Forums | MacLife

Czachorski · 2006-01-13 10:22 pm

I went up to the Apple store tonight to buy the iLife '06, and while I was there, I thought I would run some benchmarks on the Quad G5 PowerMac. While doing that, an Apple guy started talking to me and mentioned that they had 3 iMac intel machines that they just set up today. Perfect - I was running video benchmarks anyway, I would just go over and run them on the new iMac. Here are the results (including some tests I did for a Mac Mini article a few months back):

Render 10 Second Ken Burn's Effect
15 seconds, Mac Mini G4 - 1.25 GHz 512 MB RAM
14 seconds, PowerMac G4 - dual 867 MHz 1.5 GB RAM
14 seconds, iMac G5 - 1.8 GHz 256 MB RAM
9 seconds, PowerMac G5 dual 2.0 GHz 512 MB RAM
6 seconds, PowerMac G5 QUAD 2.5 GHz 2.5 GB RAM (processor never above 25%)
13 seconds, iMac Intel Duo, 2.0 Ghz, 512 MB RAM (5 second delay/stuttor on the "import" of the photo)

Render Six 2 Second Cross Dissolve Simultaneously
60 seconds, Mac Mini G4 - 1.25 GHz 512 MB RAM
57 seconds, PowerMac G4 - dual 867 MHz 1.5 GB RAM
31 seconds, iMac G5 - 1.8 GHz 256 MB RAM
20 seconds, PowerMac G5 dual 2.0 GHz 512 MB RAM
20 seconds, PowerMac G5 QUAD 2.5 GHz 2.5 GB RAM (processor never above 25%)
24 seconds, iMac Intel Duo, 2.0 Ghz, 512 MB RAM

Export 1 Minute of Video to QuickTime using for CD-ROM Setting (H264)
70 seconds, Mac Mini G4 - 1.25 GHz 512 MB RAM
53 seconds, PowerMac G4 - dual 867 MHz 1.5 GB RAM
35 seconds, iMac G5 - 1.8 GHz 256 MB RAM
22 seconds, PowerMac G5 dual 2.0 Ghz 512 MB RAM
20 seconds, PowerMac G5 QUAD 2.5 GHz 2.5 GB RAM (processor never above 25%)
94 seconds, iMac Intel Duo, 2.0 Ghz, 512 MB RAM

Create Disc Image in iDVD using Travel Cards theme and 10 Minutes of Video
28 minutes, Mac Mini G4 - 1.25 GHz 512 MB RAM
25 minutes, PowerMac G4 - dual 867 MHz 1.5 GB RAM
12 minutes, iMac G5 - 1.8 GHz 256 MB RAM
9 minutes, PowerMac G5 dual 2.0 GHz 512 MB RAM
5.1 minutes, PowerMac G5 QUAD 2.5 GHz 2.5 GB RAM (processor never above 35%)
10 minutes, iMac Intel Duo, 2.0 Ghz, 512 MB RAM

Overall, I was disappointed with the iMac duo performance. When I first walked up to the machine and started using it, it felt sluggish right away. Of course, I had just walked from the quad, but still, the apps lagged for a second when they were launched, dashboard stuttered when I first hit it, and I got the beach ball a few times right away when launching iMovie, the clock and activty monitor. Actually, it felt clunky and more like being on a PC the way that it seemed to hesitate for just a moment when doing things. But I was the most disappointed with the render speeds, which were only slightly faster than single processor iMac G5 1.8 Ghz on most test, and half the speed on anything with h264. The h264 export result was so suprising that I tried it a couple of different times with a few different setting, all with the same result - all very slow - about half the speed of the iMac G5. Could this be a case of much of the code still not being optimized for the intel?

Another thing that is still baffling me is the processing CPU % utilized on the PowerMac Quad G5 during rendering. It is still not getting up above 25% to 35% when rendering in iMovie or iDVD. When I exported from iMovie to h264, it took 59 seconds for a 1 minute clip (1000 kbit/s, 720 x 480, 30 fps), and the processor went up to 55%. Of course, all of these test are iMovie and iDVD. I did do a test exporting from Final Cut on the quad, and it was actually slower than iMovie. It took 90 seconds with the processor running at 40%. I guess that I am still baffled what the benefit is of having all the quad power processing sitting there if it is not going to get used the most when you need it - during heavy duty rendering. For that performance, it seems I would be much better of just getting the dual G5 PowerMac, because it gives similar rendering performance as the Quad, but at a far less cost!

mo' ron · 2006-01-13 10:38 pm

I bet altivec is responsible for the Macs being so fast. Video is what it's best at. RIP, altivec.

caoimhin · 2006-01-13 10:47 pm

Question: Did the new iMac Duo have the newest iLife apps running on it? If so, the results really suck! cry

mo' ron · 2006-01-13 10:52 pm

Those results aren't really too bad. Just not as great as people were expecting.

Czachorski · 2006-01-13 11:05 pm

caoimhin wrote:
Question: Did the new iMac Duo have the newest iLife apps running on it? If so, the results really suck!

Yes, it was iLife '06.

Mo' - I agree they are not too bad. But they are not 4x faster either. Not even 2x faster. More like maybe a little faster than the iMac G5 as some things, a much slower at others (h264), which I consider to be disappointing considering this is supposed to be a dual core chip and the iMac G5 was single core.

mo' ron · 2006-01-13 11:12 pm

Jobs "2x faster" comment was based on Spec benchmarks (which is a synthetic benchmark-- that means it is an application that has no purpose, other than to run a set of algorithms to test a CPUs speeds), which is the same Spec benchmarks Mac users use to bash when PC users would say "hey, our PCs our 2x faster than your Macs". Mac users bashed them because they didn't model real-world performance, and have no meaning to most people (if you are building a cluster, and you know your application uses similar algorithms to one of Spec's tests, then it has meaning to you, otherwise, it is meaningless).

Ars had a very extensive CPU architecture writeup which noted that the Pentium-Ms (which are very similar to the Yonahs) are similar to the G5s architecturally. I expect benchmarks to hold this up, except in cases where Altivec plays a role (and i'm fairly certain that Quicktime uses altivec).

Mr. T · 2006-01-13 11:17 pm

If you happen to go back to the Apple store, go to Activity Monitor and see if ATSServer is still PowerPC. It probably wouldn't have too much to do with the benchmarks, I'm just curious.

About the benchmarks, I think ppl should note that Apple released these apps very much ahead of schedule - apps that were heavily optimized for the G5 and AltiVec. Apple's main focus atm is on quantity, just getting everything to run natively. Once they get everything native, they can go back and make it "sing" as Apple likes to put it. I'm not an Intel fanboy (dispite my avatar) and I'm not defending SSE3 (it's not as good as AltiVec), I'm just saying don't judge a book by it's cover. hmm

Last edited by Mr. T (2006-01-13 11:39 pm)

Czachorski · 2006-01-13 11:44 pm

Mr. T wrote:
If you happen to go back to the Apple store, go to Activity Monitor and see if ATSServer is still PowerPC. It probably wouldn't have too much to do with the benchmarks, I'm just curious.

About the benchmarks, I think ppl should note that Apple released these apps very much ahead of schedule - apps that were heavily optimized for the G5 and AltiVec. Apple's main focus atm is on quantity, just getting everything to run natively. Once they get everything native, they can go back and make it "sing" as Apple likes to put it. I'm not an Intel fanboy (dispite my avatar) and I'm not defending SSE3 (it's not as good as AltiVec), I'm just saying don't judge a book by it's cover.

That makes a lot of sense. Unfortunately, it is also a good reason to hold off buying the 1st revision of these. I am thinking that the quad G5 is going to be better for my video needs anyway - if only I can figure out why the encoding is not cranking the processors up.

Mr. T · 2006-01-14 12:03 am

Absolutely. If vid is your thing, and you need it now, go PM. None of the Pro apps are even native yet. If you don't mind waiting, then wait for the pro stuff to go unibin, and by then the iMacs might get revised again (they'll probably get slimer and maybe a bit faster). However, I really, really wouldn't buy a G4 under any circumstance.

pcguy · 2006-01-14 12:29 am

Just looking at those number, take into consideration of slight difference in cpu speed between the mini and latest pb, I think the new MacBook will absolutely KILL the current generation G4 pb in everything even at this early stage of sw development.

Render Six 2 Second Cross Dissolve Simultaneously
60 seconds, Mac Mini G4 - 1.25 GHz 512 MB RAM
24 seconds, iMac Intel Duo, 2.0 Ghz, 512 MB RAM

Create Disc Image in iDVD using Travel Cards theme and 10 Minutes of Video
28 minutes, Mac Mini G4 - 1.25 GHz 512 MB RAM
10 minutes, iMac Intel Duo, 2.0 Ghz, 512 MB RAM

lord funk · 2006-01-14 12:42 am

Hey Czachorski - thanks for posting your results. smile

I think the comment that worries me the most is the sluggish one. I think the iMac G5 finally satisfied me as far as snappiness went. Now we can only hope that opimization is well on its way.

mo' ron · 2006-01-14 12:49 am

pcguy wrote:
Just looking at those number, take into consideration of slight difference in cpu speed between the mini and latest pb, I think the new MacBook will absolutely KILL the current generation G4 pb in everything even at this early stage of sw development.

Render Six 2 Second Cross Dissolve Simultaneously
60 seconds, Mac Mini G4 - 1.25 GHz 512 MB RAM
24 seconds, iMac Intel Duo, 2.0 Ghz, 512 MB RAM

Create Disc Image in iDVD using Travel Cards theme and 10 Minutes of Video
28 minutes, Mac Mini G4 - 1.25 GHz 512 MB RAM
10 minutes, iMac Intel Duo, 2.0 Ghz, 512 MB RAM

Both of those times are in-line with just the clockspeed differences between the G4 and Core Duo. This implies that the Core Duo isn't faster/clock than the G4 (this is assuming about a 1.8 speedup for the dual cores, which isn't too realistic).

This doesn't diminish the fact that they are in fact faster, but it also doesn't speak too well of the Core Duo's supposed "better" architecture (at the least, I would have thought the faster bus and RAM would add to the clockspeed advantage).

pcguy · 2006-01-14 12:56 am

Just as dual processor, the difference in speed is not just cpu clockx2, but rather the ability of the cpu to do multi tread/program at same time without the typical bottleneck/hangup that a single processor/single core encounter.

jayster · 2006-01-14 1:41 am

This may explain the performance on the new Intel macs with iLife

My first mac was a mini and we were instantly hooked on iMovie and iDVD but we found the performance a little sluggish, so we added 1 gig of ram and it helped a little but still was not enough so we went back to the mac store and tried out a power mac, we found that it wasn't all that much faster then the mini when encoding in iDVD. We brought our own movie to test and found that it was only about 15% faster then our little mini (dual 1.8) and cost a lot more. We also found that the G5 imac and the power mac were so close it was hard to tell a difference. We enquired about this a was told we would need to upgrade to the pro apps to really take advantage of the speed of the power mac. The Mac Genius informed us that the iLife apps simply don't take advantage of a dual CPU system or any optimization for the G5 and simply it wasn't worth upgrading if we just wanted those to apps to run better.

This is why the quad G5 never went above 25% it was only using one core.
I think we will need to wait and see how the pro apps stack up once optimized for the Intel chips.

I could be way off on this but I'm only going off of what I was told at the apple store.

akb825 · 2006-01-14 3:13 am

As you mentioned, iLife isn't all that great at handling multiple processors. You should try something that does.

This may seem like self-promoting, but my fractal program would be a very good test. It uses nearly 100% of the CPUs, it uses 4 threads to take advantage of the quad, and is a universal binary. (I haven't had a chance to test that on the Mactels for obvious reasons, but it *should* work) Assuming you can download stuff on their demo machines, you should be able to run it. I also have some formulae that you can run. I think the longest one to render would be mandelbrot variation zoomed. (for reference, it takes 40 seconds to render on my dual 2.7 GHz G5 with 1.5 GB RAM) The link is in my sig.

avkills · 2006-01-14 4:05 am

Hey Czach,

Did you do the FCP encodings on the Quad with Compressor or just with Quicktime? Compressor is a ton faster and you get better results also.

The performance does not surprise me all that much. I pretty much knew Jobs was lying his ass off. I bet if and when Apple optimizes, that the iMac Duo maybe performs 70-80% faster than the same clocked iMac G5.

The MacBook Pro on the other hand, will probably be 2-3x faster than the G4 PB, judging from your results.

Of course you did pick video encoding, which the Intel chips have always sucked ass at, so some more results are needed to fully quatitfy the situation. big_smile

-mark

avkills · 2006-01-14 4:18 am

Another question Czach, did the iMac Duo pretty much feel the same as the others... mousing and what not?

-mark

Dlwilson · 2006-01-14 10:34 am

Correct me if Im wrong but isn't OSX 64 bit native? And if so, wouldn't running on a 32 DC processor hinder performance by OSX having to emulate 32bit? PERHAPS cause the hiccups themselves?

Just Curious about your guys take on this...

dustin

pcguy · 2006-01-14 10:47 am

I do believe app lagged when launched have "everything" to do with hd and hd setting/free memory and not so much on cpu performance, unless you are talking about huge program which I am not sure about.

toadkiller · 2006-01-14 10:56 am

512 RAM is pretty weak.

dcalfine · 2006-01-14 10:58 am

The movie export test sucked. I don't think there was any doubt in anyone's mind that the PowerPC 970 was inferior to the Core duo, but I had no idea that it would be this bad. The things a dual core, for god's sake.

avkills · 2006-01-14 11:15 am

I had doubt... because I can somewhat understand the concept of overall system design...overall Intel is weaker compared to the PowerPC 970.

If you want my 2� I think Apple should keep universal binaries for a long long time. Consumers, Laptops and servers = Intel. Pro Towers = PowerPC. shrug

Of course who knows how much Apple has optimized anything for Intel...just getting it to compile and run is one thing...optimizing is another.

-mark

Shadowless · 2006-01-14 11:16 am

akb825 wrote:
As you mentioned, iLife isn't all that great at handling multiple processors. You should try something that does.

This may seem like self-promoting, but my fractal program would be a very good test. It uses nearly 100% of the CPUs, it uses 4 threads to take advantage of the quad, and is a universal binary. (I haven't had a chance to test that on the Mactels for obvious reasons, but it *should* work) Assuming you can download stuff on their demo machines, you should be able to run it. I also have some formulae that you can run. I think the longest one to render would be mandelbrot variation zoomed. (for reference, it takes 40 seconds to render on my dual 2.7 GHz G5 with 1.5 GB RAM) The link is in my sig.

I agree with using Fractographer as a benchmark. I've played around with it on my PB while looking at Activity Monitor and it splits itself up well. Should be interesting to see what kind of results you would get on those machines using it.

MacBoy4139 · 2006-01-14 11:19 am

Dlwilson wrote:
Correct me if Im wrong but isn't OSX 64 bit native? And if so, wouldn't running on a 32 DC processor hinder performance by OSX having to emulate 32bit? PERHAPS cause the hiccups themselves?

Just Curious about your guys take on this...

dustin

OS X provides full support for 64 bit computing, but it doesn't require it.

Thanks to Czach for posting the results - I was curious. I would never do anything without at least 1.5 GB of RAM (what I have now, I still find some issues here and there).

This just further supports my beliefs:

1. I'm waiting for Merom. Yonah may be great for now, but even Intel knows it isn't going to make it far through this year alone.

2. so many years of PPC optimization have got to be worth something (agreeing with the others on that point)

dcalfine · 2006-01-14 11:31 am

avkills wrote:
If you want my 2� I think Apple should keep universal binaries for a long long time. Consumers, Laptops and servers = Intel. Pro Towers = PowerPC.

I agree entirely. Of course, the best thing imaginable is if we had AMD in the game too for gaming Macs. But other than that, Apple wouldn't have much use for them. Maybe higher-end laptops.

Forums | MacLife

#1 2006-01-13 10:22 pm

Intel iMac Benchmarks - Slow With Encoding

#2 2006-01-13 10:38 pm

Re: Intel iMac Benchmarks - Slow With Encoding

#3 2006-01-13 10:47 pm

Re: Intel iMac Benchmarks - Slow With Encoding

#4 2006-01-13 10:52 pm

Re: Intel iMac Benchmarks - Slow With Encoding

#5 2006-01-13 11:05 pm

Re: Intel iMac Benchmarks - Slow With Encoding

caoimhin wrote:

#6 2006-01-13 11:12 pm

Re: Intel iMac Benchmarks - Slow With Encoding

#7 2006-01-13 11:17 pm

Re: Intel iMac Benchmarks - Slow With Encoding

#8 2006-01-13 11:44 pm

Re: Intel iMac Benchmarks - Slow With Encoding

Mr. T wrote:

#9 2006-01-14 12:03 am

Re: Intel iMac Benchmarks - Slow With Encoding

#10 2006-01-14 12:29 am

Re: Intel iMac Benchmarks - Slow With Encoding

#11 2006-01-14 12:42 am

Re: Intel iMac Benchmarks - Slow With Encoding

#12 2006-01-14 12:49 am

Re: Intel iMac Benchmarks - Slow With Encoding

pcguy wrote:

#13 2006-01-14 12:56 am

Re: Intel iMac Benchmarks - Slow With Encoding

#14 2006-01-14 1:41 am

Re: Intel iMac Benchmarks - Slow With Encoding

#15 2006-01-14 3:13 am

Re: Intel iMac Benchmarks - Slow With Encoding

#16 2006-01-14 4:05 am

Re: Intel iMac Benchmarks - Slow With Encoding

#17 2006-01-14 4:18 am

Re: Intel iMac Benchmarks - Slow With Encoding

#18 2006-01-14 10:34 am

Re: Intel iMac Benchmarks - Slow With Encoding

#19 2006-01-14 10:47 am

Re: Intel iMac Benchmarks - Slow With Encoding

#20 2006-01-14 10:56 am

Re: Intel iMac Benchmarks - Slow With Encoding

#21 2006-01-14 10:58 am

Re: Intel iMac Benchmarks - Slow With Encoding

#22 2006-01-14 11:15 am

Re: Intel iMac Benchmarks - Slow With Encoding

#23 2006-01-14 11:16 am

Re: Intel iMac Benchmarks - Slow With Encoding

akb825 wrote:

#24 2006-01-14 11:19 am

Re: Intel iMac Benchmarks - Slow With Encoding

Dlwilson wrote:

#25 2006-01-14 11:31 am

Re: Intel iMac Benchmarks - Slow With Encoding

avkills wrote:

Board footer