You are comparing apples with oranges. 16-bit pointers can address only 64KB, which was totally insufficient when CPU's transitioned to 32-bit. Because the CPU makers were too late, apps were using all kinds of hacks to support more than 64KB. In other words, just about every programmer was clamoring for 32-bit. Furthermore, 32-bit integer calculations were frequently useful and had to be emulated by using two 16-bit integers (slow+ugly).
The current situation is far different. The 2GB that you can address in most 32-bit systems is sufficient for most software that exists today.
My how times have changed. But I don't think that software companies will balk at developing 64-bit apps. Would you want to be the one selling the old, clunky (albeit fine) 32-bit app in a 64-bit world? Look where Apple is right now. Not a pleasant spot to be in.
I think there will be advances in software (ok, probably not Word docs...) that will take advantage of 64-bit goodness. Video and Database are two key areas. I can see great leaps forward for memory intensive applications.
My how times have changed. But I don't think that software companies will balk at developing 64-bit apps. Would you want to be the one selling the old, clunky (albeit fine) 32-bit app in a 64-bit world? Look where Apple is right now. Not a pleasant spot to be in.
Given that my app (and 99.9% of the apps out there) would not benefit even in the slightest by being 64 bit (and indeed, the performance could slightly degrade, and my app would use slightly more memory), no, I wouldn't feel bad about having a 32 bit app to sell.
But hey, whatever -- selling a "64 bit" machine is something I expect Apple to leverage in their marketing campaign, and perhaps it'll work to the advantage for a change.
Even if most software won't benefit by being 64 bit, there are some apps. that apparently will. The question is how much over a 32 bit version running on an extremely x86 Intel chip. If it does not outperform the 32 bit version by a fair amount, Apple is still in a compromised situation regarding 64 bit marketing. People will be quick to point out that 64 bitness is part of the Jobsian spin job and reality distortion field. Not to say that the 970 will be any slouch. If everything is true, this chip should be equivalent to a 3.6 GHz G4 in the first release and who knows how fast that VMX unit will function with the fat memory bus IBM has placed on the chip. It seems that IBM is on an aggressive roadmap in moving the chip to a 90 nm process and scaling the frequency rapidly.
However, I do hope Apple does release a 64 bit OS fairly quickly so that applications that can benefit from accessing large amounts of memory are released in fairly short order. I hope Apple doesn't get into the pattern of releasing out of sync software and hardware. OS X in the first releases were too much for the hardware. Let's hope that the next OS release does not limit the hardware from performing to its fullest capabilities.
Even if most software won't benefit by being 64 bit, there are some apps. that apparently will. The question is how much over a 32 bit version running on an extremely x86 Intel chip. If it does not outperform the 32 bit version by a fair amount, Apple is still in a compromised situation regarding 64 bit marketing.
If you are cherry picking, you can come up with a pretty convincing program.
Example:
x= somebig64bitintegerX;
y= somebig64bitintegerY;
z=x+y;
Now to do the same thing in a 32bit processor(ignoring carry, which would only make the 32bit program longer & confuse things).
xlo= thelow32bitsofsomebig64bitintegerX;
xhi= thehigh32bitsofsomebig64bitintegerY;
ylo= thelo32bitsofsomebig64bitintegerX;
yhi= thehigh32bitsofsomebig64bitintegerY;
zlo=xlo+ylo;
zhi=xhi+yhi;
3 commands versus 6 commands. In assembly there's all sorts of loads & stores flying around too, but the results are similar (or worse).
For a _demo_, on applications where the central algorithim is fundamentally 'better' on a 64bit machine, the 970 will _crush_ the G4.
But what a lot of people are saying is that "Most programs just (a) don't go there, and (b) won't _ever_ go there."
1+1 happens roughly 10,000,000,000 times more often in EVNova than any 64-bit calculation of any sort. (Pretty sure Moki would agree, there just aren't _THAT_ many bad guys
A 64-bit program (one that really does benefit) will run well on the 970. It would run poorly on the G4, the G3, and any x86 that isn't 64bit. Comparing how it will run on a 64bit ppc versus a 64bit x86 of some sort -> we won't really know until there's actual hard benchmarks. I happen to think that the relatively easy transition on the fly on the ppc between 32bits and 64bits will be a real honest benefit, as I could see applications that have some 32-bit threads, and some 64-bit threads -> benefits of both, disadvantages of neither (or at least whichever is lesser).
The machines will be also present on AppleStore as of their advertisement, but with delivery periods from 6 to 8 weeks.
There will be 3 configurations suggested:
- mono 1,4 GHz
- mono 1,8 GHz
- A Bi 2,3 GHz. For this last, the mother chart is not yet completely finalized but should be it during the first week of April. The contractor who will have gained the invitation to tender on the mother chart single processor will manufacture also the chart biprocessor.
Thus note that there will be 2 versions of mother charts, which was foreseeable since the processor on a chart girl but is not directly posed via a ZIF on the mother chart.
Benchmarks were made on the new machines.
I acknowledge being extrèmement excited by all these annonces/rumeurs.
Return amounts! We finally will have what to fight against the Wintel world. Because let us not mislead there we. The fate of APPLE is in the center of all our concerns!
[ MÃ*J ] This update relates to the benchmarks.
After discussions with the members of the TEAM Macbidouille, we decided to us autocensurer on the figures of the benchmarks.
For the first time, we had the impression of too in knowing.
It is in our opinion well too early to show them.
We hope that you will include/understand our decision.
PS: Useless to require of us by mall to know some more, one precise instruction of silence was decided car."
Under OS X you get 4GB per process of virtual address space. That does get partially eaten with mapping the applicable libraries and frameworks, but at least those common files and frameworks only have to show up once in physical RAM.
The old 2GB limit had more to do with the peripheral accessing scheme OS 9 used, those bad old days are long gone.
I agree that MacBidouille has provided some reliable rumors in the past. However, the specs they list sound a tad optimistic. Do they really believe that Apple will introduce a DP 2.3Ghz system at WWDC in June? I can see SP and DP 1.4 and 1.8Ghz PPC 970 systems, but I expect we'll have to wait until early 2004 to break the 2Ghz-barrier.
I'm hoping to buy a new desktop Mac this year. A low-end single-processor PPC 970 1.4Ghz system at $1500 combined with the 20" ACD would be just the system for me.
You're too quick, Leonis! Darn those typos! Actually, it was a Freudian slip. I say I want a 1.4Ghz PPC 970 PowerMac and 20" ACD. But I really want a 10-inch sub-PowerBook. Of course the 10-inch PowerBook would feature DVI-out to connect to the 20" ACD at home.
Even if most software won't benefit by being 64 bit, there are some apps that apparently will. The question is how much over a 32 bit version running on an extremely x86 Intel chip. If it does not outperform the 32 bit version by a fair amount, Apple is still in a compromised situation regarding 64 bit marketing. People will be quick to point out that 64 bitness is part of the Jobsian spin job and reality distortion field.
The same thing happened to Altivec. It was a good excuse for low Mhz when the G4 still competed, but people are fooled no longer. Of course, don't forget that AMD is also moving to 64-bit. We will need a 64-bit CPU to defend us from the AMD fans, so it's not just about an offensive campaign. On the other hand, we might see an alliance between AMD and Apple zealots against Intel.
I agree that MacBidouille has provided some reliable rumors in the past. However, the specs they list sound a tad optimistic. Do they really believe that Apple will introduce a DP 2.3Ghz system at WWDC in June? I can see SP and DP 1.4 and 1.8Ghz PPC 970 systems, but I expect we'll have to wait until early 2004 to break the 2Ghz-barrier.
I'm hoping to buy a new desktop Mac this year. A low-end single-processor PPC 970 1.4Ghz system at $1500 combined with the 20" ACD would be just the system for me.
Escher
I dont think we should count on that 1500 dollar setup.. But if it comes, my family takes two!
Regarding the DP 2.3 Ghz, remember it is supposed to come out two months AFTER WWDC.. And this is not Moto, its IBM. Maybe they have high enough yields? After all, the already have reached 2.5 Ghz..
" Optimism running wild amongest the rumor sites. "
I'm very sceptical of any IBM970 release @ WWDC, especially @ 2.3GHz. Also, I would believe any 2.3GHz 970 would have to be manufactured using a 0.09µm process. Hasn't it been reported by IBM that the 1.8GHz used 40 watts? Then a 2.3 would be ??? very very hot and duals would be very very very very hot.
Having said all that, here's hoping MacBidouille really does know something and isn't just making up stuff based on other rumor sites.
" Optimism running wild amongest the rumor sites. "
I'm very sceptical of any IBM970 release @ WWDC, especially @ 2.3GHz. Also, I would believe any 2.3GHz 970 would have to be manufactured using a 0.09µm process. Hasn't it been reported by IBM that the 1.8GHz used 40 watts? Then a 2.3 would be ??? very very hot and duals would be very very very very hot.
Having said all that, here's hoping MacBidouille really does know something and isn't just making up stuff based on other rumor sites.
If memory serves: reports of the first 2ghz+ chips were on the 0.13 process, the 0.9 was expected to start at 2.5ghz
If memory serves: reports of the first 2ghz+ chips were on the 0.13 process, the 0.9 was expected to start at 2.5ghz
Yes, well kind of. The 2.5GHz 970's were reported in the now unavailable German IBM press release concerning their new blades(to be demoed @ EBIT???) and were allegedly 1.8 - 2.5 GHz using 0.13µm process. This press release was yanked quickly, never to be seen again.
As far as I know, IBM has published nothing concerning the speeds for a 970 based on 0.09µm.
I was just speculating wildly that a 2.3GHz cpu would generate a lot of heat and duals would be doubly so. Here's a wild a$$ guess, a 2.3 GHz 970 might use 60 - 70 watts each, then, 970's X's 2 = 120watts to 140 watts. That's a lot. Soooooo I speculated, that if used, a 2.3 GHz 970 would almost have to be on 0.09Mm process therefore, very unlikely for WWDC.
disclaimer: my above ramblings are from some one with absolutely no technical knowledge, insider information and is to be regarded accordingly.
I was just speculating wildly that a 2.3GHz cpu would generate a lot of heat and duals would be doubly so. Here's a wild a$$ guess, a 2.3 GHz 970 might use 60 - 70 watts each, then, 970's X's 2 = 120watts to 140 watts. That's a lot. Soooooo I speculated, that if used, a 2.3 GHz 970 would almost have to be on 0.09Mm process therefore, very unlikely for WWDC.
I understand what you're saying, makes sense...
I have suggestion though...
Maybe to keep the heat down, the dual will be dual core not dual processor...would that make a difference?
We should remember that 1.8 gig was a conservative estimate.
The fact that the IBM page was 'yanked' could well have been something to do with Apple's pending WWDC announcement?
Anyway you cut it, a 1.4 970 would make a great 'low end' tower and 1.8 a nice mid and a dual 2.3 would be a terrific PR coup for the I.A alliance.
Nicely staggered over recent years of close, 'non-existent' mhz bumps of a bit here...and a bit there. With that line up...when you pay the extra mulla, you getsda business!
PPC gets over 2 gig? That would be a real counter strike for the PPC. And it's only 200 mhz over IBM's conservative goals. I would think Apple would have special treatment on 2 gig-ish chips. And if Pentium Xeon and Athlon dual systems are out there...I don't think a dual 2.3 is outlandish by any means.
I'm quite hopeful. And their quiet excitement over benchmarks has me der-rooling. That augers well.
Im curious, Im not a programmer or chip guru by any means, but would the 970 chip basically be the often rumored G5 or whatever, or is that a wrong term. There are just too many threads to read on this here. .
Comments
Originally posted by wfzelle
You are comparing apples with oranges. 16-bit pointers can address only 64KB, which was totally insufficient when CPU's transitioned to 32-bit. Because the CPU makers were too late, apps were using all kinds of hacks to support more than 64KB. In other words, just about every programmer was clamoring for 32-bit. Furthermore, 32-bit integer calculations were frequently useful and had to be emulated by using two 16-bit integers (slow+ugly).
The current situation is far different. The 2GB that you can address in most 32-bit systems is sufficient for most software that exists today.
My how times have changed. But I don't think that software companies will balk at developing 64-bit apps. Would you want to be the one selling the old, clunky (albeit fine) 32-bit app in a 64-bit world? Look where Apple is right now. Not a pleasant spot to be in.
I think there will be advances in software (ok, probably not Word docs...) that will take advantage of 64-bit goodness. Video and Database are two key areas. I can see great leaps forward for memory intensive applications.
Originally posted by Rhumgod
My how times have changed. But I don't think that software companies will balk at developing 64-bit apps. Would you want to be the one selling the old, clunky (albeit fine) 32-bit app in a 64-bit world? Look where Apple is right now. Not a pleasant spot to be in.
Given that my app (and 99.9% of the apps out there) would not benefit even in the slightest by being 64 bit (and indeed, the performance could slightly degrade, and my app would use slightly more memory), no, I wouldn't feel bad about having a 32 bit app to sell.
But hey, whatever -- selling a "64 bit" machine is something I expect Apple to leverage in their marketing campaign, and perhaps it'll work to the advantage for a change.
However, I do hope Apple does release a 64 bit OS fairly quickly so that applications that can benefit from accessing large amounts of memory are released in fairly short order. I hope Apple doesn't get into the pattern of releasing out of sync software and hardware. OS X in the first releases were too much for the hardware. Let's hope that the next OS release does not limit the hardware from performing to its fullest capabilities.
Originally posted by herbivore
Even if most software won't benefit by being 64 bit, there are some apps. that apparently will. The question is how much over a 32 bit version running on an extremely x86 Intel chip. If it does not outperform the 32 bit version by a fair amount, Apple is still in a compromised situation regarding 64 bit marketing.
If you are cherry picking, you can come up with a pretty convincing program.
Example:
x= somebig64bitintegerX;
y= somebig64bitintegerY;
z=x+y;
Now to do the same thing in a 32bit processor(ignoring carry, which would only make the 32bit program longer & confuse things).
xlo= thelow32bitsofsomebig64bitintegerX;
xhi= thehigh32bitsofsomebig64bitintegerY;
ylo= thelo32bitsofsomebig64bitintegerX;
yhi= thehigh32bitsofsomebig64bitintegerY;
zlo=xlo+ylo;
zhi=xhi+yhi;
3 commands versus 6 commands. In assembly there's all sorts of loads & stores flying around too, but the results are similar (or worse).
For a _demo_, on applications where the central algorithim is fundamentally 'better' on a 64bit machine, the 970 will _crush_ the G4.
But what a lot of people are saying is that "Most programs just (a) don't go there, and (b) won't _ever_ go there."
1+1 happens roughly 10,000,000,000 times more often in EVNova than any 64-bit calculation of any sort. (Pretty sure Moki would agree, there just aren't _THAT_ many bad guys
A 64-bit program (one that really does benefit) will run well on the 970. It would run poorly on the G4, the G3, and any x86 that isn't 64bit. Comparing how it will run on a 64bit ppc versus a 64bit x86 of some sort -> we won't really know until there's actual hard benchmarks. I happen to think that the relatively easy transition on the fly on the ppc between 32bits and 64bits will be a real honest benefit, as I could see applications that have some 32-bit threads, and some 64-bit threads -> benefits of both, disadvantages of neither (or at least whichever is lesser).
"If you do not want to read rumours, go your way. But you will miss an occasion to feel a exaltation that the mac world did not see years since!
APPLE shifted well the dates of the WWDC because of the PPC 970. The new machines there will be presented and will be even in précommande at this time there.
The machines will be also present on AppleStore as of their advertisement, but with delivery periods from 6 to 8 weeks.
There will be 3 configurations suggested:
- mono 1,4 GHz
- mono 1,8 GHz
- A Bi 2,3 GHz. For this last, the mother chart is not yet completely finalized but should be it during the first week of April. The contractor who will have gained the invitation to tender on the mother chart single processor will manufacture also the chart biprocessor.
Thus note that there will be 2 versions of mother charts, which was foreseeable since the processor on a chart girl but is not directly posed via a ZIF on the mother chart.
Benchmarks were made on the new machines.
I acknowledge being extrèmement excited by all these annonces/rumeurs.
Return amounts! We finally will have what to fight against the Wintel world. Because let us not mislead there we. The fate of APPLE is in the center of all our concerns!
[ MÃ*J ] This update relates to the benchmarks.
After discussions with the members of the TEAM Macbidouille, we decided to us autocensurer on the figures of the benchmarks.
For the first time, we had the impression of too in knowing.
It is in our opinion well too early to show them.
We hope that you will include/understand our decision.
PS: Useless to require of us by mall to know some more, one precise instruction of silence was decided car."
LINK
Originally posted by AirSluf
Under OS X you get 4GB per process of virtual address space. That does get partially eaten with mapping the applicable libraries and frameworks, but at least those common files and frameworks only have to show up once in physical RAM.
The old 2GB limit had more to do with the peripheral accessing scheme OS 9 used, those bad old days are long gone.
That's correct. I was mistaken.
Originally posted by rickag
From MacBouille
I agree that MacBidouille has provided some reliable rumors in the past. However, the specs they list sound a tad optimistic. Do they really believe that Apple will introduce a DP 2.3Ghz system at WWDC in June? I can see SP and DP 1.4 and 1.8Ghz PPC 970 systems, but I expect we'll have to wait until early 2004 to break the 2Ghz-barrier.
I'm hoping to buy a new desktop Mac this year. A low-end single-processor PPC 970 1.4Ghz system at $1500 combined with the 20" ACD would be just the system for me.
Escher
Originally posted by Escher
......with the 10" ACD would be just the system for me.
Escher
That screen is a bit too small to my eyes
Originally posted by Leonis
That screen is a bit too small to my eyes
You're too quick, Leonis! Darn those typos! Actually, it was a Freudian slip. I say I want a 1.4Ghz PPC 970 PowerMac and 20" ACD. But I really want a 10-inch sub-PowerBook.
Escher
Originally posted by herbivore
Even if most software won't benefit by being 64 bit, there are some apps that apparently will. The question is how much over a 32 bit version running on an extremely x86 Intel chip. If it does not outperform the 32 bit version by a fair amount, Apple is still in a compromised situation regarding 64 bit marketing. People will be quick to point out that 64 bitness is part of the Jobsian spin job and reality distortion field.
The same thing happened to Altivec. It was a good excuse for low Mhz when the G4 still competed, but people are fooled no longer. Of course, don't forget that AMD is also moving to 64-bit. We will need a 64-bit CPU to defend us from the AMD fans, so it's not just about an offensive campaign. On the other hand, we might see an alliance between AMD and Apple zealots against Intel.
Originally posted by Escher
...... I say I want a 1.4Ghz PPC 970 PowerMac and 20" ACD......
Escher
Yeap. The 20 incher is very nice. Image quality is unrivaled........the bummer is that it's ADC based
Warranty.....one year....blah.....but I paid extra $80 to get the third-party 3 year warranty....
Originally posted by Escher
I agree that MacBidouille has provided some reliable rumors in the past. However, the specs they list sound a tad optimistic. Do they really believe that Apple will introduce a DP 2.3Ghz system at WWDC in June? I can see SP and DP 1.4 and 1.8Ghz PPC 970 systems, but I expect we'll have to wait until early 2004 to break the 2Ghz-barrier.
I'm hoping to buy a new desktop Mac this year. A low-end single-processor PPC 970 1.4Ghz system at $1500 combined with the 20" ACD would be just the system for me.
Escher
I dont think we should count on that 1500 dollar setup.. But if it comes, my family takes two!
Regarding the DP 2.3 Ghz, remember it is supposed to come out two months AFTER WWDC.. And this is not Moto, its IBM. Maybe they have high enough yields? After all, the already have reached 2.5 Ghz..
I did add
"
I'm very sceptical of any IBM970 release @ WWDC, especially @ 2.3GHz. Also, I would believe any 2.3GHz 970 would have to be manufactured using a 0.09µm process. Hasn't it been reported by IBM that the 1.8GHz used 40 watts? Then a 2.3 would be ??? very very hot and duals would be very very very very hot.
Having said all that, here's hoping MacBidouille really does know something and isn't just making up stuff based on other rumor sites.
Originally posted by rickag
Escher
I did add
"
I'm very sceptical of any IBM970 release @ WWDC, especially @ 2.3GHz. Also, I would believe any 2.3GHz 970 would have to be manufactured using a 0.09µm process. Hasn't it been reported by IBM that the 1.8GHz used 40 watts? Then a 2.3 would be ??? very very hot and duals would be very very very very hot.
Having said all that, here's hoping MacBidouille really does know something and isn't just making up stuff based on other rumor sites.
If memory serves: reports of the first 2ghz+ chips were on the 0.13 process, the 0.9 was expected to start at 2.5ghz
Originally posted by robster
If memory serves: reports of the first 2ghz+ chips were on the 0.13 process, the 0.9 was expected to start at 2.5ghz
Yes, well kind of. The 2.5GHz 970's were reported in the now unavailable German IBM press release concerning their new blades(to be demoed @ EBIT???) and were allegedly 1.8 - 2.5 GHz using 0.13µm process. This press release was yanked quickly, never to be seen again.
As far as I know, IBM has published nothing concerning the speeds for a 970 based on 0.09µm.
I was just speculating wildly that a 2.3GHz cpu would generate a lot of heat and duals would be doubly so. Here's a wild a$$ guess, a 2.3 GHz 970 might use 60 - 70 watts each, then, 970's X's 2 = 120watts to 140 watts. That's a lot. Soooooo I speculated, that if used, a 2.3 GHz 970 would almost have to be on 0.09Mm process therefore, very unlikely for WWDC.
disclaimer: my above ramblings are from some one with absolutely no technical knowledge, insider information and is to be regarded accordingly.
Originally posted by rickag
I was just speculating wildly that a 2.3GHz cpu would generate a lot of heat and duals would be doubly so. Here's a wild a$$ guess, a 2.3 GHz 970 might use 60 - 70 watts each, then, 970's X's 2 = 120watts to 140 watts. That's a lot. Soooooo I speculated, that if used, a 2.3 GHz 970 would almost have to be on 0.09Mm process therefore, very unlikely for WWDC.
I understand what you're saying, makes sense...
I have suggestion though...
Maybe to keep the heat down, the dual will be dual core not dual processor...would that make a difference?
Originally posted by robster
I understand what you're saying, makes sense...
I have suggestion though...
Maybe to keep the heat down, the dual will be dual core not dual processor...would that make a difference?
I think the 970 is single core only. The double core version of the 970 is IBMs Power4 wo altivec.
We should remember that 1.8 gig was a conservative estimate.
The fact that the IBM page was 'yanked' could well have been something to do with Apple's pending WWDC announcement?
Anyway you cut it, a 1.4 970 would make a great 'low end' tower and 1.8 a nice mid and a dual 2.3 would be a terrific PR coup for the I.A alliance.
Nicely staggered over recent years of close, 'non-existent' mhz bumps of a bit here...and a bit there. With that line up...when you pay the extra mulla, you getsda business!
PPC gets over 2 gig? That would be a real counter strike for the PPC. And it's only 200 mhz over IBM's conservative goals. I would think Apple would have special treatment on 2 gig-ish chips. And if Pentium Xeon and Athlon dual systems are out there...I don't think a dual 2.3 is outlandish by any means.
I'm quite hopeful. And their quiet excitement over benchmarks has me der-rooling. That augers well.
And I'll take mine in chernobyl green!
Lemon Bon Bon