Bah, I don't believe that -- anything can be broken down into triangles, and modern hardware can spit out millions per second. 2D scrolling text is not exactly ideal OpenGL content either, and yet the fastest terminal program on OSX is OpenGL based. At worst Apple could define a few addition driver entry points to support the GUI acceleration.
What, in particular, makes the PDF nature of Quartz a problem for OpenGL?</strong><hr></blockquote>
So? The CPU can do the tesselation and feed triangles to the rasterizer. Not everything is curved, and for things which are curved, single scan line fragments could be output in the worst case. Most software rasterizers do this internally anyhow. Text characters are not usually distorted, and icons don't change often so caching of images in VRAM could deliver big performance wins. I don't know how much of this Quartz already does (if any), but I do know that I haven't seen anything in the OSX GUI that should cause a 400+ MHz G4 w/ at least a Rage128Pro to run sluggishly! And if they can't make it go fast, then there should be options to turn all of it off.
BerberCarpet: I understand your point, but servers have never been an Apple specialty, esp. high-end servers. JYD was talking about graphic designers, folks who have made their living on Macs for years. The designers I know haven't made any noise at all about switching to Windows. I know anecdotes aren't evidence, but neither are anyone else's, right?
The GUI is not what makes OS X slow. I've heard this from people within Apple, and, as a programmer, I tend to agree. Notice how dragging a window with its nice dropshadow is silky smooth on even the slowest machines. Some of the sluggishness might be attributable to the design of the Window Server and the way in which current applications use it, but that's not something that needs any special hardware acceleration to fix.
Aqua is just the most visible of OS X changes. There's a lot of cruft in the core system, and it'll take a while to weed out.
I'd just thought I'd point out that, no, window dragging is not silky smooth on all computers. Try dragging even a medium sized window on a Rev A iMac and you'll see what I mean
Please site your source for the market share. I missed that in the Business Week articles and Apple's financials.</strong><hr></blockquote>
From Business Week:
"Despite Apple's considerable achievements -- and the hype -- its market share remains stalled at a decidedly modest 5% of the U.S. household computer market, according to International Data Corp. Worldwide, its share is closer to 3%, though Apple claims that number would be higher if its international education market were included."
Uh, no. Don't believe me? Then ask John Carmack. Mhz does matter for games. If it didn't then the huge fps discrepancies between Macs and PCs for UT and Q3 wouldn't exist. Throughout the last few years in which PC gaming has seen a huge rise, one general trend has been observed. The latest gen video card of it's time was always cpu limited. In other words the CPU (on both sides of the fence) has been the limiting factor in higher fps numbers. The GPU industry has stayed ahead of the curve of CPU manufacturers by generally releasing next gen GPUs with a lot of headroom. So how does one explain the discrepancy in time demo fps between a PC and a Mac each with an identical GeForce3? It's definitely not the drivers. Drivers wouldn't explain that large of a difference. The fact is that P4s and Athlons use a lot more of the GF3's headroom than a Mac.</strong>
<hr></blockquote>
that large fps gap happens when testers set all candy off on quake 3 so that the chip does most of the work. in this instance, yes, mhz is everything, because thats all you've got- you're for the most part not offloading and processing to the gpu . . . with candy on and gf3 rates drop a lot- not that they don't beat the pants off the highend macs. but maybe not by as much as you think
the game i'm waiting for is civilization 3... come on macplay... think i'll wait for the g5 though with ddr and 400mhz bus because civ3 needs at least 450 fps
"Despite Apple's considerable achievements -- and the hype -- its market share remains stalled at a decidedly modest 5% of the U.S. household computer market, according to International Data Corp. Worldwide, its share is closer to 3%, though Apple claims that number would be higher if its international education market were included."</strong><hr></blockquote>
The IDC numbers don't account for the installed base - only units sold in a specific time frame.
<strong>Tesselating a continuous b-spline (of which Bezier curves are a special case) would be just as much if not more work than just transitioning them straight into pixel color values.</strong><hr></blockquote>
Fair enough, but how much of the GUI really requires that, and should it not be rasterized once and then cached until it changes? And the special cases are the most commonly drawn things. I admit that I don't know enough about the details of Aqua to really get into a discussion of its architecture -- but that doesn't change the complaints about it being slow. These days a slow GUI is inexcusable (and I don't care if the reasons is that it is based on bezier curves, or crop circles)...
More interesting is the comment above that the perceived slowness is due to internal bottlenecks of OSX, not the actual GUI.. this is very valid -- users will complain about what they perceive as being slow, not what is actually slow, and the only thing the users perceive is the interface. Its easy to take the user complaints at face value, like I did, but that is usually wrong. Curing the perception is not a bad approach though -- smoothing out the GUI despite underlying system bottlenecks, if that is easier than removing the bottlenecks (sometimes this can be done by simply slowing the rest of the animation so that it runs at a slower but more even pace). The users will complain less because the machine "feels" faster.
Well to get back on topic, the consensus seems to be that a minor speed bump will occur this week, and G5s with new motherboards (DDR, Gigawire etc) will be released later on this year (MWNY). In all likelyhood the MacuserUK report seems accurate. Apple would not realease a completely new powermac/chip with a simple press release. The G5 is a big deal, certainly MW material. So it is logical then that the updated powermacs will use Apollo G4s. Motorolla itself announced they would be available early this year. Therefore it makes sense that the new powermacs will use these chips. I doubt that Apple would/ can release powermacs at speeds much greater than a gigahertz. Apple is at Motorola mercy, if the chips don't exist, they cannot produce them. So remaining within the realm od reason, I present not a prediction, but rather a realistic proposal for updated Powermacs:
1. All new Powermacs run at 1GHZ, at various multiprocessor configs.
2. New powermacs use same motherboard/case (QS with 133 bus/memory etc).
3. All powermacs superdirves.
4. Have at least 512MB memory
5. At least Geforce2MX with DDR memory.
So my suggestion for Apple is the following configs. The only reasons I can see why they could not do this is not enough chips and Apple's need for high product margins.
$1699 Apollo 1GHZ G4 512MB PC133 32MB Geforce2MX wDDR 60GB 7200RPM HD Superdirve
$2499 Dual Apollo 1 GHZ G4 1gig PC133 Geforce 3 100GB 7200rpm HD Superdirve
$3499 Quad Apollo 1GHZ G4 1.5 Gig PC133
Dual Monitor Geforce3 120 GB 7200RPM HD Superdrive
Optional PCI Radeon 7000 and AGP Radeon 8500 plus usaul zip, SCSI etc.
Again, this is not a prediction, but a sugestion. Apple cannot control Motorola's ability to have 2GHZ G5s ready todays. For the average Mac user (the ones that do not spend absurd amounts of time lurking on the appleinsider boards) these specs would be very appealing. Though they are not G5s, they offer a lot of memory, superdirves, fast graphics, and multiprocessing. With OSX multi processors and large amounts of memory make a dramatic difference. Will this happen? Probably not. unfortunetly Apple needs to make large margins on their Pro products. Machines like the ones I described above would clear out inventory of QS cases and MB, making ready for G5s. Most importantly however, THEY ARE POSSIBLE!!! Apple if you're listening, bight the bullet and lower your margins on powermacs and give us more bang for our buck. What you lose in margins will be made up in volume. IT CAN BE DONE!!!
[quote]The IDC numbers don't account for the installed base - only units sold in a specific time frame.<hr></blockquote>
Installed base and market share are not the same thing and are not calculated the same way. Apple's market share is under 3% and has been for well over a year. Apple's installed base is higher than that, but this as a percentage is shrinking because of the shrinking in market share. The two are not the same thing, but there is causality in their relationship, at least in the long run. In the short term, trends are harder to see.
Comments
<strong>
Bah, I don't believe that -- anything can be broken down into triangles, and modern hardware can spit out millions per second. 2D scrolling text is not exactly ideal OpenGL content either, and yet the fastest terminal program on OSX is OpenGL based. At worst Apple could define a few addition driver entry points to support the GUI acceleration.
What, in particular, makes the PDF nature of Quartz a problem for OpenGL?</strong><hr></blockquote>
Bezier curves.
<strong>Bezier curves.</strong><hr></blockquote>
So? The CPU can do the tesselation and feed triangles to the rasterizer. Not everything is curved, and for things which are curved, single scan line fragments could be output in the worst case. Most software rasterizers do this internally anyhow. Text characters are not usually distorted, and icons don't change often so caching of images in VRAM could deliver big performance wins. I don't know how much of this Quartz already does (if any), but I do know that I haven't seen anything in the OSX GUI that should cause a 400+ MHz G4 w/ at least a Rage128Pro to run sluggishly! And if they can't make it go fast, then there should be options to turn all of it off.
Aqua is just the most visible of OS X changes. There's a lot of cruft in the core system, and it'll take a while to weed out.
Alex
Well Apple's market share has dropped significantly in the past few years(to 2.5% or less right now).
[QB]Is there any actual evidence of this, or are people simply projecting their own frustrations? QB]<hr></blockquote>
Well,
we have 50 plus G4s (a total cost of approximately 150,000 bgpounds - we're in education) and the powers that be will switch us to XP. Why?
Mac hardware is "too far behind" (their words)
Photoshop, Director, Dreamweaver, Flash are not available for X
"Maya for Mac just doesn't give as much bang for bucks as Max for XP" (their words)
"You can't seriously consider them value for money given the competition out there?" (their words)
Its really sad that my son will grow up in a totally Wintel world.
My daughter will grow up in a wintel world using macs
<strong>Is there any actual evidence of this, or are people simply projecting their own frustrations?
Well Apple's market share has dropped significantly in the past few years(to 2.5% or less right now).</strong><hr></blockquote>
Please site your source for the market share. I missed that in the Business Week articles and Apple's financials.
I think it all started here at AI.
<strong>I would also like know what evidence there is that new PowerMac's are being announced.
I think it all started here at AI.</strong><hr></blockquote>
MacMinute or MacUser UK, one of the 2 broke it.
<strong>
Please site your source for the market share. I missed that in the Business Week articles and Apple's financials.</strong><hr></blockquote>
From Business Week:
"Despite Apple's considerable achievements -- and the hype -- its market share remains stalled at a decidedly modest 5% of the U.S. household computer market, according to International Data Corp. Worldwide, its share is closer to 3%, though Apple claims that number would be higher if its international education market were included."
<strong>
Uh, no. Don't believe me? Then ask John Carmack. Mhz does matter for games. If it didn't then the huge fps discrepancies between Macs and PCs for UT and Q3 wouldn't exist. Throughout the last few years in which PC gaming has seen a huge rise, one general trend has been observed. The latest gen video card of it's time was always cpu limited. In other words the CPU (on both sides of the fence) has been the limiting factor in higher fps numbers. The GPU industry has stayed ahead of the curve of CPU manufacturers by generally releasing next gen GPUs with a lot of headroom. So how does one explain the discrepancy in time demo fps between a PC and a Mac each with an identical GeForce3? It's definitely not the drivers. Drivers wouldn't explain that large of a difference. The fact is that P4s and Athlons use a lot more of the GF3's headroom than a Mac.</strong>
<hr></blockquote>
that large fps gap happens when testers set all candy off on quake 3 so that the chip does most of the work. in this instance, yes, mhz is everything, because thats all you've got- you're for the most part not offloading and processing to the gpu . . . with candy on and gf3 rates drop a lot- not that they don't beat the pants off the highend macs. but maybe not by as much as you think
<strong>
From Business Week:
"Despite Apple's considerable achievements -- and the hype -- its market share remains stalled at a decidedly modest 5% of the U.S. household computer market, according to International Data Corp. Worldwide, its share is closer to 3%, though Apple claims that number would be higher if its international education market were included."</strong><hr></blockquote>
The IDC numbers don't account for the installed base - only units sold in a specific time frame.
<strong>
The IDC numbers don't account for the installed base - only units sold in a specific time frame.</strong><hr></blockquote>
True, but I don't think I've ever seen any real definitive numbers on installed base, outside of the education market, that is.
<strong>Tesselating a continuous b-spline (of which Bezier curves are a special case) would be just as much if not more work than just transitioning them straight into pixel color values.</strong><hr></blockquote>
Fair enough, but how much of the GUI really requires that, and should it not be rasterized once and then cached until it changes? And the special cases are the most commonly drawn things. I admit that I don't know enough about the details of Aqua to really get into a discussion of its architecture -- but that doesn't change the complaints about it being slow. These days a slow GUI is inexcusable (and I don't care if the reasons is that it is based on bezier curves, or crop circles)...
More interesting is the comment above that the perceived slowness is due to internal bottlenecks of OSX, not the actual GUI.. this is very valid -- users will complain about what they perceive as being slow, not what is actually slow, and the only thing the users perceive is the interface. Its easy to take the user complaints at face value, like I did, but that is usually wrong. Curing the perception is not a bad approach though -- smoothing out the GUI despite underlying system bottlenecks, if that is easier than removing the bottlenecks (sometimes this can be done by simply slowing the rest of the animation so that it runs at a slower but more even pace). The users will complain less because the machine "feels" faster.
Anyhow, this is way off topic...
1. All new Powermacs run at 1GHZ, at various multiprocessor configs.
2. New powermacs use same motherboard/case (QS with 133 bus/memory etc).
3. All powermacs superdirves.
4. Have at least 512MB memory
5. At least Geforce2MX with DDR memory.
So my suggestion for Apple is the following configs. The only reasons I can see why they could not do this is not enough chips and Apple's need for high product margins.
$1699 Apollo 1GHZ G4 512MB PC133 32MB Geforce2MX wDDR 60GB 7200RPM HD Superdirve
$2499 Dual Apollo 1 GHZ G4 1gig PC133 Geforce 3 100GB 7200rpm HD Superdirve
$3499 Quad Apollo 1GHZ G4 1.5 Gig PC133
Dual Monitor Geforce3 120 GB 7200RPM HD Superdrive
Optional PCI Radeon 7000 and AGP Radeon 8500 plus usaul zip, SCSI etc.
Again, this is not a prediction, but a sugestion. Apple cannot control Motorola's ability to have 2GHZ G5s ready todays. For the average Mac user (the ones that do not spend absurd amounts of time lurking on the appleinsider boards) these specs would be very appealing. Though they are not G5s, they offer a lot of memory, superdirves, fast graphics, and multiprocessing. With OSX multi processors and large amounts of memory make a dramatic difference. Will this happen? Probably not. unfortunetly Apple needs to make large margins on their Pro products. Machines like the ones I described above would clear out inventory of QS cases and MB, making ready for G5s. Most importantly however, THEY ARE POSSIBLE!!! Apple if you're listening, bight the bullet and lower your margins on powermacs and give us more bang for our buck. What you lose in margins will be made up in volume. IT CAN BE DONE!!!
Installed base and market share are not the same thing and are not calculated the same way. Apple's market share is under 3% and has been for well over a year. Apple's installed base is higher than that, but this as a percentage is shrinking because of the shrinking in market share. The two are not the same thing, but there is causality in their relationship, at least in the long run. In the short term, trends are harder to see.