Radiance - Mac Performance

Dear Group

I am in the evaluation phase in my quest to purchase a new computer. From what I have been told, the Motorola G4 processor is far superior in scientific calculations as compared to x-86 based processors (i.e. Intel/AMD). If it weren't for the existing investment that I have in PC based software (particularly software that utilizes a dongle which will not work on a Mac even with virtual PC), I would probably purchase one. But because of the existing investment that I do have in PC based software, I am trying to do a cost versus benifit analysis for purchasing a Mac. Has anyone here been able to do a rough estimate as to the decrease in rendering time when using a Mac as compared to x-86 based processors. I was considering in lieu of purchasing a Mac, I would purchase a dual Pentium 4 Zeon and compile it on Linux. If I am correct, the GNU C compiler has a SSE2 optimization option. Also I have l have looked at puchasing a AMDs 64 bit Opteron based workstation. I do believe there can be some performance gains achieved if it can be used in conjunction with a 64 bit Operating System. Anyway, you all can see that i am very confused as to which direction I should take so any advice is very much appreciated.

Thanks

Marcus D. Jacobs

···

_________________________________________________________________
Add photos to your messages with MSN 8. Get 2 months FREE*. http://join.msn.com/?page=features/featuredemail

I am in the evaluation phase in my quest to purchase a new computer. From what I have been told, the Motorola G4 processor is far superior in scientific calculations as compared to x-86 based processors (i.e. Intel/AMD).

I was told similar, on this list in fact. But the thing is, while that may be true in theory, the bang for the buck king of the hill is still over at AMD. Here is my very unscientific and newbie-esque story: I learned about the wonders of OS X thru Greg Ward; I had plunked my head in the Intel/Lightscape sand for years. But after talking to him about OS X I learned that it was a user friendly UNIX, and that is what appealed to me the most, not any theoretical pipeline/gigaflop superiority. I bought a G4 PowerBook over a year ago, and was thrilled with the performance, but I was still learning, loading pretty small models.

Back in October of last year, I started using Radiance in production, and saw that the time was near that I would need to have a dedicated machine to run Radiance calcs, so I loaded RHL on an old Pentium 450 we had lying around. The calc times on that were about the same as the times on my PowerBook, which is a G4 550 with 512MB of RAM. I was told there was probably something wrong with my hardware (stop laughing at the double-entendre), but a trip to Mark Stock's Radiance timings webpage (http://mark.technolope.org/pages/rad_bench.html) will tell a similar tale. The simple fact of the matter is that you can buy more Radiance-photon-number-crunching-per-dollar/euro/drachma/sheckel-what-have-you with homebuilt PCs than you can with Apple's hardware. That's not to say that Macs are crap. I love my Mac. But where speed is king, in floating point land, I have experienced better calc-per-dollar in the AMD world than that of Apple.

If it weren't for the existing investment that I have in PC based software (particularly software that utilizes a dongle which will not work on a Mac even with virtual PC)...

Virtually Useless PC should be banned. It is too slow to run Notepad efficiently, much less anything actual humans need to use to get some work done.

I would probably purchase one. But because of the existing investment that I do have in PC based software, I am trying to do a cost versus benifit analysis for purchasing a Mac. Has anyone here been able to do a rough estimate as to the decrease in rendering time when using a Mac as compared to x-86 based processors. I was considering in lieu of purchasing a Mac, I would purchase a dual Pentium 4 Zeon and compile it on Linux. If I am correct, the GNU C compiler has a SSE2 optimization option.

I would think the Zeon would outperform the G4, dollar for dollar. But I'd bet AMD can do even better.

Also I have l have looked at puchasing a AMDs 64 bit Opteron based workstation. I do believe there can be some performance gains achieved if it can be used in conjunction with a 64 bit Operating System. Anyway, you all can see that i am very confused as to which direction I should take so any advice is very much appreciated.

Now here is a new horizon that I too am very interested in. I read with delight the tales of joy surrounding the release of the Opteron. SuSE already has a 64 bit version of Linux, and I for one am unclear on what that means for us Radiance people. I'd love to hear some insight about what the 64-bit movement means in terms of performance for the typical rpict process. =8-)

Rob Guglielmetti
[email protected]
www.rumblestrip.org

···

On Thursday, May 29, 2003, at 10:40 PM, Marcus Jacobs wrote:

I'd be very cautious about it, at this time. Greg knows a lot more than I do, of course, but my guess is that compiling Radiance with 64-bit integers and floats would probably show unexpected word-length dependencies. Even so, I'd expect computation speeds to improve because of some technical aspects of the C language.

A rewrite of rholo to make use of a 64- (or even 48-) bit address space would also probably be interesting to experiment with.

Randolph

···

On Thursday, May 29, 2003, at 08:41 PM, Rob Guglielmetti wrote:

Now here is a new horizon that I too am very interested in. I read with delight the tales of joy surrounding the release of the Opteron. SuSE already has a 64 bit version of Linux, and I for one am unclear on what that means for us Radiance people. I'd love to hear some insight about what the 64-bit movement means in terms of performance for the typical rpict process. =8-)

Hi Marcus,

Of course, I'm a big fan of Apple computers. They offer a great combination of quality (i.e., sturdiness, reliability, and aesthetics) over Wintel computers. However, I'm forced to agree with Rob that in terms of peformance vis a vis Radiance, your best value is a probably a home-built, dual-processor Linux box. I very nearly got one myself, but decided instead to pay roughly twice as much to get a dual 1.4 GHz PowerMac G4 with 1 GByte of RAM and a DVD writer for about $3K. Its performance and features are comparable to PCs I could have bought prebuilt for around $1.5K. The advantage is that I don't have to administer a Linux system, which I've found in the past to be a pain in the butt. Mac OS X, built on FreeBSD, is a snap by comparison, and what's better is I can run Word and Excel and all those other Microsoft nasties without having to reboot into another system. Better still, I get to run Photosphere, which I've spent two years developing and don't have time to port to other systems right now.

The hype about the G4 being superior for scientific applications is a PR effort by Apple to save face for a processor that despite many design superiorities, just hasn't kept pace with Intel and AMD's aggressive speed increases. To be fair, you can use the Altvec processor (a short vector SIMD on the G4) to get substantial speed boosts if you go to the extra work to use it. I haven't done any custom hardware coding in Radiance, and I'm unlikely to in the future. Others are welcome, of course.

In your case, a dual-processor Wintel machine with dual-boot into Linux or (my recommendation) FreeBSD may be the best option with your existing software investment.

I don't know what a 64-bit processor will buy in terms of Radiance performance. Other than getting past the 2 GByte memory addressing limit, there might be some improvement in floating point speed since you can load double's into registers, but 64-bit integers are not really useful as far as I'm concerned.

-Greg

···

From: "Marcus Jacobs" <[email protected]>
Date: Thu May 29, 2003 7:40:25 PM US/Pacific
To: [email protected]
Subject: [Radiance-general] Radiance - Mac Performance
Reply-To: [email protected]

Dear Group

I am in the evaluation phase in my quest to purchase a new computer. From what I have been told, the Motorola G4 processor is far superior in scientific calculations as compared to x-86 based processors (i.e. Intel/AMD). If it weren't for the existing investment that I have in PC based software (particularly software that utilizes a dongle which will not work on a Mac even with virtual PC), I would probably purchase one. But because of the existing investment that I do have in PC based software, I am trying to do a cost versus benifit analysis for purchasing a Mac. Has anyone here been able to do a rough estimate as to the decrease in rendering time when using a Mac as compared to x-86 based processors. I was considering in lieu of purchasing a Mac, I would purchase a dual Pentium 4 Zeon and compile it on Linux. If I am correct, the GNU C compiler has a SSE2 optimization option. Also I have l have looked at puchasing a AMDs 64 bit Opteron based workstation. I do believe there can be some performance gains achieved if it can be used in conjunction with a 64 bit Operating System. Anyway, you all can see that i am very confused as to which direction I should take so any advice is very much appreciated.

Thanks

Marcus D. Jacobs

Hi!

First, I have just ordered my 12" Powerbook :wink: Second: my
experience is that rendering speed depends on memory (nothing
is as stupid as a rendering process on a fast machine waiting
for data from the swap!!!), cpu frequency and - that is really
important - the optimization of the compiler you are using!

I couldn't find great imporvements with radiance on 64bit so far.

However, one thing to mention - a stable 800MHz machine doing three
renderings over night is much more performant than a 3200MHz
machine which crashes three timer per rendering :wink: In my personal
experience, the "rendering speed" is not the critical point any more
in our days, much more is the stability of machines. I put 90% into
the preparation of a rendering, and only 10% in the actual render
process :wink:

So, these were some thoughts of someone who prefers to render his
pics on an old AMD K6-2 300 :wink: CU Lars.

···

--
+++ GMX - Mail, Messaging & more http://www.gmx.net +++
Bitte l�cheln! Fotogalerie online mit GMX ohne eigene Homepage!

Lars Grobe wrote:

First, I have just ordered my 12" Powerbook :wink:

Congrats! The new 12" ones are so very adorable. The aluminum cases are georgeous; the finish on my Ti PowerBook started flaking pretty early on, I'm glad they did something about it.

Second: my

experience is that rendering speed depends on memory (nothing is as stupid as a rendering process on a fast machine waiting for data from the swap!!!)...

Oh definitely; I think we're all making "rendering comparisons" of machines that are *not* relying on virtual memory at all. At least I was always talking about relatively small models.

However, one thing to mention - a stable 800MHz machine doing three renderings over night is much more performant than a 3200MHz machine which crashes three timer per rendering :wink:

Agreed, but stability has not been an issue for me at all. The computer, I mean. =8-)

Also, Greg and Jeffrey mention one BIG thing about OS X that I did not: the OS is just so darn easy to administer!! So, that should surely weigh in to the equation. With OS X, all I did was load the Radiance binaries and start using Radiance. I was able to jump into Radiance without spelunking into the depths of /etc, learning enough unix to get by, and experimenting as time allowed. It's like unix with training wheels. It's just that for me, speed is king and money is in short supply, so I put up with Linux. Still prefer working on the Mac tho. Jeffrey, if you find any good CAD applications that run on OS X do let us know.

···

----

      Rob Guglielmetti

e. [email protected]
w. www.rumblestrip.org

Jeffrey, if you find any good CAD applications that run on OS X do let
us know.

For Radiance models, formZ is a great tool. CU Lars.

···

--
+++ GMX - Mail, Messaging & more http://www.gmx.net +++
Bitte l�cheln! Fotogalerie online mit GMX ohne eigene Homepage!

I have good reports of SketchUp 3D, though I haven't used it myself. <http://www.sketchup.com/>

Randolph

···

On Friday, May 30, 2003, at 12:24 PM, Rob Guglielmetti wrote:

if you find any good CAD applications that run on OS X do let us know.

I gave SketchUp a try as it has a free demo for OS X, and it seems fairly capable. It's certainly simple enough, and it outputs .3ds format, which you can convert to MGF using 3ds2mgf and then to Radiance with mgf2rad. It's material model is by layers, but that should work for most basic applications. I'm really not a CAD user, but watching the short online tutorial was enough that I was able to create simple shapes and join them up right away. If you're a casual as opposed to a serious CAD user, it seems like a good way to go for simple models. The fact that they give you a fully functional demo that lasts for 8 hours of use is nice as well. The license cost is around $700.

-Greg

···

From: Randolph Fritz <[email protected]>
Date: Fri May 30, 2003 3:41:32 PM US/Pacific
To: [email protected]
Subject: Re: [Radiance-general] Radiance - Mac Performance
Reply-To: [email protected]

On Friday, May 30, 2003, at 12:24 PM, Rob Guglielmetti wrote:

if you find any good CAD applications that run on OS X do let us know.

I have good reports of SketchUp 3D, though I haven't used it myself. <http://www.sketchup.com/>

Randolph

Thanks for the quick review and study of Radiance compatibility. Clarification: do you mean that its native format is 3DS? or does it export to 3DS format?

I know of a major Portland Oregon architecture firm that brought in one copy of SketchUp and, over the next few months, the whole office switched, so--at least for their type of design--it's probably a capable tool for schematic design as well as simple lighting study models.

Randolph

···

On Friday, May 30, 2003, at 08:25 PM, Greg Ward wrote:

I gave SketchUp a try as it has a free demo for OS X, and it seems fairly capable. It's certainly simple enough, and it outputs .3ds format, which you can convert to MGF using 3ds2mgf and then to Radiance with mgf2rad. It's material model is by layers, but that should work for most basic applications. I'm really not a CAD user, but watching the short online tutorial was enough that I was able to create simple shapes and join them up right away. If you're a casual as opposed to a serious CAD user, it seems like a good way to go for simple models. The fact that they give you a fully functional demo that lasts for 8 hours of use is nice as well. The license cost is around $700.

It exports to .3ds format -- the native format is its own. I tested it with 3ds2mgf and it seems to work fine. The nice thing about it is you can be as sloppy or as accurate as you like, and it seems pretty easy to go back and make changes to the model as well, since all the vertices and faces hold their shape as you adjust them.

-Greg

···

From: Randolph Fritz <[email protected]>

Thanks for the quick review and study of Radiance compatibility. Clarification: do you mean that its native format is 3DS? or does it export to 3DS format?

I know of a major Portland Oregon architecture firm that brought in one copy of SketchUp and, over the next few months, the whole office switched, so--at least for their type of design--it's probably a capable tool for schematic design as well as simple lighting study models.

Rob Guglielmetti wrote:

> I am in the evaluation phase in my quest to purchase a new computer.
> From what I have been told, the Motorola G4 processor is far superior
> in scientific calculations as compared to x-86 based processors (i.e.
> Intel/AMD).

I was told similar, on this list in fact. But the thing is, while that
may be true in theory, the bang for the buck king of the hill is still
over at AMD.

You always have to say what you're comparing to. What Apple
compares in their PR is the CPU frequencies. Since the PowerPC
is a true RISC CPU, it will process about twice as much data per
cycle. This means that eg. an 800 Mhz Mac will roughly equal
a 1600 Mhz PC running Linux (your mileage may vary with Windows).

It does *not* mean that this power will come at an equal price.
There's simply no way for Apple to compete with the mass produced
parts in a PC here, even if they have switched to use some of
those as well.

so I loaded RHL on an old Pentium 450 we
had lying around. The calc times on that were about the same as the
times on my PowerBook, which is a G4 550 with 512MB of RAM. I was told
there was probably something wrong with my hardware (stop laughing at
the double-entendre),

Given the similar frequency range, it is indeed very likely that
something is wrong here. Note that laptops (on both sides) use
different types of CPUs and probably other components, which will
slow down a lot of stuff, even when the system doesn't switch to
sleep mode in between... Just think of the hard disks for
example, which rotate relatively slow to reduce noise and power
consumption.

  The simple fact of the matter is that you can buy more
Radiance-photon-number-crunching-per-dollar/euro/drachma/sheckel-what-
have-you with homebuilt PCs than you can with Apple's hardware.

Mass production and competition between various manufacturers
do have a noticeable effect.

> the GNU C compiler has a SSE2 optimization option.

CPU specific optimizations may add another dimension to the
equation. Compilers will try to do something into that direction,
but to get the most bang for the buck, some small key routines
would have to be rewritten in assembler. Imagine a vector-matrix
multiplication finishing in less than ten cycles on an Altivec
CPU, instead of the several dozen it currently takes, and the
Mac comparison would look very different to what it does now.

> I do believe there can be some performance gains achieved
> if it can be used in conjunction with a 64 bit Operating System.

SuSE already has a 64 bit version of Linux, and I for one am unclear on
what that means for us Radiance people. I'd love to hear some insight
about what the 64-bit movement means in terms of performance for the
typical rpict process. =8-)

Probably not much. The main advantage of a 64 bit system is the
larger address range. There may be compiler options that allow you
to load several values into the same registers together, but I
wouldn't expect any wonders from that if the software isn't written
with it in mind. And other (or maybe the same) compiler options will
cause harm, if they change eg. the size of an int.

-schorsch

···

On Thursday, May 29, 2003, at 10:40 PM, Marcus Jacobs wrote:

--
Georg Mischler -- simulations developer -- schorsch at schorsch com
+schorsch.com+ -- lighting design tools -- http://www.schorsch.com/

Georg Mischler wrote:

You always have to say what you're comparing to. What Apple
compares in their PR is the CPU frequencies. Since the PowerPC
is a true RISC CPU, it will process about twice as much data per
cycle. This means that eg. an 800 Mhz Mac will roughly equal
a 1600 Mhz PC running Linux (your mileage may vary with Windows).

Well, that's exactly what just didn't add up in my (admittedly basic) tests. But Mark Stock's Radiance speed tests page (http://mark.technolope.org/pages/rad_bench.html) says the same thing. The "RISC two-for-one-special" doesn't seem to be present when comparing Radiance render times rom Apple & PC. Pillo did a bunch of tests on G3 and G4 laptops; not sure if there is a desktop Mac timing on there. Greg, maybe you wanna load up the test model on your new Mac hot rod??

Given the similar frequency range, it is indeed very likely that
something is wrong here. Note that laptops (on both sides) use
different types of CPUs and probably other components, which will
slow down a lot of stuff, even when the system doesn't switch to
sleep mode in between... Just think of the hard disks for
example, which rotate relatively slow to reduce noise and power
consumption.

Yes, but my tests were done on small models that went straight into RAM and stayed there. I turned off all power save functions as well, as I suspected that might have been the problem. Perhaps there's still something off. Dunno.

Probably not much. The main advantage of a 64 bit system is the
larger address range. There may be compiler options that allow you
to load several values into the same registers together, but I
wouldn't expect any wonders from that if the software isn't written
with it in mind. And other (or maybe the same) compiler options will
cause harm, if they change eg. the size of an int.

Thanks for this info...

···

----

      Rob Guglielmetti

e. [email protected]
w. www.rumblestrip.org