PDA

View Full Version : More Benchmarks with After Effects




MacRumors
Jun 26, 2003, 07:24 PM
xlr8yourmac.com (http://www.xlr8yourmac.com/) posted some user submitted benchmarks comparing a short After Effects benchmarking test on the new G5:

Here are my results for this test on the three computers I have available to me:

1 x 1.0 GHz G4 PowerBook 17" - ~30 minutes (3 min/frame)
2 x 2.66 GHz Pentium Xeon from Boxx - 11 min, 39 sec (1.2 min/frame)
2 x 2.0 GHz PowerMac G5 - 6 min, 1 sec (0.6 min/frame)

Remember, this is on pre-production machines with non G5-"optimized" code.



crazzyeddie
Jun 26, 2003, 07:35 PM
Sounds like we have a winner :D

Freg3000
Jun 26, 2003, 07:36 PM
It will be very interesting to see what the G5 can do when the apps are optimized for it. It already is must faster than the competition.

nagromme
Jun 26, 2003, 07:59 PM
Also serves as a reminder that the PBG4 is no slouch for a portable--especially if we get higher MHz soon.

MorganX
Jun 26, 2003, 08:28 PM
I think everyone, even Intel probably conceded the G5 bests all x86 currently available in FPU performance, save Itanium 2.

However most applications use integer. I'm more interested in across the board tests when the machines are available.

No doubt the G5 is the best thing to happen to Mac hardware in a long time, along with Hypertransport, AGP, and what not.

Seems to me the only issue here is Jobs claiming to be the fastest desktop ever. It was unnecessary and will probably turn out not to be true, when it's actually a worthless anectdote anyway.

Let the machine speak for itself, when it arrives.

Fender2112
Jun 26, 2003, 09:53 PM
What if Apple intentionally stretced the truth as a marketing ploy. They make a claim, "the worlds fastest personal computer". Immedately the PC folks are trying to debunk it and the Mac folks are defending it. Imagine all the free publicity Apple is getting. Every PC and Mac fanatic from here to infinity is trying to get the two cents in.

The fact is that numbers can be manipulated to prove any point you care to make. I'm glad Apple finally has something that is at the least equal to the Intel chips.

I suspect August is going to be a very busy month.

KEL9000
Jun 26, 2003, 10:15 PM
I will feel a lot better when there is a few upgrades past 2GHz and we continue to "kick ass" for a few years. I guess the investment in the state-of-the art production plant should be a good sign for the future of the PPC9XX.

I spoke with an Intel rep a few months ago and he told me they were more concerned with AMD than IBM, I wonder if they still feel that way.

daveg5
Jun 26, 2003, 11:38 PM
Originally posted by Macrumors
xlr8yourmac.com (http://www.xlr8yourmac.com/) posted some user submitted benchmarks comparing a short After Effects benchmarking test on the new G5:



Remember, this is on pre-production machines with non G5-"optimized" code.
do this test with a dual 1.42 g4 please thanks

hvfsl
Jun 27, 2003, 05:34 AM
Originally posted by daveg5
do this test with a dual 1.42 g4 please thanks

The dual 1.42 G4 Mac should be faster than the Xeon, although it would also be good to see it on a dual 3.2Ghz Xeon/P4.

MorganX
Jun 27, 2003, 07:55 AM
>>Immedately the PC folks are trying to debunk it and the Mac folks are defending it.<<

I think it's industry folks. You make a claim like that and everyone's ears perk up.

>>Imagine all the free publicity Apple is getting. Every PC and Mac fanatic from here to infinity is trying to get the two cents in.<<

They say bad publicity is better than no publicity. I think Apple is getting limited benefits due to their history of false or distorted claims of performance.

>>I suspect August is going to be a very busy month.

It will be interesting to see. 9 fans is a lot. And upgradeability is an issue. I think it's going to do well among the existing market, just don't know if it will cause it to grow any.

MorganX
Jun 27, 2003, 08:03 AM
>>I spoke with an Intel rep a few months ago and he told me they were more concerned with AMD than IBM, I wonder if they still feel that way. <<

IMO, based on the test results they still do. With AMD, for about $500 you can swap your P4 motherboard and move to Opteron. The motherboard business is huge. Custom PCs and upgrades, huge.

I believe Intel can put a low cost Itanium out anytime they get ready. They don't don't need to, yet. 32bit processors are still leading the pack, and can get quite a bit better with just a little more cache, and actually lower clock speed. Pentiums use little cache compared to G4s. Intel can probably keep close to the same margins and make 1MB L2 standard across the board tomorrow.

I don't think MS is worried much about Panther either. All the new features look to be catching up to XP to me. That doesn't mean it's not good news for those of us who own Macs, I just don't see much that would garner much new attention, or anything particularly innovative. Just routine evolution of a product line.

I wish there was more about the adjustable displays rumor. With 9 fans 970 iMacs are probably not a possibility. That means my next PC will be a conventional one with a sharp LCD. The adjustable monitor is very enticing.

ryan
Jun 27, 2003, 08:57 AM
Originally posted by hvfsl
The dual 1.42 G4 Mac should be faster than the Xeon, although it would also be good to see it on a dual 3.2Ghz Xeon/P4.
FWIW: There are no 3.2GHz Xeon's only P4s, and you can't have dual P4s.

cuneglasus
Jun 27, 2003, 12:39 PM
I notice several people seem to have a misunderstanding about Itanium. It is not x86,it is a type of risc architecture, and will not come to a desktop while windows and x86 apps are what defines the windows world.Maybe one day in the future x86 will finally be gone but by then the Itanium will be old news.Intel has no plans for a desktop 64 bit processor anytime in the forseeable future.

macrumors12345
Jun 27, 2003, 12:56 PM
Originally posted by MorganX
However most applications use integer. I'm more interested in across the board tests when the machines are available.

I have no doubt that the P4 may be able to match the 970 in integer math (although only because it is moving to the 90 nm process months before the 970 is - once they're both on the 90 nm process the P4 may start to fall behind even on integer math). I also have no doubt that many applications primarily use integer calculations.

However, I seriously doubt that faster scalar integer calculations are of much interest to most people when comparing processors like the G5 and the P4 and the Athlon XP/Athlon-64. I personally do not care if my word processor or my web browser or my e-mail program or my chat client runs 10% faster on a P4 than on a G5 or vice versa - they're both impressive machines, and either one is going to be MORE than fast enough to handle everyday apps like the ones you are referring to. What people do care about performance for are things like video editing and encoding, 3D rendering, audio editing and encoding, graphics work, numerical simulations, statistical packages (my own area - I look forward to seeing Stata perform on a G5!), and, oh yes, games. What do almost all of these types of applications, and others like them, have in common? They all make very heavy use of floating point and/or vector calculations - scalar integer performance is generally of second order concern.

Don't get me wrong - it is good to have decent scalar performance so that your machine can run the latest bloated version of Office or IE/Mozilla 17.0 and still feel "snappy." And unlike the G4, the PPC 970 will clearly be very competitive with the P4 in the area of scalar integer performance, which should go a long way towards ending the perception among many that OS X feels too slow in day to day use as compared to Windows on the latest hardware. However, when you start talking about compute intensive tasks, i.e. things that you actually want to benchmark for rather than simply making a subjective judgement of "this computer does/does not feel fast enough to browse the internet or use Office", then scalar floating point and vector performance are what you are really interested in (note that the G4's scalar floating point performance is terrible when compared to the P4 or Athlon, which is a primary reason as to why it falls far behind in many benchmarks). And the great news is that the PPC 970 is an absolute monster in these two areas (which is not to take away from the fact that it can also give any other processor out there a run for its money in scalar integer math, of course, but it really goes into overdrive in floating point and vector operations).

macrumors12345
Jun 27, 2003, 01:05 PM
Originally posted by KEL9000
I spoke with an Intel rep a few months ago and he told me they were more concerned with AMD than IBM, I wonder if they still feel that way.

I seriously doubt it (that they view IBM with greater concern than AMD), at least not the desktop space (I'm sure that in the high end server space the Power4 and Power5 give them fits on a daily basis, but that is a different story). Intel's bread and butter is selling processors for machines that run Windows. The PPC 970 does not currently and likely never will run Windows. The AMD Athlon series does run Windows, and hence it is a direct competitor.

Yes, the G5 could (and hopefully will) encourage more switchers. But even if Apple doubled its marketshare (incredibly unlikely in the next couple years), it would still only represent a 5% reduction in PC sales, which is pretty insignificant compared to the amount of sales that Intel could potentially lose to AMD, especially if Lateron (aka Athlon-64) takes off.

The bottom line is that code base compatibility always trumps performance when it comes to market success (a lesson which Intel is learning the hard way with Itanium - not that it is really that competitive with the Power series in performance either, though). In order for PPC 970 to really start stealing lots of sales from Intel, it would probably have to be 3 times, maybe 5 times, possibly even 10 times faster than the Pentium 4. And as great a chip as it is, that is not going to happen anytime in the forseeable future.

macboro
Jun 27, 2003, 01:10 PM
I run the AE test on my Dual 1,42=5min!!! for the first 10frames and of course 50min. for the 100frames...

MorganX
Jun 27, 2003, 01:37 PM
Originally posted by cuneglasus
I notice several people seem to have a misunderstanding about Itanium. It is not x86,it is a type of risc architecture

Pentium IV's are also a type of RISC/CISC architecture. It is not x86 compatible natively which is why it's 32-bit performance is not up to Opteron standards.

And of course Intel has 64-bit desktop plans. It's just not on their current roadmap because it doesn't need to be. Unreal 64-bit may change that if they can show significant performance gain over current 32-bit processors.

Right now there's nothing 64-bit can offer an x86 desktop that it needs. Apple desktops went 64-bit becuase that's the only available PPC that can keep up with Intel.

MorganX
Jun 27, 2003, 01:41 PM
Originally posted by macrumors12345
(I'm sure that in the high end server space the Power4 and Power5 give them fits on a daily basis, but that is a different story).

IBM is selling new Itanium server and working on new Blade servers with Intel.

cuneglasus
Jun 27, 2003, 02:11 PM
Originally posted by MorganX
Pentium IV's are also a type of RISC/CISC architecture. It is not x86 compatible natively which is why it's 32-bit performance is not up to Opteron standards.

And of course Intel has 64-bit desktop plans. It's just not on their current roadmap because it doesn't need to be. Unreal 64-bit may change that if they can show significant performance gain over current 32-bit processors.

Right now there's nothing 64-bit can offer an x86 desktop that it needs. Apple desktops went 64-bit becuase that's the only available PPC that can keep up with Intel.

I think you are missing my point.All current x86 processors including the p3,p4,athalon and opteron/amd64 have risc like backends with a front end x86 decoder that breaks the x86 instructions into smaller fixed length micro ops for execution.The p4 is not unique in this.It would be inacurate to say they are not natively x86 isa processors.

The itanium is not x86 at all it is a risc architecture called EPIC.It has no x86 front end like these other chips.(and no you couldnt just slap one on) The end of all this is that Itanium could not run current windows apps except through very slow emulation,so no matter how much cheaper it becomes it will never be in a desktop.Intel plans to stick with the p4 for desktops with some minor changes to the level 1 caches when they move to the smaller .9 process maybe at the end of the year.

Be cautious to about baseing performance expectations on spec200 scores.The benchmark is infamous for understating the real performance of the ppc.It even underreports the performance of athalon and p3 verses the p4.I even saw a link off of tomshardware that was talking about how such synthetic benchmarks have been criticised for giveing inaccurate results.

rjstanford
Jun 27, 2003, 02:50 PM
Originally posted by macrumors12345
However, I seriously doubt that faster scalar integer calculations are of much interest to most people when comparing processors like the G5 and the P4 and the Athlon XP/Athlon-64.
...
However, when you start talking about compute intensive tasks, i.e. things that you actually want to benchmark for rather than simply making a subjective judgement of "this computer does/does not feel fast enough to browse the internet or use Office", then scalar floating point and vector performance are what you are really interested in One notable exception -- database processing tends to be heavily dependant on integer performance. Assuming that you have a decent I/O subsystem, that is. That's the main reason that I (for example) get interested in SpecINT scores over floating point claims. Now, a dual 2ghz G5 in an XServe chassis ... that's getting fast enough to be really interesting to me. But I'll never turn down better integer numbers.

-Richard

MorganX
Jun 27, 2003, 02:50 PM
Originally posted by cuneglasus
The p4 is not unique in this.It would be inacurate to say they are not natively x86 isa processors.

My bad. Run on sentence. I meant Pentium IV's are CISC/RISC. Itaniums do not run x86 natively.

macrumors12345
Jun 27, 2003, 04:50 PM
Originally posted by MorganX
IBM is selling new Itanium server and working on new Blade servers with Intel.

Uh, yes. It is also true that IBM sells laptops with Intel processors. And Dell sells computers with Intel processors. And Apple sells computers with both Motorola and IBM processors. And Fujitsu sells computers with Intel and AMD processors.

All of these things are true, but none of them are very relevant to the original point being made, which is that the Power4 (and its successor, the Power5) are leading the high end server market both in terms of performance and shipments. SPARC would be second (in terms of shipments, but not performance), and Itanic has 0% market share (its performance is not bad, but it can't beat Power as a server chip, and there's little software for it, so it's no wonder that nobody is buying).

e-coli
Jun 27, 2003, 05:04 PM
That's what I've been waiting to see. Thank god, we've finally got a useable computer for broadcast animation.

Hallelujah.

macrumors12345
Jun 27, 2003, 05:07 PM
Originally posted by macboro
I run the AE test on my Dual 1,42=5min!!! for the first 10frames and of course 50min. for the 100frames...

So your Dual 1.42 is also 6 times faster than a 1 Ghz PB, even though the clock speed is only 40% higher and After Effects doesn't even make very good use of dual processors in general?? At the very least, you are clearly not running the same test that was being run on the Powerbook and the Xeon (I kind of doubt that the Dual 1.42 could match the Xeon, let alone beat it by more than 2x).

MorganX
Jun 27, 2003, 09:22 PM
Originally posted by macrumors12345
All of these things are true, but none of them are very relevant to the original point being made, which is that the Power4 (and its successor, the Power5) are leading the high end server market both in terms of performance and shipments. SPARC would be second (in terms of shipments, but not performance), and Itanic has 0% market share (its performance is not bad, but it can't beat Power as a server chip, and there's little software for it, so it's no wonder that nobody is buying).

Despite great resistance, even IBM cannot ignore Itaniums growing momentum. With IBM's new Itanium servers and blades in the works, how long do you think it will remain at 0%. Xeons started with 0% of the high end server market. Then they started beating the competition at TPC. Then IBM, HP, Dell adopted them. Similar story happening with Itanium.

Intel based supercomputers more than doubled this year on the top 500 with an Itanium 2 system at #8. And the next revision is on the way.

Out of one corner of the mouth Power4, 970, out of the other, Itanium. IBM isn't stupid.

As for software, you have 64-bit Windows and Linux, SQL, Websphere and scientific software optimized or being optimized for Itanium, even ATI has Itanium optimized drivers. Anything else you can write or have written.

Like I said, G5 is great for Macintosh, but 32-bit x86 and Itanium have huge price/performance growth potential. Intel would be idiots to go 64-bit at the desktop right now. What would they gain?

Tongue in cheek: Don't be so hard on the x86 architecture, it made the mac what it is today. (AGP, Hypertransport, DDR, PCI, IDE, SATA, USB, etc.)

g3ski
Jun 28, 2003, 04:18 AM
G5, sucka.

Kickin some butt in real world apps.

for all of you who are talkin inter vs. floating point....well web surfing can be done on an imac 233....

FCP, Maya, After Affects, mathmatica, DVD studio pro, etc need and crave horse power. Seems that the G5 has it.

I agree with previous posters: let's see 3Ghz Xeon and 1.4G4/DP

g21
Jun 28, 2003, 05:37 AM
I just ran the test (properly!) on a 1.42 DP G4.

Over the first 10 frames it averaged 65 secs per frame, as compared to the G5s 36 secs. That makes the G5 about 80% faster. The next clock speed revision will certainly be at least twice as fast, the 3Ghz machines within a year will be getting on for three times as fast.

When comparing this to other processors, bear in mind that AE is not especially well optimised for dual processor machines, but this is certainly a good test for people doing real world video work on macs.

icetraxxg5
Jun 28, 2003, 06:32 PM
Originally posted by g21
I just ran the test (properly!) on a 1.42 DP G4.

Over the first 10 frames it averaged 65 secs per frame, as compared to the G5s 36 secs. That makes the G5 about 80% faster. The next clock speed revision will certainly be at least twice as fast, the 3Ghz machines within a year will be getting on for three times as fast.

When comparing this to other processors, bear in mind that AE is not especially well optimised for dual processor machines, but this is certainly a good test for people doing real world video work on macs.


And remember that After Effects isn't 64bit optimized!

e-coli
Jun 28, 2003, 06:55 PM
Did you run this test with multithreading on or off?

That's where the other chips cream the PPC. Most of the reason why AE has been so much faster on the PC.

SumDumGuy
Jun 29, 2003, 01:28 PM
I wonder if the G5 times would be even faster if the test was performed like this (http://www.digitalpostproduction.com/2003/04_apr/tutorials/aerender030408.htm).

I posted about this method in the forums a while back and really got great results. LINK (http://forums.macrumors.com/showthread.php?s=&threadid=27783)

Gregory
Jun 29, 2003, 02:22 PM
Are you sure those after effects render times are corrected for the nightfire project, because I get 11 mins and 17 sec. on a Mac G4 2x 1.25 Machine.

Are the 2x 2.66 BOXX render times correct. . . @ 11 mins and 39 sec.

Pete_Hoover
Jun 30, 2003, 05:20 PM
This is really good news. The G5 is doing especially good for a pre-production machine. I can't wait to get my hands on a full-production model.

g21
Jul 3, 2003, 04:12 AM
Thanks very much SumDumGuy for the AE tip, I had no idea! Really useful :)

Mav451
Jul 10, 2003, 01:35 AM
Originally posted by MorganX
>>Immedately the PC folks are trying to debunk it and the Mac folks are defending it.<<

I think it's industry folks. You make a claim like that and everyone's ears perk up.

>>Imagine all the free publicity Apple is getting. Every PC and Mac fanatic from here to infinity is trying to get the two cents in.<<

They say bad publicity is better than no publicity. I think Apple is getting limited benefits due to their history of false or distorted claims of performance.

>>I suspect August is going to be a very busy month.

It will be interesting to see. 9 fans is a lot. And upgradeability is an issue. I think it's going to do well among the existing market, just don't know if it will cause it to grow any.

you're definitely right on the "bad publicity" vs. none at all. One of my friends (who is a Mac owner) was talking bout this to me for the past month--probably the only reason i even came to read and learn bout the G5 in the first place.

And with the benchmark confusion of last month, what better way to let the entire enthusiast community hear about the G5? If you recall, the news of these "benchmarks" were all over the place, both AMD and Intel enthusiast sites had this as front page news.

The fact is that people are looking/reading about it. Heck, back when the G4 was released, I didn't give a damn about it. They said it was "pentium-crushing" but that didn't matter at all to me because i was already using an AMD chip at the time. The difference at that time was that the majority of enthusiast sites didn't pay attention at all to the claim b/c it was in no way close to the G5's "Fastest desktop computer" statement.