Originally posted by arn
I'm not sure where you got that the G4s cost more... it's probably not true.
The .09 micron PPC 970fx chips are around 65 mm^2. That's quite a bit smaller than a .13 micron PPC 7447. The 7447 is slightly more than 98 mm^2. The smaller die size means that IBM can produce more chips per wafer. IBM also tends to have much better yeilds than Motorola so there is less junk per wafer.
The 970fx _should_ cost quite a bit less than the PPC 7447.
Originally posted by a17InchFuture Okay, I'll tell you why everyone here says they are overclocked at 1.5: it was said because ITS TRUE! its just a fact, jack. Anyone who has been paying attention to Apple knows that
Overclocking is, by definition, pushing the clock speed of a processor past the vendor rating. Apple would be using overclocked G4 if, AND ONLY IF, Apple clocked the processors faster than Motorola's speed rating. This ISN'T the case. If you check the label on the processor of a dual 1.42 G4, you'll find cpus that are labeled 1.42 GHz.
The whole overclocking thing came from the fact that Apple was shipping the 1.42 when Motorola was only advertising the part as a 1.33 GHz. This doesn't mean that Motorola didn't have a 1.42 sort.. it just means that they were reserving them for Apple.
Originally posted by Ja Di ksw
crud, I didn't know they had that. How close is it now to the P4? What I mean is, before, you would take a mac's processor's speed and times it by 1.5-2 to get a window' prcessor's speed (at least, had always heard it was that, b/c of the pipelines). What is it now?
woa.. it's WAY more complicated than just that. You've got pipeline length, the number if instructions that can be executed per clock, the clock speed, the system bandwidth and efficiency, the quality of the branch prediction (Intel's is about the best there is)...
If I remember correctly, the G5 pipeline is generally listed as around 20, though in truth the pipe line is a different depth depending on what processor units are involved. The P4 is 23 stages.. Prescott P4 is 30 deep.
The G5 is more powerful than the P4 at the same clockspeed because (among other things) it can do more per clock. It's a very 'wide' processor that can do a lot of work at the same time.
Originally posted by johnnyjibbs
And thus the G4 is more powerful, clock for clock, than the G5
No, the G4 isn't more powerful than a G5 at the same clock speed. The G5 has a higher IPC.. It can do more work in a given # of clocks than a G4. Some of the early benchmarks between the G4s and the G5s made the G4s look pretty good.. but that was running un-optimised G4 code. Pretty much every app thats spent a modicum of time optimising for the G5 has seen significant performance gains (except Photoshop apparently). In fact, we still haven't truely seen what the G5 can do. IBM has released the first truely optimised compiler and it's producing 30%+ faster code right out of the box. Just wait till the OS gets tuned for the G5 too.
Originally posted by dongminI think the 970fx is a big piece, but certainly not the only piece of the puzzle. Other components of the G5 motherboard (controller, memory, etc.) have to be heat-controlled as well
Apple stated at some point, wish I remembered when, that the system controllers would continue to be produced on the same process as the PPC 970. Apple should have a die shrunk chipset ready with the PPC 970fx.
Originally posted by Snow_RiverUh... shouldn't that be 170MHz? This would be a step from 1.33GHz to 1.5GHz. The 1.42GHz (which, by the way, is 80MHz less than 1.5GHz, not 90MHz) has never been in a PowerBook.
The motorola press release doesn't actually mention 1.5GHz.. only the rumor stories. The press release only mentions 1.42GHz
http://www.motorola.com/mediacenter/news/detail/0,,3861_3238_23,00.html
Originally posted by WizardThe Powetune 970's still have the issue of using excessive power. Its not just the processor but the whole package that can kill a notebook design.
The other issue is that at 1.5 GHz the 970's won't perform better either, so performance wise it is a wash. The article also points out htat this processor does have new power management capabilities. So it appears that there is more to the processor than fabrication imporvements.[sic]
The 2GHz 970fx consumes power in the 20's. A 1.5GHz 970fx should be a very efficient processor. Also remember that the PPC 970 supports other bus mulipliers. If Apple decided that they couldn't get a .09 micron system controller to work in a notebook, they could clock it down to a third or a quarter of the CPU speed. This would be perfectly acceptable option for a first Gen G5 notebook.
Also, don't think for a second that a 1.5GHz G5 notebook wouldn't be faster than a 1.5 (or 1.42GHz) G4. Don't rely on old benchmarks of G4 tuned code on a G5 to determine the relative power. Cinebench performance on a dual 2GHz is 110% faster than a dual 1.42 after Cinebench was optimized for the G5. That's over twice the performance even though the G5 is only 40% faster by clock speed.
http://www.barefeats.com/g5sum.html
Originally posted by Spinko
"The 970FX, meanwhile, consumes a mere 12.3W at 1.4GHz, paving the way for PowerBook G5s. That figure is comparable to the 7.5W a 1GHz consumption of the G4-class Motorola MPC7447 that drives the current PowerBook G4s. The 970FX's SpeedStep-style PowerTune technology will help too."
quote from "the register" http://www.theregister.co.uk/content/39/35057.html
The really interesting point here is that the 1.4 GHz 970fx is quite a bit cooler than the 1.4GHz 7447A. Check the Moto press release again..
http://www.motorola.com/mediacenter/news/detail/0,,3861_3238_23,00.html
They are only comitting to under 20watts at 1.42 GHz.
Originally posted by SiliconAddict
Sorry but I claim BS on that statement. You obviously have never use a Pentium M device. I can rip a DVD and recompress it into a DIVX movie and the bottom of an IBM T40 laptop that is 1" thick is only slightly warm to the touch. For standard use I've been getting 4 hours off of a 1.5Ghz system. God people! Get real facts before you spout this BS
hear hear. This is a shock to everyone, I know.. but Apple doesn't automatically get to make the best *everything* They aren't the spoiled kid who owns the game. P-M is a powerful processor with great thermal characteristics. There are plenty of P-M notebooks that get over 5 hours of battery life. IBM sells one (with an expanded battery pack) that gets up to 9 hours. I've used a P-M machine and it was quite fast when it was clocked up. My only complaint was that it was slow to scale from it's low power speed (1/2 the full clock speed) to the full speed under load. It made the machine stutter unless you forced it to run full speed.
Originally posted by Westside guy
Admittedly those apps weren't optimized for the G5; the question is how much improvement can be reasonably expected once they are so optimized
Well, we are seeing a 30% increase in speed between gcc 3.3 and xlc... and that's just between two G5 optimized compilers. Like I mentioned before... Cinebench was getting 110% performance increase with a 40% clock speed increase... and this is running on an OS that isn't even really optimised yet. This was before xlc was finalized and released for OSX.