<strong>300% - Don't exaggurate! It's a measly 251% in 3DMark at 1600x1200, 4x FSAA. nVidia shouldn't worry (not).
Barto</strong><hr></blockquote>
<nitpicking>
300% is not an exaggeration, although Anandtech's published tests said 251%, HardOCP's said 10000+ (I don't recall the exact numbers) marks vs. 2500+ marks, which is, oh yes, 300% faster</nitpicking>
But I agree with Aris, if one (ATi) company can do it, another (Apple) can too.
Even though it almost would be too good to be true
also if u see the positiveof it u see how steadly the g4 is improving and with 1.4 ghz processors being rumored a 400mhz increase isn't bad at all compared to what we have seen before on previous updates
yup, a 40% MHz increase has been seen before, although together with a major architectural change, increasing by the same amount and a price increase like last time (urgh!) is far from impossible, if mot has been working hard enough with the .13µ SOI 74XX.
<strong>talso if u see the positiveof it u see how steadly the g4 is improving and with 1.4 ghz processors being rumored a 400mhz increase isn't bad at all compared to what we have seen before on previous updates
all powermacs need are a fast mobo</strong><hr></blockquote>
...RUMORS...
As Cubba Gooding Jr. would say 'show me the money'
You want a faster MOBO well you can blame MOT for that one too.... Fact is all the PowerPC cpu REALLY needs is a parent who gives it some attention. Three years and MOT has done nothing. Rumors not withstanding.
<strong>One more thing about the new powerMacs I forgot to include in my original post. As far as I can tell from what I have heard they will ship with 10.2 Wether or not they come out before August 24. I am not sure if they will require 10.2 to run or not I would bet they don't need it but apple will not release new machines without 10.2.</strong><hr></blockquote>
10.2 will be declared GM long before the 24th. The 24th is the day all the boxes have to be in the stores or with the resellers. Apple has plenty of time beforehand to preload the finished OS onto new hardware.
8.5GB/s of memory throughput (enough to provide each CPU with 4.3GB/s)
And of course, "Pentium-crushing performance, but now in every benchmark."
------------------------------------------
If apple made the system described above I would be delighted. Of course I would not be able to afford it. Maybe they will do something like this on the so called Xstation (if they do make one). The new PowerMacs will include a lot of this though and faster CPUs. We will see some nice performance gains come august. In time.
A. I can't remember a time when processor speeds jumped by more than 100 MHz or so. What gives anyone the idea we will suddenly jump by 2, 3 or even 400 MHz? Use your head ... people who say this are basically "buying their own bullsh*t" because it's what they want, not because there's a logical reason to believe it.
B. Motorola's cost-savings and improved management don't seem to be paying off much, and one of their best senior executives just left the company for a job at Tyco. I think they're going down the toilet for real.
I predict within a year, parts of Motorola will be sold off to other companies and the rest will go down the tubes / be liquidated. Unless they make a dramatic financial turnout over the next couple quarters, they're finished.
C. If the next batch of machines are far superior to what we have, it won't have anything to do with Motorola's CPU inside. More like Apple will have teamed up with NVidia and others to redesign the UMA, to increase the bandwidth between critical system components. That way processors in the 1GHz - 1.3 GHz range can still at least make a showing against much faster PCs (when running design software anyway). Thus Apple will have something to sell until they can come up with a real pro hardware solution.
One note: for the guy above who can't remember an update with more than 100mhz, look no further back than the dual gig model. Back in January, we got a single 933mhz, granted, but there was also the dual 800 that was bumped to a dual 1000mhz - 200mhz jump there....
Though the point may be lost by now in this thread, I think that it is important to stress that the reason why the G4 is slow is not Altivec, but Motorola's complete inability and lack of desire to manufacture speedy chips. Don't blame the silicon design for management's incompetence.
[quote]And no, the 50 MB image does not fit in any cache.
<hr></blockquote>
BWAHAHAHA. that was funny
groverat: put the knife away, because its true. IBM isn't placing a Power4 or a Power5 into an Apple, but IBM is designing the successor to the G4. It will likely do away with the L3 cache, or cut it down to 2MB size, as the Power4 as it stands now is a rather large chip. It wont come out with P4-like clockspeeds, but will feature an honest-to-goodness modern memory subsystem.
Despite all the problems with the G4 over the years ( and yes I wish they had made more progress also ) my G4 450 seems to always keep up with my friend's AMD 700 Mgz in many tasks. That's something an Intel or AMD of the same clock speed wouldn't do. This unfortunately just changed with his purchase of a Nvidia Titanium. I guess I'll have to upgrade my Radeon to a ATI Radeon 9700 when I can.
The point being ( just to play devils advocate ) it's all relative. The G4 is a pretty good CPU it's just that Motorola doesn't give a rip about the Apple side of it's business. Pure and simple.
<strong>Though the point may be lost by now in this thread, I think that it is important to stress that the reason why the G4 is slow is not Altivec, but Motorola's complete inability and lack of desire to manufacture speedy chips. Don't blame the silicon design for management's incompetence.</strong><hr></blockquote>
The reason I took Altivec into this was Apple has a choice to make 'post G3'. They could go with 'plan a' (MOT+Altivec=G4) or 'plan b' (IBM+???) Apple chose MOT's ideas and along with it they got Altivec over whatever IBM was pushing. So it's not to say 'Altivec' is a problem... far from it... but when Apple chose Altivec they got STUCK with MOT.
i just hope that after the G4 successor comes out, apple will be able to implemet a on-chip DDR controller. without the pressure of needing duals, this may actually be feasible.
I was reading some posts at MacRumors and they were talking about the Quad rumor at mosr. Several people were implying that software already written to take advantage of dual processors would have to be re-written to take advantage of quad processors. I'm probably wrong, but I thought that once an app was written to have multithreading it would take advantage of as many processors as you have. Somebody lay some knowledge on me about this.
BTW, not that it matters, the likelyhood of quad processor macs that are remotely afordable is nil.
I was reading some posts at MacRumors and they were talking about the Quad rumor at mosr. Several people were implying that software already written to take advantage of dual processors would have to be re-written to take advantage of quad processors. I'm probably wrong, but I thought that once an app was written to have multithreading it would take advantage of as many processors as you have. Somebody lay some knowledge on me about this.</strong><hr></blockquote>
Well yes and no, you could write software that could take advantage of n processors automatically, but a lot of programmers take the shortcut of writing specifically for 2 processors since that is what the majority of multi-processor systems are.
For example when you have a task with a high degree of parallelism, if you take the shortcut of always spinning off 2 threads, rather than going to the extra trouble of figuring out how many processors are available and spinning off that many threads and adding the logic in each thread to properly divide the work no matter how many threads there are; then you end up with code that will benefit from 2 processors, but gain no additional benefit from more than that.
Comments
<strong>300% - Don't exaggurate! It's a measly 251% in 3DMark at 1600x1200, 4x FSAA. nVidia shouldn't worry (not).
Barto</strong><hr></blockquote>
<nitpicking>
300% is not an exaggeration, although Anandtech's published tests said 251%, HardOCP's said 10000+ (I don't recall the exact numbers) marks vs. 2500+ marks, which is, oh yes, 300% faster</nitpicking>
But I agree with Aris, if one (ATi) company can do it, another (Apple) can too.
Even though it almost would be too good to be true
also if u see the positiveof it u see how steadly the g4 is improving and with 1.4 ghz processors being rumored a 400mhz increase isn't bad at all compared to what we have seen before on previous updates
all powermacs need are a fast mobo
[ 07-31-2002: Message edited by: O and A ]</p>
<strong>talso if u see the positiveof it u see how steadly the g4 is improving and with 1.4 ghz processors being rumored a 400mhz increase isn't bad at all compared to what we have seen before on previous updates
all powermacs need are a fast mobo</strong><hr></blockquote>
...RUMORS...
As Cubba Gooding Jr. would say 'show me the money'
You want a faster MOBO well you can blame MOT for that one too.... Fact is all the PowerPC cpu REALLY needs is a parent who gives it some attention. Three years and MOT has done nothing. Rumors not withstanding.
D
<strong>One more thing about the new powerMacs I forgot to include in my original post. As far as I can tell from what I have heard they will ship with 10.2 Wether or not they come out before August 24. I am not sure if they will require 10.2 to run or not I would bet they don't need it but apple will not release new machines without 10.2.</strong><hr></blockquote>
10.2 will be declared GM long before the 24th. The 24th is the day all the boxes have to be in the stores or with the resellers. Apple has plenty of time beforehand to preload the finished OS onto new hardware.
Dual 1GHz PowerPC G4 processors
Velocity Engine vector processing unit
Full 128-bit internal memory data paths
Powerful floating-point units supporting single-cycle, double-precision calculations
Data stream prefetching operations supporting four simultaneous 32-bit data streams
512K on-chip L2 cache running at processor speed
8MB DDR SRAM L3 shared cache, with up to 16GB/s throughput
Dual 1GHz processor busses providing each CPU with up to 4GB/s throughput
Memory
1GB or 2GB of 4-channel 266MHz PC2100 DDR SDRAM with up to 8.5GB/s throughput
Four DIMM slots supporting up to 4GB of DDR SDRAM using the following:
256MB DIMMs (64-bit-wide, 128Mb technology)
512MB DIMMs or 1GB DIMMs (64-bit-wide, 256Mb technology)
This equals:
15 gigaflops of computational power
16GB/s L3 cache throughput (250MHz 256-bit DDR SRAM)
4GB/s of throughput available to each CPU
8.5GB/s of memory throughput (enough to provide each CPU with 4.3GB/s)
And of course, "Pentium-crushing performance, but now in every benchmark."
------------------------------------------
If apple made the system described above I would be delighted. Of course I would not be able to afford it. Maybe they will do something like this on the so called Xstation (if they do make one). The new PowerMacs will include a lot of this though and faster CPUs. We will see some nice performance gains come august. In time.
<strong>Don't lie.</strong><hr></blockquote>
Can I troll at least?
B. Motorola's cost-savings and improved management don't seem to be paying off much, and one of their best senior executives just left the company for a job at Tyco. I think they're going down the toilet for real.
I predict within a year, parts of Motorola will be sold off to other companies and the rest will go down the tubes / be liquidated. Unless they make a dramatic financial turnout over the next couple quarters, they're finished.
C. If the next batch of machines are far superior to what we have, it won't have anything to do with Motorola's CPU inside. More like Apple will have teamed up with NVidia and others to redesign the UMA, to increase the bandwidth between critical system components. That way processors in the 1GHz - 1.3 GHz range can still at least make a showing against much faster PCs (when running design software anyway). Thus Apple will have something to sell until they can come up with a real pro hardware solution.
[ 07-31-2002: Message edited by: groverat ]</p>
OUCH!!!!!!!! PAIN!!!!!!!!!!!!! HE REALLY DID IT!!!!!!!! MAN THATS A LONG KNIFE!!!!!!!!!!!!!!!
In time.
One note: for the guy above who can't remember an update with more than 100mhz, look no further back than the dual gig model. Back in January, we got a single 933mhz, granted, but there was also the dual 800 that was bumped to a dual 1000mhz - 200mhz jump there....
<hr></blockquote>
BWAHAHAHA. that was funny
groverat: put the knife away, because its true. IBM isn't placing a Power4 or a Power5 into an Apple, but IBM is designing the successor to the G4. It will likely do away with the L3 cache, or cut it down to 2MB size, as the Power4 as it stands now is a rather large chip. It wont come out with P4-like clockspeeds, but will feature an honest-to-goodness modern memory subsystem.
algo: there is no 7500. stop lying.
The point being ( just to play devils advocate ) it's all relative. The G4 is a pretty good CPU it's just that Motorola doesn't give a rip about the Apple side of it's business. Pure and simple.
<strong>Though the point may be lost by now in this thread, I think that it is important to stress that the reason why the G4 is slow is not Altivec, but Motorola's complete inability and lack of desire to manufacture speedy chips. Don't blame the silicon design for management's incompetence.</strong><hr></blockquote>
The reason I took Altivec into this was Apple has a choice to make 'post G3'. They could go with 'plan a' (MOT+Altivec=G4) or 'plan b' (IBM+???) Apple chose MOT's ideas and along with it they got Altivec over whatever IBM was pushing. So it's not to say 'Altivec' is a problem... far from it... but when Apple chose Altivec they got STUCK with MOT.
Dave
[ 07-31-2002: Message edited by: DaveGee ]</p>
I was reading some posts at MacRumors and they were talking about the Quad rumor at mosr. Several people were implying that software already written to take advantage of dual processors would have to be re-written to take advantage of quad processors. I'm probably wrong, but I thought that once an app was written to have multithreading it would take advantage of as many processors as you have. Somebody lay some knowledge on me about this.
BTW, not that it matters, the likelyhood of quad processor macs that are remotely afordable is nil.
<strong>Question.
I was reading some posts at MacRumors and they were talking about the Quad rumor at mosr. Several people were implying that software already written to take advantage of dual processors would have to be re-written to take advantage of quad processors. I'm probably wrong, but I thought that once an app was written to have multithreading it would take advantage of as many processors as you have. Somebody lay some knowledge on me about this.</strong><hr></blockquote>
Well yes and no, you could write software that could take advantage of n processors automatically, but a lot of programmers take the shortcut of writing specifically for 2 processors since that is what the majority of multi-processor systems are.
For example when you have a task with a high degree of parallelism, if you take the shortcut of always spinning off 2 threads, rather than going to the extra trouble of figuring out how many processors are available and spinning off that many threads and adding the logic in each thread to properly divide the work no matter how many threads there are; then you end up with code that will benefit from 2 processors, but gain no additional benefit from more than that.
Jul 2k - Apple pulls out the DP boxes at 2x450 2x500
Jan 01 - 466/533/2x533/667/733
<hr></blockquote>
This speed bump took us from 500 MHz to 733 MHz, a 47% speed increase.
Another 47% increaes in MHz would put the G4 at 1470 MHz, roughly 1.5 GHz.
So yes, it is entirely in line with the history of the G4 that such a large speed bump may occur.