerPC or x86, what counts most heavily is the elegance of the internal design. The G5 is elegant in certain ways (3-address operations throughout, mostly symmetric functional units), and quite inelegant in others (no integrated memory controller, obscure scheduling rules, etc).

To put it simply: I just wish that people looking for elegance would look more than "skin deep" and evaluate the architectures for what they really are. As I said, the G5 has its share of elegant design. So does the Opteron --- its integer core uses a very elegant renaming/scheduling mechanism. So does the Core Duo ---micro-ops fusion is an elegant way of increasing IPC without making the core wider.

Reply Parent Score: 1