Maxwell Render | Official public forum

I keep seeing posts from Victor saying that there is still optimization to be done, so I guess you can take that to mean that Maxwell will get faster. They still have to get the cooperative rendering working properly; I'll be very interested to see how it performs when you throw 20 or 30 processors at a single image. I am surprised that we haven't seen any speed increases since the beta was released, not to mention all of the existing problems that have to be fixed before October.

The speed increase from between the alpha and beta versions of Maxwell Render are not that significant if you consider people were requesting the ability to render larger images.

I'm assuming people are rendering larger images than before but larger images take longer to render. So, we really get the ability to increase image size in the same time as it took alpha to render a smaller sized image.

I personally do not expect massive increases in render speed. Other features have yet to be added and existing features improved so they may take their toll on render times.

For image quality though (the most important aspect to me), I consider Maxwell Render to be the best.

I agree that Maxwell has the potential to be the best render engine out there but it's got to be able to render out images at high resolutions quickly or it will never be anything but a toy to play with. I also agree that I don't think were going to see that much of an additional speed increase which leaves cooperative rendering as the only hope of being able to use Maxwell in a production environment.

speed rendering always comes against phenomena render engine. Even, with other render when rendering a realistic scene the speed becomes potential..

I doubt that they will increase the number of CPU's per license again, this issue has been covered to death and NL won't budge. As for speed I've read reply's from victor saying that M~R is slow but the addition of complex geometries to a scene doesn’t exponentially increase render time. I guess this is good news but if it takes 100 or more hours to render a 4000x4000 image and you still have noise, it doesn’t really matter if it's loaded down with geometry or not.

I really don't think (I hope I'm wrong)that maxwell final release will be much more faster than this beta. I guess there will be important optimizations but, being and unbiased render, speed is an issue.
I would like to have in maxwell an alternative biased method. Some will think that this is not the philosophy of maxwell. For me, the beauty of maxwell is the ability to do spectral calculations. If NL could develop a biased method (interpolating samples, or whatever) based on the same spectral information as an alternative method for those who can't afford big render times or for animations, it would be great. At least you could have the 2 options.

I was opposed to this at first but if it means the difference between being able to produce work with Maxwell or not being able to I'll pick the first one. I really think that if they aren't able to get the speed issue resolved there going to have to go with this option because there's no point in using Maxwell if you can't product work with it. I know the purist Maxwellieans are going to say that it has to remain a unbiased renderer no matter what, but I don't think these people are living in the real world of a production environment.

Who says there won't be no more speed improvement in maxwell?

There is other priorities for the release: materials, interface, stability, etc. But believe me, the speed improvements in 1.2 were just a minor step in comparison with the speed improvements ( unbiased, always unbiased way ) there is in the way. Algorithmic and mathematical optimizations are complex, takes time and have to be done carefully and step by step.

Best regards.

That's the best thing I've heard all day. I'm really glad to hear that speed is one of your top priorities and the best improvements are just around the corner! Thanks

And one day if you need seatbelts due to speed, I'll be selling them my friends.

Woo hoo this is good news.
I've got a render thats been rendering for 136hrs and its still not clear. I really need to set off a render at 3k and have it finished with in 24hrs or less, this 3 to 5 days is crazy.
So I hope the next version has the speed improvements.

oscarMaxwell wrote:Who says there won't be no more speed improvement in maxwell?

There is other priorities for the release: materials, interface, stability, etc. But believe me, the speed improvements in 1.2 were just a minor step in comparison with the speed improvements ( unbiased, always unbiased way ) there is in the way. Algorithmic and mathematical optimizations are complex, takes time and have to be done carefully and step by step.

Best regards.

Oscar, thanks for the reply, those are great news. But, don't you think that giving to maxwell users the possibility to choose between biased or unbiased method will increase the number of users interested in maxwell??.

I'm and architect who renders mainly stills, and I'm very happy with the amazing results in maxwell, but for people that do mostly animations its some kind of complex to output good results in a reasonable amount of time.
I though the famous "rendering system 1 (RS1) would be one of the multiple options to render a scene. Can you give us some info on that, at least some hint to barely know what is that RS1 for?

Keep up the good work!

oscarMaxwell wrote:Who says there won't be no more speed improvement in maxwell?

There is other priorities for the release: materials, interface, stability, etc. But believe me, the speed improvements in 1.2 were just a minor step in comparison with the speed improvements ( unbiased, always unbiased way ) there is in the way. Algorithmic and mathematical optimizations are complex, takes time and have to be done carefully and step by step.

Best regards.

I like the sound of these priorities!

Materials, interface and stability are the things that needed most at this time, speed is acceptable for the current Beta stage.

Hi Renato,

The world is full of biased renderings. I don't want to give a backward step in Maxwell making yet-another-biased-render. Don't think you can render faster with a biased one. Let's wait to see if Maxwell is not going to be fast enough in a future.

There is other reason why it would not ve a good idea to handle inside Maxwell a biased engine... Maxwell is designed to render anything you send it, only limited for the RAM you have. We can render very complex scenes ( for example lots of complex caustics ). Do you know how huge amount of memory a biased render would need to render it? remember that a biased engine stores the ilumination. So i can't image how enormous memory you'll need to render spectrally with pixel accuraty so high subtles details of GI in a biased mode.

Of course, I'm talking about high and final quality rendering. It does not mean that I can use fast biased ones to render fast previews for material preview or viewer fast preview. But this is a different story.

RSx are code number for different Maxwell engines, coming in a future. But don't worry too much now for this.

Best regards.

I second that!