2CPU.com » News » March 2005 » Dual Xeon vs. Dual G5... Fight!

The Inquirer linked to some comparative benchmarks over at Bare Feats that pit a dual Xeon box against a Dual G5 in Maya.

Right out of the chute, the G5 Power Mac was beaten by the Dual Xeon. If you look at the database of results on the Maya Test Center, there are faster times posted for both the Xeon and G5, though the Xeon still wins.

Not exactly an apples-to-apples (heh?) comparison, but if you're interested you can take a look.

05/20/2004 02:08 PM: Six Dual Xeon Motherboards on the Cheap by Jim_
Tom's Hardware has thrown together a nice little round-up of six dual Xeon boards from Asus, Intel, MSI and Tyan. I couldn't find an overly interesting quote to paste in for you guys, so why don't yo...

12/05/2003 01:17 PM: Ace's compare Dual & Quad Opteron with Dual Xeons. by JohNEE
Not only do Ace's Hardware have a 4-way opteron box to assess but, the corresponding 848 processors, too. 3.06 1MB Xeons are fighting the Intel corner with some anecdotal references to SGI Origin an...

07/24/2003 07:28 PM: Canterwood Supports Dual Xeons? by Jim
Several people emailed this one in, so I decided to bite the bullet and post it up. The Register (and others) are running articles stating that the i875P chipset, otherwise known as Canterwood, is cap...

01/30/2003 12:31 AM: Supermicro joins Tyan with ATX sized dual Xeon boards by dadmin
Supermicro have released a couple of standard ATX sized dual Xeon boards that will spark your interest if Tyan's Tiger i7500 appealed to you: The Super X5DAL-TG and the Super X5DAL-G. The difference b...

09/30/2002 10:12 PM: New Dual Xeon Board(s) From Tyan by Hooz
A little birdie told me today that Tyan has announced a new ATX (not eATX or SSI, but regular ATX) dual Xeon board aptly named the Tiger i7500. Just like all of the other "Tiger" boards, the...

08/08/2002 03:45 PM: Via based dual Xeon hits marketing phase by dadmin
As you may remember, The DDR Zone snapped some pictures of a Via based dual Xeon undergoing testing while in Taiwan a while ago. Well, looks like the naysayers who claimed it would never make it to pr...

07/28/2002 12:12 PM: Weta Digital buys dual Xeons in bulk by dadmin
Oldish news but still relevant. New Zealand company Weta Digital is buying a swag of dual Xeons for post-production processing of coming releases of Lord of the Rings films.' We call it a render farm;...

06/05/2002 11:51 AM: Dual Xeons on a Via chipset! by dadmin
VooDoo from The DDR Zone is in Taiwan has has few pictures of what could be music to the ears of the budget conscious. How does a dual Xeon prestonia system with a Via P4X266 chipset sound? Impossible...

02/22/2002 05:34 AM: Dual Xeon Review? Here? Ever? by Hooz
We were just chatting about this in our IRC channel, so I thought I'd post a little something to see what sort of response I got.To put it plainly, I want to do a dual Xeon (Prestonia's would be great...

01/23/2002 03:45 PM: Dual Xeon from Iwill: DP400-N by dadmin
X-bit Labs have another picture teaser for the Xeon crowd: a new Xeon board. It's the Iwill DP400-N:DP400 is a dual-processor Socket 603 mainboard based on Intel 860 chipset. It is intended for Xeon C...

nine Registered User Posts: 1313 Joined: 2004-03-11

#35311 Posted on: 03/22/2005 08:42 PM Here's another one, using After Effects as the benchmarking tool: http://www.digitalvideoediting.com/articles/viewarticle.jsp?id=31238

daniel178 Registered User Posts: 359 Joined: 2001-01-10

#35312 Posted on: 03/22/2005 09:04 PM although its virtually impossible nowadays to take a Mac and PC which are "identical" or have "matching specs" and are of near-same price, (believe me, i've tried; no matter how close you get 'em, theres someone who says "no, wait... look at that, thats not the same!") i think these two articles are pretty consistant, and clear -- not a whole lotta extra junk has been added or endlessly explained, just the main stuff that actually makes a considerable difference, and for the non-enthusiast looking to make a desicion, these are good reads. Personally, I like the "out of the box" tests. Why? Simply becuase most consumers dont endlessly tweak their systems. They want a fast machine when they buy it, not after (what could be days) of little tweaks, program and system configuration setting changes. [oh, and i have to give barefeats, a basically Mac-oriented/-advocating site, kudos this time for not swaying their results... too much ]

i_wolf labhair dom as gaelige Posts: 2097 Joined: 2002-11-19

#35313 Posted on: 03/22/2005 09:48 PM Here are some more from barefeats. To be honest, I'm not too surprised that Maya is faster on the PC. The Maya application and development on the PC platform has evolved over a longer period of time than that of the Macintosh version. The current G5 compiler support on Macintosh is also a very sad state of affairs. There could be huge number of reasons for such large discrepancies in speed difference. Either way I wouldn't be buying a Mac for its speediness in Maya anyway. From a PR point of view, more important than Maya (since its a very "niche" type application) Apple has taken a major kick to the groin recently with Doom 3. A high profile game that will run well on practically on any high end x86 hardware and run abysmally on a top of the line PowerMac. An interview with Aspyr in this case lay the blame on a number of issues... 1) Lack of a good optimized compiler, 2) problems in the PPC ISA converting float to int 3) differences in the way the scheduler work in OS X , 4) differences in the OpenGL pipeline. Personally I think that prob 2 is rubbish and is easily overcome with shadow values in code, that is how Apple recommends developers to optimize in such cases for the PPC ISA. As for the scheduler in OS X, some of the lads I have worked with in here have often complained that they get far higher GFlops when they run the same code under PPC Linux in stead of under OS X. The scheduler in OS X while great for giving a responsive many-things-going-at-once feel is not up to par when you want single app dedicated performance. Looks to me like Apple have their work cut out for them in the near future and beyond. 1) They need to provide a far better optimized compiler for the G5, 2) They need to improve their OpenGL driver performance, 3) they should provide travelling macintosh optimization workshops that move between company and company etc... 4) They could do with improving the scheduler in OS X for the benefit of single app performance. At least a few of these performance issues have started to be addressed in Tiger But at the end of the day these suggestions may improve the performance situation in Maya, however they may or may not bring performance on par with an equally clocked Athlon 64 or a high end Xeon. There will always be some apps that run better on Mac and others on WinTel/WinAMD. Maya might just be one of these ones that run better on WinTel. Hung like a donkey. Go like a horse!

rmn oh my, it's huge! Posts: 6013 Joined: 2002-01-26

#35314 Posted on: 03/22/2005 10:58 PM Yes, Maya is highly biased towards x86. They should have used 3DS MAX. RMN ~~~

AssKoala Anti-Zealot @ GATech Posts: 3309 Joined: 2002-01-02

#35315 Posted on: 03/22/2005 11:38 PM Originally posted by i_wolf 1) Lack of a good optimized compiler, 2) problems in the PPC ISA converting float to int 3) differences in the way the scheduler work in OS X , 4) differences in the OpenGL pipeline. Differences in the OpenGL pipeline? And what are those, pray tell? As for the problem with Doom3 on the Mac, the sources I've read say the problem is in relinquishing control of the graphics system to the game, slowing things down considerably (little to nothing to do with the CPU scheduling as proven below). When running with virtually no graphics to test the CPU, the Dual G5's are only slightly slower than a PC (somthing like 125fps versus 135fps). I think Macworld's review pointed this out. Me Webpage | If you always think like an expert, you'll always be a beginner. | "A handful of knowledgeable people is more effective than an army of fools" -Writing Secure Code, 2nd Ed.

i_wolf labhair dom as gaelige Posts: 2097 Joined: 2002-11-19

#35316 Posted on: 03/23/2005 12:30 AM Differences in the OpenGL pipeline? And what are those, pray tell? Differences in the OpenGL pipeline meaning that the OS X OpenGL stack is quite young in OS X and is still evolving. Some things in the OpenGL stack are quite slow and in some cases there are alternative and quicker ways of achieving the same result. Glenda Adams from Aspyr media had the following to say over at ars' forums. He has replied back to a few people here Originally posted by Glenda Adams over at ars' forums: When I originally said the game was feeding the driver as fast as it could handle, that was a (maybe not perfectly clear) way of saying the game wasn't CPU bound on the app side. At the end of our optimizations we hit a point where the game was pretty much stalling on GL calls, so further speed ups either had to come from inside GL or in the video drivers, or a massive re-engineering of the Doom 3 rendering pipeline. Hung like a donkey. Go like a horse!

AssKoala Anti-Zealot @ GATech Posts: 3309 Joined: 2002-01-02

#35317 Posted on: 03/23/2005 03:25 AM Those aren't differences in the OpenGL pipeline, those are OS X's inefficient OpenGL pipeline, the pipeline itself is the same, the speed at which it does things is different. Regardless, it is an issue with OS X itself, not the G5 Hardware. Me Webpage | If you always think like an expert, you'll always be a beginner. | "A handful of knowledgeable people is more effective than an army of fools" -Writing Secure Code, 2nd Ed.

Ballkula 3D designer Posts: 124 Joined: 2002-04-21

#35318 Posted on: 03/27/2005 02:17 PM We've read all the excuses now for 2 years, when it comes to a direct comparison, G5 vs Xeon/Opteron. I guess what matters is real-world performance, here, today - not what could have been. A fast PC will almost always be faster than a Mac at the same price, the competition between AMD and Intel, nVidia and ATi and so forth, will make that happen. On the other hand, the performance gap is not necessarily a huge one, so if your heart beats for OS X and your applications are provided for, then you can easily go for the Apple-alternative. ----------------------------------------------- "Even a battlesuit wont protect you from the void..." -----------------------------------------------

i_wolf labhair dom as gaelige Posts: 2097 Joined: 2002-11-19

#35319 Posted on: 03/27/2005 09:51 PM We've read all the excuses now for 2 years, when it comes to a direct comparison, G5 vs Xeon/Opteron. I guess what matters is real-world performance, here, today - not what could have been. A fast PC will almost always be faster than a Mac at the same price, the competition between AMD and Intel, nVidia and ATi and so forth, will make that happen. On the other hand, the performance gap is not necessarily a huge one, so if your heart beats for OS X and your applications are provided for, then you can easily go for the Apple-alternative. And as always.... for the past two years , it has been argued and agree that cross platform performance comparisons largely depend on what you are doing. There are and will be applications which will be much faster on a G5 than an Athlon 64 or Xeon and vice versa. Making generalisations like x is always faster than y, is rediculous and is only providing free marketing for Apple, Intel or AMD. X will only be faster than Y in certain situations and in certain applications and this hasn't changed over the past 2 years regardless what AMD, Apple or Intel would have us believe! Hung like a donkey. Go like a horse!

rmn oh my, it's huge! Posts: 6013 Joined: 2002-01-26

#35320 Posted on: 03/27/2005 11:07 PM Originally posted by i_wolf There are and will be applications which will be much faster on a G5 ...we just haven't been able to find them yet. RMN ~~~

i_wolf labhair dom as gaelige Posts: 2097 Joined: 2002-11-19

#35321 Posted on: 03/28/2005 09:31 AM ...we just haven't been able to find them yet. How hard were you looking!..... There is OS X , iLife, FCP and DVDstudio Pro for a start !!!!!!! :P joking aside, in most areas (probably with the exception of 3d modelling) the Macintosh platform has equally well performing native Macintosh apps which do more or less the same job as the big name windows equivalents. However I for one reckon that probably the biggest gaping hole performance wise for the macintosh platform is in the area of 3d modelling and design. Apple will loose out on this market because with the exception of maybe one or two apps (Luxology's Modo is particularly fast on the G5 as is Bryce 3d) the current hardware/software combo cannot compete (unless you only use Modo and Bryce! :rolleyes: ). Again I don't necessarily think this is so much a hardware issue as a software issue. There is no point in having this fancy hardware if you do not have the software or tools to take advantage of it. .....and on the topic of performance .... "the Amiga will rule the world!" Hung like a donkey. Go like a horse!

nine Registered User Posts: 1313 Joined: 2004-03-11

#35322 Posted on: 03/28/2005 10:13 AM I hate to get all software snob, but................Bryce 3d? *snicker*

Ballkula 3D designer Posts: 124 Joined: 2002-04-21

#35323 Posted on: 03/28/2005 02:14 PM Originally posted by i_wolf And as always.... for the past two years , it has been argued and agree that cross platform performance comparisons largely depend on what you are doing. There are and will be applications which will be much faster on a G5 than an Athlon 64 or Xeon and vice versa. Making generalisations like x is always faster than y, is rediculous and is only providing free marketing for Apple, Intel or AMD. X will only be faster than Y in certain situations and in certain applications and this hasn't changed over the past 2 years regardless what AMD, Apple or Intel would have us believe! I've used Macs for ages, and currently uses a DP2GHz and a DP2.5GHz, mostly because some people forced Final Cut on us (it's a great app, I know..) I am convinced that in a real business, you can so easily measure performances between different systems, and so far the Mac is not making anyone impressed. There are certain filters and functions in various programs which can be really, really fast, but overall our x86-systems outruns the Macs easily. So, what some people claim that the G5 can do in "theory" is not worth anyting, imho, cuz it doesn't help me or anyone else outside the benchmarking/forumspeculating world Also: the noiselevels on the current top-Macs are just over the top, any tips to make them sound less like a 747? ----------------------------------------------- "Even a battlesuit wont protect you from the void..." -----------------------------------------------

LRSeriesIII Aspiring Rocket Scientist Posts: 1128 Joined: 2002-08-29

#35324 Posted on: 03/28/2005 07:42 PM Originally posted by Ballkula Also: the noiselevels on the current top-Macs are just over the top, any tips to make them sound less like a 747? Are you talking about the dual 2.5GHz machines? It was my understanding that the liquid cooling system on that model was supposed to lower the noise output compared to the dual 2GHz model. The one G5 I worked on (a dual 2GHz, and I worked on it as in it was sick, not as in using it) did not seem terribly loud when the fans were at their low speed, though one of the symptoms of the problem this mac was having was that the fans would gradually spool up while it was on, even though it was not doing anything. As you may have found out, at full tilt those fans are loud, even compared to most high-end PC workstations I have been around. ->Computers ->Folding for team 3074

i_wolf labhair dom as gaelige Posts: 2097 Joined: 2002-11-19

#35325 Posted on: 03/28/2005 09:54 PM Well you can and you can't measure real world performance. Photoshop for example, has some filters which render fastest on the G5, some which render fastest on Athlon 64 and obviously some which render fastest on the Xeon. Which platform is fastest.... well if you use the filters primarily that the Xeon works best with then the Xeon is fastest. The point is that you use what is fastest for your trade. Obviously then for you and your business you have found that whatever applications you use are much slower than an x86 equivalent or there isn't an adequate macintosh native equivalent app. Thats your decision made. Don't buy a mac simple... it doesn't fit your usage requirements ! There are enough people that buy a macintosh for its performance in the areas that they require that you have to know that its performance is not merely theoretical. Part of the reason cross platform comparisons are difficult is because as I already said, a test isn't truly cross platform unless you use the same application on both platforms. However particularly if you are doing a Mac V PC type analysis, if you use the windows ported version of the app on mac , in a lot of cases this doesn't represent the typical usage of your average macintosh user. Most macintosh users prefer to use native optimized macintosh applications which will do a similar job to a ported windows app only much faster than that also available on mac 'ported-to-mac' app. Another problem I have seen when people make cross platform task performance comparisons is that they dont compare like with like. For example. When I originally bought my Opteron 246 rig , a friend of mine had a dual Xeon 3GHz+ rig. I was telling him how long it was taking me to convert my music collection to mp3 for my new jukebox (iriver H110 at the time). He laughed and told me his machine was far faster.... that I should have bought another Xeon (have a dual 2.8 as well). He was encoding at 4 to 5 times the speed I was. He was encoding with music match and I believe the Xing mp3 encoder. I was encodign with EAC + Lame. Yes my setup did do the same job, and it did it much slower, however the resulting sound quality from each file would be of a far higher quality and audio fidelity with the method I used. The same principle can be applied to many applications , filters , compressors that are cross platform and particularly not cross platform. When doing a cross platform task comparison, it is probably most 'real world' to compare based on the same job. Pick a job that you want accomplished on both platforms and then pick the best application on both platforms to accomplish that job. Ultimately taking the speed AND final output quality into account when deciding on the best and fastest platform.The point is that often there is a difference between a cross platform test to see which is quickest at running an application and a test to see which is the best/quickest platform for a task. A subtle but important difference none the less (BTW I am not endorsing any platform with that statement... just a matter of fact in my opinion). Finally since speed and quality are often based on a ratio of one to the other ultimately the decision of 'best platform for a job' is a matter of personal preference... how much speed are you willing to sacrifice for quality and vice versa. The point of this rant is that making sweeping generalisations that X is faster than Y is just so rediculous. Yes maybe in your area thats possible, but there will always be somebody in another area that will argue the exact opposite. You use whatever tool you deem best suited for the job at hand and that is that. We work with image analysis and processing of CT scans. For us, the same code runs far far faster on the G5 to any x86 platform. However there is probably somebody out there writing a similar package that would argue the exact opposite. My own personal findings are that like Itanium, I would also realise that the G5 is not for everyone. But if you are willing to put the time and effort into writing good optimized software for it or running well optimized software on it , it can pay off in a large way. Having said that often times you will not and do not have a choice. ... and then there are always funny situations such as this ; I have seen Microsoft Office on Mac run faster under virtual PC than the actual native Microsoft Office X for mac! With regard to your noise issues, again I have experience with a gen 2 (1.8 and 2.5 Powermac) and a gen 1(2Ghz Powermac).... the problems you describe are not indicative of a healthy PowerMac. The 2.5 should be quieter than the 2Ghz version plain and simple. Something is very wrong if it is much noisier. I would take it back and insist that it be fixed. When running (processor usage set to highest in energy saver) it should be exceptionally quiet. This is relative anyway to the two 747's I own. One a dual 2.8 Xeon rig with stock Intel wind tunnel heatsink/fan combo and a dual Opteron 246 with stock heatsink/fan combo. One potential reason that your powermac 2.5 is noisier than normal could be that you may have a blanket of dust inside the case. Try getting an air duster and blow a lot of the dust out from inside the case. One major disadvantage of the PowerMac design is that they are vacuum cleaners for dust! Hung like a donkey. Go like a horse!

Ballkula 3D designer Posts: 124 Joined: 2002-04-21

#35326 Posted on: 03/28/2005 11:23 PM The reason to why I complain about the noiselevels from our newest Macs is simple: we're not used to Macs being noisy If I buy a dual Xeon 3.6GHz with state-of-the-art graphics, then I expect to be forced to put the machine in a cabinet together with the servers in a separate room. However, better fans, better cabinets on the pc-side have improved things lately, but it's still a b*tch :P Also: When it comes to 3d-rendering and visualisation, we've heard for a good 2 years now that "the next version will equalize things" - but nothing has really happened. Couldn't it just be that a G5 just aint faster than current applications show? ----------------------------------------------- "Even a battlesuit wont protect you from the void..." -----------------------------------------------

i_wolf labhair dom as gaelige Posts: 2097 Joined: 2002-11-19

#35327 Posted on: 03/29/2005 05:26 AM The reason to why I complain about the noiselevels from our newest Macs is simple: we're not used to Macs being noisy If I buy a dual Xeon 3.6GHz with state-of-the-art graphics, then I expect to be forced to put the machine in a cabinet together with the servers in a separate room. However, better fans, better cabinets on the pc-side have improved things lately, but it's still a b*tch :P Well to be honest the PowerMac *should* be a lot quieter than a Xeon or Opteron rig. A while back (maybe a year ago) I sent Jim in some sound clips of my two home machines (Opteron 246 with stock cooling and a Xeon 2.8 (533fsb) with stock cooling) compared with a Powermac I took home from work for testing. While not exactly scientific, I used the same external microphone and recorder on my personal rigs and the PowerMac (2GHz in this case). The PowerMac was most certainly the quietest of the lot by a considerable margin. The 2.5's are quieter still. Why not do a search of the threads here and perhaps have a listen for yourself? There are other reports about the net and practically everywhere there is a review of the PowerMac there is rarely any criticism of noise. The noisiest thing I have ever heard in a PowerMac is the graphics card. What graphics card do you have in your PowerMac ? The ATI 9800pro is noisey enough but nowhere near as noisey as the 6800 Ultra. However that definately sounds like a sick puppy you are working on. Also: When it comes to 3d-rendering and visualisation, we've heard for a good 2 years now that "the next version will equalize things" - but nothing has really happened. Couldn't it just be that a G5 just aint faster than current applications show? Well I don't have any affiliation with or know Apple's reasoning but here is how I see things. I don't think this is a hardware issue at all. The hardware is at least up there with x86. In a lot of cases performance is at the very least competetive with x86 performance across a wide range of applications. The biggest notable exceptions are 3d modelling and games. If the hardware was not up to par then it would mean that we would not be seeing the types of things we see apps like Modo do. Modo according to Brad Peebler of luxology runs fastest on the Mac. Apps like Motion and FCP show that there is lots and lots of horsepower there on tap for heavy processing jobs if necessary. In raw GFLOPS or TFLOPS, presently there isn't a supercomputer in the top 10 that has the same price/performance of Virginia Tech's monster... not a single x86 supercomputer with the same per node performance/price. So it definately looks to me like a software issue. While not a 3d content creation app, I found the Doom3 benchmarks very telling. Even when using the same graphics card as a PC the mac lags so far behind that it couldn't possibly be a hardware issue (they are using the same graphics card after all). Practically all problems with regard to 3d performance are software issues. Personally I think there are a number of factors coming into play. 1) OpenGL pipeline on OS X needs a lot of work (Tiger supposedly brings approximately 30% performance increase for OpenGL apps). Presently the API calls are very slow in Panther. 2) Graphics card drivers need more work. 3) The sceduler in OS X Panther is not great for the G5. This has had a lot of work done on it for Tiger. To get an idea of what I'm talking about if you were to run for example a RC5 client on PPC Linux for G5 and the same client for OS X you would see that the PPC Linux version produces a far higher number of keys on the same hardware. 4) GCC3.3 is brutal and does not provide autovectorisation for altivec nor adequate G5 optimizations full stop. Most macintosh apps written today tend to be written in XCode with GCC3.3 unfortunately. The scheduler in this version of GCC does not lend itself well to the instruction grouping and despatch mechanism of the PPC970. 5) As already spoken about before ... many cross platform mac apps have been merely ported from a windows version without much thought given towards optimization of the mac version. This involves more than just handtuning certain aspects of code. Unfortunately ported apps do not lend themselves particularly well to being well optimized. The problem with (5) particularly is that the mac is a niche market. Financially I can understand why many companys will not spend significant amounts of time optimizing for a platform that has such a small market share. Because of this Apple has to make sure it is as easy as possible for developers to get performance easily. They could do this by rectifying points 1 -4 that I made. If they did this then it would be easier for developers to get more performance with the minimal of work (after all its a niche market). Then naturally more people would consider the mac platform. The more people considering the mac platform, the more incentive there is for companys to optimize their macintosh versions. I think Tiger shows massive improvements across the board in terms of performance. OpenGL performance has improved considerably on todays apps. They have provided major performance technologies such as Core Image, Core Video and Core Audio which are extremely simple to leverage and already have large functionality built right in for developers with fantastic perforance. In other areas Tiger has a new scheduler that is far more G5 optimized. The compiler in XCode 2 has full autovectorisation support for altivec and can generate much faster code for the G5 processor than GCC 3.3. With regard to 3d Content Creation I think the new compiler, improved OpenGL pipeline and overall OS optimization for the G5 should deliver much higher performance on the same hardware over Panther. But it is still only half the battle. Apple really need to get their 3d drivers up to a professional level. If Apple wants to use ATI and nVidia consumer parts (which are by and large idential to the profesional parts except for the drivers) they need to either provide professional x86 level of quality drivers for these consumer parts or actually offer an ATI FireGL, nVidia Quadro or 3d Labs Wildcat as a build to order purchase option. I often wonder if perhaps Apple should enter the 3d content creation market with their own app as they did with FCP and some of the other apps in the digital media market??? Hung like a donkey. Go like a horse!

rmn oh my, it's huge! Posts: 6013 Joined: 2002-01-26

#35328 Posted on: 03/29/2005 07:28 AM Originally posted by i_wolf I found the Doom3 benchmarks very telling. Even when using the same graphics card as a PC the mac lags so far behind that it couldn't possibly be a hardware issue (they are using the same graphics card after all). If I put the same graphics card on an 800 MHz PIII and on an Athlon 64 4000+, I suspect the performance will be slightly different, despite both systems using "the same graphics card after all". Anyway, 3D rendering is (still) all about the CPUs, not the graphics card. I doubt Adobe wastes much time with platform-specific optimisations in either case; if After Effects is that much slower on a Mac, it's either because the hardware is slower or the compilers are terrible. If you compare software that is optimised (ex., Shake, Digital Fusion, etc.), the x86 version is still significantly faster (which probably explains why Apple decided to kill the x86 versions of the applications they bought - if the Mac version was competitive, they wouldn't have needed to do that, the performance would determine the clients' decision). For most "workstation" work, however (Photoshop, modelling, compositing, etc.), most of the CPU cycles are wasted waiting for human input, so the overall difference is irrelevant. Is it really that annoying if a filter takes 1.5 seconds instead of 1 to apply? Anyway, Macs are no longer aimed mainly at professional DCC; Apple were smart enough to realise the way the wind was blowing, and avoid going the same way as the Amiga. They are "friendly home PCs" that happen to come with a decent selection of DCC software, and are fast enough for most users. Arguing that Macs aren't faster than PCs at extreme tasks like 3D rendering or video compression or image processing is a bit like arguing that MiniDV handycams are crap because they don't have XLR sound output - they're not meant to. RMN ~~~

Mr. Hahn Unregistered

#35329 Posted on: 03/29/2005 07:50 AM Doesn't (or if not, shouldn't?) ibm have a PPC compiler that produces faster code than gcc. I mean I love gcc for portability, but after all it is pretty bad for code optimization compared to say... the people that make the processor and the compiler.

Mr. Hahn Unregistered

#35330 Posted on: 03/29/2005 07:55 AM Originally posted by rmn If I put the same graphics card on an 800 MHz PIII and on an Athlon 64 4000+, I suspect the performance will be slightly different, despite both systems using "the same graphics card after all". Anyway, 3D rendering is (still) all about the CPUs, not the graphics card. I doubt Adobe wastes much time with platform-specific optimisations in either case; if After Effects is that much slower on a Mac, it's either because the hardware is slower or the compilers are terrible. If you compare software that is optimised (ex., Shake, Digital Fusion, etc.), the x86 version is still significantly faster (which probably explains why Apple decided to kill the x86 versions of the applications they bought - if the Mac version was competitive, they wouldn't have needed to do that, the performance would determine the clients' decision). For most "workstation" work, however (Photoshop, modelling, compositing, etc.), most of the CPU cycles are wasted waiting for human input, so the overall difference is irrelevant. Is it really that annoying if a filter takes 1.5 seconds instead of 1 to apply? Anyway, Macs are no longer aimed mainly at professional DCC; Apple were smart enough to realise the way the wind was blowing, and avoid going the same way as the Amiga. They are "friendly home PCs" that happen to come with a decent selection of DCC software, and are fast enough for most users. Arguing that Macs aren't faster than PCs at extreme tasks like 3D rendering or video compression or image processing is a bit like arguing that MiniDV handycams are crap because they don't have XLR sound output - they're not meant to. RMN ~~~ Just because it isn't as fast a graphics workstation doesn't make it a bad workstation system. It is a great little system for any unix user in need of a good desktop/workstation unix os. I only speak of my experience though.

AssKoala Anti-Zealot @ GATech Posts: 3309 Joined: 2002-01-02

#35331 Posted on: 03/29/2005 08:02 AM Originally posted by Mr. Hahn Just because it isn't as fast a graphics workstation doesn't make it a bad workstation system. It is a great little system for any unix user in need of a good desktop/workstation unix os. I only speak of my experience though. He didn't say it was bad. He said it was a "friendly home PC". Leave Workstation work to those systems more suited to the tasks if performance is an issue. For example, the Itanium2 is the king of this (and all else) according to the resident looney. Me Webpage | If you always think like an expert, you'll always be a beginner. | "A handful of knowledgeable people is more effective than an army of fools" -Writing Secure Code, 2nd Ed.

rmn oh my, it's huge! Posts: 6013 Joined: 2002-01-26

#35332 Posted on: 03/29/2005 08:39 AM On the contrary, Macs can make great "interactive" workstations - they don't need to be terribly fast; just fast enough to not be annoying. Benchmarking how long it takes to apply 50 filters is irrelevant for the real world, where people apply one every 5 minutes, and spend the rest of the time looking and making adjustments. And if there's one area where the Itanic doesn't even leave the harbour, that's DCC. The Itanium's FPU might be very impressive on paper, and might be very good for some specific tasks, but I've never heard of anyone building or using an Itanium render farm. Some renderers (ex., Mental Ray) were ported to IA64, but no-one uses them. And, since most big CGI firms develop their own renderers, legacy software support isn't really the issue. Even Intel has come out and said that the Itanium "isn't really aimed at the workstation market". RMN ~~~

Mr. Hahn Unregistered

#35333 Posted on: 03/29/2005 09:09 AM Originally posted by rmn On the contrary, Macs can make great "interactive" workstations - they don't need to be terribly fast; just fast enough to not be annoying. Benchmarking how long it takes to apply 50 filters is irrelevant for the real world, where people apply one every 5 minutes, and spend the rest of the time looking and making adjustments. And if there's one area where the Itanic doesn't even leave the harbour, that's DCC. The Itanium's FPU might be very impressive on paper, and might be very good for some specific tasks, but I've never heard of anyone building or using an Itanium render farm. Some renderers (ex., Mental Ray) were ported to IA64, but no-one uses them. And, since most big CGI firms develop their own renderers, legacy software support isn't really the issue. Even Intel has come out and said that the Itanium "isn't really aimed at the workstation market". RMN ~~~ What do you mean by " interactive"? do you just not interact with your windows pc or something? I mean I don;t use the thing because the dock bar is pretty.

AssKoala Anti-Zealot @ GATech Posts: 3309 Joined: 2002-01-02

#35334 Posted on: 03/29/2005 09:20 AM Originally posted by rmn And if there's one area where the Itanic doesn't even leave the harbour, that's DCC. The Itanium's FPU might be very impressive on paper, and might be very good for some specific tasks, but I've never heard of anyone building or using an Itanium render farm. Some renderers (ex., Mental Ray) were ported to IA64, but no-one uses them. And, since most big CGI firms develop their own renderers, legacy software support isn't really the issue. Even Intel has come out and said that the Itanium "isn't really aimed at the workstation market". RMN ~~~ I guess I should have noted the sarcasm Me Webpage | If you always think like an expert, you'll always be a beginner. | "A handful of knowledgeable people is more effective than an army of fools" -Writing Secure Code, 2nd Ed.

i_wolf labhair dom as gaelige Posts: 2097 Joined: 2002-11-19

#35335 Posted on: 03/29/2005 10:40 AM If I put the same graphics card on an 800 MHz PIII and on an Athlon 64 4000+, I suspect the performance will be slightly different, despite both systems using "the same graphics card after all". You are quite right. However considering that on paper the PPC970 is at least in the same ball park theoretical performance wise as x86 hardware, I would doubt that the processor is the bottleneck for Doom3 as a P3 800 would be ! Furthermore it has ample bandwidth so again I doubt that causes a bottleneck on the PowerMac. The point I was trying to allude to was that with Doom 3 most likely the biggest bottleneck will be the graphics card. The graphics cards are the same hardware wise on both x86 and PPC sides of the fence and this was the point exactly. Only software engineering issues such as usage of a bad compiler, bad drivers and/or a slow OS OpenGL stack could account for the terrible performance in Doom 3 under OS X. By the same token, Doom 3 runs much slower under linux (on the same x86 hardware) than it does under Windows. The Doom example was just that, an example I was using to show how software engineering practices can account for massive performance differences from platform to platform and even within the same hardware platform. I doubt Adobe wastes much time with platform-specific optimisations in either case; if After Effects is that much slower on a Mac, it's either because the hardware is slower or the compilers are terrible. Barefeats seamed to think that their nightflight render test was fastest on the G5 in after affects. I see however that charlie white thinks otherwise in his tests. I say tomato... you say.. eh... nevermind. You are right Adobe does not spend much time optimizing for the Mac platform anymore. Apple has made so many inroads into their DCC territory I doubt its in Adobe's interest. I would imagine that in a few revisions time Motion will be the 'after effects' version of choice on the mac platform. Today its very limited in what it can do compared with AE but the things it does do that are also accomplished in after effects tend to be much faster in Motion on Mac... If you compare software that is optimised (ex., Shake, Digital Fusion, etc.), the x86 version is still significantly faster (which probably explains why Apple decided to kill the x86 versions of the applications they bought - if the Mac version was competitive, they wouldn't have needed to do that, the performance would determine the clients' decision). Actually I have heard some people say that 3.5 flies along much faster than 3 or 2.5 on the Mac. I thought that 3.5 finally has Shake at the same levels as 2.5 was at on x86. Again I say this out of genuine surprise to hear that it is still significantly slower. Can you provide a link to some benchmarks please of the 3.5 version against say the most recently sold version on the PC(2.5 I believe it was). This is hardly a proper comparison I know.. but I would still be fascinated to see for myself? Personally I think it will be interesting to see what comes out of NAB this year. Last year provided only small updates to Shake and FCP providing small G5 optimizations here and there but no major from the ground up reworkings etc.. I think it was thinksecret that were saying that the reason there were such limited token updates last year was because major work was going on behind the scenes on the pro app line up which are due to be introduced at NAB this year. http://www.thinksecret.com/news/0502fcp5.html and http://www.thinksecret.com/news/0503nab.html Apparantly most pro apps announced at NAB will require Tiger !?!?! I would imagine that at the very least Apple's next pro line up will benefit heavily from Tiger and at that I would say Shake should get a big speed up at the very least. I've a few friends who work in film and broadcasting and they all complain about the lack of direction in Shake for the past while. Apple seams to be letting it stagnate without adding anything major or important to it. As for digitalfusion is this even available for OS X and the G5 ? I checked over at eyeonline.com but I could find no mention on their requirements page for digitalfusion http://www.eyeonline.com/Web/EyeonWeb/Products/system_requirements/system_requirements.aspx Arguing that Macs aren't faster than PCs at extreme tasks like 3D rendering or video compression or image processing is a bit like arguing that MiniDV handycams are crap because they don't have XLR sound output - they're not meant to. Thats an interesting point you have raised. There was a time that the PowerMac line was very heavily weighted and sold primarily as a DCC product. However Apple seams to sell it as a highend 'home pc' and also as a highend 'professional workstation'. Can it be both ? I don't see why not, after all in recent times we have seen a major convergence between the technology we have under the desk in our living room and the workstation we use in a studio. Renderman is supposedly (according to the steve jobs/Pixar endorsed promotion video for the G5) the fastest platform for renderman... what are the odds! :rolleyes: It probably is at this stage and you can bet that thats because Steve Jobs asked (read: demanded) that a lot of effort be put into optimizing the Macintosh version. Still you are correct 3d rendering is an area that is badly represented on the Mac. The Mac platform needs to have more than just one high performing renderer available. Maya on Mac just doesn't cut the mustard when you compare the Mac version with the x86 version. There needs to be a high speed renderer available for the mac besides renderman. This is strange because the altivec units should lend the G5 class machines particularly well to this type of work under Maya etc... Modo is a nice surface modelling app which shows that when properly used the G5 certainly isn't a slouch in other aspects of the 3d content creation. Probably an exception to the rule in the 3d content creation market is that Modo runs fastest under G5 class hardware, however I would imagine there is probably an equally fast application that will do the same job under x86.... horses for courses. As for the Mac's aren't faster or PC's aren't faster argument.... I think this is something that will go on for years and it will always hit the same answer.... use what is best for your current requirements. Doesn't (or if not, shouldn't?) ibm have a PPC compiler that produces faster code than gcc. I mean I love gcc for portability, but after all it is pretty bad for code optimization compared to say... the people that make the processor and the compiler. Yes thats right. XLC and XLF are available under OS X from IBM or within a nice IDE from Absoft. I believe this is what Virginia Tech used to get every last drop of performance out of their supercluster. Generally speaking it is much faster than GCC particularly for vector and floating point under the G5. Integer work also gets big speed ups but not in the same magnitude as floating point and vector. Funnily enough it also speeds up G4 code. It doesn't have proper autovectorisations yet though. You know looking back at my reply, it just strikes me that there is absolutely no way Apple could afford to pass up on the Cell processing achitecture. It just would have so many performance advantages for the 3D rendering and content creation markets. Hung like a donkey. Go like a horse!

2CPU.com » News » March 2005 » Dual Xeon vs. Dual G5... Fight!