Macworld Official g5 assesment

[BevvyB]

OK FOLKS

THIS IS THE FIRST 'PROPER' ASSESMENT FROM MACWORLD

http://www.macworld.com/2003/09/revi...g5testresults/

I don't know about you lot, but I'm really looking forward to 250+ fps when I'm playing Quake III :)

[BevvyB]

Hope 10.3 brings the extra muscle and helps unlock the hidden potential which I'd like to think we're not seeing just yet

[BevvyB]

I am a recent convert to Altiverb. I want to run lots.

How many do reckon you can run on a G5 dual 2ghz? Compared to dual pc etc

[Mind-Over-Midi]

Hay man, just freeze em.

[BevvyB]

Not so cool on a bus send

Favourite would be to run a few as bus sends

[tendril]

If you ignore the useless Photoshop and Quake benchmarks

I'm curious why you think the PS benchmarks are useless?

if anything the Cinema 4D scores are useless because that application is not optimized for the G5 yet.

Perhaps, but nothing tweaked in the OS is going to increase the number of available Altiverbs.

Not sure what you mean. But here's some numbers from Photoshop 7 on the same machine running 10.2.6 vs Panther beta:

g4 dual 500 (10.2.6):
90 Degree Rotate - 2.3
9 Degree Rotate - 8.9
.9 Degree Rotate - 7.3
1 Gaussian - 3.2
3.7 Gaussian - 5.4
87 Gaussian - 8.4
1 pix USM 4.1
3.7 pix USM 6.1
10 pix USM 8.1
Despecle 3.6 1.7
RGB-CMYK 9.8
60% Reduction 3.3


g4 dual 500 (panther beta 7b39):
90 Degree Rotate 1.5
9 Degree Rotate 5.6
.9 Degree Rotate 7.1
1 Gaussian 2.9
3.7 Gaussian 3.9
87 Gaussian 5.8
1 pix USM 2.6
3.7 pix USM 4.7
10 pix USM 7.9
Despecle 3.3
RGB-CMYK 9.8
60% Reduction 2.9

So Panther looks like it will bring improvements, but maybe not enough for an additional AltiVerb depending on what else you're running.

Of course, the OS Apple is shipping the G5s with is patched to run on them. Likely not very well optimized either.

[tendril]

Originally posted by bombguy
Graphics and audio are different enough, and garage plug-in manufacturers are different enough from Adobe, Inc. that these numbers don't mean much to people using audio apps.

Curious. So you arrived at your 25-30% number presumably by looking at iTunes MP3 encoding, since that's the only audio app tested?

Interesting methodology there.

If there were convincing performance to be had in audio, Apple would be publishing Emagic benchmarks (a company they own) instead of fluffed up nonsense from their soon to be ex-buddy Adobe.

I say, ignore everything Apple publishes as benchmarks, for audio or anything else. Benchmarks that come from the manufacturer of the product being tested are never reliable, from Apple, from nVidia, Intel, or anyone else...is this news to anyone?

[tendril]

Originally posted by bombguy
I've watched several generations of chips come and go from various manufacturers.

In other words, you're making predictions about G5 performance based on the past. This is fine, because there truly aren't any reliable benchmarks out yet for the G5, even the MacWorld ones are suspect because they say very little about their methodology, and because the 'speedmark' numbers are complete BS (take another look at the G5 1.6 vs G4 1.42, and then look at the other tests, something isn't right). And lastly they used Cinema 4D, which is not optimized for the G5 yet. But, you appeared to be using the MW benchmarks as proof that the G5 is 25-30% faster, when they simply aren't worthy of that kind of statement.

Regardless, it's certainly not going to smoke a Pentium running at double the clock speed across general purpose tasks.

We agree here. I don't think anyone can realistically expect the G5 to 'smoke' a P4 at 3Ghz, but I do think it will give it a run for it's money. We just can't say for sure yet.

[Robotnik]

Originally posted by bombguy
I've watched several generations of chips come and go from various manufacturers.

Historically, CPU speed relates linearly to audio plug-in performance. This makes sense, because CPU speed indicates how fast instructions can be run. RISC/CISC and other architectural wars washed in practice. Things like SIMD/3DNOW/Altivec offer little spikes along the way but they haven't found wide appeal and don't seem to apply to general purpose mathematical tasks such as the ones used in many plug-ins.

To whit: a G4/400 gave almost exactly 30% more plug-ins than a G3/300. It didn't become a "supercomputer" just because they put G4 instead of G3 in front of the number that actually matters.

People are expecting some sort of magical boost with the G5 and it's not coming. Maybe you'll get 70% more oomph from a G5/1.6GHz than a G4/1.0GHz instead of the 60% raw clock speed boost. So wow, now it's like a "G4/1.7" instead of a "G5/1.6".

Regardless, it's certainly not going to smoke a Pentium running at double the clock speed across general purpose tasks.

The fact that it comes close if you put two of them in a computer and run a specific Photoshop macro is nifty, but it doesn't mean much to audio users.

--Erik

"Many aspects of audio applications will profit from the new G5 Frontside Bus; for example, especially when large amounts of audio data must be streamed, or moved in memory. This includes audio inputs and outputs plus audio busses within an application's internal mixer."

[BrianT]

At 1GHz FSB for the G5 2GHz, the Mac goes from slowest to fastest FSB overnight. That's a factor of ~6x over the G4s.

The real issue here regarding all the various CPU, FPU, FSB, etc numbers being tossed around, and their relative impact on audio, is deceptively simple. One question:

"Where is the bottleneck?"

That question has different answers for different applications, so none of these benchmarks, or even a collection of them, will ever tell it all for a specific user, his app, and how he uses it.

Most audio use is not FSB, or even memory speed limited. Granted, a factor of 6x has to show some impact, but not as much as it will in other applications. Far and away, audio is CPU and hard disk bound, and in the case of traditionally coded plugins, specifically FPU bound. That is, the FSB and main memory subsystem can deliver more than enough bandwidth to keep the CPU fully fed. The system is waiting on the CPU to pump out Native plugin DSP, assuming the HD is swingin'. This scenario varies in a multifunction app like a DAW, obviously, but for most people, plugin count while maintaining a responsive GUI is where the buck stops.

This is obvious in the case where an app's specific measured performance area of interest (as in Native plugin count) scales linearly with CPU clock speed. Linear scaling with CPU clock speed demonstrates that it's the CPU speed (actually primarily FPU unless Altivec or SSE are used) that is the limiting factor. In that case, for instance, increasing the FSB speed while maintaining the same CPU speed would have marginal impact. Certainly less than 10%.

Anyway, the point is that there are enough variables that the only real tests for DAWs are DAWs. Some benchmarks have some bearing, but nothing else streams a large number of seperate and fragmented HD files while doing heavy number crunching in realtime while substantially loading the PCI bus and being required to keep the GUI unaffected, in the same way as a DAW.


Regards,
Brian T

[hmurchison]

So it's Macworld. .so what. Speedmarks...Cinema4d Benches?


All pretty much worthless right now.

The PPC970 is MUCH different from the 603/604/G3/G4 lineage.

I've followed Apple for years watching them move from Moto 68k CPUs to PPC chips. Even then it took over a year for apps to be written to support the new PPC 601 chips.

Today is no different. If you are happy with your computer now then keep it. If you must have a faster computer now then buy all means buy a G5. It will only get faster as apps are tweaked.

We have Dual Integer and FPU units to work with now. We have a CPU that , in order to maximize performance, requires a different mode of thinking for programming.

The Benchmarks aren't lying...they're just outdated.

Support of the G5 is dependent on Developers supporting the features of the processor.

Things are improving rapidly. Apple/IBM are making swift work on the GCC 3.3 Compiler and IBM currently has a high end PPC 970 Compiler which will offer even better applicaton code. Will this affect Music apps? That remains to be seen but I can guarantee you that the benches you're seeing now are almost all misrepresenting what the G5 can "really" do.

[BevvyB]

MaccAddicts first crack at it

http://www.macaddict.com/news/news_007.html

[BrianT]

Originally posted by hmurchison

Support of the G5 is dependent on Developers supporting the features of the processor.

Things are improving rapidly. Apple/IBM are making swift work on the GCC 3.3 Compiler and IBM currently has a high end PPC 970 Compiler which will offer even better applicaton code. Will this affect Music apps? That remains to be seen but I can guarantee you that the benches you're seeing now are almost all misrepresenting what the G5 can "really" do.

I agree, in principle. A quick look backwards at the progress of the P4 as SSE2 was gradually implemented will confirm the concept.

IBM's improved version of Altivec on the PPC970 is better technology than SSE2 on the P4, and could make a big difference....in theory. In practice, a great deal was made of Altivec years ago with the G4 intro, but precious little benefit has materialized for Mac DAW users in years of time gone by.

Two issues, IMO. Lack of enthusiastic developer support, and the untidy fact that Altivec has limited impact in apps that are multitasking in nature, like DAWs. It hates being interrupted by non-Altivec instructions to the CPU. That is why it can speed up an app like the RC5 client by a factor of fully 2.5!!!! while having marginal impact in Logic. Unless IBM has compensated for that by design, Altivec will never have the impact on a DAW in the realworld that it
"should". I suspect that has much to do with the lack of audio developer enthusiasm.

But I still think the G5 is cool.


Regards,
Brian T

[tendril]

Most audio use is not FSB, or even memory speed limited. Granted, a factor of 6x has to show some impact, but not as much as it will in other applications. Far and away, audio is CPU and hard disk bound, and in the case of traditionally coded plugins, specifically FPU bound.

How does this change, though, as you add more RAM? It seems like a G5 with a lot of RAM should lessen the dependence on the HD, this is where the increased FSB bandwith should shine.

Anything FPU bound will see a great shot in the arm on a G5 as it's dual FPU units (or 4 on the DP machines) are stormers.

The weakness of the G5 is integer performance. Altivec is about a wash with the G4, losing in some areas gaining in others, but the clock difference makes the G5 pull ahead here.

[tendril]

Well, it may not be audio, but it looks like photoshop performance skyrockets when you have 2Gb of RAM. MacAddict has some benchmarks from photoshop:



Not sure if I can believe that, it's a huge leap in performance, almost too good. That's scary...

EDIT: after further review, they don't say how much RAM the other machines had, except the Dual 800 had 512MB of RAM...so the test was skewed toward the G5 with 2GB.

Still waiting for someone to do proper benchmarks...

[BrianT]

Actually, the graphs highlight the relative incompetence of their benchmarking. The order of magnitude improvement from adding substantial ram is inadvertently demonstrating why their other benchmarks aren't showing what they expected.

The RAM added is sufficient to move the test off the HD and into RAM. Huge difference, not in the CPU performance, but in the speed difference between RAM and hard disk. And that is why the other benchmarks are clustered so tightly, not by CPU type, but by the hard disk technology in use at the time the computer was manufactured. Look at the 1.8GHz single G5 vs the 2GHz dual G5. Often very close. Why? They use the same hard disk. Most of the benchmarks in the Macaddict article are simply showing the performance difference between different vintage hard disks. Quite useless benchmarking for the most part, really. The only useful conclusion is that RAM is faster than hard disks. I already knew that.

Actually, based upon the quality of this article, these people have no business attempting to "educate" anyone about computer performance. The benchmarks are worse than useless to a Mac DAW user. They are misleading.......mostly in the G4's favor. Getting all of the tested computers to accomodate the test file within RAM would show what's really going on, and I would predict the G4's would see nowhere near the speed gains that the G5 does. Most of the Mac sites could use some help on the hardcore tech side of things.


Regards,
Brian T

[BrianT]

There is a big change coming in the way pros work on DAWs. I've had a chance to experience it, and it is very good.

I'm going to write a little mini-article on it in a week or so.


Regards,
Brian T

[tendril]

Yes, that's a good way of looking at it: the longer graphs essentially measure the HD speeds.

My speculation is that even if you put 2GB of RAM in the G4 that the speed gains would be nothing amazing...alas, the test needs to be done and they really should be taken to task for that graphic because it's misleading.

I was under the impression that streaming wasn't always used by DAWs or plug-ins if RAM is available. In any event, this should really help softsynths and to a lesser extent native processing plug-ins.

[FOURTHTUNZ]

Originally posted by tendril


My speculation is that even if you put 2GB of RAM in the G4 that the speed gains would be nothing amazing...alas, the test needs to be done and they really should be taken to task for that graphic because it's misleading.

This is what I've been wondering about and waiting for, that is testing of G4's and G5's with gobs of ram, 2 gig- 8-gig would be interesting.
OSX with dual G5's, would be cool to be doing a bounce then jump out to toast and burn a cd while its doing the bounce
As far as the Mac addict testing I thought it was interesting to see how much better the G5 did with lots of ram, and I find it to be as useful as any of the other "tests" posted on the internet.
daniel

[BevvyB]

I'm getting this from a German website -

They have been testing Logic Audio on a 1.6ghz G5 and comparing it to a dual 1ghz G4

Altiverb instances using "Philips Hall (Stereo To Far Omnis):

G4 dual 1ghz - 3-4 instances
G51.6ghz 10-11 instances

Platinumverb on default setting:

G4 dual 1ghz - 24-25 instances
G5 1.6ghz 52 instances

http://www.mac-pro-audio.de/email-bi...erG5-test.html

[mpr3]

Originally posted by BevvyB
Altiverb instances using "Philips Hall (Stereo To Far Omnis):

G4 dual 1ghz - 3-4 instances
G51.6ghz 10-11 instances

That is HUGE.

On my dual 1.42 I can get 5 of that particular IR (Phillips) with the buffer set to 128. And can get 7 with a 2048 buffer.

[Mind-Over-Midi]

[Killafornia]

Originally posted by BevvyB

Platinumverb on default setting:

G4 dual 1ghz - 24-25 instances
G5 1.6ghz 52 instances

http://www.mac-pro-audio.de/email-bi...erG5-test.html

Hmm. 2.2ghz Athlons can run around 70+ Platinum Verbs on Logic 5.

[hmurchison]

Originally posted by Killafornia
Hmm. 2.2ghz Athlons can run around 70+ Platinum Verbs on Logic 5.

The G5 gives up 600Mhz and Logic isn't optimized yet. I'd say it's doing pretty damn good.

[Killafornia]

Yeah its doing pretty good. Not bad (Price, and release date aside). Its definetly huge gain of power for mac users. I dont think it can knock me out of my room through the wall into a tree tho .

[BevvyB]

New Photoshop tests done PROPERLY with the Adobe g5 optimiser plug

http://www.chaosmint.com/benchmarks/...c-g5-ps7bench/

Also, look here - these are the final scores for Photoshop across the board

2x 2000 G5 OSX 10.2.7 (G5 plugin) 547 (energy settings highest perf)
2x 3060 Xeon (OC'd 2400) 488
2x 2930 Xeon (OC'd 2400) 471
3200 P4 (800MHz) 427
3000 P4 (800MHz) 405
3495 P4 (OC'd 3.06) 386
3060 P4 XP Pro (533 FSB) 358 HT
2x 2200 Xeon PC 800 RDRAM CPQ Evo 357 HT
2x 1500 G4 (OC'd 1420) 348
2x 1333 G4 DDR OS9.2 (oc'd 1.25) 346
1800 G5 OSX 10.2.7 (G5 plugin) 344 (energy settings highest perf)
2x 1420 G4 OSX 10.2.4 338
2x 2400+Athlon MP 338
2x 1250 G4 OS 9.2.2j 337
3200+Athlon XP 332
1800 Opteron(dual-chnlDDR 333) 332
2x 1333 G4 DDR OSX10.2.2(oc 1.25) 326
1800 OPteron(singl-chnlDDR333) 320
3000+Athlon XP 318
2x 1250 G4 OSX 10.2.5 318
2x 1250 G4 DDR OSX 10.2.1 316
2x 1800 Athlon MP 312
2800+Athlon XP Barton 298
2x 2000 P4 Xeon 286
2x 1200 G4Powerlogix(867MHzG4/QS) 285 upgraded
2x 1533 Athlon MP 285
2x 1533 Athlon MP 283
2530 P4 mobile (OC'd 1400) 282
2700 P4B (OC 2400, 600 MHz FSB)280
2x 1466 Athlon XP 279
1600 G5 OSX 10.2.7w/G5 Plugin 276 *MacNNscores
2666 P4 (DDR 333) 269
2x 1000 G4 DDR 10.2 267
2400+Athlon XP 262
2x 1000 G4 OS9 260
2x 1000 G4 OSX 10.1.5 254
2400+Athlon 252
2400 P4B (800MHz) 251
2400b (sis 648 DDR400) 251
1600 Centrino IBM T40 250
2400 P4 (533MHz bus) 249
2400 P4 B 241
2340 P4 (overclock) 239
1600 Centrino Dell D800 236
2400 P4 234
1800+Athlon XP (1533 MHz) 226
1577 oc'd Athlon XP (Lestat) 221
2x 1000 G4 OSX 10.2.2 (upgraded) 218 ?!(dual 533 logic board)
1548 Athlon XP 214
1670 Athlon XP (2000+) 213
1667 Athlon XP 211
1400 Athlon XP 1600+ xp pro 200
1x 1533 Athlon MP 197
1300 Centrino Sony VAIO Z1A 196
1000 G4 17" Powrbk OSX 10.2.6 196
2000 P4 Xeon 194
1400 Athlon XP 1600+'98SE 191
1000 G4 OSX TiPbk 10.2.2 185
2x 533 G4 OSX 10.1.5 175
2x 533 G4 OS 9.2.2 174
1800 P4 173
1200 AthlonMP 168
1508 Celeron (overclock) 167
1400 PIII Tualatin 160 **?
2x 550 G4 OSX 10.2.3 (OC Cube) 160 **?
2x 500 G4 OSX 152
2x 450 G4 OS9 151
1333 Athlon TBird 147
2x 450 G4 OSX 10.1.5 143
800 G4 Pbook OSX 1MB L3 135
733 G4 (miro7) 134
667 G4 PBk OS9 noL3 127
667 G4 PBk OSX 10.2.3 no L3 125
466 G4 OS9 123
667 G4 OSX TiPBk 10.1.5 noL3 121
866 PIII 114
466 G4 OSX 133 MHz bus 112
550 G4 Powrbk OS9* 104
500 G4 Pbook (OC'd 400) 103
1x 450 G4 OSX 100 MHz bus 101
1000 Athlon TBird (PS6.01) 100
550 G4 Powrbk OSX* 95
933 Transmeta Crusoe Sony 78
700 G3 iBook 74
600 G3 iBook OS 9.2.2j 70
233 PII 30

Don't ask me why they haven't got the LATEST PC chips on this list

Also, at the Paris conference, Steve jobs said that 3ghz G5's will be available this time next year, so at least we have a chip that's going somewhere quickly

Still haven't seen much about Logic Audio/Native Audio Apps. Soon!

I wonder when Al is getting his hands on one....

[FOURTHTUNZ]

Originally posted by BevvyB
New Photoshop tests done PROPERLY with the Adobe g5 optimiser plug


Don't ask me why they haven't got the LATEST PC chips on this list

Yeah that would give a better idea as far as real world but at least the new macs are back in the game hardware wise.
They're still tweaking the os so it'll be interesting to see where it is 2 months.
Hey I'm still loving my Mac QS 733 until
Ptools looks more solid on X.

daniel

[tendril]

Here's the top 15 cleaned up a bit:

2x 2000 G5 OSX 10.2.7 (G5 plugin) 547 (energy settings highest perf)
2x 3060 Xeon (OC'd 2400) _ _ _ _ _488
2x 2930 Xeon (OC'd 2400) _ _ _ _ _471
 _ 3200 P4 (800MHz) _ _ _ _ _ _ _ 427
 _ 3000 P4 (800MHz) _ _ _ _ _ _ _ 405
 _ 3495 P4 (OC'd 3.06) _ _ _ _ _ _386
 _ 3060 P4 XP Pro (533 FSB) _ _ _ 358 HT
2x 2200 Xeon PC 800 RDRAM CPQ Evo 357 HT
2x 1500 G4 (OC'd 1420) _ _ _ _ _ _348
2x 1333 G4 DDR OS9.2 (oc'd 1.25) _346
 _ 1800 G5 OSX 10.2.7 (G5 plugin) 344 (energy settings highest perf)
2x 1420 G4 OSX 10.2.4 _ _ _ _ _ _ 338
2x 2400+Athlon MP _ _ _ _ _ _ _ _ 338
2x 1250 G4 OS 9.2.2j _ _ _ _ _ _ _337
 _ 3200+Athlon XP _ _ _ _ _ _ _ _ 332
 _ 1800 Opteron(dual-chnlDDR 333) 332

the G5 is a Photoshop monster.