PowerMac G5 Reports and Benchmarks

[MacRumors]

PowerMac reports continue to trickle in to MacBytes.com (http://www.macbytes.com/).

- Barefeats has posted PowerMac G5 2GHz Cinebench 2003 (http://www.barefeats.com/g5sum02.html) benchmarks.
- Japan G5s to be delayed (http://host.wallstreetcity.com/wsc2/Autoflag.html?Button=Get+Story&DB=SQL&SID=234x5705&Symbol=AAPL) until September.
- G5 1.8GHz Cinebench (http://www.postforum.com/forums/read.php?f=6&i=87424&t=87424) scores.
- Virtual PC refuses to run on the PowerMac G5 (http://www.spymac.com/forums/showthread.php?s=&threadid=37234).

Surprisingly few new reports of 1.6GHz G5s have emerged, and no detailed Photoshop results are yet available.

The author of Xbench clarified to xlr8yourmac.com (http://www.xlr8yourmac.com) that the current version of Xbench is not well equipped to benchmark the G5 machines:

Xbench 1.0 uses the vec_dst instruction, in the altivec test, and the memory stream tests, which causes a big performance hit on the G5. I talked to some Apple engineers at WWDC about the issue and am working on a 1.1 release that avoids using vec_dst when running on the G5.
The new release should be out real soon.
-- author of Xbench

[Haberdasher]

No Virtual PC 6? But what about the incredible comparative benchmarks?!? :D

[Powerbook G5]

Nuts to VPC, I say! The G5 is too good to be running Windows :cool: Well, actually, I am sure there will just need to be a patch to get it working before too long. I cannot wait to see some G5 optimized benchmarking done to see just what these babies can do.

[mymemory]

Irony:

Getting a G5 Dual 2Ghz with 8GB of Ram
Run Virtual PC
Then play Solitarie

[idea_hamster]

Originally posted by Powerbook G5
I am sure there will just need to be a patch...
Oh goody! A Microsoft patch! It'll be my first.

I always felt so, you know, left out...marginalized...different -- and just because everyone I knew would be all "Hey, did you get that patch?" and I'd be all dejected, like, "No, my computer didn't need any patch."

Maybe MS will include VPC patches in their nightmare opt-out auto-patch system. Then we'll really be part of the club!

You gotta be -- what does Strongbad say? -- jockin' me. :rolleyes:

(Not a flame on PBG5 -- just not looking forward to the "patch" lifestyle.)

[evolu]

Originally posted by idea_hamster
Oh goody! A Microsoft patch! It'll be my first.

I always felt so, you know, left out...marginalized...different -- and just because everyone I knew would be all "Hey, did you get that patch?" and I'd be all dejected, like, "No, my computer didn't need any patch."

Maybe MS will include VPC patches in their nightmare opt-out auto-patch system. Then we'll really be part of the club!

You gotta be -- what does Strongbad say? -- jockin' me. :rolleyes:

hilarious

[tazznb]

...around waiting until the software's optimized before rushing out there to do tests / benchmarks. This may prove to be misleading for the computer illiterate, or plain old FUD / FODDER for the ones that have an unreasonable bias against the platform.

Sad thing is that people that don't know any better will think it to be as true as the word of god.

Hopefully I'll get my dual G5 among the first since I'm in education.

[twinturbo]

Why's everyone running Cinebench on the new G5's, esp. when it's not even optimized for the platform? Photoshop's optimzed now, and no one is running PSBench7 or anything like that to see just how these systems stack up. I'm finding that very agrivating, cause there's a lot of anti-mac bantering going on (and I must admit I've been a bit disappointed in the results myself) around these benchmarks. So how about an optimized Photoshop benchmark to kick things in gear? Anyone?

[NavyIntel007]

VPC doesn't work? Not surprising. Watch Microsoft not even release a patch to make it work. Certainly MS doesn't want you to buy a Mac, especially when it's faster than a PC.

edit: or if they do I imagine it will be full of...

while(1);

so that you think your Mac is really slow but it's really just lousy MS crap.

[CaptainScarlet]

Personally, I think all of the Benchmarks so far, are a bunch of sh**!!!!

And yes, we all know they are all not optimized for the G5, well then don't post them or make a big deal about them...

Give us really testing...How is Mail, how is Illustrator, how is PHOTOSHOP!?!?!? How long does iTunes take to rip a CD with 15 songs, to 256 kbps AAC files!!

That's what people really want to see....User testing...not R&D testing....


I'm still buying one, no matter what the benchmarks say....


CS....out

[etoiles]

yes, those cinebench numbers are not overly impressive... I got 110 on an 'out of the box' P4 2.4 which costs much less than a 1.6 G5 (which scored 158). So the P4 and the G5 would render about at the same speed at equal mhz. I hope optimization will improve things a bit in favor of the G5...

Personally, I couldn't care less for VirtualPC...

[NavyIntel007]

VPC is really a waste. I mean, you could buy a dirt cheap PC for $200 if you needed that one archaic acounting program to work... I really think that people are buying VPC for one or two or at max 5 programs that if they did their homework, they'd find other, cheaper equivilants.

[CaptainScarlet]

Originally posted by etoiles
Personally, I couldn't care less for VirtualPC...

I totally agree....The only reason I would even think of running this would be for Microsoft Access. That's it, if any....

Games?!?! Well I also have a PC in the house, just for that. And lets not forget about the P2 either....


CS.....out

[MrMacMan]

Waaiittt a second...


Didn't I see xBench Scores already on the G5?

Um....

:confused:

[xtekdiver]

Originally posted by NavyIntel007
VPC is really a waste. I mean, you could buy a dirt cheap PC for $200 if you needed that one archaic acounting program to work... I really think that people are buying VPC for one or two or at max 5 programs that if they did their homework, they'd find other, cheaper equivilants.

Actually I just bought VPC; and I need it too. You see, I need to be able to write .Net code, but I don't want to buy a PC laptop, so I will run VPC on a Powerbook (WHENEVER THEY COME OUT!!!!) Sorry. Just a bit frustrated. Anyway, VPC will come in handy when my mean boss tells me I have to write in .Net instead of Java :mad: That being said, it will be interesting to see if MS will actually make the effort to support the G5 and just how quickly they will get an update out. So the moral of my story is: if the Powerbook updates turn out to be G5s I'm screwed!!!:eek: :eek:

[themadchemist]

Nice to get all the links to benchmarks in one post! Thanks arn.

[NavyIntel007]

Originally posted by xtekdiver
Actually I just bought VPC; and I need it too. You see, I need to be able to write .Net code, but I don't want to buy a PC laptop, so I will run VPC on a Powerbook (WHENEVER THEY COME OUT!!!!) Sorry. Just a bit frustrated. Anyway, VPC will come in handy when my mean boss tells me I have to write in .Net instead of Java :mad: That being said, it will be interesting to see if MS will actually make the effort to support the G5 and just how quickly they will get an update out. So the moral of my story is: if the Powerbook updates turn out to be G5s I'm screwed!!!:eek: :eek:

I'm sorry... I totally forgot about software development. VPC is definitely a great tool to port aps to PC. Probably even better for web designers to make sure pages render right in IE.

[arn]

Xbench 1.1 is now available.

Optimized for G5.

Waiting for new results...

arn

[dguisinger]

Originally posted by NavyIntel007
I'm sorry... I totally forgot about software development. VPC is definitely a great tool to port aps to PC. Probably even better for web designers to make sure pages render right in IE.

I as well have been looking at buying a G5 to replace my last desktop PC (Dell P4 1.5GHz). I do tons of MS .NET Development (I live and swear by .NET....), and thus need a windows enviroment to code. I know the G5 will run my Windows code faster than my current Dell, the only thing holding me back is the $2k i still owe on the machine (ouch...its 2 1/2 years old!).

[MrMacMan]

Is it me or do you guys think there should be 2 benchmarks:

Optomized for the G5 And
Not Optimized for the G5.

Look, right now most products are *not* optomized for the G5, so optomizing benchmarking is almost pointless.

Its giving the advantage to the G5 almost un-fairly.

You must expect they don't optomize to have a fair benchmark.

[DeusOmnis]

Can anyone tell me what the vecLib FFT test is? My computer scored EXTREMELY low in it... a 23.97. I'm running a dual 1.0 ghz G4 MDD and I got wasted by a Dual 800 mhz Quicksilver machine.

[1stunna]

hey guys, ill have a xbench score for a single 1.6 tomorow.

[windwaves]

well I guess a surprisingly large number of people realized that the 1.6 is likely not a great leap forward from the dual G4s. And in any case I am not surprised at all since it seems somewhat clear to me that very very few have been shipped/delivered.

[daRAT]

Originally posted by tazznb
...around waiting until the software's optimized before rushing out there to do tests / benchmarks. This may prove to be misleading for the computer illiterate, or plain old FUD / FODDER for the ones that have an unreasonable bias against the platform.

Sad thing is that people that don't know any better will think it to be as true as the word of god.

Hopefully I'll get my dual G5 among the first since I'm in education.

What you said is true, this link http://www.theandyzone.com/Computer/shootout.html found on http://www.xlr8yourmac.com/ shows the AMD machine chewing up the Mac's G5 included.

However, if I hadn't read here that the Cinebench used was the old one, not optimized for the G5 I would be stunned at the results.

If you have the time and money, anyone can build a faster machine. I am waiting for real users to say what they think of the speed of the new machines ;]

[daRAT]

Originally posted by windwaves
well I guess a surprisingly large number of people realized that the 1.6 is a piece of c**p. Which, in turn, surprises me - I always thought that "people" never got it right :)

And what technical data or hands on experience do you base this informative opinion on?

The lack of "reports" ? I know if I had got one, the first thing on my mind would be to spend a weekend enjoying the new machine ;] Not running benchmarks. Give the users a week and we will see more.

[windwaves]

daRAT, no technical basis really. Pure speculation. In fact I think it was too early in the morning for me to really say much. I have to edit that message.

What was on my mind is that the reason we have not seen many home made benchmarks is because not many of these new macs have been shipped/delivered at all. Thereby I somewhat disagree with your thought about people not being concerned with benchmarking; on the contrary I have a feeling many new owners will be doing that first thing.

But my main thought is the fact that Apple has said +100k have been ordered (although I believe the vast majority were duals) but probably the only ones who got them so far in numbers are Apple's corporate clients.
And I bet no individual will even see a dual before September.

[Baja2k]

Ok, maybe this question has been asked but the relative trickle of information about the G5 is killing me.

Does anyone on this site have an "Expected Ship Date" prior to 8/29?

I guess a few 1.6 machines seem to have made their way to the public. But has anyone received the dually?

My "Expected" ship date is 8/29 (and has been for a month) and I'm startin' to get the shakes. Please Apple don't be late!

[Mad Baggins]

Originally posted by daRAT
I know if I had got one, the first thing on my mind would be to spend a weekend enjoying the new machine ;]

But the G5s and old G4s all run the same OS X and software. Really, the biggest difference is the speed of the G5s--in which case "enjoying" the new machine is equivalent to benchmarking it. :D

[macrumors12345]

Originally posted by twinturbo
Why's everyone running Cinebench on the new G5's, esp. when it's not even optimized for the platform? Photoshop's optimzed now, and no one is running PSBench7 or anything like that to see just how these systems stack up. I'm finding that very agrivating, cause there's a lot of anti-mac bantering going on (and I must admit I've been a bit disappointed in the results myself) around these benchmarks. So how about an optimized Photoshop benchmark to kick things in gear? Anyone?

Yes, Photoshop would be a much better benchmark for comparison (since it has been at least somewhat optimized for the G5). However, people are running Cinebench because all you have to do is download a single file and run it. With PS, you need to actually own the application, either that or separately download the demo version (which is not as easy to find as Cinebench), the G5 patch (may or may not actually work with the demo version!), and the benchmark file/scripts. It is much harder. So that, unfortunately, is why people are resorting to Cinebench.

Incidentally, my friend at Adobe tells me that some engineers ran a massive Photoshop benchmark (it tooks hours to run!), and the G5 really was *double* the speed of the G4. Also, a Maxon developer has posted that a recompiled version of Cinebench scores around 265 CB for the single CPU test with the 2 Ghz G5. I believe that the Dual 1.42 Ghz scores around 135 CB for the single CPU test, so you can see that the 2 Ghz G5 is *twice* as fast as the fastest G4 in this test. Do you see a pattern forming? I hope it holds for other real world apps too (when they have been recompiled)... ;-)

Incidentally, this G5 score for the single CPU bench would be comparable to an Athlon 3200+ or a P4 2.66 Ghz. For whatever reason the P4 does especially well at Cinebench (as you can see since it handily beats the fastest Athlon...and that is without even using Hyperthreading), so losing to the fastest Xeon (3.06 Ghz) by 10% in this test is not a problem in my view, given that it is basically the worst case scenario for the G5.

[daveL]

Originally posted by DeusOmnis
Can anyone tell me what the vecLib FFT test is? My computer scored EXTREMELY low in it... a 23.97. I'm running a dual 1.0 ghz G4 MDD and I got wasted by a Dual 800 mhz Quicksilver machine.
That's the type of result that makes me skeptical of Xbench, in general. I ran it twice on my 1GHz TiBook. The first time the FFT test score was terrible, similar to your results. I ran it again and got a score 5x higher! And, yes, I had quit all other apps. On top of it all, 1.1 *still* says my CPU is 667 MGz, instead of 1 GHz.

On the topic of G5 optimized software builds: Don't you think Apple will recompile OS X and all their apps for the G5? Of course they will. Photoshop has already optimized for G5. Any major app on the Mac will want the G5 boost, most of which will come simply by running with the tweaked OS X libraries / frameworks.

[garymm]

Originally posted by daveL
On top of it all, 1.1 *still* says my CPU is 667 MGz, instead of 1 GHz.

do you think it's possible that it's giving you the current cpu speed? I'm pretty sure OS X automatically throttles down the CPU on a laptop if it's not being used.
But if it were reporting the cpu speed incorrectly, it wouldn't surprise me. I've heard a lot of complaints about it.

[Gelfin]

Originally posted by daveL
That's the type of result that makes me skeptical of Xbench, in general. I ran it twice on my 1GHz TiBook. The first time the FFT test score was terrible, similar to your results. I ran it again and got a score 5x higher!

This is actually the kind of result that should make you skeptical of benchmarks in general, not just Xbench, but particularly benchmarks which are designed to be dropped onto a given machine and run by amateurs.

Just quitting all other user applications is not sufficient to get consistent results out of a benchmark. Modern OSes have so much going on under the hood that closing other apps is just the tip of the iceberg.

And here we get to the fundamental catch-22 of benchmarking. In order to generate consistent, reproducable results, you must get the system into a state that in no way resembles the typical running state of the machine, and thus the numbers don't really inform the discussion of real-world performance. On the other hand, benchmarks which allege to gauge real-world performance are a joke because there is no consistency, and there's always room for argument on the basis of the individual system context.

It seems like it ought to be easy to just get an answer to the question, "how fast is it?" I've done this professionally, though, and it's nothing like "easy." I hate even reading about benchmarks nowadays, because the results are almost meaningless.

[1stunna]

test1:

Results 124.04
System Info
Xbench Version 1.1
System Version 10.2.7
Physical RAM 256 MB
Model PowerMac7,2
Processor PowerPC 970 @ 1.60 GHz
L1 Cache 64K (instruction), 32K (data)
L2 Cache 512K @ 1600 MHz
Bus Frequency 800 MHz
Video Card GeForce FX 5200
Drive Type ST380013AS
CPU Test 90.16
GCD Recursion 72.93 2.85 Mops/sec
Floating Point Basic 210.14 747.32 Mflop/sec
AltiVec Basic 78.29 1.14 Gflop/sec
vecLib FFT 49.00 755.06 Mflop/sec
Floating Point Library 262.85 10.52 Mops/sec
Thread Test 80.49
Computation 59.27 477.51 Kops/sec, 4 threads
Lock Contention 125.37 1.57 Mlocks/sec, 4 threads
Memory Test 241.11
System 188.79
Allocate 329.91 111.23 Kalloc/sec
Fill 136.04 1082.91 MB/sec
Copy 181.52 907.60 MB/sec
Stream 333.53
Copy 330.19 1346.57 MB/sec [G5]
Scale 312.42 1297.81 MB/sec [G5]
Add 342.88 1446.91 MB/sec [G5]
Triad 351.25 1445.69 MB/sec [G5]
Quartz Graphics Test 154.81
Line 179.17 4.56 Klines/sec [50% alpha]
Rectangle 172.06 12.10 Krects/sec [50% alpha]
Circle 171.77 3.96 Kcircles/sec [50% alpha]
Bezier 147.58 1.60 Kbeziers/sec [50% alpha]
Text 120.38 1.96 Kchars/sec
OpenGL Graphics Test 178.12
Spinning Squares 178.12 124.65 frames/sec
User Interface Test 138.33
Elements 138.33 47.07 refresh/sec
Disk Test 105.63
Sequential 114.51
Uncached Write 142.09 56.56 MB/sec [4K blocks]
Uncached Write 137.22 53.56 MB/sec [256K blocks]
Uncached Read 76.73 12.15 MB/sec [4K blocks]
Uncached Read 132.03 53.35 MB/sec [256K blocks]
Random 98.02
Uncached Write 84.92 1.21 MB/sec [4K blocks]
Uncached Write 102.71 23.17 MB/sec [256K blocks]
Uncached Read 95.49 0.63 MB/sec [4K blocks]
Uncached Read 113.35 23.33 MB/sec [256K blocks]



test2

Results 128.84
System Info
Xbench Version 1.1
System Version 10.2.7
Physical RAM 256 MB
Model PowerMac7,2
Processor PowerPC 970 @ 1.60 GHz
L1 Cache 64K (instruction), 32K (data)
L2 Cache 512K @ 1600 MHz
Bus Frequency 800 MHz
Video Card GeForce FX 5200
Drive Type ST380013AS
CPU Test 115.66
GCD Recursion 72.79 2.84 Mops/sec
Floating Point Basic 212.47 755.61 Mflop/sec
AltiVec Basic 74.60 1.08 Gflop/sec
vecLib FFT 131.71 2.03 Gflop/sec
Floating Point Library 263.97 10.57 Mops/sec
Thread Test 80.64
Computation 59.35 478.12 Kops/sec, 4 threads
Lock Contention 125.77 1.58 Mlocks/sec, 4 threads
Memory Test 242.09
System 189.38
Allocate 304.24 102.58 Kalloc/sec
Fill 131.01 1042.84 MB/sec
Copy 203.21 1016.07 MB/sec
Stream 335.44
Copy 332.89 1357.60 MB/sec [G5]
Scale 314.62 1306.98 MB/sec [G5]
Add 343.47 1449.40 MB/sec [G5]
Triad 353.27 1453.97 MB/sec [G5]
Quartz Graphics Test 151.82
Line 172.21 4.38 Klines/sec [50% alpha]
Rectangle 172.08 12.11 Krects/sec [50% alpha]
Circle 172.93 3.99 Kcircles/sec [50% alpha]
Bezier 143.64 1.56 Kbeziers/sec [50% alpha]
Text 116.68 1.90 Kchars/sec
OpenGL Graphics Test 173.26
Spinning Squares 173.26 121.25 frames/sec
User Interface Test 136.51
Elements 136.51 46.45 refresh/sec
Disk Test 105.58
Sequential 114.08
Uncached Write 145.76 58.03 MB/sec [4K blocks]
Uncached Write 137.39 53.63 MB/sec [256K blocks]
Uncached Read 76.66 12.14 MB/sec [4K blocks]
Uncached Read 126.90 51.27 MB/sec [256K blocks]
Random 98.25
Uncached Write 87.25 1.25 MB/sec [4K blocks]
Uncached Write 103.36 23.31 MB/sec [256K blocks]
Uncached Read 94.82 0.63 MB/sec [4K blocks]
Uncached Read 110.76 22.80 MB/sec [256K blocks]



test3:

Results 135.21
System Info
Xbench Version 1.1
System Version 10.2.7
Physical RAM 256 MB
Model PowerMac7,2
Processor PowerPC 970 @ 1.60 GHz
L1 Cache 64K (instruction), 32K (data)
L2 Cache 512K @ 1600 MHz
Bus Frequency 800 MHz
Video Card GeForce FX 5200
Drive Type ST380013AS
CPU Test 118.69
GCD Recursion 73.95 2.89 Mops/sec
Floating Point Basic 198.78 706.93 Mflop/sec
AltiVec Basic 78.23 1.14 Gflop/sec
vecLib FFT 143.08 2.20 Gflop/sec
Floating Point Library 263.18 10.53 Mops/sec
Thread Test 81.63
Computation 60.51 487.47 Kops/sec, 4 threads
Lock Contention 125.39 1.57 Mlocks/sec, 4 threads
Memory Test 234.46
System 180.08
Allocate 297.39 100.27 Kalloc/sec
Fill 131.40 1045.97 MB/sec
Copy 175.84 879.22 MB/sec
Stream 335.90
Copy 332.45 1355.82 MB/sec [G5]
Scale 313.41 1301.93 MB/sec [G5]
Add 345.68 1458.75 MB/sec [G5]
Triad 355.03 1461.23 MB/sec [G5]
Quartz Graphics Test 153.34
Line 175.22 4.46 Klines/sec [50% alpha]
Rectangle 171.20 12.04 Krects/sec [50% alpha]
Circle 173.84 4.01 Kcircles/sec [50% alpha]
Bezier 144.02 1.56 Kbeziers/sec [50% alpha]
Text 119.57 1.95 Kchars/sec
OpenGL Graphics Test 179.16
Spinning Squares 179.16 125.37 frames/sec
User Interface Test 136.37
Elements 136.37 46.40 refresh/sec


as you can see, it does better the more times you run it.

still doesnt seem right.

[tazznb]

Originally posted by 1stunna
test1:

Results 124.04

test2

Results 128.84

test3:

Results 135.21

as you can see, it does better the more times you run it.

still doesnt seem right.

This is why I think we should all AVOID using unproven benchmarks (which happens to be ALL OF THEM in my opinion).

Someone should just write to Adobe, and request a prescripted Photoshop / Illustrator salvo for download.

Contact reliable 3D software makers and get identical 3D files to render for the 3D enthusiasts, etc.

Other than that the reporting of benchmarks is just a HUGE waste of everyone's time.

[1stunna]

yes, i would have much rather run psbench, or some optimized g5 cinebench. but as you said, the adobe people require you to install SW, something the asses at my school wont allow. and cinebench has not yet been updated.

if there is any other standalone bench thats been optimized, id be more than happy to run it.

[macrumors12345]

Originally posted by tazznb
Someone should just write to Adobe, and request a prescripted Photoshop / Illustrator salvo for download.


Boy...Illustrator. My friend is a developer on the Illustrator team...I dunno if that app is even optimized for the 7450 core yet, let alone the G5. Well, maybe that is an exaggeration, but suffice it to say that it is not one of Adobe's more streamlined apps. I don't even know if it takes advantage of Altivec yet.

[henryblackman]

I think what we are going to find is that the G5 and Panther will bring the most visible speed improvements. Panther has had some really nice reengineering which means thread handling is vastly improved... which particularly benefits dual proc machines, but also benefits us all with slower 800 Mhz iMacs (boo hoo).

Apple have said in their documentation for ADC members that optimising for the G5, will likely optimise for all processors.

What I'd like to see is Cinebench run on a G5, with Panther (even if it is a developer version). I think we'd see a great improvement on scores then!

[daveL]

Originally posted by garymm
do you think it's possible that it's giving you the current cpu speed? I'm pretty sure OS X automatically throttles down the CPU on a laptop if it's not being used.
But if it were reporting the cpu speed incorrectly, it wouldn't surprise me. I've heard a lot of complaints about it.
Actually, I'm on wall power and have my power settings set to go full bore all the time. It's just Xbench.

[etoiles]

Originally posted by tazznb

Contact reliable 3D software makers and get identical 3D files to render for the 3D enthusiasts, etc.


well, this is exactly what cinebench is, a few standardized scenes running through the Cinema4D engine... of course, the engine is not optimized yet, but that is pretty much what you get if you are doing 3D using Cinema4D NOW...

[MOM]

Originally posted by macrumors12345
Boy...Illustrator. My friend is a developer on the Illustrator team...I dunno if that app is even optimized for the 7450 core yet, let alone the G5. Well, maybe that is an exaggeration, but suffice it to say that it is not one of Adobe's more streamlined apps. I don't even know if it takes advantage of Altivec yet.

Yes, despite the tweaking Adobe did to PS when dualies first came out on the Mac, I think the general assumption that Adobe software is well written for the Mac is FALSE. I hope I'm wrong with the latest PS plugin for the G5. I saw some comparisons of software running on dual processor machines done by CreativeMac (this was probably over a year ago) and Adobe barely got any benefit from two processors. However, Studio Artist (which makes no particular claims to dual processor optimization) took much better advantage of two processors.

I know my examples are a bit dated, but I haven't seen any reason to beleive things have change. This goes triple for Adobe video apps (duh, why would they want to help Apple on this front). Again, I hope that with the G5 that Adobe puts a little more effort into making their Mac software optimized.

[Sun Baked]

Originally posted by 1stunna
test1:

Results 124.04
System Info
Xbench Version 1.1
System Version 10.2.7
Physical RAM 256 MB
Model PowerMac7,2
Processor PowerPC 970 @ 1.60 GHz
L1 Cache 64K (instruction), 32K (data)
L2 Cache 512K @ 1600 MHz
Bus Frequency 800 MHz
Video Card GeForce FX 5200
Drive Type ST380013AS
CPU Test 90.16
GCD Recursion 72.93 2.85 Mops/sec
Floating Point Basic 210.14 747.32 Mflop/sec
AltiVec Basic 78.29 1.14 Gflop/sec
vecLib FFT 49.00 755.06 Mflop/sec
Floating Point Library 262.85 10.52 Mops/sec
Thread Test 80.49
Computation 59.27 477.51 Kops/sec, 4 threads
Lock Contention 125.37 1.57 Mlocks/sec, 4 threads
Memory Test 241.11
System 188.79
Allocate 329.91 111.23 Kalloc/sec
Fill 136.04 1082.91 MB/sec
Copy 181.52 907.60 MB/sec
Stream 333.53
Copy 330.19 1346.57 MB/sec [G5]
Scale 312.42 1297.81 MB/sec [G5]
Add 342.88 1446.91 MB/sec [G5]
Triad 351.25 1445.69 MB/sec [G5]
Quartz Graphics Test 154.81
Line 179.17 4.56 Klines/sec [50% alpha]
Rectangle 172.06 12.10 Krects/sec [50% alpha]
Circle 171.77 3.96 Kcircles/sec [50% alpha]
Bezier 147.58 1.60 Kbeziers/sec [50% alpha]
Text 120.38 1.96 Kchars/sec
OpenGL Graphics Test 178.12
Spinning Squares 178.12 124.65 frames/sec
User Interface Test 138.33
Elements 138.33 47.07 refresh/sec
Disk Test 105.63
Sequential 114.51
Uncached Write 142.09 56.56 MB/sec [4K blocks]
Uncached Write 137.22 53.56 MB/sec [256K blocks]
Uncached Read 76.73 12.15 MB/sec [4K blocks]
Uncached Read 132.03 53.35 MB/sec [256K blocks]
Random 98.02
Uncached Write 84.92 1.21 MB/sec [4K blocks]
Uncached Write 102.71 23.17 MB/sec [256K blocks]
Uncached Read 95.49 0.63 MB/sec [4K blocks]
Uncached Read 113.35 23.33 MB/sec [256K blocks]



test2

Results 128.84
System Info
Xbench Version 1.1
System Version 10.2.7
Physical RAM 256 MB
Model PowerMac7,2
Processor PowerPC 970 @ 1.60 GHz
L1 Cache 64K (instruction), 32K (data)
L2 Cache 512K @ 1600 MHz
Bus Frequency 800 MHz
Video Card GeForce FX 5200
Drive Type ST380013AS
CPU Test 115.66
GCD Recursion 72.79 2.84 Mops/sec
Floating Point Basic 212.47 755.61 Mflop/sec
AltiVec Basic 74.60 1.08 Gflop/sec
vecLib FFT 131.71 2.03 Gflop/sec
Floating Point Library 263.97 10.57 Mops/sec
Thread Test 80.64
Computation 59.35 478.12 Kops/sec, 4 threads
Lock Contention 125.77 1.58 Mlocks/sec, 4 threads
Memory Test 242.09
System 189.38
Allocate 304.24 102.58 Kalloc/sec
Fill 131.01 1042.84 MB/sec
Copy 203.21 1016.07 MB/sec
Stream 335.44
Copy 332.89 1357.60 MB/sec [G5]
Scale 314.62 1306.98 MB/sec [G5]
Add 343.47 1449.40 MB/sec [G5]
Triad 353.27 1453.97 MB/sec [G5]
Quartz Graphics Test 151.82
Line 172.21 4.38 Klines/sec [50% alpha]
Rectangle 172.08 12.11 Krects/sec [50% alpha]
Circle 172.93 3.99 Kcircles/sec [50% alpha]
Bezier 143.64 1.56 Kbeziers/sec [50% alpha]
Text 116.68 1.90 Kchars/sec
OpenGL Graphics Test 173.26
Spinning Squares 173.26 121.25 frames/sec
User Interface Test 136.51
Elements 136.51 46.45 refresh/sec
Disk Test 105.58
Sequential 114.08
Uncached Write 145.76 58.03 MB/sec [4K blocks]
Uncached Write 137.39 53.63 MB/sec [256K blocks]
Uncached Read 76.66 12.14 MB/sec [4K blocks]
Uncached Read 126.90 51.27 MB/sec [256K blocks]
Random 98.25
Uncached Write 87.25 1.25 MB/sec [4K blocks]
Uncached Write 103.36 23.31 MB/sec [256K blocks]
Uncached Read 94.82 0.63 MB/sec [4K blocks]
Uncached Read 110.76 22.80 MB/sec [256K blocks]



test3:

Results 135.21
System Info
Xbench Version 1.1
System Version 10.2.7
Physical RAM 256 MB
Model PowerMac7,2
Processor PowerPC 970 @ 1.60 GHz
L1 Cache 64K (instruction), 32K (data)
L2 Cache 512K @ 1600 MHz
Bus Frequency 800 MHz
Video Card GeForce FX 5200
Drive Type ST380013AS
CPU Test 118.69
GCD Recursion 73.95 2.89 Mops/sec
Floating Point Basic 198.78 706.93 Mflop/sec
AltiVec Basic 78.23 1.14 Gflop/sec
vecLib FFT 143.08 2.20 Gflop/sec
Floating Point Library 263.18 10.53 Mops/sec
Thread Test 81.63
Computation 60.51 487.47 Kops/sec, 4 threads
Lock Contention 125.39 1.57 Mlocks/sec, 4 threads
Memory Test 234.46
System 180.08
Allocate 297.39 100.27 Kalloc/sec
Fill 131.40 1045.97 MB/sec
Copy 175.84 879.22 MB/sec
Stream 335.90
Copy 332.45 1355.82 MB/sec [G5]
Scale 313.41 1301.93 MB/sec [G5]
Add 345.68 1458.75 MB/sec [G5]
Triad 355.03 1461.23 MB/sec [G5]
Quartz Graphics Test 153.34
Line 175.22 4.46 Klines/sec [50% alpha]
Rectangle 171.20 12.04 Krects/sec [50% alpha]
Circle 173.84 4.01 Kcircles/sec [50% alpha]
Bezier 144.02 1.56 Kbeziers/sec [50% alpha]
Text 119.57 1.95 Kchars/sec
OpenGL Graphics Test 179.16
Spinning Squares 179.16 125.37 frames/sec
User Interface Test 136.37
Elements 136.37 46.40 refresh/sec


as you can see, it does better the more times you run it.

still doesnt seem right. code works ;)

[henryblackman]

Originally posted by MOM
Yes, despite the tweaking Adobe did to PS when dualies first came out on the Mac, I think the general assumption that Adobe software is well written for the Mac is FALSE.

I think your right. PS has a *lot* of legacy code in it, which probably hasn't been changed for years. It's not multi-threaded, it was tweaked to work with Carbon on X, and generally is MUCH slower than similar applications doing the same thing.

The optimising that Adobe has done for the G4 and Altivec and the G5 will likely only effect some operations, albeit perhaps commonly used, time-consuming ones.

The user interface needs a heck of a rewrite to even make it usable if you ask me.

[XnavxeMiyyep]

Adobe hardly optomized Photoshop for OS X. My Dads 400 MHz Graphite iMac w/ 384 MB of RAM on OS 9.2 opens Photoshop 4x faster than my Powermac MDD Dual 867 w/ 768 MB of RAM.

[ultrafiel]

Sunbaked and 1Stunna, I would suggest you get some more RAM for your systems. Even if it is fast RAM, more than 256 will really help you out. OS X loves that stuff. Since you have to put it in pairs get two 256 Dimms and your set, if you can afford more even better.

[MOM]

More RAM is sure to be better for all real world things. But I would like to mention tha on AppleInsider someone ran Xbench with 1GB and 2GB or RAM and the results were similar as reported here. Actuaaly the 2GB Ram did worse. Then again, who cares about XBench.

[Sun Baked]

>ultrafiel

I was just quoting and modifying the thing with the {[ code ] .... [ / code ]} so I could read 1Stunna's test.

[a1291762]

Originally posted by 1stunna
test1:
Results 124.04
test2
Results 128.84
test3:
Results 135.21

I'm really hoping the 1.6Ghz G5 is comparable to a similarly priced PC but I don't grok benchmark numbers like this. What do the numbers *actually* mean? Are there places that give the numbers for PCs? Is it even worth comparing the benchmark on different platforms?

I just want to know if the 1.6Ghz G5 (which will cost me over $3,000 AUD) is going to be comparable to a $2,000 AUD PC (which I'm told is what I can expect to pay for a decent PC - presumably a P4 2.4-6 Ghz or similar AMD with a similar setup to the Mac). I can tolerate a certain discrepancy in the price due to the excellent software available on Mac OS X (compared to whatever I can get running under Linux/FreeBSD on the PC).

Of course, if the G5 isn't going to keep up, I'll get a PC and use my lowly iMac DV 400 for iMove/iTunes/Appleworks etc. I want to get a new Mac but here in Australia, the "Macs cost the same as PCs" argument doesn't work. One day Apple may fix the price difference here but I'm not holding my breath.

Link

[1stunna]

more ram would help, but its not my comp. its the schools.

[acj]

Originally posted by henryblackman
I think your right. PS has a *lot* of legacy code in it, which probably hasn't been changed for years. It's not multi-threaded, it was tweaked to work with Carbon on X, and generally is MUCH slower than similar applications doing the same thing.

The optimising that Adobe has done for the G4 and Altivec and the G5 will likely only effect some operations, albeit perhaps commonly used, time-consuming ones.

The user interface needs a heck of a rewrite to even make it usable if you ask me.

Photoshop it multithreaded.
You probably want photoshop elements.

[Dont Hurt Me]

since this is a benchmark topic i thought i would post my single g4 1.4 for comparison purpose for those looking at those numbers

Results- 114 cpu test 163 quartz 135 open gl 170 i ran the test only once.

[John Q Public]

I hope those scores are only because of non-optimised Cinebench and OS...CINEBENCH 2003 v1

this is the Dual 1.8 G5...
****************************************************

Tester :

Processor :
MHz :
Number of CPUs : 1
Operating System :

Graphics Card :
Resolution : <fill this out>
Color Depth : <fill this out>

****************************************************

Rendering (Single CPU): 188 CB-CPU
Rendering (Multiple CPU): --- CB-CPU


Shading (CINEMA 4D) : 227 CB-GFX
Shading (OpenGL Software Lighting) : 605 CB-GFX
Shading (OpenGL Hardware Lighting) : 936 CB-GFX

OpenGL Speedup: 4.12

****************************************************

this is mine...

CINEBENCH 2003 v1
****************************************************

Tester :

Processor : RSFastMac
MHz : 1250 (PowerMac G4 v3.3)
Number of CPUs : 2
Operating System : 10.2.6

Graphics Card : GeForce4 Ti
Resolution : 1920x1200
Color Depth : Millions

****************************************************

Rendering (Single CPU): 120 CB-CPU
Rendering (Multiple CPU): 218 CB-CPU

Multiprocessor Speedup: 1.82

Shading (CINEMA 4D) : 147 CB-GFX
Shading (OpenGL Software Lighting) : 406 CB-GFX
Shading (OpenGL Hardware Lighting) : 800 CB-GFX

OpenGL Speedup: 5.45

****************************************************
and this is my laptop (Compaq 2145)

CINEBENCH 2003 v1
****************************************************

Tester :

Processor : Compaq 2145US
MHz : 1729 (Athlon-M 2400+ Barton Core)
Number of CPUs : 1
Operating System : WindowsXP Professional

Graphics Card : ATi Mobility U1 AGP
Resolution : 1440x1050
Color Depth : 32bpp

****************************************************

Rendering (Single CPU): 218 CB-CPU
Rendering (Multiple CPU): --- CB-CPU

Shading (CINEMA 4D) : 219 CB-GFX
Shading (OpenGL Software Lighting) : 584 CB-GFX
Shading (OpenGL Hardware Lighting) : 329 CB-GFX

OpenGL Speedup: 2.67

****************************************************

like I said...I HOPE those numbers are only because of non-optimised OS and App...because I expected it to spank my G4 (by more than it did)...but to have a CPU score less than my laptop running at a slightly lower speed???...and considering laptops run slow in comparison to a desktop machine (I haven't run CineBench on any of the desktops here yet)

[stid]

Probably we need to wait for optimized tester.

I advise you all that xbanch v1.1 is not so good for testing OpenGl and Grphic performance.

I have a G4800DP with GeForce ti 4600 and I get same performance as G4800DP GeForce2Mx.

CINEBENCH 2003 v1 as the author said is not optimized for G5 at all.

The problem I see as a mac user is another... out there there are about 50% of the software that are just not optimized or bad optimized for G4. Don't expect G5 optimization so quickly in the future.

Just my opinion.

[John Q Public]

then again...there is the issue of a non-optimised OS also...yes it has some 64bit libraries...but I'd heard 10.4 SHOULD be optimised for the G5 (hopefully they do what Micro$haft did with NT4...multiple platforms on the same CD (or in this case...DVD)

...seeing that "Panther" will be dropping support for Beige G3's (Gossamer and Artemis)...it's likely "Lynx" will drop everything below "Sawtooth" (or even Digital Audio)...dunno 'til it happens I guess...

[stid]

About 2 years ago I have bought the fastest machine of the world, a PPC G4 800 DP with 1,25MB ram, Delta Audiophile 24/96 sound card, GeForce 4Ti 4600, 2 HD 120GB two monitors and some other nice things. I have spend about 6000$ for it.

It was a G4 with an altivec and a new system on it called OSX (I fantastic one after 10.2 I have to admit). I remember the first year was a frustration because
not all the software was ported to OSX, most of them are ported but not optimized. I remember PC people talk about 160 fps in the Game, I remember PC people using Cubase SX and 10 VST plug-in without any problems, people using 3D and 2d Graphic software that run as a hell, I saw PC people with flash plug-in that support their browser, I saw PC people using game coded in Shockwave and using it fast.

But I have the fastest machine of the world!... I need just to wait... they will optimize for it, we need to wait for altivec support, they will do it, I'm shure.

I'm still waiting...

But one day, my G4 is became slow and slow... most software was not optimized for it yet and the one that was, Game, Pro Audio, Graphics, are phased on the medium, I repeat "MEDIUM" PC counterpart. My G4 is not enough for it, it is too slow. Game like Medal of Honor, WWII Online get about 10-12 fps. PC user get 60-80-120, audio software drain my CPU down and I can use 1-2 VST plug-in and most of them like Moog, Live, Cubase sometimes are too slow also if I use it lonely. For most graphics software the story is the same.

I'm still waiting...

But one day, things became true!!!!!

Apple have just released a new machine. It's name is G5. It's the fastest computer of the world!!!. It will cost me about 6000$ (I talk about a complete configuration, with RAM, 2 drive, The best videocard, A sound card and some other nice things).

I will buy it... and I have just to wait... they have to port the software for It. They have to upgrade the system. They have to optimized it and recompile for 64bit architecture.

I'm still waiting... I'm still waiting... I'm still waiting...

or...

I can buy a PC. Have a stupid system (not so stupid as the old one but not nice as OSX)... With 6000$ I can buy a so fast stupid PC machine that I will play Game at 300 fps, play Ableton Live and Reason together as rewire slave in Cubase and put on it 10 VST plug-in and have my cpu used at 50% (I saw it), use 3D software, Flash, Adobe without any problems. I can buy a Matrox DV editing accelerator card for 500$ and have the driver for my system that support it, I can do all this. I was sure that software house and hardware house will support me, they reply to my mail, I will have the latest driver, the latest software all optimized for my machine. I can have support forums with developers that reply me and not just ignore me because I'm one of the xxx stupid damn Mac users that pretend Altivec optimization for software that they have paid exactly as the PC user have. I can have all this. I can have all this Now!

I will don't have the fastest machine of the world, but men... I don't know why but I have stop to waiting.

So now kill me, open a flame and burn my house... offend me and tell what you want. But I know this is exactly my story, my money, my experience. I have supported Apple for about 15 years, from the old Mac Plus... I want to believe that G5 is the solution to all that, but I can't convince myself of it, no more. My next machine will be a Stupid, unusable, handmade, IRQ setuppuble kazaa style like PC. With a stupid, carboncopymac windows XP system. But I'm stop waiting... now.

Please, convince me I'm wrong... let me spend this next 6000$ right, on Apple. Please... convince me I'm wrong.

Sorry for my bad English.

[Dont Hurt Me]

rather then spend 6 grand why dont you upgrade the cpu, your geforce 4 blows away the fx 5200 & 9600 pro, you got the memory why start over? you just need a cpu boost.

[AngryAngel]

In the reviews I've seen, the 9600 128MB on a PC beats a GF4 Ti on many tests. OK, so the Mac version is 64MB, but I wouldn't say the Ti blows away the 9600.

...but I don't think stid's post meant he was about to spend another six grand on a mac.

stid: you make very valid points. All I can say is that I am willing to pay double the price of a PC for a Mac (I just have, on my G5 1.6), just because of the OS.

[Xero]

im not sure how or when it came in, but id just like to declare that University-Wisconsin, Madison had a 1.6 g5 in their tech store when i went to buy my tablet today, hooked up to a 23" cinema display nonetheless, and i was pretty impressed to say the least. man everything about that machine is so fine tuned and nit-picky perfect! i love how the cd rom drive bay dorr slides down right under the front mesh grill. slick. and it was quite as hell.

[John Q Public]

Originally posted by stid
Please, convince me I'm wrong... let me spend this next 6000$ right, on Apple. Please... convince me I'm wrong.

Sorry for my bad English.

fear not stid...you need not spend $6000 for better performance...but a mere 10% of that...

http://eshop.macsales.com/MyOWC/Upgrades.cfm?model=157&type=Processor&TI=%23TimeFormat%28Now%28%29%2C+&shoupgrds=Show+Upgrades

the people at MacSales can help you get more power until you feel the G5 performs to your satisfaction...not to mention MUCH cheaper than "Un-Switching" to the Micro$haft World...

[stid]

Originally posted by AngryAngel

...stid: you make very valid points. All I can say is that I am willing to pay double the price of a PC for a Mac (I just have, on my G5 1.6), just because of the OS.

You know... I have think and think about it. And at the end, as the last 15 years... I will buy a mac.

I can't use a PC. It is also not very important if PC run a little bit faster. I just hope that some kind of pro good software will be optimized a little bit in the future. G5 result now show that we can use actual not optimized software at about 180-200% of their performance of G4.

I will buy a damn MAC G5. But this time I will wait for second generation one. I hope in December or Jenuary.

I know I'm crazy :)

[tazznb]

Originally posted by stid
You know... I have think and think about it. And at the end, as the last 15 years... I will buy a mac.

I can't use a PC. It is also not very important if PC run a little bit faster. I just hope that some kind of pro good software will be optimized a little bit in the future. G5 result now show that we can use actual not optimized software at about 180-200% of their performance of G4.

I will buy a damn MAC G5. But this time I will wait for second generation one. I hope in December or Jenuary.

I know I'm crazy :)

Take a look at this... not too bad in my opinion.

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

[Tim Flynn]

With virtual PC no working on a G5 and it requiring a lot of work to fix. I wonder how long (if) Microsoft will take to get it going?

We have some Orange PC or is Orange micro cards here. I can envision a x86 addon card perhaps a P4 or what might be a better fit is an AMD Opteron or Athlon 64. This could plug into one of those new fast PCI-X slots in a G5.
Nobody could argue that you don't have the fastest computer! (Actually someone will always argue)
:)

[henryblackman]

Originally posted by Tim Flynn
With virtual PC no working on a G5 and it requiring a lot of work to fix. I wonder how long (if) Microsoft will take to get it going?



Apparently about a year, and it will arrive at the same time as Office 11. I think I read that at The Register.

[daveg5]

Originally posted by stid
Probably we need to wait for optimized tester.

I advise you all that xbanch v1.1 is not so good for testing OpenGl and Grphic performance.

I have a G4800DP with GeForce ti 4600 and I get same performance as G4800DP GeForce2Mx.

CINEBENCH 2003 v1 as the author said is not optimized for G5 at all.

The problem I see as a mac user is another... out there there are about 50% of the software that are just not optimized or bad optimized for G4. Don't expect G5 optimization so quickly in the future.

Just my opinion.
barefeats reported thet cinebench dosnt use the video card, but is almost entirely cpu/memory/bus based

[daveg5]

Originally posted by Tim Flynn
With virtual PC no working on a G5 and it requiring a lot of work to fix. I wonder how long (if) Microsoft will take to get it going?

We have some Orange PC or is Orange micro cards here. I can envision a x86 addon card perhaps a P4 or what might be a better fit is an AMD Opteron or Athlon 64. This could plug into one of those new fast PCI-X slots in a G5.
Nobody could argue that you don't have the fastest computer! (Actually someone will always argue)
:) absolutely great idea if it could use the agp card for graphics aceeleration 2-3d and cost cheaper then a equivalent pc, i sy 6-900.