Portions of the part about IBM are true. They have set astonishing records in GHz. using Germanium, but as I recall and I'm too lazy to look it up, introduction won't be for a few years.
Good Gawd. I about to wet myself. I haven't been this excited since the allenmcjones, kormac, dorsal rumors. Can you imagine a 2-3GHZ Mac? Oh man, oh man, I just p*ssed on myself
Not flaming Aramas, I don't have enough knowledge/expertise to dispute any of his claims, BUT, usually when something sounds TOO GOOD TO BE TRUE, it usually is.
On the other hand I'm looking forward to my DUAL IBM G4's at 100GHz by the end of the year.
Well, I agree it does sound to good to be true but it definitely sparks your interest and indicates that there are options on the horizon. The DSP stuff agrees with what Moki and others have said and the different machines they have tried agrees with Dorsal posts from two years ago (DDR, etc) and indicates that Moto is trying to do something with regards to desktop alternatives, which, from a business and PR standpoint makes sense. I also think the points about Apple expecting MOTO to have a chip with a DDR controller on board made it so Apple didn't try to make mods themselves but problems develop in the design of anything an then Xserve.
Time will tell, the fact is that faster machines will eventually come out and the question is when.
On further review, I'm not waiting on the 100 GHz/ dual G4 Powermacs, "Why",you ask.
You think the current G4's are band width choked, holy smoke, a dual 100GHz G4 with a 133MHz FSB would blow so many bubbles you'd think Lawrence Welk was performing a concert in you computer. :eek:
This stuff makes a lot of sense. Apple's high-end software offerings are enabling them to tap into a new market in which the cost of hardware is basically no object provided the performance and productivity benefits are real. A high-end Mac workstation running Shake, FCP, Reason, Silicon Grail et al at speeds satisfying to professionals (along with Photoshop, MS Office, IE et al for productivity when necessary) would be one-stop shopping for a lot of professional film/audio production houses, and I think even if they ended up costing 10 grand each Apple could sell a boatload of them.
Another benefit is that these kinds of buyers are a lot more tech-saavy than the average consumer. Apple could build a system with, say, a 1Mhz G5+ (or whatever) and a memory/bus architecture so insanely efficient that it ran rings around systems with three times the clock speed, and there'd be no need to go on about the "Megahertz Myth"; these professionals don't need to be told that clock speed isn't the be-all and end-all of performance. But like I said, the only problem would be that these systems really would need to be fast, much faster than any current Macintosh, and they'd need to be fast in ways that can be quantified and benchmarked (not "I'm a lot more productive on my PowerMac, even if it takes me six times as long to render my footage").
Can Apple do it? Would they be buying all these high-end software products if they didn't think they could?
I'd be interesting in reading what he thinks... </strong><hr></blockquote>
Nothing really revolutionary in there -- it sounds like he's been reading this forum for the last 6 months. What was it in there that you found particularly amazing? There are a bunch of things that don't make any sense, and he's been listening to some of the reasonably optimistic rumours. If you're a pessimist you'll dismiss all that just as you would have dismissed it when you read it here. The guy has no credibility to make his statements more believable, and he makes several mistakes (technical and grammar!) which makes him questionable.
OSX is VERY tempting when you run a wintel machine and you have to reset it DAILY just to keep it running correctly. and you have to REFORMAT almost on a yearly basis when the BSOD's start nockin on your doorstep.
</strong><hr></blockquote>
re-format yearly? Ever tried developing for the windows platform? try reformatting every 1-3 months for any non-win2K machine (6 mos for NT4). I have no opinion on winXP, as I haven't used it for development yet.
Its that damn shared .dll architecture that makes system maintenance a nightmare. I started out developing on 95, then 98 and reformatting was just part of the job. With NT4 it got better, and better still with 2K (Im working on 9 months with my current installation, only my 2nd in 2 years).
Ive heard not so good things about XP stomping on its system files, but its relatively immature compared to 2K right now.
I LOVE OSX, since I had no legacy software or devices I punted OS9 off my iBook and never looked back. BRING ON JAGUAR
"and I thought you said I never offered anything "
Well, credit where it's due...
(Mind you...the length of that read/link equals the total sum of useless one liners you put in here over the last gawd knows how long...)
I'd happily take more constructive contributions like this! It kinda had me laying eggs...(excitement...)
You've upped the anti with that link!
Hmmm. Authentic or not...in one big dollop...it makes an enticing read.
Programmer. I take your point(s). However, the 'tone' of said post made me realise just what Apple might have been up against. ie Motorola may have dropped the ball like they did with the original G4 debut...so...Apple 'do it themselves' with regard the 'X-serve'. Hoping that whatever problems dogged the 'genuine' DDR solution/bus will be resolved by Moto' for a late summer/fall release. Could it be a case that they were trying to be 'too clever' with the 'intended' motherboard...? ie In light of reducing the 'perceived' performance gap they aimed too high and tripped up. ie Overcomplicated instead of getting the basic thing out there? ...and now will probably end up with a 'hack' or 'basic' DDR solution just to finally get it out the door...
In the meantime, the DDR 'straight through...' solution will be worked on and hopefully yields will be improved for a Rio/7500 launch early next year. I take it this is Programmer's favoured 'controller' performance booster. With a 7500 no less!? A stretched pipeline taking us up to 2.4 gig?!?! If they can do that...then Moto'/Apple will be the comeback kid.
What I viewed as 'apathy' or 'negligence' in getting to more modern specs...seem like they are genuine engineering problems with PPC's/motorola etc. Spending ages on a DDR board then yields are crap...or there are bugs...or it get canned due to one or more reasons.
Okay. I can perhaps understand why we don't have DDR...but still, you'd hope that the combined might of Apple/IBM and Motorola would be able to come up with a DDR motherboard solution? Taiwanese motherboard makers are spitting them out left right and centre...and faster than 266!
Many of my posts have been frustrated (much like the response of many to no 'power'Macs at New York...) but the link gave me an insight (true or not...) on what the hell has been going on with the 'lack' of progress on PPC.
With Intel revealing their roadmap plans well in advance...it'd be nice if Apple just came clean and told us what the hell is happening with PPC. I, for one...could plan my buying accordingly.
I heard many moons ago that the 7500 was the G5. Turns out...this is a superstretched G4 apollo on Rio architecture. That's what we'll get next year. Well the Register said that too? The guy on that board seems to say the same. But is the 7500 on Moto's site?
The 7470, on the other hand, is what we'll get next month? 1-1.5 Gig on X-serve DDR at least? Nice catch up. But by the time it ships...I'd rather hang on for the Rio '7500'. The first time I heard about the G5 anyhow (some years ago...)..the info' seemed to indicate it was merely a pipeline stretched G4 Apollo on superior throughput. I think I could buy that machine.
Fact or fiction. I dunno. But it kinda put everything in perspective re: PPC over the next year re: the lack of delivery and WHAT may be coming and when in a nutshell. Altivec II sounds nice.
Things aren't looking as bad as I thought...as long as they deliver.
Just a thought: This guys' said that the specs he's thrown around are all publically available. Sure enough, some of the G5 specs (64 bit, 2.4GHz, etc.) have been made publically available... at the Register.
Armas, what a character! Obviously he doesn't work for Apple...if he ever did then he's fired by now. He breaks NDA every other sentence in his post.
One thing that was very interesting was his bit about the reason Apple is stuck with PC 133; that they are waiting on Moto's next G4 revision that will have ATA controller, DDR333 controller, and more all on die. Is this something that is likely? Anyone here know for sure?
If so, then not only would it explain Apple's lag in this area, but it would suggest that when Apple finally does "fix" their mobos, they will leapfrog everything on the Wintel side.
<strong>Armas, what a character! Obviously he doesn't work for Apple...if he ever did then he's fired by now. He breaks NDA every other sentence in his post.
</strong><hr></blockquote>
... the other big problem, is this guy's written verbal style is so distinctively messed up, that anybody'd who'd worked with him and read anything of his could practicaly smell his identity thru that prose.
If this guy indeed works for Apple, his secret identity is no secret.
He should leak his stuff to somebody else, and get <i> them </i> to write it.
DDR motherboards for current PCs are nowhere near as efficient as Apple's memory architecture -- Apple is getting 75-85% efficiency out of the SDRAM, whereas many of the PC boards are lucky to get 50%. Rather than spending the effort to get a fairly minor incremental improvement, it wouldn't surprise me at all to learn that Apple decided to "skip" a generation and go for a bigger leap. They've done that in the past, and it is a reasonable course of action. Engineering being what it is, however, means that schedules can slip... so they've had to scramble in the meantime.
The Xserve is an intermediate solution, albeit a decent one for a server. It may also be a chipset that Apple bought rather than developed in-house. Or last year they simply took their existing chipset and extended it by the most expedient method possible to add the capabilities that the Xserve needs. They might have done this regardless of the processor plan because it makes more sense for a server machine. If the processor-based plan isn't ready yet they may be forced to use the Xserve chipset in the PowerMac just to keep up until their "leap frog" can finally happen.
TheRegister's 7500 rumour represents a huge change to the processor -- it is perfectly reasonable to call it a G5. The Arman fellow got the "AltiVec-II" from the fact that the 8540 has SIMD instructions in its integer registers, which is a bad misinterpretation IMHO. We have no concrete signs of a change to the AltiVec programming model, and I don't really think there needs to be any changes to it... they should just focus on making the current model go faster. This avoids forcing code changes and avoids fragmenting the market based on processor difference. The G3/G4 gap is bad enough -- we don't want a G3/G4/G4v2/G5 mess. That starts feeling like the x86 world! PowerPC has always been better at this than the x86 guys, and I hope that doesn't change. Apple can't afford it.
When the new "leap frog" solution arrives it will make for much faster machines, rather than the constant nickle'n'dime approach of the x86 crowd. People complain about how Apple always does small bumps, but really they typically have fewer significant design changes -- a major architecture change happens, followed by a series of speed bumps until the next big change happens. To me this is better than a constant series of not-so-major-but-annoying architecture changes. It is a bit more frustrating, especially when combined with Apple's tight lipped policy.
<strong>Armas, what a character! Obviously he doesn't work for Apple...if he ever did then he's fired by now. He breaks NDA every other sentence in his post.
One thing that was very interesting was his bit about the reason Apple is stuck with PC 133; that they are waiting on Moto's next G4 revision that will have ATA controller, DDR333 controller, and more all on die. Is this something that is likely? Anyone here know for sure?
If so, then not only would it explain Apple's lag in this area, but it would suggest that when Apple finally does "fix" their mobos, they will leapfrog everything on the Wintel side.</strong><hr></blockquote>
Well, from what I was told the new MOT G4 that has a
166mhz process bus, allowing for 333mhz DDR and we should get up to 1.4ghz if MOT can deliver them. If not, then maybe at least they'll be announced.
To have access to such confidential information at Apple ("in house CPU development in Building 2"), this guy would have to be working on the project himself. This would mean that he is an experienced CPU designer. A statement he makes casts doubt on credentials.
[quote] ( all things being equal, double the pipline = double the clock speed, with no increase in intructions/sec )
<hr></blockquote>
No. This is not true, the whole benefit of pipelining is that with the increased clock frequency that is possible (due to reduced propagation delay per stage), you have instructions being completed more frequently (ideally).
If doubling the number of pipeline stages with double the clock frequency gave someone *NO* effective speed benefit; then why pipeline?
Yes, you have your branches and inter-instruction dependency problems. And too much pipelining without decent branch prediction or instruction reordering will harm you; but by and large the original statement is false.
And doubling the number of pipeline stages will not automatically give you a the ability to double the frequency of the clock. Otherwise Motorola could just slap on 7 more "dummy" stages at the end of the 7450's pipeline that basically do nothing but store the instruction in an instruction register for a cycle; and they automatically get to double the clock frequency. This is ludicrous! Clock frequency is determined by the propagation delay in the longest stage of the pipeline. If it takes 5 nanoseconds for an instruction to go through the longest stage; adding 50 1 nanosecond stages will not allow you to increase the clock frequency.
Furthermore, if Apple were doing in house CPU development; it sure has heck wouldn't be done in Infinite Loop 2. It would be done somewhere on Bandley or Valley Green; or off campus entirely.
if the new powermacs to be introduced in august are still using the g4, don't expect it to hit 1.6 ghz. i think the machine is a proto for some other chip instead of the g4.
Comments
Ah heck, I looked it up anyway<a href="http://www.ibm.com/news/2001/06/25.phtml"; target="_blank">100 GHz pie in the sky</a>
<img src="graemlins/lol.gif" border="0" alt="[Laughing]" />
<strong>Bigc
Not flaming Aramas, I don't have enough knowledge/expertise to dispute any of his claims, BUT, usually when something sounds TOO GOOD TO BE TRUE, it usually is.
On the other hand I'm looking forward to my DUAL IBM G4's at 100GHz by the end of the year.
[ 07-26-2002: Message edited by: rickag ]</strong><hr></blockquote>
Well, I agree it does sound to good to be true but it definitely sparks your interest and indicates that there are options on the horizon. The DSP stuff agrees with what Moki and others have said and the different machines they have tried agrees with Dorsal posts from two years ago (DDR, etc) and indicates that Moto is trying to do something with regards to desktop alternatives, which, from a business and PR standpoint makes sense. I also think the points about Apple expecting MOTO to have a chip with a DDR controller on board made it so Apple didn't try to make mods themselves but problems develop in the design of anything an then Xserve.
Time will tell, the fact is that faster machines will eventually come out and the question is when.
<strong>Hmm. That link made my head spin.
I'd hope even a grain of it comes true in the next year...
Makes my wait for a POWERMac system even more excruciating.
Der-rool, der-rool.
Has Programmer read that link yet?
I'd be interesting in reading what he thinks...
Great link.
Lemon Bon Bon
and I thought you said I never offered anything
<img src="graemlins/smokin.gif" border="0" alt="[Chilling]" />
You think the current G4's are band width choked, holy smoke, a dual 100GHz G4 with a 133MHz FSB would blow so many bubbles you'd think Lawrence Welk was performing a concert in you computer. :eek:
<strong>What if in fact there really were two separate lines soon for the "PowerMacs?"
Deskmacs:
<a href="http://www.deskmac.com"; target="_blank">http://www.deskmac.com</a>;
(this was just registered on the 22nd by the way)
</strong><hr></blockquote>
Very intriguing.
This stuff makes a lot of sense. Apple's high-end software offerings are enabling them to tap into a new market in which the cost of hardware is basically no object provided the performance and productivity benefits are real. A high-end Mac workstation running Shake, FCP, Reason, Silicon Grail et al at speeds satisfying to professionals (along with Photoshop, MS Office, IE et al for productivity when necessary) would be one-stop shopping for a lot of professional film/audio production houses, and I think even if they ended up costing 10 grand each Apple could sell a boatload of them.
Another benefit is that these kinds of buyers are a lot more tech-saavy than the average consumer. Apple could build a system with, say, a 1Mhz G5+ (or whatever) and a memory/bus architecture so insanely efficient that it ran rings around systems with three times the clock speed, and there'd be no need to go on about the "Megahertz Myth"; these professionals don't need to be told that clock speed isn't the be-all and end-all of performance. But like I said, the only problem would be that these systems really would need to be fast, much faster than any current Macintosh, and they'd need to be fast in ways that can be quantified and benchmarked (not "I'm a lot more productive on my PowerMac, even if it takes me six times as long to render my footage").
Can Apple do it? Would they be buying all these high-end software products if they didn't think they could?
<strong>Has Programmer read that link yet?
I'd be interesting in reading what he thinks... </strong><hr></blockquote>
Nothing really revolutionary in there -- it sounds like he's been reading this forum for the last 6 months. What was it in there that you found particularly amazing? There are a bunch of things that don't make any sense, and he's been listening to some of the reasonably optimistic rumours. If you're a pessimist you'll dismiss all that just as you would have dismissed it when you read it here. The guy has no credibility to make his statements more believable, and he makes several mistakes (technical and grammar!) which makes him questionable.
He still is posting , just refresh the ars forum.
He meant 80 Ghz and the 100 Ghz , and apparently now 110 Ghz , he provided links also
<strong>
OSX is VERY tempting when you run a wintel machine and you have to reset it DAILY just to keep it running correctly. and you have to REFORMAT almost on a yearly basis when the BSOD's start nockin on your doorstep.
</strong><hr></blockquote>
re-format yearly? Ever tried developing for the windows platform? try reformatting every 1-3 months for any non-win2K machine (6 mos for NT4). I have no opinion on winXP, as I haven't used it for development yet.
Its that damn shared .dll architecture that makes system maintenance a nightmare. I started out developing on 95, then 98 and reformatting was just part of the job. With NT4 it got better, and better still with 2K (Im working on 9 months with my current installation, only my 2nd in 2 years).
Ive heard not so good things about XP stomping on its system files, but its relatively immature compared to 2K right now.
I LOVE OSX, since I had no legacy software or devices I punted OS9 off my iBook and never looked back. BRING ON JAGUAR
I'll pay
Well, credit where it's due...
(Mind you...the length of that read/link equals the total sum of useless one liners you put in here over the last gawd knows how long...)
I'd happily take more constructive contributions like this! It kinda had me laying eggs...(excitement...)
You've upped the anti with that link!
Hmmm. Authentic or not...in one big dollop...it makes an enticing read.
Programmer. I take your point(s). However, the 'tone' of said post made me realise just what Apple might have been up against. ie Motorola may have dropped the ball like they did with the original G4 debut...so...Apple 'do it themselves' with regard the 'X-serve'. Hoping that whatever problems dogged the 'genuine' DDR solution/bus will be resolved by Moto' for a late summer/fall release. Could it be a case that they were trying to be 'too clever' with the 'intended' motherboard...? ie In light of reducing the 'perceived' performance gap they aimed too high and tripped up. ie Overcomplicated instead of getting the basic thing out there? ...and now will probably end up with a 'hack' or 'basic' DDR solution just to finally get it out the door...
In the meantime, the DDR 'straight through...' solution will be worked on and hopefully yields will be improved for a Rio/7500 launch early next year. I take it this is Programmer's favoured 'controller' performance booster. With a 7500 no less!? A stretched pipeline taking us up to 2.4 gig?!?! If they can do that...then Moto'/Apple will be the comeback kid.
What I viewed as 'apathy' or 'negligence' in getting to more modern specs...seem like they are genuine engineering problems with PPC's/motorola etc. Spending ages on a DDR board then yields are crap...or there are bugs...or it get canned due to one or more reasons.
Okay. I can perhaps understand why we don't have DDR...but still, you'd hope that the combined might of Apple/IBM and Motorola would be able to come up with a DDR motherboard solution? Taiwanese motherboard makers are spitting them out left right and centre...and faster than 266!
Many of my posts have been frustrated (much like the response of many to no 'power'Macs at New York...) but the link gave me an insight (true or not...) on what the hell has been going on with the 'lack' of progress on PPC.
With Intel revealing their roadmap plans well in advance...it'd be nice if Apple just came clean and told us what the hell is happening with PPC. I, for one...could plan my buying accordingly.
I heard many moons ago that the 7500 was the G5. Turns out...this is a superstretched G4 apollo on Rio architecture. That's what we'll get next year. Well the Register said that too? The guy on that board seems to say the same. But is the 7500 on Moto's site?
The 7470, on the other hand, is what we'll get next month? 1-1.5 Gig on X-serve DDR at least? Nice catch up. But by the time it ships...I'd rather hang on for the Rio '7500'. The first time I heard about the G5 anyhow (some years ago...)..the info' seemed to indicate it was merely a pipeline stretched G4 Apollo on superior throughput. I think I could buy that machine.
Fact or fiction. I dunno. But it kinda put everything in perspective re: PPC over the next year re: the lack of delivery and WHAT may be coming and when in a nutshell. Altivec II sounds nice.
Things aren't looking as bad as I thought...as long as they deliver.
Lemon Bon Bon
[ 07-26-2002: Message edited by: Lemon Bon Bon ]</p>
Keep your salt shaker handy.
One thing that was very interesting was his bit about the reason Apple is stuck with PC 133; that they are waiting on Moto's next G4 revision that will have ATA controller, DDR333 controller, and more all on die. Is this something that is likely? Anyone here know for sure?
If so, then not only would it explain Apple's lag in this area, but it would suggest that when Apple finally does "fix" their mobos, they will leapfrog everything on the Wintel side.
<strong>Armas, what a character! Obviously he doesn't work for Apple...if he ever did then he's fired by now. He breaks NDA every other sentence in his post.
</strong><hr></blockquote>
... the other big problem, is this guy's written verbal style is so distinctively messed up, that anybody'd who'd worked with him and read anything of his could practicaly smell his identity thru that prose.
If this guy indeed works for Apple, his secret identity is no secret.
He should leak his stuff to somebody else, and get <i> them </i> to write it.
The Xserve is an intermediate solution, albeit a decent one for a server. It may also be a chipset that Apple bought rather than developed in-house. Or last year they simply took their existing chipset and extended it by the most expedient method possible to add the capabilities that the Xserve needs. They might have done this regardless of the processor plan because it makes more sense for a server machine. If the processor-based plan isn't ready yet they may be forced to use the Xserve chipset in the PowerMac just to keep up until their "leap frog" can finally happen.
TheRegister's 7500 rumour represents a huge change to the processor -- it is perfectly reasonable to call it a G5. The Arman fellow got the "AltiVec-II" from the fact that the 8540 has SIMD instructions in its integer registers, which is a bad misinterpretation IMHO. We have no concrete signs of a change to the AltiVec programming model, and I don't really think there needs to be any changes to it... they should just focus on making the current model go faster. This avoids forcing code changes and avoids fragmenting the market based on processor difference. The G3/G4 gap is bad enough -- we don't want a G3/G4/G4v2/G5 mess. That starts feeling like the x86 world! PowerPC has always been better at this than the x86 guys, and I hope that doesn't change. Apple can't afford it.
When the new "leap frog" solution arrives it will make for much faster machines, rather than the constant nickle'n'dime approach of the x86 crowd. People complain about how Apple always does small bumps, but really they typically have fewer significant design changes -- a major architecture change happens, followed by a series of speed bumps until the next big change happens. To me this is better than a constant series of not-so-major-but-annoying architecture changes. It is a bit more frustrating, especially when combined with Apple's tight lipped policy.
<strong>Armas, what a character! Obviously he doesn't work for Apple...if he ever did then he's fired by now. He breaks NDA every other sentence in his post.
One thing that was very interesting was his bit about the reason Apple is stuck with PC 133; that they are waiting on Moto's next G4 revision that will have ATA controller, DDR333 controller, and more all on die. Is this something that is likely? Anyone here know for sure?
If so, then not only would it explain Apple's lag in this area, but it would suggest that when Apple finally does "fix" their mobos, they will leapfrog everything on the Wintel side.</strong><hr></blockquote>
Well, from what I was told the new MOT G4 that has a
166mhz process bus, allowing for 333mhz DDR and we should get up to 1.4ghz if MOT can deliver them. If not, then maybe at least they'll be announced.
Next year, will be the big year for Apple tho.
[quote] ( all things being equal, double the pipline = double the clock speed, with no increase in intructions/sec )
<hr></blockquote>
No. This is not true, the whole benefit of pipelining is that with the increased clock frequency that is possible (due to reduced propagation delay per stage), you have instructions being completed more frequently (ideally).
If doubling the number of pipeline stages with double the clock frequency gave someone *NO* effective speed benefit; then why pipeline?
Yes, you have your branches and inter-instruction dependency problems. And too much pipelining without decent branch prediction or instruction reordering will harm you; but by and large the original statement is false.
And doubling the number of pipeline stages will not automatically give you a the ability to double the frequency of the clock. Otherwise Motorola could just slap on 7 more "dummy" stages at the end of the 7450's pipeline that basically do nothing but store the instruction in an instruction register for a cycle; and they automatically get to double the clock frequency. This is ludicrous! Clock frequency is determined by the propagation delay in the longest stage of the pipeline. If it takes 5 nanoseconds for an instruction to go through the longest stage; adding 50 1 nanosecond stages will not allow you to increase the clock frequency.
Furthermore, if Apple were doing in house CPU development; it sure has heck wouldn't be done in Infinite Loop 2. It would be done somewhere on Bandley or Valley Green; or off campus entirely.
[ 07-26-2002: Message edited by: PipelineStall ]</p>