PDA

View Full Version : FR or Maxwell- It's all about speed NOW!


soccerrprp
12-09-2005, 01:28 PM
Greetings!

I know that there has been a great deal of disappointment with Maxwell's renderer. But, I would like to keep it as an option until they get their act together.

1. Provided that Maxwell gets it together, has it's renderer been promising? I mean earlier versiion for c4d users?
2. I know that there are some who have and are using the latest version, how does it compare to FR?

Clearly, if Maxwell cleans up its latest version, the issue will not be quality for me. I've seen some remarkable stills with both, but the issue of SPEED will be an issue. I don't have multiple cpu's to my disposal (only 1 PIV) and would like to know which of these will give me the fastest renders on average. Also, how they render animations....

Any insight? Help?

Thanks all. Will need to make a decision by late january or early february to purchase.

duderender
12-09-2005, 01:59 PM
People expect everything instantly and its not going to be that way. You have to remember that the philosphy behind Maxwell and finalRender are opposite. MW wants to go you powerful materials with little interaction, where finalRender is about all control.

So, that being said where is the tradeoff?? Well, consider the following. If you were to add up all the time from beginning to end of applying materials, tweaking and rendering do you really think it would be that far apart? Obviously, it may be if you used canned materials... however because Maxwell is using brsf curves for materials, when you want aluminum it looks like it so your tweaking isn't necessarily so involved.

So, because maxwell takes 30 hrs to render, you just spend 30 hrs perfecting materials in native c4d. I hope I illustrated this well.

I have both, and while NL really goofed up on their development path I do think it has huge potential but right now I'll concentrate on finalRender.

Here's an example using Standford's 'dragon' at 7 million polygons and skydome GI with an area light. I wanted to push c4d and fR with poly limits. For more high poly goodness: http://www-graphics.stanford.edu/data/3Dscanrep/

http://www.jase.ca/content/forumpost/finalrender/stanford_dragon_01.jpg

soccerrprp
12-09-2005, 02:19 PM
So, Maxwell simply uses the c4d render engine better, correct? Or more efficiently? Whereas, FR is a separte program that provides additional functionality?

Maxwell should be easier to use, but what of speed? Does this mean that Maxwell's speed is no greater than AR's? The images are only better?

FR, from what I have seen/read/heard is faster, correct?

paulselhi
12-09-2005, 02:20 PM
I don't see why that dragon needs 7 million polygons, there realyy is not that much detail, ZBrush could do a much more impressive job with around a million polys


The render is also very flat if there is more details then we are seeing then the render is lacking

moka.studio
12-09-2005, 02:27 PM
Hi Duderenedr,
nice rendering! Did you build the shaders with the shader tree or as C4d materials?


I am not sure I understand your point...
Are you saying that although MAxwell takes longer to render, it makes up for this by needing less tweaking of materials ?
I think regardless of what render engine you are using, if you are using it in production you will have a library of materials that will cover many of your needs ( though not all).
I am slowly starting to build up a library wit FR shaders for exemple.
The same is true for light rigs, if you are working on similar kinds of projects.

moka.studio
12-09-2005, 02:29 PM
I don't see why that dragon needs 7 million polygons, there realyy is not that much detail, ZBrush could do a much more impressive job with around a million polys


The render is also very flat if there is more details then we are seeing then the render is lacking

What does Zbrush have to do with this??
He did not post it as an exemple of modelling, but as an exemple of rendering high polygon counts with FR.

JamesMK
12-09-2005, 02:30 PM
So, Maxwell simply uses the c4d render engine better, correct?

- No, it's got nothing to do with the C4D render engine. It's an engine all by itself.

Or more efficiently?

- See above

Whereas, FR is a separte program that provides additional functionality?

- FR is a separate renderer (too), and plugs into C4D as a render effect.

Maxwell should be easier to use, but what of speed?

- The Maxwell method is inherently slower than FR

Does this mean that Maxwell's speed is no greater than AR's?

- On most scenarios, Maxwell is a lot slower than any other render engine.

The images are only better?

- Since Maxwell uses full spectral path tracing (or something to that effect) they do tend to produce an unsurpassed lighting quality. But for a price, and the price you pay is very long rendertimes.

FR, from what I have seen/read/heard is faster, correct?

- It's several magnitudes faster than Maxwell, and most often a few times faster than AR too (with a few minor exceptions). If you do GI with blurry effects for instance, FR runs circles around AR. If you do ordinary GI, FR is still faster than AR on average, and most of the time it's easier to get a good solution without having to fiddle around a lot.

soccerrprp
12-09-2005, 02:30 PM
I don't see why that dragon needs 7 million polygons, there realyy is not that much detail, ZBrush could do a much more impressive job with around a million polys


The render is also very flat if there is more details then we are seeing then the render is lacking

Paulselhi,

In your opinion, does zbrush renders rival what you have seen from maxwell and/or fr? Or are you talking about the dragon render specifically.

I must admit, price is also a consideration and if zbrush can render nicely and quicker, so be it.

moka.studio
12-09-2005, 02:34 PM
Paulselhi,

In your opinion, does zbrush renders rival what you have seen from maxwell and/or fr? Or are you talking about the dragon render specifically.

I must admit, price is also a consideration and if zbrush can render nicely and quicker, so be it.

huh, no.
BUt (see above) I think he was refering to the modelling.

One more point. If you are looking to buy a render engine now, and want to do renderings with GI, then get FR.
It is there and is functional.
Waiting for Maxwell can mean waiting fora very long time.

AdamT
12-09-2005, 02:36 PM
>So, Maxwell simply uses the c4d render engine better, correct? Or more
>efficiently? Whereas, FR is a separte program that provides additional
>functionality?

No, absolutely not. Both are standalone render engines that are accessed through Cinema. But fR is integrated extremely well into Cinema, so it almost seems like a different native engine. Maxwell is not integrated very well (launches separate render program, requires Maxwell material and emitter tags, still doesn't support animation, etc.).

>Maxwell should be easier to use, but what of speed? Does this mean that
>Maxwell's speed is no greater than AR's? The images are only better?

M~R should be easier to use--in theory. But the jury is still *way* out on this one. In the beta version it was easy to set up lights because they worked like real lights, e.g., set a table lamp to 75W and it would produce the correct lighting. In RC 1-3 this is no longer the case. Also, materials in the RC are more complex and less predictable than they were in the beta.

One might *assume* that they will get the RC back to where the beta was eventually, but you have to consider that the internals of the RC are almost completely rewritten--because NL simply couldn't solve some important issues with the approach they were taking with the beta (clipmaps, sunlight through transparent materials, etc.). Therefore, comparing M~R beta to M~R RC is very much like comparing apples to oranges. So far the RC renders a lot faster than the beta--but still much slower than fR or AR--but image quality is seriously lacking compare to anything.

>FR, from what I have seen/read/heard is faster, corr
ect?

Yes, it's faster than AR when it comes to GI, and lightyears faster than M~R.

At this point I would recommend that people stay far far away from Maxwell. It may eventually be a worthwhile solution, but it's got a long way to go before it gets there and it's arrival is in no way certain. Just have a look at the M~R forums, which have become more or less one long rant against NL, its policies, procedures, and the RC(s).

soccerrprp
12-09-2005, 02:39 PM
One more point. If you are looking to buy a render engine now, and want to do renderings with GI, then get FR.
It is there and is functional.
Waiting for Maxwell can mean waiting fora very long time.

I kind of thought so. Yah, I've been reading the Maxwell forum at Maxwell and it appears that "they" are making some progress (literally daily) with the new renderer. I'm really hoping that "they" will get it working SOON so that speed comparisons can be made.

Thanks Moka-j.

Oh, crud. Thanks AdamT. Good information. Your post came up after i responded to Moka-j. I have a client that is interested in both and has asked for my humble and limited opinion. I've been "listening on" to discussions about both and FR seems to be the sure bet for now.

paulselhi
12-09-2005, 02:47 PM
What i meant was why would you want to render a 7 million poly count dragon anyway seeing as the amount of detail could be far more impressive with a seventh of that poly count, i mean FR can render 7 million polys, well on my paltry amd 1.6 Ghz 256 MB RAM i can easily render 1 million polys with AR

here is a very rough example.. i was just doodling one day and it could take far more detail with the poly count as is (about 700.000+)

The point is AR can easily handle this

http://www.black-and-white-to-color.com/stuff/zbalien.jpg

soccerrprp
12-09-2005, 02:50 PM
Thanks Paulselhi!

moka.studio
12-09-2005, 03:05 PM
What i meant was why would you want to render a 7 million poly count dragon anyway seeing as the amount of detail could be far more impressive with a seventh of that poly count, i mean FR can render 7 million polys, well on my paltry amd 1.6 Ghz 256 MB RAM i can easily render 1 million polys with AR

here is a very rough example.. i was just doodling one day and it could take far more detail with the poly count as is (about 700.000+)

The point is AR can easily handle this



I still don't get your point.
1. It is fair to assume that you have more experience setting up scenes with C4d, than Duderender does with FinalRender. I don"t think his rendering was meant to show the capacities of FR to deliver Crisp detail. It is only a test.

2. You are using Normal Maps I assume with your Model?
FR can use Normal Maps as well to get better results. Again, It has a lot to do with familiarity with the render engine.

3 In general, I don\t think someone made the point that AR cannot handle large amounts of detail.

4 Why would someone want to render 7 milion polys?
well why not?
It is a test after all, and the model is part of the Princeton Repository ( see link), so that it is an accessible high-poly model for this purpose.

paulselhi
12-09-2005, 03:14 PM
No normal maps..pure geometry.

Again the point that AR can handle the necessary poly count so what is so impressive about FR handling large poly counts ? since my 1.6 Ghz 256 MB ram handles 1 mill polys well i would imagine that all the big boys with their supa doopa machines would breeze through several million polys with AR

Agreed FR is very fast for GI but AR is still a very good renderer

duderender
12-09-2005, 03:16 PM
Wow, so many posts in such a short time that I will try and answer all them...

@paulselhi: I think you're being far more critical than you need to be. You're the first to not like my render, but if you choose to and then follow up with a 'flat' looking render to counter... that's not so cool. My exercise was not one of efficiency. Of course no one would want a 7million polygon dragon, however my test was to exercise the memory constraints of such a dense polygon scene to see the memory management between c4d and finalrender. Normal mapping IS NOT detail either. Its simply a fakery and does not cast shadows, etc.. etc...

@moka-j: the render is fR materials and the shadows is using GI and 1 area light.

@soccerrprp: both maxwell and finalrender are 'external' rendering engines to replace c4d's native renderer and AR. finalrender is exceptionally well integrated within the c4d interface so it is just like it being native to c4d. Maxwell on the other hand is more standalone in nature but with the plugin it does integrate well.

duderender
12-09-2005, 03:19 PM
No normal maps..pure geometry.

Again the point that AR can handle the necessary poly count so what is so impressive about FR handling large poly counts ? since my 1.6 Ghz 256 MB ram handles 1 mill polys well i would imagine that all the big boys with their supa doopa machines would breeze through several million polys with AR

Agreed FR is very fast for GI but AR is still a very good renderer

I will tell you exactly why. I deal with CAD converted geometry for rendering and remeshing is very time consuming and runs the risk of errors. Therefore its not unusual to see scenes get into the millions of polygons.

And no, there is limit to what AR can do on a 32bit console. There is talk of fR and C4D doing some geometry caching to mimic that of a 'gigapoly' core ala XSI.

williamsburroughs
12-09-2005, 03:24 PM
Nice test duderender...glad to see you pushing the threshold of what fR can do in a 32bit environment. I know CAD people will be pleased with your tests and findings.

Keep them coming!

Cheers,
-policarpo

duderender
12-09-2005, 03:27 PM
Nice test duderender...glad to see you pushing the threshold of what fR can do in a 32bit environment. I know CAD people will be pleased with your tests and findings.

Keep them coming!

Cheers,
-policarpo

Cheers dude.

Here's one coming from CAD, with straight GI... no materials. Model is about 60% done, so note its at 1.8 million polys:
http://jase.ca/content/forumpost/finalrender/fr2_prms_test_07.jpg

williamsburroughs
12-09-2005, 03:31 PM
Ho ho mamma! That looks friggin sweet. What was the render time on that one?

moka.studio
12-09-2005, 03:38 PM
Nice one!
btw, you can add additional variables to the FR stamp,
adding %rendertime will show the render time, which can be usefull for tests.

soccerrprp
12-09-2005, 03:40 PM
Sorry guys, someone mentioned that MW does not support animation as of yet? Did I read right?

duderender
12-09-2005, 03:53 PM
Ho ho mamma! That looks friggin sweet. What was the render time on that one?

Going by memory, it was a minute or two. Bear in mind this is using 3 or 4 processors.

@moka-j: yes I usually edit them, but sometimes I forget :P

@soccerrprp: you can do them, but its ridiculously painful without the plugin.

paulselhi
12-09-2005, 05:16 PM
Cheers dude.

Here's one coming from CAD, with straight GI... no materials. Model is about 60% done, so note its at 1.8 million polys:


Again .. why the need for 1.8 million polygons for this model ? Just beacuse you can ? I am sure this could be modelled and rendered with a fraction of the poly count
Now if you were tlaking about something similar to vue's ecosystem with billions of polygons justified by heaps of individual objects then that would be a different matter but why would anyone want to model something like you have shown with 1.8 million polygons ?

lllab
12-09-2005, 05:17 PM
well as a professional i guess it is not so much a war between which is better or faster etc.

it is more as a photographer having more than one camera. each is good for a special task.

maxwell is good for really testing light simulations- it is not a renderer it is a simulator in fact. but i think it is missleading to say it will be easy to handle- not really i guess- it's still the user - like with photography again:-)

FR is great with GI, AA, very deep customization, some poeple might think it is what we wished AR should be, but still it misses some other features AR has.

AR is a great overall workhorse, with only a few weaknesses, thats missing buckterendering(+DR) and sloooow/old GI. but it has excellent area light very good shaders excellent workflow, and very powefull, integrated tools.

to be not limited to only one "camera" is a very nice thing for c4d to have, and in my opinion one major step toward the big ones ( maya, max etc).

and to be honest i guess for all who can afford it i think you need all 3 of them, at least the people making their living from it will be happy to buy all three of them, they are no alternatives, they all complement each other well...

cheers
stefan

duderender
12-09-2005, 05:30 PM
Again .. why the need for 1.8 million polygons for this model ? Just beacuse you can ? I am sure this could be modelled and rendered with a fraction of the poly count
Now if you were tlaking about something similar to vue's ecosystem with billions of polygons justified by heaps of individual objects then that would be a different matter but why would anyone want to model something like you have shown with 1.8 million polygons ?

*yawn* please go about what you were doing.. you missed the point. Thanks for looking.

paulselhi
12-09-2005, 05:33 PM
*yawn* please go about what you were doing.. you missed the point. Thanks for looking.

no it is you who are in dream world.. the post started as a maxwell thread then moved on to the wonders of FR..these wonders were being shown as the ability to render multi million polygon scenes which AR can do just as well

lllab
12-09-2005, 05:36 PM
the "wonders" of FR are to have state of the art GI and AA and MB...

not less not more.

in regards of polycount i guess c4d 64bit is the winner of all at the moment.

stefan

duderender
12-09-2005, 05:46 PM
the "wonders" of FR are to have state of the art GI and AA and MB...

not less not more.

in regards of polycount i guess c4d 64bit is the winner of all at the moment.

stefan

fR will be 64bit ready according to Edwin at Cebas. Not sure when, as they are finalizing fixes for 9.52 and Mac version.

However, Edwin has mentioned another geometry processing method but the current C4D SDK limits what they can do at the moment, but the idea was to render with 'gigapoly' on a 32bit platform.

However, I'm curious with 64bit because Maxon state 20% improvement with render speed. 64bit doesn't make things faster per se but instead allows for more to be processed at a given time so in the long run you see workflow gains.

I have heard that high poly count scenes give Maxwell a bit of a hard time, maybe Adam can comment as I've not used Maxwell for that.

Srek
12-09-2005, 05:47 PM
Please keep it civil guys.

Cheers
Björn

paulselhi
12-09-2005, 05:56 PM
Agreed Gi is much faster with FR, not sure about MR, and motion blur and dof are well sorted in FR. But the polycount abilities are really something else. I just don;t see the point in showing us a zillion poly model rendered in FR that would just as good with several hundred thousand polys

For example this model of just some thing i was toying with and gave up on simply because i did not consider the poly count which i am sure i could massively reduce. it comes in at about 36000 polys..does it really need x miilion polys to show the wonders of the rendered used ?

It renders in AR with radiosity at about 3 mins of my single processor super duper box

http://www.black-and-white-to-color.com/stuff/apugi.jpg

lllab
12-09-2005, 06:21 PM
all civil here at least :-)

well i give up, some dont want undertstand...doena tmatter anyway...

ar64 is best at polycount

fr is best with beautiful and fast GI not mattert what polycount

mr is most real, it is very efficent with polycount NL claimes. meaning it has almost same speed with few or millions of polys.- well that has to be prved first-. but to be fair mr is the only one not finished yet. so lets wait to judge it too early.

so conclusion- ALL of them are very good renderengines- each in their areas- as i said before:-) you might want all 3 of them. there is really no need for this kindergarden-like "myrenderengineisbetter" stuff.

my3cents
srefan

williamsburroughs
12-09-2005, 06:24 PM
Agreed Gi is much faster with FR, not sure about MR, and motion blur and dof are well sorted in FR. But the polycount abilities are really something else. I just don;t see the point in showing us a zillion poly model rendered in FR that would just as good with several hundred thousand polys

For example this model of just some thing i was toying with and gave up on simply because i did not consider the poly count which i am sure i could massively reduce. it comes in at about 36000 polys..does it really need x miilion polys to show the wonders of the rendered used ?

It renders in AR with radiosity at about 3 mins of my single processor super duper box



I think you're missing the point. DudeRender is experimenting how to use fR with CAD based derived models. CAD=High poly counts, because they tend to be used for engineering and rapid prototyping (ie: Virtual Models of the Real thing).

Hope that makes sense.

Anyhoo, keep up with the posts DudeRender...loving the examples.

Cheers,
-policarpo

paulselhi
12-09-2005, 06:41 PM
So is it standard practice for cad modellers to render a cube with several million polys when 6 would suffice?

why not show FR for what it can really do ? If poly count gets you bashing the bishop then why not have a render of a model that justifies million + polys insterad of something that could be well modelled with a few (hundred ) thousand ?

lllab
12-09-2005, 07:55 PM
please come down and dont bother other users...

stefan

williamsburroughs
12-09-2005, 08:00 PM
So is it standard practice for cad modellers to render a cube with several million polys when 6 would suffice?


Well now that's just plain silly.

It would be triangulated so 6 sides wouldn't really work.

:rolleyes:

Cheers,
-policarpo

Srek
12-09-2005, 08:18 PM
So is it standard practice for cad modellers to render a cube with several million polys when 6 would suffice?

Maybe you missed the point that the dragon is not modeled but generated from 3D Scanner Data. Thats why it has that insane high polycount. It's not realy intentional.
Cheers
Björn

JamesMK
12-09-2005, 08:43 PM
However, I'm curious with 64bit because Maxon state 20% improvement with render speed. 64bit doesn't make things faster per se but instead allows for more to be processed at a given time so in the long run you see workflow gains.
You are correct in that 64 bit does not offer any speed increase per se - however, when compiled, they can remove some of the compatibility code otherwise needed to ensure that C4D can run on, say a run of the mill P4. And by removing that code, the compiler can use only the faster calls that are available on more recent processors, hence the speed increase.

.

Ernest Burden
12-09-2005, 09:16 PM
Well now that's just plain silly.
It would be triangulated so 6 sides wouldn't really work.

It would be 6 quads if I were working with it.

Sometimes large counts come from having large numbers of medium complexity objects in a scene. I may have a fairly simple room that is jammed full of ornate tables and chairs, maybe a vase of flowers on each table--no one object is all that big in polycount, but hundreds of them? I have one client that loves to build these glass rooms with walls of wine bottles--that adds up quick.

moka.studio
12-09-2005, 10:28 PM
paul ,
sometimes you work with what you get, I have had to do renderings of scenes with 100,000's polys, which I would have modelled using at most 20,000 polys or so....
Just there is often not enough time to re-model things....

Anyway that is going away from the point of the thread, whic was to try to compare ( objectively) the advantages and weaknesses of render engines...

jabbermacy
12-09-2005, 10:48 PM
So is it standard practice for cad modellers to render a cube with several million polys when 6 would suffice?

why not show FR for what it can really do ? If poly count gets you bashing the bishop then why not have a render of a model that justifies million + polys insterad of something that could be well modelled with a few (hundred ) thousand ?

I don't agree with the tone, but agree with the message. I think the point was to see how many polys it takes to crash the renderer, not the efficiency of the modelling. Just take a hit and chill my man :)

Primitiv
12-09-2005, 11:50 PM
I am with Paulsehi on this one. Good work habits and poly count economy is a must. Handling zillions of polys is one thing, but generating well constructed models is just as important, if not more.

Maybe the point was not the number of polys, but it was the point made by Duderender anyway: "Look how well FR handles this 7 million polys dragon". Yes, well, it is nice. But this model could be done far more economically. Why waste the render abilities on something that does not need to be so huge to begin with? And I also agree that the detail is just not there given the poly count. Look at what that guy here does with quite low poly count and good mapping in zBrush: http://www.mattcioffi.com/cgsamples.htm

If you have to make a point, then make a good one.

umblefugly
12-09-2005, 11:58 PM
You also should remember that this is a scanned object and therefor your ZB arguments are garbage. Give the guy a break. Its a polygon count test not a detail test.Soon ill have the cash after this project to go to 9.5 and have left over cash which ill be spending on FR2. Maxwell is not a choice for me no matter the hype. Cheers

OOps cross posted with Bjorn.

williamsburroughs
12-10-2005, 03:26 AM
I am with Paulsehi on this one. Good work habits and poly count economy is a must. Handling zillions of polys is one thing, but generating well constructed models is just as important, if not more.
http://www.mattcioffi.com/cgsamples.htm

If you have to make a point, then make a good one.

To know what you do is one thing.
To do what you know is one thing.

But to know what it is you do, this is the thing to strive for.

Cheers,
-policarpo

Ernest Burden
12-10-2005, 01:11 PM
I am with Paulsehi on this one. Good work habits and poly count economy is a must. Handling zillions of polys is one thing, but generating well constructed models is just as important, if not more.

And a point for another thread.

I have been modeling in 3D daily since 1987. I know how to model with an economy of polygons. You have to with 16Mhz processors. Those habits stick with you. But models sometimes get huge because my scene is a very large area filled with many, many economicly built objects. How about a parking area with hundreds of cars--each car can be only a few hundred polys (though most people wouldn't go much below 1000 polys for a car), but as a group, it adds up.

So I don't really care whether the dragon should have been 2000 polys, if fR2 can render it at several million, then it could render a scene of mine if it got that big despite my low-poly approach.

That's the point of the thread, isn't it?

lllab
12-10-2005, 02:12 PM
exactly!

i absolutly second earnests post.
there are fields where you just have to deal with high polycounts, one of it is architectural scenes. sometimes they can happen to be very big, no matter how efficient modelling was done...

cheers
stefan

duderender
12-10-2005, 05:05 PM
Ernest, Stefan, Policarpo, Bjorn and the like... thanks for actually reading the posts I made and the point I making.

The thing I learned from the example is that on a 32bit console, with 2GB of RAM I was able to push a scene to 7 million polygons and allow finalrender to push that scene across a few computers over the network.

To the original poster, I was showing that with finalrender and a large polycount scene and a decent HD res image, you could achieve something in an hour. Not so with Maxwell.

I think its important to find the limits of what you're working with. So in this case I find something with a large amount of polygons. I've already pushed AR2 to 5 million so I wanted more.

Obviously, in this thread I do not do a common thing and that is working with existing CAD models to generate visualizations for clients. So what's important? Simple, detail and lots of it. In addition, you have to be able to zoom right up close to an object to illustrate an edge, and if it was low poly on a curved surface then you are pretty much screwed.

Please get back to topic, if you are interested more about CAD rendering then we can create a thread for it and that way we don't necessarily have to entertain those ignorant on the subject.

Now that I have the plugin for RC3 (need to check if RC4 is available) I can test this same scene in Maxwell.

Stay tuned...

lllab
12-10-2005, 06:09 PM
i would wait a few days. rc3 is not ery good in rendering at the moment. maybe in rc4 or 5 it will be better again.

cheers
stefan

soccerrprp
12-10-2005, 06:19 PM
Thank God! We seem to be getting back on topic. Duderender, I look forward to your comparisons (in speed) with maxwell and FR. [my original topic!]. :applause: :D

shetland
12-11-2005, 05:15 AM
Taking this back on topic (slightly),

I've noticed the doing GI with AR on the mac (G5 -dual 2.0) is outperformed by a std 3.0 Pentium. Not by a lot (a typical scene I'm working with - 20-30%). Side by side with std. raytracing, then the edge goes to the mac, but by only 30%. So, either GI does not lend itself to much threading, or it is significantly more optimized for the PC architecture. By no means are my tests comprehensive, but in general, I have found this to be true. I wish it were not so. (given my investment in the G5 vs the pc).

Would it be a fair assumption or simply outlandishly speculative to assume that if one holds two cinema licenses, one mac, one pc, one should purchase FR for the pc? Would the bucket rendering still work with a mixed network - or is this per homogenous systems, e.g., all pentiums, all athlons, etc?

I ask this knowing that FR is not yet available for OS X in any case. Apologize in advance if I have erected a lame light on my head!

Srek
12-11-2005, 08:46 AM
Taking this back on topic (slightly),

I've noticed the doing GI with AR on the mac (G5 -dual 2.0) is outperformed by a std 3.0 Pentium. Not by a lot (a typical scene I'm working with - 20-30%). Side by side with std. raytracing, then the edge goes to the mac, but by only 30%. So, either GI does not lend itself to much threading, or it is significantly more optimized for the PC architecture. By no means are my tests comprehensive, but in general, I have found this to be true. I wish it were not so. (given my investment in the G5 vs the pc).

It's not that surprising. Only very few algorithms can make optimal use of multithreading.
The 2GHz dual G5 has a single Cinebench of 219 and multi 401. For the 3 GHz P4 it's 310 and 365. So if the algo in question would be optimal for MT the G5 would have an advantage of 10%. For Single Threading the P4 has an advantage of 40%.
The more surprising part is that in your case the mac is up to 30% faster then the PC in some cases ;)

Cheers
Björn

duderender
12-11-2005, 10:19 AM
i would wait a few days. rc3 is not ery good in rendering at the moment. maybe in rc4 or 5 it will be better again.

cheers
stefan

Am I missing something or does RC3 not have the server.exe? To be fair, I'd like to have MW rendering over multiple CPU's.

-j

lllab
12-11-2005, 12:04 PM
server is includet in main renderwindow now, but is diabled in rc3.
too you have to wait for next releases-

cheers
stefan

shetland
12-11-2005, 04:20 PM
It's not that surprising. Only very few algorithms can make optimal use of multithreading.
The 2GHz dual G5 has a single Cinebench of 219 and multi 401. For the 3 GHz P4 it's 310 and 365. So if the algo in question would be optimal for MT the G5 would have an advantage of 10%. For Single Threading the P4 has an advantage of 40%.
The more surprising part is that in your case the mac is up to 30% faster then the PC in some cases ;)

Cheers
Björn


Thank you Bjorn. I appreciate the thoughts & insight.

See, I think you are also correct in your statement, that the occasional results in my tests which show the G5 outperforming should be considered suspect: - this becomes understandable when one realizes the context of the testing; when I exercise the pentium in this regard, I set cinema's priority to low, run multiple applications in the foreground, including Apple's well known, "G5 Benchmark optimizer" :applause:

kind regards.

shelley

duderender
12-11-2005, 05:54 PM
server is includet in main renderwindow now, but is diabled in rc3.
too you have to wait for next releases-

cheers
stefan

Thanks dude.

However, Maxwell couldn't render the dragon :(
I need to wait for my other workstation to arrive, maybe with 3gb ram. I will then move over to xp64 to try it as xp64 will allow 4gb per 32bit process. I was trying a 1million poly object but RC3 has this stupid 'invalid normal' error which I can't figure out.

-j

duderender
12-13-2005, 02:12 PM
Some Stanford Buddha goodness:
http://jase.ca/content/forumpost/finalrender/stanford_happy_01.jpg

moka.studio
12-13-2005, 05:28 PM
Duderender,
is this the fastskin node,?

duderender
12-13-2005, 05:35 PM
Duderender,
is this the fastskin node,?

yes it is used.

soccerrprp
12-13-2005, 05:40 PM
NICE!

Questions:

1. How long did it take to render?
2. What exactly are your computer specs? RAM, CPU?
3. Is there a CLEAR improvement in rendering time with RAM greater 1GB? 2GB?

fluffouille
12-13-2005, 06:21 PM
Hi image took a little bit over one hour to render (remember, it's a high rez model, so it's heavy polygon-wise), I don't know his specs, though.
The ram won't really increase render time (except when you reach your memory limit and it starts to use the virtual ram), it will allow you to work with and render bigger files.

duderender
12-13-2005, 06:42 PM
NICE!

Questions:

1. How long did it take to render?
2. What exactly are your computer specs? RAM, CPU?
3. Is there a CLEAR improvement in rendering time with RAM greater 1GB? 2GB?

1. This was just over an hour if I recall correctly. I will rerender when I get home
2. This was on a AMD X2 3800+, 2GB RAM
3. YES!! The more RAM you can afford the better. With both the dragon and the buddha C4D hit over 1GB, where the dragon pushed it right to 2GB. Considering XP uses up to 256MB, if you have 1GB in your machine it leaves C4D and any textures you want with 768MB of RAM and that's not a lot.

Moving forward, my slaves will have 1GB minimum and workhorses 3GB minimum for 32bit and 4GB for 64bit.

soccerrprp
12-13-2005, 08:45 PM
Thanks duderender!

In search of additional DDR Ram now!

Canadianboy
12-14-2005, 12:24 PM
the only thing im interested in right now is to see how this thread turns out lol

andronikos916
12-20-2005, 06:23 AM
if speed is what you looking go for FRst2... I do not understand why you even started that thread?

if you want to compare similar render engines Compare Vray vs FRst2. I hope by the end of your comparison not to switch to Max.

For me AR2.5 is fast (need to be faster IMO) and Maxwell for specific quality renderings.

cy,
Andronikos

sketchbook
12-21-2005, 07:14 AM
Thanks dude.

However, Maxwell couldn't render the dragon :(
I need to wait for my other workstation to arrive, maybe with 3gb ram. I will then move over to xp64 to try it as xp64 will allow 4gb per 32bit process. I was trying a 1million poly object but RC3 has this stupid 'invalid normal' error which I can't figure out.

-j

this is news to me. are you saying that you can access up to 4GB of memory for cinema using c4d 32bit? this would be good news.

Srek
12-21-2005, 07:34 AM
this is news to me. are you saying that you can access up to 4GB of memory for cinema using c4d 32bit? this would be good news.
Yes, thats correct, CINEMA 4D 32 Bit can use up to 4GB when run on XP 64 Bit.
Cheers
Björn

lllab
12-21-2005, 10:08 AM
if you read the maxwell forum there was something about the invalid norma error. it happens with imported geometry sometimes. try another export format.
its a known error.

cheers
stefan

ooo
12-21-2005, 10:28 AM
Just a few Maxwell questions here,

Although I had given up hope in Maxwell already, I mailed with Richard (CEO) and he promised me an OSX version in the next couple of days (begin next week?). A "good" RC5 for Windows is also promised in the next days. I know, yet another promise, but ok, I give it another week before I decide what to do.
My question is: will I be able to use this version (RC?) for C4D renderings? What about the C4D-plugin etc. Sorry for my ignorance on this but I really don't want to invest my time in an unfinished product just for the fun of it. The Maxwell-forums are a confusing place to be so I ask it here. So if OSX-version arrives as promised and turns out to be stable, will it be usable for a C4D-user or does it involve a complex procedure?

Thanks, odo

lllab
12-21-2005, 10:55 AM
time will show- nobody knows beside the real beta testers...

cheers
stefan

AdamT
12-21-2005, 01:02 PM
Just wait and see. One thing I've learned is that "next days" in NL-speak can mean anything from a couple of days to never.

ooo
12-21-2005, 01:18 PM
Yes, I know all about this waiting thing. I paid for this vapourware in April so it's about time I think :eek:

If a stable RC5 for windows appears, does it mean that most problems are solved and then the waiting is on Final release 1, or is also the plugin still in alfa? So what does a C4D-user need, to be able to use this comfortably in production?

odo

ooo
12-30-2005, 03:38 PM
Hell yes, today another announcement from Maxwell (see maxwell forum). No RC5 is not ready yet, but next week it will be. I don't know how many next day/weeks announcements there have been but they really think it will be there now, really. How many final chances have I given these guys?

odo :hmm:

tcobb
12-30-2005, 06:50 PM
Well, this is the first I've heard of a Mac RC release.. so that's something. I have a good feeling about this... I think.
Where's my Mac FR news now? come ON!!

CGTalk Moderation
12-30-2005, 06:50 PM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.