That article appears less informed than your _average_ AI poster. A couple lines made me question the validity of the entire thing.
Quote:
However, sources said, Q37 won't ship with a 64-bit version of Mac OS X, limiting OS performance gains in the initial release.
The authors obviously think that the 'bitness' of a program determines a program's speed and that a 64bit OS will run faster than a 32bit OS.
The rest of the article is fairly sound but only rehashes the same rumors found on every board... heheh 'according to sources'. I guess the article isn't terrible, just nothing new.
However, sources said, Q37 won't ship with a 64-bit version of Mac OS X, limiting OS performance gains in the initial release.
There's a good way to smoke out a source. Each group gets a different project number (Q1, Q2, ... Qn). The number that pops up identifies the group, if not necessarily the person.
Probably not so, but Tom Clancy wannabes can fantasize, right?
This is starting to make a lot of sense, and I think we will be verrryyy happy in a couple of weeks.
Quote:
From the above-linked article:
Apple's goal for Smeagol is to deliver Mac OS X performance at least "on par" with what Jaguar could achieve on Motorola G4 chips running at the same speed...
...Even before Panther ships, early adopters of the new Mac system should also apparently be able to take advantage of the new processor's fast new front-side bus and cache.
I think the author is attributing way too much importance to 64-bitness for performance, way too little to the faster bus, doesn't even mention that the 970 should get more work done per clock cycle, and also fails to mention that the 970s (if we do indeed get 1.8 GHz models) should be running higher clocks speeds the current G4s.
From what I understand, OS X, and all of our current apps running under OS X, should be able to make great leaps forward in performance, even without Pather's 64-bit capabilities -- much better than merely being "on par".
That article appears less informed than your _average_ AI poster. A couple lines made me question the validity of the entire thing.The authors obviously think that the 'bitness' of a program determines a program's speed and that a 64bit OS will run faster than a 32bit OS.
The rest of the article is fairly sound but only rehashes the same rumors found on every board... heheh 'according to sources'. I guess the article isn't terrible, just nothing new.
Hmm, I think you're less average than you think you are...
The authors suggest very little about the 'bitness' in the article. They suggest a great deal about the relationship between gcc versions and the 970, which would seem to be spot on.
Two things we can be certain of:
1) gcc will produce increasingly better, faster, smaller 970 code with time.
2) 10.3 will be faster than 10.2 on all hardware, not just the 970s.
My take on the article is that a version of 10.2 will be produced for the new machines (which goes against what I thought Apple would do, and clearly is not making Steve a happy boy) and not only will it not be as optimized as 10.3 due to the fact that it's simply not 10.3, it might even be *relatively* slower than 10.2 for the G4 architecture due to gcc not supporting the 970 at the same level. The statement that they hope it runs as fast as 10.2 would on a comparably clocked G4 is, well, a serious disappointment since it suggests that until 10.3 ships, we should ignore any benefits the 970 might deliver beyond Mhz.
You can expect to hear Steve say: "and wait until you see how fast it is running Panther..." as a caveat to the announcement. Also worth noting is that the article suggests that Smeagol won't be ready until late July/August, so don't expect to buy that shiny 970 in 2 weeks...
I wonder what Apple will ship to reviewers. New hardware that is only linearly and predicatably faster than the G4 will be not looked upon favorably. Expect a lot of "Still much slower than a P4" statements. Of course, they might ship an incomplete Panther, which might play better since they'd blame any problems on the OS being unfinished rather than the hardware being slow. Due to this, Apple might have been forced to release a preview Panther at WWDC.
The authors suggest very little about the 'bitness' in the article. They suggest a great deal about the relationship between gcc versions and the 970, which would seem to be spot on.
Two things we can be certain of:
1) gcc will produce increasingly better, faster, smaller 970 code with time.
2) 10.3 will be faster than 10.2 on all hardware, not just the 970s.
I'm sure having an optimized compiler will be a Good Thing, but I'm suspicious that the lack of compiler optimization would nearly cancel out so many of the benefits of the 970 vs. the G4.
If this were true, Mac software companies would be in a bit of a bind: Sell two different versions of every OS X app, or sell one compromise version for both G4s and 970s.
Questions for anyone who knows: What's typically done for G3 vs. G4? Other than Altivec issues, which can be dealt with like "if G4 do this fast stuff, else do this other slow stuff", are there conflicting optimizations for G3 vs. G4? If so, how do most software vendors deal with this?
My gut feeling is that 970 systems will easily outdo G4 systems on nearly all tasks, compiler optimizations or not, 64-bit OS or not.
'Smeagol' is from JRR Tolkien's Lord of the Rings.
It is the actual (original) name of the character most people know as 'Gollum', who was known to refer to the One Ring as "My Precious".
history info link
- - -
FWIW, insofar as 'Q37', I was just thinking about Q-Ships today.
(to preemptively explain another likely obscure reference, Q-Ships was a military term for warships disguised as mere freighters. They were designed to look like sitting ducks to lure attacks from German U-Boats and then pound the snot out of them.
The term was later used to describe high performance sports cars that were similarly "disguised", although more in Europe than in the USA, where we called them "Sleepers".
A modern example of a Sleeper or Q-Ship automobile would be something like the Audi RS4 Quattro Avant. It looks like a normal compact station wagon, but its engine has been worked over by Porsche: 0-60mph is <5 sec, 0-100ph is <12sec, and 0-125 is 17sec. Coming back down, you find that Porsche's worked on the suspension/braking systems too: 70mph-0 in 155ft.
It would be nice to have a Macintosh Q-Ship
-hh
Last edited by -hh on 06-09-2003 at 09:32 PM
I thought the 'Q' explanation was insightful...whether true or not. The idea of a 'sleeper' that 'pounds the snot' out of 'em works well for me in the PPC vs Pentium 4 context. Get ready, Intel... It may have some workstation/switcher market connotations...
As for Compiler readiness. Yeah. I expect that. If you don't optimise software you cpu doesn't run to its full potential. Well, duh.
The 970 should, in my view run at least as well as a 1.8 gig G4 on much improved bandwidth. (Don't see any G4s with extra fpu, that bandwidth or 1.8 gig. Do you?) If it takes 'Panther' to get to it's full 'twice as fast' claims. Okay. I can live with that. It needs compiler 3.3 and we all sodding know its for Panther and optimised for Panther. So, what's the big deal?
You're still going to get a faster machine from IBM/Apple than you would anytime from Motorola in the next year or so.
Imagine getting a much faster machine. Then one that gets faster still when you get your free update to Panther. Which should turn the 970 into 'Mr. Hyde'. Ouch time for Mr. Pentium.
Personally, Lemon BOn Bon has just laid out the smack on Adobe Design Collection 6, MX Studio and Final Cut Pro 3 so I won't be able to buy a 970 if they ship in August. It'll be post 970. By then dual 1.8s...and a Rev B coming down the pipe.
I'll live.
Either way. It's a new era, folks. There's so many 970 smoke signals saying the same thing...it's just a matter of waiting. The WWDC should be a stormer.
Comments
We need a time machine that will send us to june 23rd.
I agree. Just to go forward in time for about a 30 second summary of what happened.
Then we can come back and determine if this discussion needs to continue or not.
This is starting to make a lot of sense, and I think we will be verrryyy happy in a couple of weeks.
. Apple and IBM are also reportedly tailoring the chip for a new high-frequency, point-to-point Mac bus dubbed ApplePI
mmmm...... pie..........
However, sources said, Q37 won't ship with a 64-bit version of Mac OS X, limiting OS performance gains in the initial release.
The authors obviously think that the 'bitness' of a program determines a program's speed and that a 64bit OS will run faster than a 32bit OS.
The rest of the article is fairly sound but only rehashes the same rumors found on every board... heheh 'according to sources'. I guess the article isn't terrible, just nothing new.
However, sources said, Q37 won't ship with a 64-bit version of Mac OS X, limiting OS performance gains in the initial release.
There's a good way to smoke out a source. Each group gets a different project number (Q1, Q2, ... Qn). The number that pops up identifies the group, if not necessarily the person.
Probably not so, but Tom Clancy wannabes can fantasize, right?
Originally posted by blue2kdave
more on smeagol
This is starting to make a lot of sense, and I think we will be verrryyy happy in a couple of weeks.
From the above-linked article:
Apple's goal for Smeagol is to deliver Mac OS X performance at least "on par" with what Jaguar could achieve on Motorola G4 chips running at the same speed...
...Even before Panther ships, early adopters of the new Mac system should also apparently be able to take advantage of the new processor's fast new front-side bus and cache.
I think the author is attributing way too much importance to 64-bitness for performance, way too little to the faster bus, doesn't even mention that the 970 should get more work done per clock cycle, and also fails to mention that the 970s (if we do indeed get 1.8 GHz models) should be running higher clocks speeds the current G4s.
From what I understand, OS X, and all of our current apps running under OS X, should be able to make great leaps forward in performance, even without Pather's 64-bit capabilities -- much better than merely being "on par".
Rothenberg and dePlume have collaborated on articles before, and this is the second Ciarelli / Rothenberg article.
Originally posted by dfiler
That article appears less informed than your _average_ AI poster. A couple lines made me question the validity of the entire thing.The authors obviously think that the 'bitness' of a program determines a program's speed and that a 64bit OS will run faster than a 32bit OS.
The rest of the article is fairly sound but only rehashes the same rumors found on every board... heheh 'according to sources'. I guess the article isn't terrible, just nothing new.
Hmm, I think you're less average than you think you are...
The authors suggest very little about the 'bitness' in the article. They suggest a great deal about the relationship between gcc versions and the 970, which would seem to be spot on.
Two things we can be certain of:
1) gcc will produce increasingly better, faster, smaller 970 code with time.
2) 10.3 will be faster than 10.2 on all hardware, not just the 970s.
My take on the article is that a version of 10.2 will be produced for the new machines (which goes against what I thought Apple would do, and clearly is not making Steve a happy boy) and not only will it not be as optimized as 10.3 due to the fact that it's simply not 10.3, it might even be *relatively* slower than 10.2 for the G4 architecture due to gcc not supporting the 970 at the same level. The statement that they hope it runs as fast as 10.2 would on a comparably clocked G4 is, well, a serious disappointment since it suggests that until 10.3 ships, we should ignore any benefits the 970 might deliver beyond Mhz.
You can expect to hear Steve say: "and wait until you see how fast it is running Panther..." as a caveat to the announcement. Also worth noting is that the article suggests that Smeagol won't be ready until late July/August, so don't expect to buy that shiny 970 in 2 weeks...
I wonder what Apple will ship to reviewers. New hardware that is only linearly and predicatably faster than the G4 will be not looked upon favorably. Expect a lot of "Still much slower than a P4" statements. Of course, they might ship an incomplete Panther, which might play better since they'd blame any problems on the OS being unfinished rather than the hardware being slow. Due to this, Apple might have been forced to release a preview Panther at WWDC.
Hmm.
Originally posted by johnsonwax
Also worth noting is that the article suggests that Smeagol won't be ready until late July/August, so don't expect to buy that shiny 970 in 2 weeks...
Buying is one thing - getting it home is another
Originally posted by johnsonwax
The authors suggest very little about the 'bitness' in the article. They suggest a great deal about the relationship between gcc versions and the 970, which would seem to be spot on.
Two things we can be certain of:
1) gcc will produce increasingly better, faster, smaller 970 code with time.
2) 10.3 will be faster than 10.2 on all hardware, not just the 970s.
I'm sure having an optimized compiler will be a Good Thing, but I'm suspicious that the lack of compiler optimization would nearly cancel out so many of the benefits of the 970 vs. the G4.
If this were true, Mac software companies would be in a bit of a bind: Sell two different versions of every OS X app, or sell one compromise version for both G4s and 970s.
Questions for anyone who knows: What's typically done for G3 vs. G4? Other than Altivec issues, which can be dealt with like "if G4 do this fast stuff, else do this other slow stuff", are there conflicting optimizations for G3 vs. G4? If so, how do most software vendors deal with this?
My gut feeling is that 970 systems will easily outdo G4 systems on nearly all tasks, compiler optimizations or not, 64-bit OS or not.
--hh
macrumors regular
Registered: Jul 2001
Location: NJ Highlands
quote:
--------------------------------------------------------------------------------
Originally posted by arnette
Am I the only one not getting the 'Smeagol' reference??
--------------------------------------------------------------------------------
'Smeagol' is from JRR Tolkien's Lord of the Rings.
It is the actual (original) name of the character most people know as 'Gollum', who was known to refer to the One Ring as "My Precious".
history info link
- - -
FWIW, insofar as 'Q37', I was just thinking about Q-Ships today.
(to preemptively explain another likely obscure reference, Q-Ships was a military term for warships disguised as mere freighters. They were designed to look like sitting ducks to lure attacks from German U-Boats and then pound the snot out of them.
The term was later used to describe high performance sports cars that were similarly "disguised", although more in Europe than in the USA, where we called them "Sleepers".
A modern example of a Sleeper or Q-Ship automobile would be something like the Audi RS4 Quattro Avant. It looks like a normal compact station wagon, but its engine has been worked over by Porsche: 0-60mph is <5 sec, 0-100ph is <12sec, and 0-125 is 17sec. Coming back down, you find that Porsche's worked on the suspension/braking systems too: 70mph-0 in 155ft.
It would be nice to have a Macintosh Q-Ship
-hh
Last edited by -hh on 06-09-2003 at 09:32 PM
I thought the 'Q' explanation was insightful...whether true or not. The idea of a 'sleeper' that 'pounds the snot' out of 'em works well for me in the PPC vs Pentium 4 context. Get ready, Intel... It may have some workstation/switcher market connotations...
As for Compiler readiness. Yeah. I expect that. If you don't optimise software you cpu doesn't run to its full potential. Well, duh.
The 970 should, in my view run at least as well as a 1.8 gig G4 on much improved bandwidth. (Don't see any G4s with extra fpu, that bandwidth or 1.8 gig. Do you?) If it takes 'Panther' to get to it's full 'twice as fast' claims. Okay. I can live with that. It needs compiler 3.3 and we all sodding know its for Panther and optimised for Panther. So, what's the big deal?
You're still going to get a faster machine from IBM/Apple than you would anytime from Motorola in the next year or so.
Imagine getting a much faster machine. Then one that gets faster still when you get your free update to Panther. Which should turn the 970 into 'Mr. Hyde'. Ouch time for Mr. Pentium.
Personally, Lemon BOn Bon has just laid out the smack on Adobe Design Collection 6, MX Studio and Final Cut Pro 3 so I won't be able to buy a 970 if they ship in August. It'll be post 970. By then dual 1.8s...and a Rev B coming down the pipe.
I'll live.
Either way. It's a new era, folks. There's so many 970 smoke signals saying the same thing...it's just a matter of waiting. The WWDC should be a stormer.
I can't wait.
Lemon Bon Bon
My gut feeling is that 970 systems will easily outdo G4 systems on nearly all tasks, compiler optimizations or not, 64-bit OS or not.
There's alot of crap and voodoo mind games put out there by the Apple team I'm sure. Only when Steve Jobs takes to the stage will we truly know.
I agree with you in short. We shouldn't be worried even minus Panther...it should well and truly blow the dual G4 out the water.
Lemon Bon Bon
Originally posted by JLL
Hmm, Nick Ciarelli - is that the real name of Nick dePlume?
Rothenberg and dePlume have collaborated on articles before, and this is the second Ciarelli / Rothenberg article.
Is Nick dePlume even his real name? I always thought it was just a play on the term "nom de plume" (a pen name).