3D Render Times--G5 1.6 Ghz-vs- G4 & PC's

[SpaceMusic_Guy]

OK folks, I went into Micro Center (which finally has a G5 1.6 Ghz tower on display) and did some 3D Renders of a Bryce 5 still image. Over the past 4 months, I've rendered the exact same image on a number of PC systems and G4 systems as well. I have all the render times written down and documented.

Although Bryce 5 is not a high-end 3D app., it's still a pretty good indicator of Raw CPU speed with NO optimizations for any specific CPU (PC or Mac). So, here is a comparison of the G5 1.6 Ghz compared to other systems rendering the same Bryce 5 image:

G5 1.6 Ghz -- 1:21

2.66 Ghz Pentium 4 -- 1:22

2.8 Ghz Pentium 4 -- 1:13

3.06 Ghz Pentium 4 -- 1:11

Athlon XP 2400+ -- 1:09

1 Ghz G4 (single) -- 2:27

Dual 1.25 G4 (1 processor used) -- 1:58

Dual 1.42 G4 (1 processor used) -- 1:43

It will be interesting to time the 1.8 and Dual 2.0 as well when those units become available. So far, the G5 it is not quite as fast as I thought they would be, although an improvement over the G4 for sure. However, the G5 1.6 Ghz was only 22 seconds faster than a single G4 1.42 on the render test.

The Athlon XP 2400+ processor seems to kick some butt on all of the processors tested so far (haven't had a chance to test faster Athlons at this time).

Well, any comments or observations on this people ??

 

[themadchemist]

Re: 3D Render Times--G5 1.6 Ghz-vs- G4 & PC's

Originally posted by SpaceMusic_Guy
OK folks, I went into Micro Center (which finally has a G5 1.6 Ghz tower on display) and did some 3D Renders of a Bryce 5 still image. Over the past 4 months, I've rendered the exact same image on a number of PC systems and G4 systems as well. I have all the render times written down and documented.

Although Bryce 5 is not a high-end 3D app., it's still a pretty good indicator of Raw CPU speed with NO optimizations for any specific CPU (PC or Mac). So, here is a comparison of the G5 1.6 Ghz compared to other systems rendering the same Bryce 5 image:

G5 1.6 Ghz -- 1:21

2.66 Ghz Pentium 4 -- 1:22

2.8 Ghz Pentium 4 -- 1:13

3.06 Ghz Pentium 4 -- 1:11

Athlon XP 2400+ -- 1:09

1 Ghz G4 (single) -- 2:27

Dual 1.25 G4 (1 processor used) -- 1:58

Dual 1.42 G4 (1 processor used) -- 1:43

It will be interesting to time the 1.8 and Dual 2.0 as well when those units become available. So far, the G5 it is not quite as fast as I thought they would be, although an improvement over the G4 for sure. However, the G5 1.6 Ghz was only 22 seconds faster than a single G4 1.42 on the render test.

The Athlon XP 2400+ processor seems to kick some butt on all of the processors tested so far (haven't had a chance to test faster Athlons at this time).

Well, any comments or observations on this people ??

Regarding the single 1.42 G4. Your table suggests that you tested a DUAL, in which case the G5's performance is commendable:

1.6 GHz G5 beats 2.84 GHz G4 (well 1.42x2, not really 2.84, but for argument's sake) by 22 seconds. I was also impressed that it beat the Intel 2.6 GHz, which is a full GHz faster than it.

I think these are very good results.

The Dual 2 GHz should cream everything you tested, but then again, it's total clock speed should suggest that anyway.

In barefeats tests, it got a score of 22, as opposed to 29 for the 1.6 GHz.

If you look at it proportionately, your test should render a result of about 1:01 for the Dual 2.0, which bests the competition.

The real question is how well Opterons and Xeons do...In this case, maybe better than the G5.

 

[ianimate3d]

It clearly states 1 cpu used on the dual machines. Bryce is not dual aware.

 

[SpaceMusic_Guy]

Re: Re: 3D Render Times--G5 1.6 Ghz-vs- G4 & PC's

Originally posted by themadchemist
Regarding the single 1.42 G4. Your table suggests that you tested a DUAL, in which case the G5's performance is commendable:

Yes, it was a Dual 1.42.........however Bryce is not ported to take advantage of 2 processors......it only uses 1 processor. So the render time is from just a single 1.42. Since that is the case, the time from the G5 1.6 Ghz is not overly impressive.

1.6 GHz G5 beats 2.84 GHz G4 (well 1.42x2, not really 2.84, but for argument's sake) by 22 seconds.

No, it was just the single 1.42 that it beat, it didn't use both processors.

I was also impressed that it beat the Intel 2.6 GHz, which is a full GHz faster than it.

It beat the 2.66 Pentium 4 by 1 second. Yes it's a Ghz faster, but from a raw CPU standpoint, how much would you pay for a 2.66 Ghz Pentium 4 compared to the G5 1.6 Ghz ?? For just straight CPU power (for Rendering) your paying about $1,200 more for the G5 1.6 for roughly the same power.

I think these are very good results.

I'd say good, not great results.

The Dual 2 GHz should cream everything you tested, but then again, it's total clock speed should suggest that anyway.

When I eventually do the Bryce Render test on the Dual 2.0 Ghz G5, it will only use 1 processor, so the render time will reflect what a single 2.0 Ghz G5 can do.

In barefeats tests, it got a score of 22, as opposed to 29 for the 1.6 GHz.

If you look at it proportionately, your test should render a result of about 1:01 for the Dual 2.0, which bests the competition.

We shall see......



The real question is how well Opterons and Xeons do...In this case, maybe better than the G5.

I think they will do better than the G5 in this case. And I haven't even tested the faster Athlon's yet.........which might beat all of the single processor Mac's and PC's.

 

[macrumors12345]

Re: Re: Re: 3D Render Times--G5 1.6 Ghz-vs- G4 & PC's

Originally posted by SpaceMusic_Guy
I think they will do better than the G5 in this case. And I haven't even tested the faster Athlon's yet.........which might beat all of the single processor Mac's and PC's.

Maybe, but I wouldn't bet on it. The Athlon 2400+ runs at something like 1.9 Ghz, so it is actually performing a bit slower clock for clock than the G5. The Opteron maxes out at 2 Ghz and is heavily based on the Athlon core. It does have superior memory bandwidth and floating point, but if this test were really memory constrained or made heavy use of FP, it is unlikely that the G4/1.42 would be anywhere near the P4/2.66 at all (but it is). If I had to guess, I would be that the 2 Ghz G5 and the Opteron 246 will perform about the same in Bryce 5, but who knows for sure. Regardless, it's a very old application...it's probably compiled for the PIII core (which is why the Athlon does pretty well).

 

[actionslacks]

This is the most unscientific experiment I have seen regarging the G5s. if you want real results you should use Shake, Renderman, After Effects, etc. these programs are what count because they are profesional apps, have PC counterparts, and will eventually be compiled specifically for the G5.

 

[jelloshotsrule]

Originally posted by actionslacks
This is the most unscientific experiment I have seen regarging the G5s. if you want real results you should use Shake, Renderman, After Effects, etc. these programs are what count because they are profesional apps, have PC counterparts, and will eventually be compiled specifically for the G5.

doesn't matter how "professional" an application is. the fact is, he did these tests. and maybe not every factor was controlled besides the processors, but these results are what he's found. no reason to shoot down his efforts

 

[SpaceMusic_Guy]

Originally posted by actionslacks
This is the most unscientific experiment I have seen regarging the G5s. if you want real results you should use Shake, Renderman, After Effects, etc. these programs are what count because they are profesional apps, have PC counterparts, and will eventually be compiled specifically for the G5.

Umm........Bryce 5 HAS a PC counterpart.........that's how I was able to compare the different systems. As far as the higher end Pro App's being compiled specifically for the G5, I'm sure that will happen.........but it will probably be a while before we see that.

I don't know that I would call the test all that "un-scientific." It's just one application but it is a descent start to getting an idea of Raw CPU Speed.

Granted, Bryce 5 is not optimized for the G5........but never the less........it's REAL results for a non-optimized application.

 

[legion]

3 Questions about the still you're rendering and the systems:

1) How large is the still image in KB or MB?

2) How much RAM did each system have?

3) What OSs were on each machine?

I have no problem with the test being "un-scientific" seeing as all the tests up until now are just as scientific. At least in this case the same individual is running the same test on each machine.

(of course, everyone will have their own excuse for why their preferred platform was beat, but at least this adds to the general knowledge base and SpaceMusic_Guy seems unbiased-- unlike tests run by companies who have a stake in the results (Apple, IBM, Intel, AMD and the software makers who are beholden to each of these companies))

Thanks for the info SM_G

 

[MacAztec]

Originally posted by SpaceMusic_Guy
Umm........Bryce 5 HAS a PC counterpart.........that's how I was able to compare the different systems. As far as the higher end Pro App's being compiled specifically for the G5, I'm sure that will happen.........but it will probably be a while before we see that.

I don't know that I would call the test all that "un-scientific." It's just one application but it is a descent start to getting an idea of Raw CPU Speed.

Granted, Bryce 5 is not optimized for the G5........but never the less........it's REAL results for a non-optimized application.

Thats cool that it is unoptimized for the CPUs, but the thing is, you CAN optimize it for the G5, and there is not much to optimize for the P4.

I don't see how you can be unimpressed with the results. Did you expect a 1.6GHz G5 to be like a 3.2GHz P4? I didn't.

I really want to see some controlled experiments with the G5. Like, a G5 with 1GB RAM vs a P4 with the same. And not some hardrive that has been fooled around with by thousands of people installing different crap things on it.

This is an excellent start. The G5 at 1.6GHz (about 10 percent fast MHz, and, the G4 was a Dual, whether or not Bryce is optimized, the system is) turned out about 20% faster.

 

[MacAztec]

Originally posted by legion

(of course, everyone will have their own excuse for why their preferred platform was beat, but at least this adds to the general knowledge base and SpaceMusic_Guy seems unbiased-- unlike tests run by companies who have a stake in the results (Apple, IBM, Intel, AMD and the software makers who are beholden to each of these companies))

Thanks for the info SM_G

Apple's tests were done by Veritest (sp?) and were not controlled by Apple.

 

[Catfish_Man]

Originally posted by MacAztec
Thats cool that it is unoptimized for the CPUs, but the thing is, you CAN optimize it for the G5, and there is not much to optimize for the P4.

Never heard of MMX/SSE/SSE2? Or the lack of a free FXCH on the P4? Or the funky memory setup, or the double clocked simple integer units and weak floating point unit? The P4 is one of the chips that really only does well when apps are optimized for it. In the worst case a 1.5GHz P4 performed like a 200MHz Pentium Pro when running a certain emulator. That's rather extreme though. The thing is, most of the P4 tweaks are already done, and the G5 is just starting. The G5 has plenty of stuff to tweak for, as do the P4 and G4. The Athlon does pretty well with just about any code you throw at it since it doesn't really do anything too weird. Just big L1 caches and lots of execution units.

 

[Mav451]

Re: Re: Re: Re: 3D Render Times--G5 1.6 Ghz-vs- G4 & PC's

Originally posted by macrumors12345
Maybe, but I wouldn't bet on it. The Athlon 2400+ runs at something like 1.9 Ghz, so it is actually performing a bit slower clock for clock than the G5. The Opteron maxes out at 2 Ghz and is heavily based on the Athlon core. It does have superior memory bandwidth and floating point, but if this test were really memory constrained or made heavy use of FP, it is unlikely that the G4/1.42 would be anywhere near the P4/2.66 at all (but it is). If I had to guess, I would be that the 2 Ghz G5 and the Opteron 246 will perform about the same in Bryce 5, but who knows for sure. Regardless, it's a very old application...it's probably compiled for the PIII core (which is why the Athlon does pretty well).

2400+ runs @ 2.0ghz (15x 133fsb dual), or 1995mhz +/- in variation.

I'm not so sure that the Opteron is maxed out at 2ghz. With Tom at Amdzone.com (love him or hate him) he has got it to 2.2ghz on AIR cooling. Furthermore, with the Athlon64 debuting at 2ghz, i doubt that the opterons will stay at 2ghz for much longer. Regardless, the clock for clock comparison is useless in the real world where money matters. The 2400+ is a STEAL at 78 dollars. It is pretty much a budget processor, like the 2100+. Both overclock well on air to about 2.3ghz. When i say overclock, that means raise voltage slightly and forget it (like that weird oven infomercial)

I expected the G5 to do very well as its a new core out--obviously optimizations will do alot (the P4 would be nothing w/o those sse2 optimizations against the athlon).

 

[Rezet]

Re: 3D Render Times--G5 1.6 Ghz-vs- G4 & PC's

Originally posted by SpaceMusic_Guy
OK folks, I went into Micro Center (which finally has a G5 1.6 Ghz tower on display) and did some 3D Renders of a Bryce 5 still image. Over the past 4 months, I've rendered the exact same image on a number of PC systems and G4 systems as well. I have all the render times written down and documented.

Although Bryce 5 is not a high-end 3D app., it's still a pretty good indicator of Raw CPU speed with NO optimizations for any specific CPU (PC or Mac). So, here is a comparison of the G5 1.6 Ghz compared to other systems rendering the same Bryce 5 image:

G5 1.6 Ghz -- 1:21

2.66 Ghz Pentium 4 -- 1:22

2.8 Ghz Pentium 4 -- 1:13

3.06 Ghz Pentium 4 -- 1:11

Athlon XP 2400+ -- 1:09

1 Ghz G4 (single) -- 2:27

Dual 1.25 G4 (1 processor used) -- 1:58

Dual 1.42 G4 (1 processor used) -- 1:43

It will be interesting to time the 1.8 and Dual 2.0 as well when those units become available. So far, the G5 it is not quite as fast as I thought they would be, although an improvement over the G4 for sure. However, the G5 1.6 Ghz was only 22 seconds faster than a single G4 1.42 on the render test.

The Athlon XP 2400+ processor seems to kick some butt on all of the processors tested so far (haven't had a chance to test faster Athlons at this time).

Well, any comments or observations on this people ??

Yeah, I have a comment... a question rather...
Do you have a personal bias towards AMD?
I have seen many top of the line P4s and Athlons go at it in different tests. In around 80% of all tests Intel wins.
If you are really saying that Athlon 3200 is faster than Pentium 3.2, I have nothing more to add... as this is a total waste of time.

 

[Mav451]

yes, the 3.2 P4 C beats the 3200+, that's a pretty well known fact.

However, the older P4 B models are very evenly matched, if not beaten by the Athlon XP lineup, matched with the appropriated Nforce2 motherboard (NF7-S, Gigabyte, or Asus )

By pricing, AMD wins hand down, but performance wise, when processors are above 450 dollars, then yes, P4's win.

Take the 2.6ghz P4-B (533 fsb) - 183 dollars.

The 2600+ (333 fsb) is clocked at 2.083ghz, but can compete very well agains the 2.6. The 2600+ is, however, only 96 dollars.

For the C processors, i would rate the 2.8C to be about the same as the 3000+

The pricing is actually pretty accurate for these 2 as well, with the 3000+ coming in at around $256, while the P4 2.8C comes in at $270.

 

[pianojoe]

Hm. I wonder if Bryce is altivec enhanced for the G4 anyway?

 

[SpaceMusic_Guy]

Re: Re: 3D Render Times--G5 1.6 Ghz-vs- G4 & PC's

Originally posted by Rezet
Yeah, I have a comment... a question rather...
Do you have a personal bias towards AMD?
I have seen many top of the line P4s and Athlons go at it in different tests. In around 80% of all tests Intel wins.
If you are really saying that Athlon 3200 is faster than Pentium 3.2, I have nothing more to add... as this is a total waste of time.

No, I don't have a bias towards AMD. However, in my Bryce 5 rendering tests that I've done over the past 3-4 months, it seems to be faster than their Pentium counter parts. The Athlon XP 2400+ rendered faster than the 2.8 Ghz Pentium 4:

Athlon XP 2400+ -- 1:09

2.8 Ghz Pentium 4 -- 1:13

I've NEVER owned a PC. I've only owned Macs. What I've noticed by doing these render tests is that the Mac's have been behind in the Horsepower Department (by a good bit) to my utter Dismay !! However, I'm thinking of getting a PC in the near future for Rendering 3D Stills and Animations as well. For straight Raw CPU power for Rendering, getting an AMD (or maybe a Pentium) is more bang for the buck by far than the Mac (which I'm not happy about).

However, after my render tests that I've done, I would get an AMD PC. In my experience, the Athlon XP's seem to be a little snappier than the Pentium 4's.

Now having said that, I will be getting a new G5 sometime in the future to do almost everything else that I do.

As far as the Athlon 3200 goes, I never said that it was faster than a P4 3.2. I've never tested either of them yet with Bryce 5.

 

[SpaceMusic_Guy]

Originally posted by pianojoe
Hm. I wonder if Bryce is altivec enhanced for the G4 anyway?

I'm almost positive that it isn't. It's really stupid that most of the 3rd party Mac Software out there isn't even optimized for Altivec..........which is supposed to be the big advantage of the G4. What good is having the Altivec if 80% of the software isn't written for it.......damn stupid if you ask me. Or just plain LAZY on the part of Software Companies not to compile for it.

Either way, you end up having no advantage over the PC processors because of that (which infuriates me to no end) !!!

 

[SpaceMusic_Guy]

Originally posted by legion
3 Questions about the still you're rendering and the systems:

1) How large is the still image in KB or MB?

2) How much RAM did each system have?

3) What OSs were on each machine?

I have no problem with the test being "un-scientific" seeing as all the tests up until now are just as scientific. At least in this case the same individual is running the same test on each machine.

(of course, everyone will have their own excuse for why their preferred platform was beat, but at least this adds to the general knowledge base and SpaceMusic_Guy seems unbiased-- unlike tests run by companies who have a stake in the results (Apple, IBM, Intel, AMD and the software makers who are beholden to each of these companies))

Thanks for the info SM_G

1. The image that was rendered was "800 x 425" in Pixel dimensions.

2. All the PC systems had 512 Mb of ram in them (which is the stock configuration out of the box). All the Mac systems had 256 Mb of ram except for the Dual 1.42 Ghz which had 512 Mb of ram (stock configurations out of the box).

3. All the PC systems were running Windows "XP Pro" except for the Athlon 2400+ which had "XP Home", and all the Macs used OS X.

Basically, I just go into Micro Center and test all the machines in their stock configurations.

 

[actionslacks]

Originally posted by SpaceMusic_Guy
Umm........Bryce 5 HAS a PC counterpart.........that's how I was able to compare the different systems. As far as the higher end Pro App's being compiled specifically for the G5, I'm sure that will happen.........but it will probably be a while before we see that.

I don't know that I would call the test all that "un-scientific." It's just one application but it is a descent start to getting an idea of Raw CPU Speed.

Granted, Bryce 5 is not optimized for the G5........but never the less........it's REAL results for a non-optimized application.

Ummm...

1. Using display machines at Micro Center is not exactly a controlled environment.

2. These are real results according to whatever you scribled down in your notepad while at Micro Center.

3. I applaud your efforts, but we work in Maya and Shake and until I begin testing Renderman on my machines, I am not going to put any stock in numbers collected from these kinds of armchair comparisons.

 

[actionslacks]

Originally posted by jelloshotsrule
doesn't matter how "professional" an application is. the fact is, he did these tests. and maybe not every factor was controlled besides the processors, but these results are what he's found. no reason to shoot down his efforts

I am not shooting down his efforts. That was not my intention nor the point of my post. The topic of this thread is "3D Render times-g5 1.6 Ghz vs G4 & PC's".

If this is the app that he uses and he is looking to purchase machines that will perform the best for it, then that's great.

The point is that no one should really take too much stock in the reults until REAL tests are completed and there are benchmarks to compare for 3D apps.

It does matter how "professional" a program is. The programs I mentioned are used considerably more often to create content for TV/film. They are designed to work best on certain machines. Renderman, for example, is new for OSX and frankly that means I might be switching to G5s/xserves and I need to know specifically how this app works with this hardware. I want machines and software that are optimized to work together.

 

[macrumors12345]

Re: Re: Re: Re: Re: 3D Render Times--G5 1.6 Ghz-vs- G4 & PC's

Originally posted by Mav451
2400+ runs @ 2.0ghz (15x 133fsb dual), or 1995mhz +/- in variation.

I'm not so sure that the Opteron is maxed out at 2ghz. With Tom at Amdzone.com (love him or hate him) he has got it to 2.2ghz on AIR cooling. Furthermore, with the Athlon64 debuting at 2ghz, i doubt that the opterons will stay at 2ghz for much longer.

Ah, sorry, you misinterpretted what I meant when I said "maxes out." I didn't mean that it is impossible to get an Opteron to run past 2 Ghz, or that no Opteron will ever run past 2 Ghz - I simply meant that the fastest Opteron that you can buy today runs at 2 Ghz, just as the fastest G5 that you can buy today runs at 2 Ghz. Obviously, the G5 too can presumably be overclocked (but I wouldn't risk my $3,000 investment), and it too will scale much higher in the future, but neither of those things means too much today.

 

[macrumors12345]

Re: Re: Re: 3D Render Times--G5 1.6 Ghz-vs- G4 & PC's

Originally posted by SpaceMusic_Guy
For straight Raw CPU power for Rendering, getting an AMD (or maybe a Pentium) is more bang for the buck by far than the Mac (which I'm not happy about).

Not really. Unlike Bryce 5 (which I sincerely hope is NOT what you are planning to run), most modern 3D rendering programs are multithreaded (unlike Bryce), so they take full advantage of both processors on a Dual 2 Ghz G5. Now obviously it varies somewhat by application (for example, Cinema4D seems a bit slower on the G5 than on the high end Xeon - at least until more optimizations are made beyond a simple recompile - while Renderman seems significantly faster on the G5 than on the high end Xeon), but it is safe to say that the Dual G5 is at least roughly equivalent to the Dual Xeon 3.06, if not faster. But a Dual Xeon/3.06 from Dell will set you back a good $3,500 or more when similarly equipped to the Dual G5, so I wouldn't call it "more bang for the buck." Likewise, a Dual Opteron 246 can probably compete pretty well with the Dual G5 and the Dual Xeon in terms of performance, but it's definitely not going to come in under $3,000 (the two processors alone will cost over $1,600), unless you are planning to build the whole box yourself from spare parts you have lying around in your garage!

Again, if you are running a specific program that happens to perform better than average on the P4 (e.g. Cinema 4D...although even with that, who knows what could happen now with the IBM XL compilers being released for OS X), then it can be more cost effective to buy a Wintel machine. But it would be highly inaccurate to make a general statement that "for rendering, getting and AMD/Intel is more bang for the buck." Pixar certainly disagrees strongly with that statement...but I'm sure THEY don't know ANYTHING about rendering, right? ;-)

 

[Mav451]

hmm i know that in a magazine i read @ border a while ago (some kind of 3d rendering magazine, forgot the name) that mentioned employing an EXTREMELY cost-efficient render farm using amd 1600xp's (aka old palominos).

Those things were 50 bucks or less in mid 2002. Though rite now, the 1700's would be better as they produce anywhere from 5-10degrees C on avg LESS heat (1.5vcore vs. 1.75vcore ugh).

The render farm was used for the popular or "unpopular" (depending on your opinino) summer movie "Hulk".

Macrumors: Yeah i see what you mean. In terms of purchasing, the opteron is maxed at 2ghz. And 1600 dollars is no small feat ; neither is 3000 dollars for the dual G5.

Other than the apparent rarity of ECC registered ddr400 ram (Estimate $300+/- for 2 sticks of 512 ddr400), the rest of the system (board, psu, sata hd's, video card fans) does not exceed 700 dollars.

SK8N is $230 ; quality psu $80, 160gb seagate sata $170; panaflos x 4 ($25); 9600 pro $135 .

The total "miscellaneous" parts come out to $645, +/- variations in pricing. That's not even the cost of a single 246 cpu (which shows how new those things are )

I believe the 246 systems "as a whole" are definitely going to run $3000-4000 for sure though, but these systems are not built with the "consumer level" Sk8N, which is hundreds cheaper than the MSI/Tyan counterparts.

 

[visor]

What a drag.

Well, what do you expect from a completely not optimized Software on a new Processor. Of course it only scales with Frequency.How could it be any different? As long as you don't at least recompile it with IBM's new compiler, or at least gcc3.3 with G5 Optimisation, G5's will most likely even be a bit slower than the G4 when looking at a proceeeses/MHz because the G5 has a longer pipeline.

Remember - longer Piplines is what i386's have. It's supposed to be real bad and slow, that's why the G4 always was far supperiour to Pc's