PDA

View Full Version : Mac Mini and LW8 - anyone doing it?



Ade
02-03-2005, 05:45 AM
I'm thinking of wanting a mac mini and know the video card is only 32 ddr. Just wanted to know feedback if it is as bad as I think or not that bad for average tasks?

Id get the 1.42 g4 and 1 gig ram.
Is the video card soldered in or can it be replaced?

ingo
02-03-2005, 08:25 AM
The videocard is soldered onto the motherboard, so no change possible. I work still on an old G4 450 with the good ol' Ati Rage Pro with 16 MB, it works, but you have sometimes to wait a long time for the display of a wireframe scene. So i think the MiniMac is only usable for small scenes or for rendering animations.

Johnny
02-03-2005, 01:28 PM
Maybe the mac mini will be this anemic only for this revision...the first iPod was pathetic comparied to what you can get now, but when iPod sales took off, Apple saw fit to feed it some vitamins, and today's pod is a blood-drinking killer compared to rev. 1

I'm hoping mac mini sales will merit a similar vitamin treatment. Who here wouldn't buy an unimpaired mac mini with a G5 under the hood, esp if sales volume is enough to keep the price at $499-ish?

J

BeeVee
02-03-2005, 04:20 PM
If you want the Mac Mini to go a bit faster there's always overclocking...

http://dialspace.dial.pipex.com/prod/dialspace/town/pipexdsl/q/aqza37/macmini/index.html

B

Ade
02-03-2005, 04:27 PM
im happy with the 1.42 speeds but display is a more important factor, with 32 megs thats poor..I think ill give it a miss.

Ryhnio
02-05-2005, 08:40 PM
the mac mini is great for an entry computer, not extensive modeling and animations.
if you want to learn 3d get a mac mini, if you want to create real 3d get a real mac-


ryhnio-

cresshead
02-06-2005, 08:28 PM
well i'm just about to get one!
and think it's a real bargin....i'm going for this config.

1.42
super drive
512 ram

yeah the v card is just 32 mb....but seeing as my other windows pc's are a sony laptop with 11mb, a dell 700mhz pIII with a 64mb geforce card and a SGI 320 dual 450 pII with a crossbar video which i set to 32mb and find okay i feel that this "wee beasty" will be fine for what i want it for...which is finally getting a mac [my first] and it being my fastest pc!....plus cheap and capable of making dvd's with zero effort!....hit the burn button!

of course i'll also be running lightwave on it!

and corel draw 11

plus i-life 05
and possibly i-work and final cut express HD...

i'll also connect and use my iomega 250 gig hard drive [usb2 and firewire]

to view it i'll be using a 19" crt

also like to add that i'm currently looking thru reviews of it...to see how it compare to a g5 imac as an alternative...nowt's set yet...though it does look okay for most usage...not a leader for sure but capable performer with 512 ram.

synist
02-08-2005, 05:36 AM
I have been thinking of getting a mac mini to use as a render machine.
I can network it to my powermac and control it with remote desktop.

But i am wondering if there are any heat problems with leaving this machine running for days and weeks at a time?

I remember the G4 cube having problems with heat when left on.
Has anyone had any problems with overheating?

Wouldn't a tower of Mac Minis be a cool render farm.

Johnny
02-08-2005, 08:32 AM
I have been thinking of getting a mac mini to use as a render machine.
I can network it to my powermac and control it with remote desktop.

But i am wondering if there are any heat problems with leaving this machine running for days and weeks at a time?

I remember the G4 cube having problems with heat when left on.
Has anyone had any problems with overheating?

Wouldn't a tower of Mac Minis be a cool render farm.


None of the reviews I've read online have mentioned overheating as an issue, and these have been hard-core reviews, even detailing how to putty-knife open the case to stuff more RAM or a new Drive into the Mini...

it looks well-vented, tho dense inside..

J

Captain Obvious
02-08-2005, 08:54 AM
If the Mac mini gets too hot, the fan will start spinning faster. As far as I've heard, that hardly ever happens. It's as quiet, and cool, as can be.

trison
02-09-2005, 08:35 AM
I bought a Mac Mini the day it came out and I love it for most of the normal stuff. (web surfing, watching DVDs, etc) but so far I've used Photoshop, Illustrator and After effects on it and it has choked on all of them with files that weren't that complex. I don't think it's really a good work machine. also, I have only turned it off once in the time I've had it and I don't have any heat problems and it runs really quiet.

BTW I got the 1.4 off the shelf so the specs are:

1.4 ghz G4
256mg DDR
32mg Video MEM
80gig HD
Combo drive

BeeVee
02-09-2005, 08:54 AM
This is an interesting article positing the point of the Mac Mini by noted computer ubergeek Robert X Cringley:

http://www.pbs.org/cringely/pulpit/pulpit20050120.html

B

cresshead
02-09-2005, 09:57 AM
trison, you really must go n upgrade your memory in your mac mini to 512 all of the reviews state that this a "needs must" as with 256 the mac mini is strangled by the lack of ram...with 512 it frees up the mac to really deliver with good quality performance as good as a emac or powerbook and better than most of those in actuallity plus was a good runner against a imac 1.6 as well on some of thise tests in the review.

256 ram is just not good enough for day to day usage with the i life product or such like...same would be true of lightwave i think though no one has tested it with that as yet from what i've read in reviews.

steve g

cresshead
02-09-2005, 09:59 AM
i'v not commited to a mac mini as yet...i really want to see a review or user feedback from someone running the mac mini with 512 ram and lightwave 3d.

once i see a review i'l either get the mac mini or a imac.

trison
02-09-2005, 11:07 AM
I plan on upgrading the RAM in the future. I just bought the mini on credit so I want to take care of that first :)

Captain Obvious
02-10-2005, 06:37 AM
I bought a Mac Mini the day it came out and I love it for most of the normal stuff. (web surfing, watching DVDs, etc) but so far I've used Photoshop, Illustrator and After effects on it and it has choked on all of them with files that weren't that complex. I don't think it's really a good work machine. also, I have only turned it off once in the time I've had it and I don't have any heat problems and it runs really quiet.

BTW I got the 1.4 off the shelf so the specs are:

1.4 ghz G4
256mg DDR
32mg Video MEM
80gig HD
Combo drive

Don't blame it on the mini, blame it on the 256 megs of RAM. OS X needs at least 512. If you do a lot of Photshop and such, it is more than worth the money to upgrade to 1024.

Johnny
02-10-2005, 09:06 AM
Don't blame it on the mini, blame it on the 256 megs of RAM. OS X needs at least 512. If you do a lot of Photshop and such, it is more than worth the money to upgrade to 1024.


yeah..ALL of the reviews I've read say the same thing. 512MB is a BARE minimum for all OS X systems.

Captain Obvious
02-10-2005, 09:17 AM
yeah..ALL of the reviews I've read say the same thing. 512MB is a BARE minimum for all OS X systems.

I have an original iMac with 192 megabytes, and trust me when I say that the 233MHz processor is my LEAST concern. ;) Constant beachballing is not the processor's fault, it's the RAM (or lack thereof).

Marcia
02-10-2005, 09:19 AM
Since the mini tops out at 1 gig RAM, and OS X snarfs up a good bit of that, it seems you would be awfully limited in what you could do, LW-wise.

Johnny
02-10-2005, 10:51 AM
Since the mini tops out at 1 gig RAM, and OS X snarfs up a good bit of that, it seems you would be awfully limited in what you could do, LW-wise.


true...'your mileage may vary.' Most of my scenes snarf up 512MB per chip or less, so 1GB in a mac mini would be fine for the vast majority of my stuff..

I have a 600 mhz G3 iBook with 640 MB of RAM, and that little beast kicks butt...

Insufficient RAM is always a bottleneck worth fixing

synist
02-10-2005, 01:38 PM
[QUOTE=trison]I bought a Mac Mini the day it came out and I love it for most of the normal stuff. (web surfing, watching DVDs, etc) but so far I've used Photoshop, Illustrator and After effects on it and it has choked on all of them with files that weren't that complex.

I agree with what everyone else says 512MB minimum.

I've noticed that photoshop uses around 100MB RAM just standing there.

Has anyone gone for 1GB RAM?

Captain Obvious
02-10-2005, 02:31 PM
I bought a Mac Mini the day it came out and I love it for most of the normal stuff. (web surfing, watching DVDs, etc) but so far I've used Photoshop, Illustrator and After effects on it and it has choked on all of them with files that weren't that complex.

I agree with what everyone else says 512MB minimum.

I've noticed that photoshop uses around 100MB RAM just standing there.

Has anyone gone for 1GB RAM?

When it's just sitting there, Photoshop generally uses about a meg or two for me. :p

synist
02-10-2005, 03:12 PM
When it's just sitting there, Photoshop generally uses about a meg or two for me. :p

I don't understand why mine uses so much?
I only drew a few sqiggles.

Captain Obvious
02-10-2005, 03:51 PM
I don't understand why mine uses so much?
I only drew a few sqiggles.

"Just sitting there" meaning nothing open. ;)

synist
02-10-2005, 04:13 PM
"Just sitting there" meaning nothing open. ;)

No, just sitting there doing nothing. :}

Captain Obvious
02-11-2005, 03:42 AM
I rarely have anything open in Photoshop without actually doing anything with it at the moment. :p

salomon
02-14-2005, 11:37 AM
minimac is a very good deal, the only mess might be the hard drive speed, slow saving each time you go to save your work. For the rest everything is perfect, ok don't forget you job.

harlan
02-14-2005, 12:13 PM
The Mac Mini is a great system, especially for Modeling work. Pack it up with RAM, at least 512, and if you've got booku dinero laying around head over to Fryz and pick up a 2 Gig stick (and a Pallete Knife).

LW's Modeler on the Mac, for whatever reason, has a number of OGL and OGL Performance issues, but you can still get great work done in it - even on a Mini! :)

In reference to the Photoshop RAM usage - Photoshop should only be using that much RAM if you have a document open (ie: a blank document, a document with some squibbles, etc...). If you fire up Photoshop without opening a document, it should "idle" fairly efficiently with minimal RAM usage.

ingo
02-14-2005, 12:28 PM
Hmm, a 2GB stick, i thought the mainboard doesn't support it ???

Otherwise i just read a review in the german MacWelt, a daughtermagazine of MacWorld, and in the benchmarks it looks good compared to the 1.6 GHz iMac, even in the HD test, BUT except in the Cinebench test, where it was far behind the iMac. Could it be that Cinebench is too G5 optimized ??

harlan
02-14-2005, 02:18 PM
It'll support a 2GB stick - in fact, Apple is supposed to be "authorizing" the use of 2GB sticks in the Mini. Speaking from personal experience, they work just fine.

The Mini will certainly hold its own in terms of performance, it's a great little unit (especially considering its size and cost).

As far as benchmarking goes, I consider Cinebench to be no more thorough or accurate than placing a computer on a treadmill to see how fast it runs - its completely worthless (as are most benchmarks). The only true benchmark of performance, in my opinion at least, is to sit down and put the computer through its paces.

I don't have an iMac G5 here to compare it with, but it runs remarkably close to my 17" PowerBook 1.5Ghz (also with 2GB ram). All my other systems are Dual PowerMac G5's, and the Mini runs comparable to the PowerMac on a single CPU.

In regards to the HD speed issue, you could always just strap in a 7200RPM 2.5" Drive in you need the extra boost - but the HD speed issue has been negligible in terms of everyday use.

The Mini works quite well as an iDVD box. The past few days I've been kicking out a buttload of DVD tests for this project I've been working on, and the Mini does a phenomenal job at it - with the new iDVD you can also burn out ISO images to send to another unit for burning if you don't have a Superdrive.

Captain Obvious
02-14-2005, 05:08 PM
Hmm, a 2GB stick, i thought the mainboard doesn't support it ???

Otherwise i just read a review in the german MacWelt, a daughtermagazine of MacWorld, and in the benchmarks it looks good compared to the 1.6 GHz iMac, even in the HD test, BUT except in the Cinebench test, where it was far behind the iMac. Could it be that Cinebench is too G5 optimized ??
Cinebench is G5 optimized, yes.

harlan
02-14-2005, 05:11 PM
Actually, there are two Cinebench sh!tmarks available, one is G5 optimized while the other isn't.

ingo
02-15-2005, 02:50 AM
So i guess they used the G5 optimized benchmarks since all other tests are similar in speed to the 1.6 G5 iMac. I'm still thinking if i should upgrade my old G4 450 or get a new miniMac....

BeeVee
02-15-2005, 02:53 AM
But isn't it a G4 too? Do you mean for a render farm?

B

salomon
02-15-2005, 03:00 AM
just wondering if LW is optimized for 64 G5 use? a minimac renderfarm still be a great idea and that's the way i'll go to try with this solution.

Captain Obvious
02-15-2005, 07:03 AM
Correct me if I'm wrong, but doesn't LW do most of its magic on Velocity Engine? Rendering, anyway. The G5 has vastly superior floating point performance compared to the G4, but G4s and G5s are fairly equal in LW rendering, clock for clock. It would make sense, then, if it's done on Velocity Engine. If that is true, G5 optimizations wouldn't really be all that needed.

Lightwolf
02-15-2005, 07:28 AM
Correct me if I'm wrong, but doesn't LW do most of its magic on Velocity Engine? Rendering, anyway.
Nope. Most of LWs internal calculations require double precision floating point, which can't be natively handled by AltiVec / VE (there are libraries out there that allow for it though, I'm not sure if they make sense on a G5 though). So these will be handled by the FPU.
This is for example why a P4 with a relatively weak FPU, but SSE2 with double fp support, can outpace (in some cases) an Opteron which has a fast fpu, but slow SSE2...

Cheers,
Mike

eblu
02-15-2005, 07:35 AM
Correct me if I'm wrong, but doesn't LW do most of its magic on Velocity Engine? Rendering, anyway. The G5 has vastly superior floating point performance compared to the G4, but G4s and G5s are fairly equal in LW rendering, clock for clock. It would make sense, then, if it's done on Velocity Engine. If that is true, G5 optimizations wouldn't really be all that needed.

you are wrong.
LW uses the Velocity engine for lens flares and a few other non-essential rendering plugins.

sorry to inform you of this I know its a sad thing to learn.

the reason LW is mhz sensitive with both the g4 and g5, is that it is not in any way tuned for either chip and relies on a brute force approach to rendering speed increases. ie: many of the killer features of both chips are wasted on Lightwave's rendering engine.

Captain Obvious
02-15-2005, 07:46 AM
Pardon my french, but, uh, that sucks. A lot. The G5 is vastly superior to the G4 on non-vectored floats, what with the two powerful and fully pipelined FPUs. On the other hand, if it is true, some G5 optimizations to the renderer and the PowerMacs would be pretty much the rendering platform for LW. They're fairly compeditive to x86 as it is, I mean.

CPU bandwidth cannot be the issue since the G4 is fairly equal to the G5. Hmm. Odd.

Lightwolf
02-15-2005, 07:50 AM
On the other hand, if it is true, some G5 optimizations to the renderer and the PowerMacs would be pretty much the rendering platform for LW.
Well, if you look at how much the competition squeezed out with a G5 optimized rendering benchmark, they did gain roughly +15% of performance by hand optimizing their code... So I wouldn't really expect any big jumps in that area soon. Now, a decent vector unit that supports doubles :D ...

Cheers,
Mike

Captain Obvious
02-15-2005, 10:43 AM
Well, if you look at how much the competition squeezed out with a G5 optimized rendering benchmark, they did gain roughly +15% of performance by hand optimizing their code... So I wouldn't really expect any big jumps in that area soon. Now, a decent vector unit that supports doubles :D ...
Hello mr IBM executive. Can we please have VMX2? Pretty please? :p

Anyway, theoretically, the G5 should be maybe twice as fast per clock cycle as the G4 on brute force floats, right?

Lightwolf
02-15-2005, 11:36 AM
Anyway, theoretically, the G5 should be maybe twice as fast per clock cycle as the G4 on brute force floats, right?
Yeah, pretty much. I don't know what the relation of floating point ops per cycle is on either of them (faster clocks means longer pipelines, which in turn can lead to some instructions taking _more_ cycles to execute).

Cheers,
Mike - who still thinks Apple using the Opterons was the coolest rumour ever... *sigh*

eblu
02-15-2005, 12:30 PM
my understanding was that the g5 had the equivalent of dual floating point units, compared to the g4, above and beyond the altivec. it was hinted that this was a direct response to the poor floating point performance of the g4, after altivec was generally ignored. It was further hinted that to take advantage of this extra floating point unit, you'd have to do some extra work.

i don't even pretend to be up on the tech behind the processor, everything I know about the chip is watered down info, but I was lead to believe that the reason Lightwave seemed to Not get the benefits of the dual floating point units, was that Lightwave was never updated to take advantage of them.

Captain Obvious
02-15-2005, 01:12 PM
Yeah, pretty much. I don't know what the relation of floating point ops per cycle is on either of them (faster clocks means longer pipelines, which in turn can lead to some instructions taking _more_ cycles to execute).

Cheers,
Mike - who still thinks Apple using the Opterons was the coolest rumour ever... *sigh*

It does have a longer pipeline, yes, but the time it takes to execute an instruction probably isn't as important as the number of instructions it can keep churning out. The G5 can do two floats per cycle, the G4 one. Pipeline bubbles probably aren't much of a problem with rendering, and either way, the G5 has a very powerful branch predictor that probably makes the whole pipeline depth irrelevant.

So why isn't the G5 faster? :confused:



my understanding was that the g5 had the equivalent of dual floating point units, compared to the g4, above and beyond the altivec.
Yes, that's right. The G5 has two units, the G4 one. Even if it had had only one unit, the G5 would still be better at floats, since the unit is much more powerful.


It was further hinted that to take advantage of this extra floating point unit, you'd have to do some extra work.
Probably, yeah, depending on how the code looks. Chances are they haven't done any work on it to do so.


So, theoretically, if it really is the true that LW only takes advantage of one of the G5's two FPUs, it should be (again, theoretically) possible that they could double the rendering speed in LW by optimizing the code. That would make the G5 so darned fast it barely even makes sense. ;) And it makes even less sense that they have not done the necessary optimizations yet, so there must be some other problem.

Lynx3d
02-15-2005, 02:40 PM
Uh ye old "double number means double performance" misconception...

ILP (instruction-level parallelism) is something that actually happens inside the CPU (for typical superscalar CPUs), a huge amount of the CPU logic is only there to decide which execution order is the best. Btw, one reason dual-core CPUs start spreading everywhere is mainly because you already need a LOT more logic for just a little better ILP that results in better IPC, the actual execution untis already shrunk to a minor part of todays CPUs and merely adding another one doesn't give you much, so it's easier to just put a whole second core on the die.

No programmer can really dictate or even suggest the CPU to do two instructions in parallel (only with SIMD/vector instructions you can pack several data sets into one instruction), he can only try to minimize dependencies that would prevent parallelism in first place.
And it's the compiler that optimizes the the machine instructions created from high-level language source code, and many experts think compilers should even take away most of the the complexity from CPUs of re-ordering the instructions (that's where VLIW, EPIC and also cell comes in, you can for example tell Itanium which instructions can and should be done in parallel, but that makes compiler a LOT more complex and important yet again, RISC in general was a step to put complexity from CPU into the compiler too)

And honestly, i'm sure there are much more efficient ways on the algorithmic side to speed up LW by a much greater factor than start tweaking the code on the instruction-level...

cresshead
02-15-2005, 03:02 PM
so in essence yur saying that the "double speed" capability in theory will remain just that..."a theory"....which means that for rendering a stack of mac mini's makes more sense than a batch of g5 towers....for the cost of a single dual 2.5 g 5 you can get 5 mac minis with a total gighz of near 7gighz for the same price of a 5 gig g5 mac....all with 512 ram in them.......

Lightwolf
02-15-2005, 03:13 PM
And honestly, i'm sure there are much more efficient ways on the algorithmic side to speed up LW by a much greater factor than start tweaking the code on the instruction-level...
Which is pretty much the point in all software design / programming. Tweaking compiler options should only be done with optimized algorithms, and then you might start thinking about further, assembly level optimizations in hotspots.
Mind you, Maxon (who afaik use the same compiler that LW uses on OSX) could only release a manually G5 patched CineBench (then again, Maxon started out with a C++ compiler on the Amiga), so in this case you can blame both companies for not having ported to XCode, Apple for not makeing up their minds in the past years on APIs, and Metrowerks for not optimizing their compiler. Pick whatever you prefer ;)
Cheers,
Mike

Captain Obvious
02-15-2005, 05:05 PM
So, essentially, it's not NewTek's fault that LW doesn't use both units, but the compiler's?


Uh ye old "double number means double performance" misconception...
In this case, it's not really that big a misconception, though. No more than that double processors is twice as fast as a single processor, anyway. ;)


So, uh, NT, how is the porting to Xcode coming along? ;)

Apple, IBM and the other developers of GCC for PPC, how is that auto-vectorization coming along? ;)

Lightwolf
02-15-2005, 05:26 PM
So, essentially, it's not NewTek's fault that LW doesn't use both units, but the compiler's?
Put it this way ... if the compiler NT uses doesn't support the fpu properly, then you can't really blame them for not hand-coding all critical parts in assembler, especially with an XCode port coming up...

Not that I'm trying to make up excuses for them, but the development situation, especially for 50% legacy code like LW, on the Mac is extremely short of glamorous.
I guess at one time they'll have to make a cut and say "OS 10.4 or newer only and scr*w your old plugins" (I bet Chuck would phrase it differently though ;) ).


Apple, IBM and the other developers of GCC for PPC, how is that auto-vectorization coming along? ;)
Well, that ain't gonna help much as long as AltiVec only does singles. the only parts of LW using single floats are basically colour / buffer based calculations, everything else is doubles (everything connected with geometry).

AFAIK the IBM compiler allows for an extra 10-15% over GCC, at least in benches...

But, as Lynx3D wrote, there is still _plenty_ of headroom in the basic algorithms, and this will even profit _all_ platforms. I mean, you want LW to be faster overall, not "faster than a PC" ... even if that means it is dog slow on both platforms to start with ;)

Cheers,
Mike

Captain Obvious
02-15-2005, 06:39 PM
Put it this way ... if the compiler NT uses doesn't support the fpu properly, then you can't really blame them for not hand-coding all critical parts in assembler, especially with an XCode port coming up...
Well, no, not really. It would probably be better if they spent that time on improving the general math behind it instead of spending hundreds of programming hours on tweaking the assembler code.


I guess at one time they'll have to make a cut and say "OS 10.4 or newer only and scr*w your old plugins" (I bet Chuck would phrase it differently though ;) ).
The sooner they do that, the better! Err, well, I guess Apple can actually release Tiger first, but... Yeah. One of the things that makes OS X good is that Apple decided to do away with full legacy support.


Well, that ain't gonna help much as long as AltiVec only does singles. the only parts of LW using single floats are basically colour / buffer based calculations, everything else is doubles (everything connected with geometry).
Oh, I know, I didn't mean it just for LW! ;)


AFAIK the IBM compiler allows for an extra 10-15% over GCC, at least in benches...
So why not use that instead of Xcode/GCC? Or can you compile with IBM's and still use Xcode?


But, as Lynx3D wrote, there is still _plenty_ of headroom in the basic algorithms, and this will even profit _all_ platforms. I mean, you want LW to be faster overall, not "faster than a PC" ... even if that means it is dog slow on both platforms to start with ;)
Maybe I'm just strange, but I don't think LW's renderer is all that slow. Of course, I don't render 100-million-polygon scenes, so what do I know? ;) But as far as I've been able to compare it to Cinema 4D's and Maya's, it really doesn't seem all that slow.

Lightwolf
02-15-2005, 06:42 PM
So why not use that instead of Xcode/GCC? Or can you compile with IBM's and still use Xcode?

Hiya,
my last reply before I hit the sack ;)
I think you can use IBMs compiler as a drop in replacement for GCC, and it is source code and binary compatible (to an extent I guess).
I guess someone would have to evaluate that at NT though, especially if the benefits are worth the (potential) trouble. Then again, the uses intels icc on the other side of the fence...
Cheers,
Mike

Captain Obvious
02-15-2005, 08:17 PM
I should probably go to sleep, too... It is 4:15 am... :eek:

Anyway, thanks for clearing some things up! I always appreciate the opportunity to learn more, especially about interesting stuff like how many bits there are in LW's renderer and such. ;)

As for the trouble... Well, they're switching compilers regardless. Why not switch to the better? Chances are they'll run into trouble regardless, and the ~15% extra performance you get out of IBM's couldn't hurt, could it? ;)

Lynx3d
02-16-2005, 10:01 AM
Btw, GCC 4.0 is planned to be released on April. 15th according to Mark Mitchell...
I really think i should get one of those Mac minis once LW is ported to XCode...then i could finally do the Mac-compiles myself. Wouldn't really care about performance...

Whether LW is slow or not also depends on what you do with it, but i think it's beyond question that meanwhile there are GI and volumetric algorithms that can be magnitudes faster in many situations than what LW offers, plus a whole bunch of other techniques to enhance quality without ridiculous AA settings etc.

Lightwolf
02-16-2005, 10:09 AM
Raytracing seems to be another part where LW is very slow, and everything that involves it... If you look at the amount of rays something like FPrime can fire into a scene in any given time, with virtually no pre-processing ... wow.

The G4 and the laptop innards are putting me off the miniMac, but I'll probably have a G5 sooner or later anyhow, for XCode ports and compatibility testing of CDs.

Cheers,
Mike

marinello2003
02-16-2005, 10:13 AM
I currently use a G4 Powermac 1.4 GHZ with a Radeon 9800 pro card and 768 MB Ram, and LW runs very fast on that. The mini has the same processing power, but its graphics card is very limited. That will be where things slow down for you with LW. If you were looking for an entry level Mac for LW, I would suggest getting a G4 Tower or the new iMac G5. If budget is a concern, check out the refurbished iMac G5's on the Apple website here.

http://store.apple.com/1-800-MY-APPLE/WebObjects/AppleStore.woa/72501/wo/Qk4p3urI3hxA2o4KVHQhGOKv4CG/1.0.0.11.1.0.6.7.1.0.1.1

Apple Certified iMac G5
iMac G5 1.6GHz/ 256MB/ 80GB/ Combo/ 56K/ 17-inch - Apple Certified
Learn More
• Save 16% off the original price
Original price: $1,299.00
Your price: $1,099.00
Free Shipping

Then I would Add 512MB Ram and you are good to go with LW!

-Brent

marinello2003
02-16-2005, 10:32 AM
I bought a Mac Mini the day it came out and I love it for most of the normal stuff. (web surfing, watching DVDs, etc) but so far I've used Photoshop, Illustrator and After effects on it and it has choked on all of them with files that weren't that complex. I don't think it's really a good work machine. also, I have only turned it off once in the time I've had it and I don't have any heat problems and it runs really quiet.

BTW I got the 1.4 off the shelf so the specs are:

1.4 ghz G4
256mg DDR
32mg Video MEM
80gig HD
Combo drive

Add more ram and you are good to go! It is the 256MB that is slowing your machine down the most. You will see a drastic improvement if you add 512 for a total of 768MB. 256 is barely enough to run OSX alone. I have a G4 with 768 and it works really well with LW.

-Brent

eblu
02-16-2005, 10:37 AM
Hiya,
my last reply before I hit the sack ;)
I think you can use IBMs compiler as a drop in replacement for GCC, and it is source code and binary compatible (to an extent I guess).
I guess someone would have to evaluate that at NT though, especially if the benefits are worth the (potential) trouble. Then again, the uses intels icc on the other side of the fence...
Cheers,
Mike

what I heard about IBM's is that its built for speed over accuracy. alot like intel's compilers (the ones they use for benchmarks). the obviously mac biased individuals reporting this made the point that IBM hadn't screwed the ratio (speed versus accuracy) nearly as much as Intel had (as in: you wouldn't want to trust your data to anything compiled with intel's benchmark compiler, but Ibms might be ok), but thats all hearsay. If I were Newtek I'd want to be on the "solid" platform, that means the defacto mainstream compiler for each platform. IBM's isn't, GCC is.

Johnny
02-16-2005, 10:49 AM
The mini has the same processing power, but its graphics card is very limited. That will be where things slow down for you with LW. If you were looking for an entry level Mac for LW, I would suggest getting a G4 Tower or the new iMac G5.


Hi...I think you raise good, factual points, but I think that the main thrust of discussion about the Mini is whether it's a good render node machine..Most of the folks here already have hardware at least as capable as machines you cite, but everyone's looking for cost-effective render farm solutions..not everyone can afford a stable of Dual 2.5Ghz G5's, or G5 iMacs, but it seems that the mini is a capable, affordable render node, a group of which offers better renderpower per buck than a G5. and that sub-par card it comes with doesn't impact rendering..

From what I've read here, when it comes to rendering, Mhz matters, not video cards or fast HDs. My plan would be to do all my modeling and tweaking on my Dual G5, while those minis would spend their lives, 24/7 whipping out the frames. with RFC, I have the flexibility to take the G5 on and offline, either processor, whenever I want so that I can always have my "serious" machine available to me. What I've noticed after nearly 3 weeks of round-the-clock screamernetting, is that my G5 *still* has plenty of juice left for other things, including FCP, photoshop, and LW modeling and rendering.

J

Lynx3d
02-16-2005, 11:02 AM
You will see a drastic improvement if you add 512 for a total of 768MB. 256 is barely enough to run OSX alone. I have a G4 with 768 and it works really well with LW.

If i'm not mistaken you can only put one RAM module in the Mac mini, so 768 would definitely be impossible.


what I heard about IBM's is that its built for speed over accuracy. alot like intel's compilers (the ones they use for benchmarks). the obviously mac biased individuals reporting this made the point that IBM hadn't screwed the ratio (speed versus accuracy) nearly as much as Intel had (as in: you wouldn't want to trust your data to anything compiled with intel's benchmark compiler, but Ibms might be ok), but thats all hearsay.

Help, yet more hearsay :D
You know, there are strict ISO standards for what a Compiler has to do to be ISO conform. You can allow most compilers (including GCC) to break certain ISO rules to improve speed, and unless you're doing scientific computations you might not even give a **** about whether you get a NaN or some other useless result (and most rendering software deliberately uses these tweaks i guess...)
But apparently some compilers are actually tuned to mainly give impressive SPEC results...

Johnny
02-16-2005, 01:22 PM
I have no coding or programming expertise to add to part of this discussion, but isn't fPrime a monument to what can be done with code that's perfected bit by bit, over and over?

I seem to recall the author saying that's just what he did, a little bit, every day, until he produced a product which was to LW'ers on both platforms like chum in shark-infested waters.

J

Captain Obvious
02-16-2005, 03:41 PM
But apparently some compilers are actually tuned to mainly give impressive SPEC results...
Remember the whole issue about Apple's SPEC scores with the G5 vs Pentium 4? My guess is they used GCC on both platforms. The Pentium 4 gets like twice the score with Intel's compiler. Sure, it probably is better than GCC, but not that much better.

Johnny
02-18-2005, 10:08 AM
Interesting article about the Mini: http://www.macworld.com/news/2005/02/18/mini/index.php


Sort of suggests that Apple might have tapped into the zeit geist as far as peoples' wanting lean, mean and inexpensive computing UNITS, rather than just the big mambo expensive killer machines alone.

Now, if we could only get everyone wired up with XGrid, and make all apps XGrid savvy...

J

Lightwolf
02-18-2005, 10:26 AM
Remember the whole issue about Apple's SPEC scores with the G5 vs Pentium 4? My guess is they used GCC on both platforms. The Pentium 4 gets like twice the score with Intel's compiler. Sure, it probably is better than GCC, but not that much better.
I was heavily participating in the discussion here, of course I remember. The whole point back then was the fact that vendors tend to use the best optimizing compiler they can afford, this is why LW is compiled using the icc on the PC, and not gcc (the de-factor standard on the PC would be msvc anyhow, not gcc). And obviously, since the P4 has a weak fpu, but very strong SSE2, a compiler that doesn't vectorize won't really help, and does make things a bit unfair. Heck, that is like running Power601 code on a G5, doable but pointless if you benchmark. (and heck, I remember my 68060 being faster than the Power601 back then :) ).

Cheers,
Mike

cresshead
02-18-2005, 11:16 AM
"splosh!"...yup..guess wot?...ahh ha!...i've finally take the plunge and ordered my very first apple mac computer!

went with a mac mini 1.42 with 512 ram and a super drive...it may take 4-6 weeks though as they are really in demand right now!

i trialed the one in the shop with lightwave de ad all seemed okay with it except not sure how to sort out the dock obscuring the bottom buttons of modeler and layout...me new at this so any hints would be good other than making the dock hide from the preference menu......

:D

Captain Obvious
02-19-2005, 04:28 PM
I was heavily participating in the discussion here, of course I remember. The whole point back then was the fact that vendors tend to use the best optimizing compiler they can afford, this is why LW is compiled using the icc on the PC, and not gcc (the de-factor standard on the PC would be msvc anyhow, not gcc). And obviously, since the P4 has a weak fpu, but very strong SSE2, a compiler that doesn't vectorize won't really help, and does make things a bit unfair. Heck, that is like running Power601 code on a G5, doable but pointless if you benchmark. (and heck, I remember my 68060 being faster than the Power601 back then :) ).
Well yeah. It would be pretty darned stupid of them to compile with something that does all floating point action on the P4's rather limited FPU (Intel have been pretty clear about them wanting you to do all that on SSE2) just to "make things fair." But in order to make it slightly less unfair, they really need to move to GCC or IBM's compiler on PPC. From what I've gathered, CodeWarrior is very dated.

Oh, and as for 68k vs PPC, I'm pretty sure there was some 68k code that actually ran faster when emulated on a PPC. ;) But I know what you mean. Of course you should try to optimize. A processor is never really better than the code compiled for it.



i trialed the one in the shop with lightwave de ad all seemed okay with it except not sure how to sort out the dock obscuring the bottom buttons of modeler and layout...me new at this so any hints would be good other than making the dock hide from the preference menu......
Right click (ctrl click) the "handle" in the Dock, Turn Hiding On. Problem solved! ;)

Lightwolf
02-20-2005, 05:13 AM
Well yeah. It would be pretty darned stupid of them to compile with something that does all floating point action on the P4's rather limited FPU (Intel have been pretty clear about them wanting you to do all that on SSE2) just to "make things fair."
As a small trivia note, x86-64 Windows XP doesn't support the FPU at all, only SSE2. So all 64bit windows app will have to use SSE2 for fp (a bit of a bummer for AMD I guess, but looking at where technology is heading with the Cell... more paralellism, it makes sense. Also, it means that task switches will be faster, since the fpu registers don't have to be stored/retrieved for every task).

Cheers,
Mike

Captain Obvious
02-20-2005, 07:56 AM
As a small trivia note, x86-64 Windows XP doesn't support the FPU at all, only SSE2. So all 64bit windows app will have to use SSE2 for fp (a bit of a bummer for AMD I guess, but looking at where technology is heading with the Cell... more paralellism, it makes sense. Also, it means that task switches will be faster, since the fpu registers don't have to be stored/retrieved for every task).

The Opteron and Athlon64 has SSE2, so it's not that big a bummer for them. But it's too bad for the old AthlonXP, I guess. Either way, I'd say it's nice that they (Microsoft) are trying to force some progress. SSE2 is better than the FPU, so why not use it?

Lightwolf
02-20-2005, 08:01 AM
But it's too bad for the old AthlonXP, I guess.
Not really, since it wont run 64bit XP anyhow. And all x86-64bit processors do support SSE2 (unfortunately not all of them support SSE3 though :( ).
Cheers,
Mike

Lynx3d
02-20-2005, 08:25 AM
Windows XP x64 won't run on Athlon XPs anyway, it will only run on 64bit CPUs obviously...

I read AMD basically uses the same execution units for x87 and SSE instructions operating on floats anyway, that's why x87 and scalar SSE(2) performance is almost ideantical. That and the fact that Intel heavily focuses on SIMD to achieve higher throughput is why there are still only 8 80bit x87 registers while there are 16 128bit SIMD registers.
That SSE2 is "better" however depends on what you need, for scientific computations you perhaps would still opt to use the extra internal precision that 80bit floats provide (something that is pretty nonstandard however, PPC doesn't have something anyway), but you'd lose the contents on every context switch *g*

Captain Obvious
02-20-2005, 01:20 PM
Doh! :o I knew I missed something! ;)

cresshead
02-21-2005, 06:10 PM
we've seemed to have drifted off topic to a windows 64 technical overview of instruction set data....you lost me about 8 paragraphs ago! :eek:

Andyjaggy
02-26-2005, 01:14 PM
Soooooooo........... Has anyone used LW8 on a mac mini yet? Or do we need to start another thread.

cresshead
02-26-2005, 01:20 PM
mine's on order!
...may take 6 weeks though!!!!!
it WILL have liw 8 on it!....well 8.2!

i've ordered a 1.42gighz with 512 ram and super drive

Johnny
02-26-2005, 03:12 PM
I thot those units were already shipping? mebbe different for outside US?? I'm weeks or less from ordering 2 1.42Ghz minis myself..

J

Captain Obvious
02-26-2005, 04:16 PM
They are shipping... just not very quickly. ;)