Regarding the 970 and pentium comparisment: That benchmark doesn´t use altivec does it? What abou the pentium? Does that have anything that isn´t mesaured in the test?
No the Spec benchmark doesn't use altivec. I don't know about the pentium, but they have that SSE thingy. I think altivec is far superior to SSE though.
Regarding the 970 and pentium comparisment: That benchmark doesn´t use altivec does it? What abou the pentium? Does that have anything that isn´t mesaured in the test?
Correct, Spec doesn't use altivec, or compare any SIMD-technologies. This is where our PowerPC processors would kick serious x86-ass (not because the diff. between PPC and x86, but because of the differences between the techs developed for these ISA's). Altivec is much better than SSE&SSE2&MMX the Intel&AMD processors use.
Haven't been around much lately (or not posting much anyway), but there were a couple of comments that needed to be made.
1/ Comparing Apple to AMD and Intel is ridiculous. They are different companies, and produce entirely different products.
Intel/AMD produce chips....that's it...they therefore can release roadmaps. Motorola and IBM can/are free to do the same.
Apple, like Dell/HP/*insert preferred Wintel manu here*, cannot do so, as they are not in a position to know when there will be sufficient volume of chips. Most Wintel manufacturers do not say "the Pentium 3.5g will be out in November...here's the computer that will be using it".
2/ I have a sneaking suspicion that the main reason that Apple won't come right out and say "we're gonna use the 970", aside from killing sales (a valid point IMHO), is that they don't know for sure. What if they do so, and last minute errata delay the chip for another year?
Twould suck, I know, but figure they do know that performance is hurting sales, and they likely are sitting on alternatives......
3/ Hehe. I'm a little surprised this hasn't been mentioned yet.
Could well be (considering the other rumours floating around), that "there will come a time" means "you'll hear it from the court reporters". Wouldn't surprise me if "the particular shareholder who asked the question would be invited, if he desired" means "we'll be using shareholders as witnesses".
Regarding the 970 and pentium comparisment: That benchmark doesn´t use altivec does it? What abou the pentium? Does that have anything that isn´t mesaured in the test?
No, doesn't measure altivec. And I don't think SSE/SSE2 are accounted for either, but my understanding is altivec blows those away still (hey, I could be wrong). For Altivec to boost SPEC scores, IBM or Apple would have to have an auto-vectorizing compiler, which I don't think is out there (and if it was would it be any good?)
But also those are just the projected numbers for the 970 @ 1.8 GHz, while the P4 numbers are real. So the 970 could be faster or slower, depending on the compiler and other things.
Oh, and here are those opteron numbers:
Pre-release estimated Opteron @ 1.8 ghz
SpecInt base: 1048
SpecFP base: 998
Actuall shipping Opteron @ 1.8 ghz
SPECint: 1170
SPECfp: 1219
So the actual number can vary from before and after. So they might also do the same for the 970...
Does anybody know something about IBM's track record when it comes to publising pre-production/release performance numbers? I can't really see IBM as a company that brags about high Spec-numbers and then don't deliver when the processor is actually shipped.
If Apple release a roadmap of any sort it would be suicide.
Apple is not like Intel or AMD where they lisence out their chips to different manufacturers.
Intel can have a public roadmap because these manufacturers do not do business with each other, they compete.
Apple has no competition (in-terms of Macintosh based computers) so they relay on secrecy, otherwise their sales would be dead in the water and people would wait until the latest and greatest came around, or waited until speeds were good enough to themselves.
It seems Apple can not learn to manage their inventory either. They repeatedly have nothing to sell when new models are introduced and then wind up with enough excess inventory that they can't move and so can't introduce the new models until it is gone. This continues to be a serious management shortcoming. At one point this last year Apple had in excess of 15 weeks of inventory of iMacs collecting dust. Then of course potential customers are (rightly) concerned about "old" equipment which affects sales, too.
No the Spec benchmark doesn't use altivec. I don't know about the pentium, but they have that SSE thingy. I think altivec is far superior to SSE though.
The SSE instruction set for PCs has the same problem that Altivec has...until software developers write code utilitzing it, it just as well might not be there. The same thing is true of hyperthreading and multiprocessor configurations in the PC world and, of course, multiprocessors don't do much in a Mac until there is software to take advantage of them. That is one reason that the Mac needs a more powerful single processor.
Comments
Originally posted by Anders the White
Thats not funny
Regarding the 970 and pentium comparisment: That benchmark doesn´t use altivec does it? What abou the pentium? Does that have anything that isn´t mesaured in the test?
No the Spec benchmark doesn't use altivec. I don't know about the pentium, but they have that SSE thingy. I think altivec is far superior to SSE though.
Originally posted by Anders the White
Thats not funny
Regarding the 970 and pentium comparisment: That benchmark doesn´t use altivec does it? What abou the pentium? Does that have anything that isn´t mesaured in the test?
Correct, Spec doesn't use altivec, or compare any SIMD-technologies. This is where our PowerPC processors would kick serious x86-ass (not because the diff. between PPC and x86, but because of the differences between the techs developed for these ISA's). Altivec is much better than SSE&SSE2&MMX the Intel&AMD processors use.
1/ Comparing Apple to AMD and Intel is ridiculous. They are different companies, and produce entirely different products.
Intel/AMD produce chips....that's it...they therefore can release roadmaps. Motorola and IBM can/are free to do the same.
Apple, like Dell/HP/*insert preferred Wintel manu here*, cannot do so, as they are not in a position to know when there will be sufficient volume of chips. Most Wintel manufacturers do not say "the Pentium 3.5g will be out in November...here's the computer that will be using it".
2/ I have a sneaking suspicion that the main reason that Apple won't come right out and say "we're gonna use the 970", aside from killing sales (a valid point IMHO), is that they don't know for sure. What if they do so, and last minute errata delay the chip for another year?
Twould suck, I know, but figure they do know that performance is hurting sales, and they likely are sitting on alternatives......
3/ Hehe. I'm a little surprised this hasn't been mentioned yet.
Could well be (considering the other rumours floating around), that "there will come a time" means "you'll hear it from the court reporters". Wouldn't surprise me if "the particular shareholder who asked the question would be invited, if he desired" means "we'll be using shareholders as witnesses".
Originally posted by Anders the White
Thats not funny
Regarding the 970 and pentium comparisment: That benchmark doesn´t use altivec does it? What abou the pentium? Does that have anything that isn´t mesaured in the test?
No, doesn't measure altivec. And I don't think SSE/SSE2 are accounted for either, but my understanding is altivec blows those away still (hey, I could be wrong). For Altivec to boost SPEC scores, IBM or Apple would have to have an auto-vectorizing compiler, which I don't think is out there (and if it was would it be any good?)
But also those are just the projected numbers for the 970 @ 1.8 GHz, while the P4 numbers are real. So the 970 could be faster or slower, depending on the compiler and other things.
Oh, and here are those opteron numbers:
Pre-release estimated Opteron @ 1.8 ghz
SpecInt base: 1048
SpecFP base: 998
Actuall shipping Opteron @ 1.8 ghz
SPECint: 1170
SPECfp: 1219
So the actual number can vary from before and after. So they might also do the same for the 970...
Originally posted by Jared
If Apple release a roadmap of any sort it would be suicide.
Apple is not like Intel or AMD where they lisence out their chips to different manufacturers.
Intel can have a public roadmap because these manufacturers do not do business with each other, they compete.
Apple has no competition (in-terms of Macintosh based computers) so they relay on secrecy, otherwise their sales would be dead in the water and people would wait until the latest and greatest came around, or waited until speeds were good enough to themselves.
It seems Apple can not learn to manage their inventory either. They repeatedly have nothing to sell when new models are introduced and then wind up with enough excess inventory that they can't move and so can't introduce the new models until it is gone. This continues to be a serious management shortcoming. At one point this last year Apple had in excess of 15 weeks of inventory of iMacs collecting dust. Then of course potential customers are (rightly) concerned about "old" equipment which affects sales, too.
Originally posted by NETROMac
No the Spec benchmark doesn't use altivec. I don't know about the pentium, but they have that SSE thingy. I think altivec is far superior to SSE though.
The SSE instruction set for PCs has the same problem that Altivec has...until software developers write code utilitzing it, it just as well might not be there. The same thing is true of hyperthreading and multiprocessor configurations in the PC world and, of course, multiprocessors don't do much in a Mac until there is software to take advantage of them. That is one reason that the Mac needs a more powerful single processor.