How much faster than a PB G4 would a PB G5 really be?

[Naimfan]

I know the question is somewhat unanswerable until we actually see G5 Powerbooks. But I've just invested some time looking at Barefeats, and from the data there I suspect the difference won't be quite as night and day as some might suspect.

For example, the Photoshop 7 non-MP aware filters test shows the G5 1.6 PM to be roughly 20% faster (score of 74 for the G5 and 92 for the 1.25 PB). The difference is smaller on MP aware filters--62 for the G5 to 71 for the 1.25 PB. The differences are more dramatic on the Cinebench CPU render, with the G5 1.6 scoring 158 to 227 for the 1.25 PB. (Lower is better.) I left out the gaming scores because they're not relevant to me.

Two thoughts occur to me. First, I realize it's problematic to compare scores across a platform. That said, the differences can be partially explained by the different clock speeds--increasing the G4 to 1.6 GHz (yes, I know we can't--just offering a hypothetical) appears to close most of the gap. Putting a faster HD in the PB would also help close the gap--I'm assuming the PB and PM were both stock--if Barefeats explains otherwise I'm not seeing it.

Second, the above comparison assumes Apple can get the same performance out of a G5 1.6 chip in a laptop that they do in a desktop. From everything I've read, I'm not at all sure that's a given--many people have pointed out how much more difficult it is got get great performance out of a laptop.

So I conclude that a G5 PB would be faster than the existing G4s, but not by the margin many appear to expect. From the Barefeats data, I'd expect a G5 PB to offer less of a performance advantage than the G5 PM (assuming a 1.6 GHz G5 in the PB). So as an estimate, a G5 PB might offer a 10-20% improvement in performance over the existing PBs. This is frankly much lower than I would have thought prior to looking at Barefeats.

Other thoughts/comments?

Best,

Bob

 

[thehuncamunca]

the much faster bus, faster L2 cache will really help to speed things up
once programs are more optimized for the G5 you'll also see even more performance gains

I know the question is somewhat unanswerable until we actually see G5 Powerbooks. But I've just invested some time looking at Barefeats, and from the data there I suspect the difference won't be quite as night and day as some might suspect.

For example, the Photoshop 7 non-MP aware filters test shows the G5 1.6 PM to be roughly 20% faster (score of 74 for the G5 and 92 for the 1.25 PB). The difference is smaller on MP aware filters--62 for the G5 to 71 for the 1.25 PB. The differences are more dramatic on the Cinebench CPU render, with the G5 1.6 scoring 158 to 227 for the 1.25 PB. (Lower is better.) I left out the gaming scores because they're not relevant to me.

Two thoughts occur to me. First, I realize it's problematic to compare scores across a platform. That said, the differences can be partially explained by the different clock speeds--increasing the G4 to 1.6 GHz (yes, I know we can't--just offering a hypothetical) appears to close most of the gap. Putting a faster HD in the PB would also help close the gap--I'm assuming the PB and PM were both stock--if Barefeats explains otherwise I'm not seeing it.

Second, the above comparison assumes Apple can get the same performance out of a G5 1.6 chip in a laptop that they do in a desktop. From everything I've read, I'm not at all sure that's a given--many people have pointed out how much more difficult it is got get great performance out of a laptop.

So I conclude that a G5 PB would be faster than the existing G4s, but not by the margin many appear to expect. From the Barefeats data, I'd expect a G5 PB to offer less of a performance advantage than the G5 PM (assuming a 1.6 GHz G5 in the PB). So as an estimate, a G5 PB might offer a 10-20% improvement in performance over the existing PBs. This is frankly much lower than I would have thought prior to looking at Barefeats.

Other thoughts/comments?

Best,

Bob

 

[Naimfan]

I agree with you on the optimized applications, but as far as the faster bus etc. go, aren't they already optimized on the G5 1.6? And thus the G5 1.6 represents the objective for PB performance (assuming a 1.6 chip).

Best,

Bob

 

[thehuncamunca]

i just meant that compared to the G4, the G5 has a much faster bus

I agree with you on the optimized applications, but as far as the faster bus etc. go, aren't they already optimized on the G5 1.6? And thus the G5 1.6 represents the objective for PB performance (assuming a 1.6 chip).

Best,

Bob

 

[dbauer]

Good question

You bring up some very good points and the question is excellent. I agree with you that a 1.6 G5 may not be a night and day speed increase but I think Apple will put the 1.8 G5 970FX processor in the PM's because they are not as hot. I remember reading someplace that IBM has a processor that can adjust speed according to the task being performed (and it may be the 970FX but I can't recall) so this gap may be even more difficult to see in the real world. I don't put much stock in benchmarks anyway because they often test against software I don't have and would never use. I would like to see a benchmark taking 50 photos from iPhoto and exporting them to a Quicktime movie. That is something I do all the time and it takes a long time on my G4/433. It seems that most site only benchmark the latest machines but I think that most people that would buy the latest machines are using an old beige G3 or even older than that.
Anyway, its difficult to measure a PB against a PM because the PM has a 7200 RPM 8MB cache hard drive the the PM has a 4200 rpm one (i think). That alone is 80% faster. If the PB had the faster hard drive there may not be any difference between them. I know the G5 has more bandwith and all but that doesn't mean you can see the speed difference in everyday apps.

 

[Opteron]

The G5 chip has a bus half the speed of the actual CPU, hance 2GHz CPU has a 1GHz FSB. The G4 on the other hand hit a brick wall at 167, This alone it the crippling factor of the G4.

Hence the G% bassed powerbooks when they arrive will be substancially faster.

 

[Naimfan]

The G5 chip has a bus half the speed of the actual CPU, hance 2GHz CPU has a 1GHz FSB. The G4 on the other hand hit a brick wall at 167, This alone it the crippling factor of the G4.

Hence the G% bassed powerbooks when they arrive will be substancially faster.

Opteron--

I understand the bus difference, but that's already taken into account--the benchmarks test the performance of the whole system, do they not? And based on that, I don't think the G5 PBs when they come out (whenever that might be) will offer as big a performance jump as many here seem to expect (barring, of course, apps optimized for the G5).

It's instructive to note that the performance advantage of the dual G5 2.0s over dual G4 1.42s are very close to the clock advantage...


Best,

Bob

 

[wordmunger]

I think most of the *massive* performance increases we saw with the G5 had to do with the 8GB RAM support. In memory-hogging apps, this made a huge difference. Obviously, this wouldn't be a factor on laptops.

The main reason I'm waiting for a G5 is not some quantum leap in speed, but support for future apps. My TiBook is over 3 years old now, and I want my next laptop to last at least as long.

 

[QCassidy352]

bob,
you bring up an excellent point. IMHO, people are way too critical of the G4. At this point, people won't buy another PB G4 just on principle - it doesn't even matter to them if they need a G5, or what a G5 would do. They just think G5 = new, G4 = old --> I need a G5.

The G4 is not at all a bad chip. The main problem with it is that Motorola stopped development on it, and it never hit high enough clockspeeds. But if the G4 could hit 2 Ghz or so, it would still be a very viable chip.

And very interesting point about the 2.0 G5 compared to the 1.42 G4.

 

[lind0834]

Anybody still stuck using a G3 (like me) will tell you that the performance benchmarks aren't everything. I know that my iBook is probably as fast at the first gen G4 PM, but that doesn't mean that DVD Studio Pro, or Garage Band would work as well on my machine as anything with a G4.

Right now, Apple considers the G4 standard for all thier pro apps (as they should). In a few years.. the requirement for FCP6 will be G5. Then it will make all the difference in the world.

 

[Stolid]

There will, of course, be gains as compilers and assembly authors learn how to best schedule for the G5, but more importantly (to me at least), is that I know I'll have 64-bit insides if I get a PBG5.
Last thing I want is to have some piece of software that requires a G5 because they compiled it on -march G5 -O3 (to pseudo-GCC it) which means it won't run on a G4.
I remember back in the early 90s my 386 could run software a 286 outright COULDN'T. Now in the PC world almost every piece of software from 386-today *can* run on a 386 (albeit not well), but I foresee that ending with the jump to 64 bit. So for me its a future usage issue, not a speed issue (though I'd love to see some very high speed G5s in a Book)

 

[aswitcher]

There will, of course, be gains as compilers and assembly authors learn how to best schedule for the G5, but more importantly (to me at least), is that I know I'll have 64-bit insides if I get a PBG5.
Last thing I want is to have some piece of software that requires a G5 because they compiled it on -march G5 -O3 (to pseudo-GCC it) which means it won't run on a G4.
I remember back in the early 90s my 386 could run software a 286 outright COULDN'T. Now in the PC world almost every piece of software from 386-today *can* run on a 386 (albeit not well), but I foresee that ending with the jump to 64 bit. So for me its a future usage issue, not a speed issue (though I'd love to see some very high speed G5s in a Book)

Yep, this factor is a big thing for my reason to wait for the G5. That and the speed bonus for optomised software, giving the processor a better chance of keeping up for the next 3-4 years.

 

[Naimfan]

I'm really surprised to read that people are waiting for G5 PBs not because of speed but for futureproofing. That seems to be shifting the debate over G5 PBs some.... And that is one of my concerns in buying a new 1.25 G4--even though it would be fast enough for anything I'm likely to do, my concern is that down the road Apple won't be supporting the G4 (for understandable business reasons). I'm using a G3/400 now, and although it seems to do well with Panther, my understanding is that many new apps won't run on a G3 at all.

In speed terms, I've now reviewed benchmarks from a number of other sources, and experimented with some different laptops, both Mac and PC. And I think that when like is compared with like, available performance v. price is broadly comparable (save for specialized/technical "power" apps, which I've not experimented with).

I'm hankering after a new PB and am sorely tempted to take advantage of an educational offering that gives me $150 off any iPod (not mini) when I buy a PB with Applecare. But I also don't want to receive it and then have Apple release new PBs the week after... Given that the speed update, whether from a G5 or new G4, is likely to be nominal, I'm not very concerned that a new PB would be that much faster--certainly not for what I'll be using it for. So perhaps what I'll do is order one the very last day of the promotion (3-27), and if new PBs are released, I'll just cancel the order and order a new one... We shall see...

Best,

Bob

 

[hjhhjh]

the fsb is faster, and theres larger cahces, but the fact that the g4 has almost 3 times less the pipleline stages than the g4, meaning a 1.33 ghz g5 is roughly equal to that of a 1.33 ghz g5

 

[earthtoandy]

certainly wont be worth the extra power consumption and heat!

 

[cube]

bob,
The G4 is not at all a bad chip. The main problem with it is that Motorola stopped development on it, and it never hit high enough clockspeeds. But if the G4 could hit 2 Ghz or so, it would still be a very viable chip.

G4 will hit over 2 GHz and dual core. Freescale is also working on a 64-bit core scaling over 3 GHz.

But the single core 7448 (at least 1.5 GHz, 1MB L2) won't sample until H1 2005.
Dual core won't sample until H2 2005.

Let's see if now that they have a new name, they're able to match their schedules.

 

[Blue Velvet]

...my concern is that down the road Apple won't be supporting the G4 (for understandable business reasons).

That day will surely come but not for a long while yet.

If G3s are still partially supported now, I would expect the G4s to be supported for at least another 3-4 years, if not longer. Apple are still producing G4 laptops & eMacs and they won't be entirely left behind... prob until X1.

Well, I'm hoping anyway.

 

[V.A.Toss]

the fsb is faster, and theres larger cahces, but the fact that the g4 has almost 3 times less the pipleline stages than the g4, meaning a 1.33 ghz g5 is roughly equal to that of a 1.33 ghz g5

Yeah its true the G4 has shorter pipelines. But general consensus is that the G5 has better branch prediction. So it evens itself out.


Thankyou Lind0843 and stolid. Hit the nail slap ****ing bang on the head.
You will see the performance gain in a few years.


The FSB on a G4 absolutely cripples its performance, it just gets saturated.

And benchmarks arent a very good way to judge speed.
If you judge speed by benchmarks then you will find that a P4 will keep up with the G5. No chance.
You will also see the P4 will keep up with an equivalent Athlon, again untrue.
If you use a powermac G4, then you will see that it gets hammered by a G5. Particularly anything remotely memory intensive.


Benchmarks show the advantage the G5 has over the G4 is roughly the equivalent of the clock advantage. That is actually very good when considering; You will find when looking at the benchmarks of a 1.25Ghz G4 Vs a 1.0 Ghz G4, that the performace gain is LESS than the gain of clock speed.

By the way lind, dont worry about being stuck on a G3 ibook. I too am stuck with a G3, rev.a imac. Feel free to laugh.

 

[Mechcozmo]

The G3 went from being introduced at around 233Mhz and finished up at 900Mhz, in the last G3 iBooks. At least that is what I remember...

The G4 was introduced at around 350Mhz in the original PCI-graphics "Yikes!" model... I think...

So, it seems that the G4 will continue for quite a while, at least to 1.8 Ghz. It will not fade quickly... we will have an iBook G4 still in a couple years.

The G5 PowerBooks will not be a million times better. And so I am not worried, because my PowerBook will last quite a while.

 

[JFreak]

clock-for-clock the G4 is faster than the G5, that has been proved many times. in protools audio work 1.5GHz G4 powerbook is equal to single 1.6GHz G5 powermac, even with the great fsb difference of 167MHz vs 800MHz. this of course assumes the protools powerbook system has an external firewire hard drive installed.

(but at that point, the G4 fsb is fully saturated and the G5 fsb could still move more data were the cpu fast enough to feed it. so, it's not ok to say G4 performance scales by the clock speed - we are now at the point where it doesn't matter if the G4 chip operated at 10GHz, if the fsb is not improved, there's not going to be any kind of performance advantage at all. from now on the performance of G4 systems will only improve if the fsb can be made faster, but otherwise that's it.)

point being the G4 is a good chip design, possibly better than the G5. it's just that the cpu is not the whole system, and it's the whole system that counts...