Are you trying to start a credentials measuring contest with me? Okay.. I'll bite.
-21 years as a hardware tech (17 as Apple Certified) with a specialty in Mac based super computer cluster setup and maintaining the health of the hardware and software on multi-year contracts since 1998. I have setup everything from Beige G3 tower clusters to modern day Mac Pro clusters.
-24 years as a BSD user.
-OpenBSD developer for both PowerPC and x86.
Your credentials don't change the fact that you don't seem to understand real world performance. All you have really done is list your personal specs and name a technology or two. How about some real world examples and definitions rather than tech babble?
I prefer to remain anonymous as sometimes I deal with insider information on here and I'd like to stay in Apple's good graces and not have things traced back to me.
But if it makes you feel better, I'm also Apple certified, and I'm a developer dealing with multi core, GPU programming, and SSE (would have been Altivec on PowerPC.)
Using the word "huge" to explain the gains does not reflect the clock speeds the busses run at. I'm not saying the bus doesn't outperform the G4 because it does but in the real world it is no where near the increase the clocks speeds are. The 1 GHz G5 bus is over 5.9x the speed of the fastest G4 bus (167MHz) but in the real world it only works about 2.5 - 3x faster and not everything in real world computing is directly dependent on bus speed.
A head researcher at one of the labs I do work in and I even did a very thorough real world number crunching test. The results showed that in the very bus dependent calculations we were running that the 1 GHz G5 bus only equated to about a 300-400MHz G4 bus if one existed. This test was done with a 34x dual 2 GHz G5 (dual 1 GHz bus) cluster vs a cluster of 40x dual 1 GHz G4 Quicksilver's (133MHz bus) and 5x dual 500MHz GE G4's (100MHz bus). The G5 cluster was capable of about 603 gigaflops vs. the G4 cluster that came close at about 532 gigaflops.
I'm going to call bunk on this. The G5 is slightly faster clock for clock than the G4, and your G5s have 4x the clock rate, yet they're only able to eak out barely a tie?
That's not a system bus problem. The G5s in your numbers are underperforming despite their great system bus.
I will admit that bus speed makes a much bigger difference in x86 hardware generally.
The system bus makes the exact same difference. x86 and PowerPC made different design decisions, but PowerPC is not made of pixie dust that allows it to violate the laws of physics and move data to the processor using magic.
My direct experience tells me the G4 could have taken away the need for the G5 if it was able to get the bus up to just the 300MHz range.
This is pretty much what I've been saying. So we're in agreement?
The entire reason the G5 exists is pretty much because the G4 had a slow system bus. Either the G4's system bus was a problem, or it wasn't and Apple released the G5 for no reason. You can't have it both ways.
My single 7448 on a 100MHz bus will smoke any single G5 and is about on par with a dual 1.5GHz 7455. Please use your number philosophy to explain to me how a single G4 on a 100MHz bus can noticeably surpass a single 1.8G5 with a 900MHz bus? Please
Dual processor machines had a system bus for each CPU. Dual
core chips share the same system bus for both cores.
You're comparing a dual processor design to a dual core design when they're not the same thing. A dual core G4 would have not been functionally equivalent to a dual processor G4.
Thats a dual 7447 (note the 512K L2) in that test which is a gutless chip vs the 7455 or 7448 yet it still scores only about 200 less in the overall score (which actually reflects the real world a bit) than a dual 1.8GHz G5 with a dual bus thats 5.39x faster and a next gen chip.
Any dual 1.25GHz+ MDD owner was silly to buy those Sonnet MDD upgrades as a 7455 tends to be the same speed as a 7447 thats 400MHz+ faster.
Again, that's great and all, but the mistake you're making is comparing dual processors to dual cores.
The fact is: For awhile, PowerPC processors were superior to Intel and AMD. Up until the G5. The G3 and G4s matched or at times outperformed the PIIs and PIIIs they were up against. The G5, while certainly powerful, wasn't anywhere close to being ready to be put into a notebook, and Apple, along with the rest of the industry, knew that notebooks were a greater source of revenue than desktops.
It was a good race, but it honestly went downhill when Motorola had trouble breaking 500 mhz. That's when the clock rate troubles started. The G4's efficiency let it stay relevant for a while, but it really needed to stay at least somewhat competitive with Intel's clock rate to trounce Intel.
I kind of miss the PPC days; they were unique and the hardware wasn't just a fancy PC like it is nowadays. But at the same time, Intel has some great processors and it is understandable why Apple moved on.
The one thing I don't miss is Apple's PowerPC chipsets. Like the PowerMac 6500 which had tons of compatibility issues. Or my Power Mac G3 which would randomly corrupt drives over 10 gigs (but advertised support for drives up to 100 gigs.)
Sorry to say Apple's strength has never really been custom designed hardware internals. iPod went pretty well for the most part, and the iPhone has been strong because so much of the design has been outsourced.
Intel Macs have bee totally solid in the meantime. My Mac Pro is the best machine I've had since my old Power Mac 7300. Not a single issue aside from a dead 8800 (produced by NVidia).
Don't be surprised if Apple moves to ARM next, but at the same time, don't be surprised if Apple stays with Intel. Intel has really picked up it's game against ARM, especially in the energy saving department.
The Mac will probably die before there is any chance of it moving to ARM. ARM doesn't provide any advantages at this time over Intel for desktops.
x86 is actually moving into ARM's space, which will be interesting to watch. Traditionally, x86 has not worked well for cell phone size devices.
Actually the dual core version of the G4 had a system bus that would run at 600MHz & use DDR2 RAM.
http://www.freescale.com/webapp/sps/site/overview.jsp?code=DRPPCDUALCORE
And I totally agree that the system bus is HUGE reason the G4 really fell out of ranks later on in life. When you consider how the CPU ranged from 350MHz to 1.67GHz but system bus speeds only increased from 100MHz to 167MHz you really begin to see where the G4 began to struggle for bandwidth. Had the G4 made it up to 400MHz or even 333MHz in bus speed, later PowerBooks and iBooks may have not felt so long in tooth by the end.
This is interesting. If I'm reading it right, they moved the memory controller onto the processor eliminating the need for a system bus (much like the G5.)
It's still clock for clock and clock per watt way way behind what Intel had, but it would have solved the G4s biggest problem at the time. Trying to find a release date on it.
Edit: The earliest release references I can find for the dual core G4 with the integrated system bus was in 2006, after the first Intel Macs started to ship with Core processors at dual 2.1 ghz and 4 core desktops at 3 ghz. Note that the MPC7448 did not have the fixed system bus, only the MPC8641.
.
