Slashdot Mirror

Hello, Mr. Anonymous Coward. Apparently you don't understand the difference between a render farm and a cluster.

A cluster is something that appears to software as a single computer, and does its own load distribution. Assembling a cluster requires some specialised hardware and software, which are anything but trivial. You can either develop your own (like Cray are currently doing with Opteron CPUs) or you can use ready-made components and just adapt them to your configuration, like Virginia Tech did with their G5 cluster. The fact that the G5 is actually an IBM PPC970 enabled them to use existing technology for the clustering, cutting costs dramatically (compared to a x86 cluster).

A render farm, however, is something completely different. It doesn't require any special hardware or software; just a simple network connection. It's up to the renderer to detect and distribute the tasks by the various nodes. Which any decent 3D rendering package can do. In this situation, the price is just the price of the base system (plus a network card, if one isn't incuded). And this means that, besides being faster, x86 hardware is also cheaper.

If you are interested in the relative performance of different systems in 3D rendering, I recommend visiting this site. It compares over 50 different CPUs, all rendering exactly the same (10) scenes. I suggest comparing a dual G5 to a dual Xeon. And before you complain that Lightwave "isn't properly optimised for Altivec", let me assure you that it is. It just happens to be "properly optimised" for SSE2, too, which I'm sure you agree is only fair.

However, to make the G5 a viable alternative for rendering, Apple (IBM, to be more precise) doesn't need to make it faster. They (Apple) just need to make it cheaper. Unfortunately, the dual Xserve G5 is still overpriced (and not even available yet).

> benchmark using Adobe AfterEffects, but that app does
> not use both processors on the Mac,

So you're claiming it's unfair because somehow it uses both processors on the single-processor PC...?

> and is not Altivec optimized,
> but AE is optimized for Intel.

Is it? Then why was the dual Mac equally crushed by the Athlon, in the previous test? Let me remind you the Athlon does not support SSE2, so it has no Altivec equivalent.

> He further stacked the deck by running the benches
> on dual processors, where a fair test would
> have benched a single-proc app on single-proc macs
> and PCs.

So you're saying a single-processor Mac performs better than a dual-processor Mac? Now I'm definitely confused. He pitches a Mac with two processors against a PC with one processor and you say that's biased towards the PC...?

I agree that it wasn't fair. Personally I think he should have used a dual- or quad-Xeon, instead of a single-CPU "consumer" Pentium 4.

> He used codecs that are also optimized
> poorly on the Mac,

Could you please make it clear what codecs you're talking about?

> I suggested he do the benches with a program
> that is equally optimized for both platforms,
> like Cleaner 6 or Shake.

Cleaner is about the slowest, crappiest encoder ever created (this applies to both the PC and Mac version). Shake (as you well know) is no longer being sold for the PC. And neither of those programs is in the same market as After Effects. If you want an alternative in a close (though higher-end) segment, you have Discreet's Combustion 2.

Personally, I would have liked to see a comparison of 3D rendering, too. Since 3DS MAX doesn't run on Macs, they could use Lightwave, for example. BTW, you can see tons of Lightwave benchmarks here.

> In response to my polite letter,

If your letter was anything like your post above, then, it wasn't "polite", it was "deranged".

> why would anyone believe this idiot?

Hm... tough one... I got it! Because it's true...? Because anyone can get the files he used and run his or her own benchmarks? Because Photoshop is the most important image editing program in the market (including the Mac market)?

I have something very important to say: My GeForce2 MX is the fastest graphics card in the world. People who benchamrk cards using Quake III or AutoCAD are biased because those programs are not properly optimized for my GeForce2 MX. If anyone tells you that ATI's AiW 9700 Pro or nVidia's GeForce4 Ti4600 are faster than (or in any way superior to) my GF2 MX, they are either idiots, or liars, or both.

Thats is what you sound like.

RMN
~~~

Oh boy, you can't even distinguish between FP and SIMD... what am I saying, you can't even distinguish between your head and your ass!

Hm... maybe there's a reason for that.

If you want to compare Altivec (Mac SIMD) code running on PPC with code running on x86, you should compare it to SSE2 code (x86 SIMD). A quick look at Lightwave benchmarks should make things pretty clear to you:

http://www.blanos.com/benchmark/

If you're going to compare generic FPU performance with SIMD performance, you might as well say Pentiums are faster than themselves (because P4 SSE2 is much faster than P4 FP). Duh!

And if SIMD is so relevant in the real world, howcome the Athlon XP (which does not support SSE2) is the most used CPU in intensive computation, such as climate simulations, genetic research, etc.? Why don't these guys use Macs, hm...?

Are they idiots too, like everyone except you?

And you confuse reality with advertising. The only advantage the G4 has over the Athlon is the Altivec unit. But if you take a Pentium 4, for example, you have SSE2, which is similar to the Altivec (and runs faster). And very few programs actually make good use of the Altivec. AMD will add SSE2 support to the Opteron, BTW.

If you compare the MHz, the G4 is indeed much faster than a Pentium 4, and perhaps even slightly faster than an Athlon (5-10%). But the problem is, the G4 is still stuck at 1 GHz. Athlons now run at 2 GHz and P4s at 3 GHz. And, for each MHz, the G4 is definitely not 3 times faster than a P4. Memory and bus bandwidth are also much bigger in PCs, especially if you use RDRAM or dual-channel DDR333 / DDR400.

Digital Video Editing magazine ran some tests on After Effects some time ago. I submitted the link to Slashdot but apparently it wasn't considered relevant, so they never posted it. Lots of other sites did. You can find it here. You can also see plenty of Lightwave benchmarks here.

Apple is simply no loger competitive in markets that require top processing power (ex., rendering, video, scientific research). That doesn't mean Macs can't be good home computers (especially for non-techies), but there's no point in making such obviously false statements as "Macs are faster than PCs". It only makes Mac users sound like a bunch of fanatics.

Macs are not faster than PCs. But in most cases they don't need to be.

Ok, so clock frequency != performance. We know that. However, by any reasonable performance metric (no, Photoshop tests don't count because Photoshop is highly AltiVec optimized but not very well SSE/MMX/SSE2 optimized), the Athlon/P4 creams the PowerPC. Look up the benchmarks for yourself. Artificial benchmarks, real world benchmarks. It doesn't matter. The G4 always pulls ahead in one or two benchmarks but loses in most of them.

Compare here or here (or here! to here ..and no, there's simply no single-CPU Config for the big Workhorses anymore! ;-)

Ok, so clock frequency != performance. We know that. However, by any reasonable performance metric (no, Photoshop tests don't count because Photoshop is highly AltiVec optimized but not very well SSE/MMX/SSE2 optimized), the Athlon/P4 creams the PowerPC. Look up the benchmarks for yourself. Artificial benchmarks, real world benchmarks. It doesn't matter. The G4 always pulls ahead in one or two benchmarks but loses in most of them.

Compare here or here (or here! to here ..and no, there's simply no single-CPU Config for the big Workhorses anymore! ;-)

Ok, so clock frequency != performance. We know that. However, by any reasonable performance metric (no, Photoshop tests don't count because Photoshop is highly AltiVec optimized but not very well SSE/MMX/SSE2 optimized), the Athlon/P4 creams the PowerPC. Look up the benchmarks for yourself. Artificial benchmarks, real world benchmarks. It doesn't matter. The G4 always pulls ahead in one or two benchmarks but loses in most of them.

Compare here or here (or here! to here ..and no, there's simply no single-CPU Config for the big Workhorses anymore! ;-)

Ok, so clock frequency != performance. We know that. However, by any reasonable performance metric (no, Photoshop tests don't count because Photoshop is highly AltiVec optimized but not very well SSE/MMX/SSE2 optimized), the Athlon/P4 creams the PowerPC. Look up the benchmarks for yourself. Artificial benchmarks, real world benchmarks. It doesn't matter. The G4 always pulls ahead in one or two benchmarks but loses in most of them.

Compare here or here (or here! to here ..and no, there's simply no single-CPU Config for the big Workhorses anymore! ;-)

>> So next you're going to say I can't compare a 1.8 GHz Athlon with a 2.53
>> GHz Pentium because the Athlon has a lower clock speed...?

> Nope. Never even implied that.

You said:

>> They tested software (After Effects) that does not take advantage
>> of dual processors on a dual G4 of slower individual processor speed.

How could they have tested it on a system of higher (or the same) processor speed (I presume you meant clock speed) if they don't make G4s any faster than 1 GHz...?

> Adobe hasn't been that helpful to Apple is quite a while.

Why should they be "helpful" to Apple when Apple has become one of their main competitors in Mac software?

It's up to Adobe's customers to pressure them to improve their software. If there's room for improvement, and if the clients let Adobe know that those improvements will increase their sales, Adobe will surely invest more time and money optimising their programs.

> The tests invariably come out the way the review wants them to.

Well, I've seen thousands of tests that pitch Pentiums against Athlons and ATI against NVidia and so on, and they all seem to agree on the results (within 10%). The article I linked to was a reasonably fair comparison between a Dual G4 and an Athlon MP workstion (although it wasn't the fastest Athlon available at the time, and although they halved the amount of RAM in the Athlon, and although the G4 was more epxensive).

But you say that test is no good because After Effects doesn't use the second G4 CPU properly (but do you know if it uses the second Athlon CPU properly?).

So I ask you: what program do you suggest one should use to compare both plaftorms? Lightwave? Combustion? Both are quite well optimised for SMP workstations, so they should be a good test, right?

I didn't find any data for Combustion, but here's a page with a ton of Lightwave benchmarks:

http://www.blanos.com/Benchmark/

> "Strangely enough"?? It was used to support my
> statement that the second processor is barely
> used by After Effects. It had nothing to do
> with other platforms.

Yes, but it's strange that in that article they say "After Effects runs slower on the G4 than on the Athlon because it's not properly optimised", and then they mention a series of programs that (they say) are optimised for the G4 and... don't compare them to the same programs on an Athlon (or Xeon). Don't you find that a bit strange...? How hard would it have been to run the Combustion benchmarks on a PC?

> Your message was nothing but an Apple-bashing.

No offence, but I really couldn't care less about Apple. I don't sell PCs or Macs, and I'll use whichever hardware and software get the job done better (be it PC, Mac, SGI, etc., and Linux, Irix, Windows, MacOS, etc.). Here's my original message again:

If you "do 3D" (or compositing) professionally you will be using the platform that lets you meet deadlines.

True or false? Does a frame rendered on a Mac look any different from the same frame rendered by the same program on a PC or SGI? If the resulting quality is the same, the difference boils down to speed.

And the G4 is definitely not it.

True or false? Aren't the fastest G4s slower than the fastest Xeons / Athlons? Try this with any program that runs on both platforms. The 1 GHz G4 may be able to keep up with a 1.7 GHz P4. But it cannot keep up with a P4 running at 2.5 GHz.

Unless the G5 manages to at least keep up with the Pentium 4 / Athlon / Hammer, Apple doesn't really stand much of a chance in the high-end.

The P4 is likely to scale up to 3 GHz until the end of the year. Athlons are likely to scale up to 2 GHz at least. AMD's new CPU (Hammer / Opteron) will be released early next year, and will work in up to 8-way SMP (that's eight CPUs in a single workstation). It will be rated at around 3400+ (about 1.5 times faster than the Athlon XP 2000+). Will a 1 GHz G4 be able to compete with these CPUs? I think not. Apple needs a fast G5.

If they continue to support Linux render nodes it's not because they like Linux, it's because that's the only way people will buy the (slower, more expensive) Apple workstations.

Who do people buy Linux render nodes? Because they're faster than the Windows versions? Not really (sometimes they are, but we're talking less than 5%). They buy them because they're cheaper. You only need to pay for the hardware; no "Microsoft tax". No-one is going to make a render farm of Macs. Even if they weren't slower than the x86 systems, it would simply be too expensive. If Apple killed the Linux versions, they wouldn't sell the software at all, except possibly to very small companies and home users. This way they manage to sell 3 or 4 Mac workstations plus 20 or 50 licenses for Linux render nodes.

Again: my message was on-topic and accurate. I work in the field, I know what I'm talking about. I don't care if you're sentimentally attached to Apple or Microsoft or Intel or AMD or SGI or Transmeta or Linux or whatever. That may influence home users, but it doesn't influence the high-end industry. It's all about "bang for the buck". And right now, bang for the buck means Athlons and Linux.

RMN
~~~