PowerPC 975 or Not?

[NNO-Stephen]

ok, now, since I was saying replace the mid-range powerbook with the 1.6, and replace the 1.5Ghz with the 1.8Ghz.

redo your math and you see this would increase the performance of the PowerBooks, plus, move it to an architecture that is not almost entirely out of steam.


5. Rendering with Final Cut Pro is a CPU intensive operation. Adding "gobs" of memory or super fast disk arrays will not speed up rendering. If you want to render faster, get a faster CPU -- preferably two of them -- with a faster front side bus.


also, it's not the HD or tons of memory that make the PowerMacs faster, it's the CPU, and the non- "circa 1998" bus speed.

 

[JFreak]

actually, there's nothing wrong with G4 cpu:s itself, but the bus is really a limiting factor.

put in current 1.5GHz G4 and 1.6GHz G5 chips to a benchmark and you'll see that the G5 wins purely beca

use of the bus speed (throughput!) of the G5 that is about 5X of the G4 bus speed. apple once had the L3 cache in G4 chips and that helped a lot; in my personal experience i can say that a 667MHz 15" dvi powerbook is really equal in power to a 12" 867MHz powerbook - and the only major difference is that the 15" dvi pb has a 1MB L3 cache and the 12" pb doesn't have it. that L3 cache is a clear indicator about the benefit of the L3 cache, and i personally cannot see why on earth apple chose to drop the cache away in designing the al-books.

if apple chooses to continue using the G4 chip, they better re-introduce the L3 cache. it really helped a lot to hide the bottleneck of the crippled front-side-bus. naturally, it would be far better if the G4 could have a competitive fsb, but as that is not likely to happen, the G4 will need the L3 cache.

does anybody remember the "supercomputer" ads about the G4 chip? that hasn't gone anywhere, th G4 chip design is still more than ok. it is only the FSB that isn't feeding the cpu with enough data. so APPLE IF YOU'RE LISTENING, please, re-introduce the L3 cache or give a faster fsb if you're going to stick with the G4 with the powerbook line... give us protools users someting usable for a change. the current powerbooks are not cutting it.no mattr if the G4 could be run at 10GHz, the throughput isn't enough. fix it, period.

 

[NNO-Stephen]

It wasn't Apple that decided to drop the L3 cache from the 7447A CPU, it simply doesn't support it, that's why it's not there. Apple claims it doesn't need it because it now has 512k of L2 Cache.

the bus speed, also not going to change. again, not due to Apple, but the design of the G4 itself.

 

[Henriok]

if apple chooses to continue using the G4 chip, they better re-introduce the L3 cache.

Apple does not manufacture nor design the G4 processors, Motorola/Freescale do. They have for some years done some serious refocus on the embedded market and large and fast L3 caches are no priority here.

Freescale have released some compellig roadmaps though with two new cores that will replace the G4 sometime in the future. e600 and e700 targets the high end embedded and computing market and will probably reintroduce L3 caches, as well as on die RapidIO based memory controllers.

The question is when.

 

[thatwendigo]

There's actually a pretty fundamental flaw in how you're looking at those results and that is there is pretty much no meaning to the argument you are putting forward based on the numbers. Having numbers in an argument doesn't benefit it unless there is some meaning behind using them.

The meaning is thus, since you missed it: Changing the processor to a hotter one doesn't really benefit (7447A to 970), whereas the redesign[/o] of an existing line became more efficient (7455 to 7447A).

First things first, the 7455 to 7447 sees an increase in the level 2 cache along with a better graphics cards and other motherboard improvements. You'd find the level 2 cache is almost entirely responsible for the improvement hence the reason the jump is more substantial.

Motorola product data sheet

MPC7455 - 1.0ghz max frequency; 32.5w typical, 50w maxl; 1.3, 1.6, 1.8v core; 32k L1 Data, 32k L1 Int; 256k L2 Internal; 2MB max L3
MPC7447 (not A revision) - 1.267ghz max frequency; 18.3w typical, 25.6 max; 1.1, 1.3v core; 32k L1 Data; 32k L2 Int; 256k L2 Internal; no L3

And, pulled from MPC7447AEC Rev. 0, 2/2004 - MPC7447A RISC Microprocessor Hardware Specifications (a Motorola PDF):

MPC7447A - uncertain max frequency; 12.5w typical. 25w max (DFS enabled); v; 32k L1 Data; 32k Int; 512k L2 Internal

There is a discrepancy between the PDF (which is pre-release) and the later Motorola parts catalog that is linked above. The catalog claims 256k on the 7447, but the PDF says 512k on both 7447 and 7447A. That makes it a little hard to say for certain. The 7447A does have a thermal diode and a new feature called Dynamic Frequency Switching that keeps the processor from running anything close to maximum heat when its enabled, unless you're pushing it with full CPU usage all the time.

The G4 was a potentially good chip but it's currently crippled and scaling it further won't gain anything. They either need to improve the G4, and there are plenty of areas to improve, or swap it out for a G5 if they realistically want to continue the line otherwise performance will just go nowhere.

Which is one reason for the upcoming FreeScale e600/e700 lines of processors. The e600 is supposed to come in around 2.0ghz and climb from there, while supporting CMP (chip multiprocessing), which is a way to say that it's going to have more than one core on the die. It's also supposed to come in with a 400mhz FSB instead of the older 166mhz bus, on-die DDR control, and other niceties, with a target heat of 25 watts for the dual-core 2.0ghz part.

Oh, and it's instruction set compatible with the G4, so if your code runs on the G4, it runs just fine on the e600. For those who doubt that this can be done, FreeScale has already brought two new processors to market. The e300 and e500 are rolling as we speak.

A 1.6 GHz G5 will have about the same power consumption as a 1.5 GHz G4 and Apple is using a higher bus speed and higher frequency, for reference the 7447A has a max power consumption of 30W at 1.42 GHz or 5W higher than 1.33 GHz and I suspect 1.5 GHz would be very close to 40. Where the differences lie would be the supporting architecture and cost.

Actually, if you read the specifications in the PDF I mentioned, the DFS feature, when enabled, allows the MPC7447A to run at around 12.5 watts typical. This is because of the dynamic power management features in the chip, which allow it to actually beat the Centrino on low-powered state consumption at a mere 4 watts in sleep mode.

 

[Dont Hurt Me]

All i can say is thatwendigo has a lot of faith in Moto/freescale roadmaps not based on actual shipping cpu's such as current shipping G5. I have much more faith in IBM and what they can do rather then Moto's years of screwing Apple and Apple's customers with last place G4(unless the test is on powerused rather then performance. The current G5 was a rush job to get something to market to make up for lackluster G4. I think the new G5(WWDC) will show a more polished G5 (970 970fx 975 )or whatever but when i see 1 g5 1.8 matching and exceeding what 2 G4s 1.42 can do then you should know the gig is up with G4. way past time.

 

[thatwendigo]

All i can say is thatwendigo has a lot of faith in Moto/freescale roadmaps not based on actual shipping cpu's such as current shipping G5.

Not shipping? Crolles2 is on target and moving chips in volume, and FreeScale has been consistently providing what they said they would, incvlcuding bringing new and competitive processors out in the e300 and e500 lines. They never said that the e600 would be available right now, and neither have I, aside from a few wishful comments about wanting one in the PowerBookj.

I have much more faith in IBM and what they can do rather then Moto's years of screwing Apple and Apple's customers with last place G4

Ah, right... IBM's vastly superior because they've put out a lot of 970 consumer revision... Oh, no, it seems that we're still waiting on the chips to be delivered. Really, DHM, it doesn't make any sense to be cheerleading IBM when they've yet to truly prove that they're going to do things differently than the old Motorola. If they show an update at WWDC that's significant, then that goes a long way towards making a name for them with the mac world, but it still leaves a yearlong gap.

At the end of the day, we're all operating on a thin layer of fact and a lot of speculation. It's senseless to be in either the IBM or FreeScale camp until we see just what Big Blue can do.

The current G5 was a rush job to get something to market to make up for lackluster G4. I think the new G5(WWDC) will show a more polished G5 (970 970fx 975 )

This is one of the rare things that I've seen you say that I might agree with, though it's still not exactly how I would put things. We might or might not see a "polished" G5 at the conference, and if it's nothing more than slightly bumped 970s or 970FXs, then I'm going to be a little more than lightly pissed at IBM.

That's putting us back in that "slow" boat that you keep calling the G4 and blaming on Motorola.

or whatever but when i see 1 g5 1.8 matching and exceeding what 2 G4s 1.42 can do then you should know the gig is up with G4. way past time.

Unlikely, but possible. We'll see what the conference brings.

 

[Dont Hurt Me]

Question is will Apple wait for WWDC to update the powermac? WWDC will tell the story of what Apple is doing and where they are going. Apple has known a lot longer then you or I about Freescales/moto plans and wouldnt have signed on to IBM unless they were going this route. as of today there is no G4 that can take on a G5 and you know it. To sell computers you have to move forward and roadmaps and promises dont cut it. G5 is real,e600 e700 are not but i will concede that why bother with a e600 at all unless it was going to be used by Apple? we can speculate forever no one knows what Apple does until they do it. Fact is G5 kills G4 and Apple says so themself. Selling G4 for years hasnt helped Apple marketshare and i would say that G4 is one of the main reasons so many left Apple.

If Apple wants to act like top dog and price for it then the Cpu they are using better kick the crap out of the other guys and G4 hasnt done that in years. Solution? G5s my next machine will have one.