Fryrender demo is up!

[CoolColJ]

The above scene probbaly favours Fry, due the different mix of materials, because I have a simple light lit scene with glass spheres, and the render times seem similar

[CoolColJ]

Did you model to scale? I know it's a big factor of light intensity in Maxwell.

(Mmmh, the demo download link seems to be down. I always get bounced back to the registration screen. Oh well...)

should be scale correct, I'll check

[lwan]

I think Fry has some "cheat" going on sorta like the light cache with Vray when using path tracing...

not AT ALL.

when you click on each object's properties, they can be set to 1 of 3 light modes - full radiance map, merged radiance map, per-vertex radiance map. I think this is for the virtual reality engine
you can also view the radiance map in the main Fry render program. The bright lit areas are red and so on to cooler colours for areas with less light
and each object can be set a visual importance rating. It defaults to all even. But I think this may determine the sampling allocation ...pretty handy along with the selectable area rendering

not AT ALL. this is indeed for realtime engine export and fry core don't deal with this. These paremeters are NEVER used by the renderer, NEVER. no cheats are going one here, just as unbiased as you would expect.


and BTW, dev team for the core is one person

Regards,
lwan

[BbB]

Good to know you're keeping an eye on the competition Iwan. By the way, your download link is down. Doesn't work for me in any case...

[matsta]

hey dude.

i have quite a nice test scene for you. would be an interesting comparison between the two. here it is in .blend format. tell me if u need it in a different format though.

greetz

[psor]

I have to say, it looks like a great piece of software! At first sight I like the
interface and the Aeronoise feature ... hehe. I'll evaluate it later, have not
that much time right now. ;o))

nb: And it's darn fast!



take care
psor

[zsouthboy]

So i'm playing with Fryrender now.

It looks like it changes its sampling strategy after each "sample level" is reached. For example, until the first "pass", the noise is very pre-Indigo 8, and to the next "pass", it looks like indigo's mutation strategy (i.e. RQMC), then for the next "pass", it changes again. Kind neat to watch.

The ability to just select an area that needs "help" (shadow areas, caustic areas) is a HUGE plus.

The default lens "vignetting" looks good. I like it.

No Blender plugin = for the lose. I won't be using fryrender seriously or considering buying it until there is such a plugin, or .obj import support, or something. I could at least play with Maxwell because of that - and AFAIK they still do not have a Blender plugin available either.

[suvakas]

I rendered the standard cornell scene with both. I have to say, that according to this Fry is not that fast. Also what is up with the AA? Check out the border area of rectangular meshlight on Fry image.


Render time 15 minutes each (on quad)

[edit] 100w light used. Works fine. Don't know what's up with your scene Cool.

[zsouthboy]

I have a question:

What is the comparable number of max consec rejections for indigo, to Fry?

Fry definitely doesn't make the fireflies like indigo does at say, 1000 - and we've established in the past that a lower number clears faster. (but is a biased result)

Any conjecture?

@suvakas:

Yep, Fry looks like it has the same "softness" as Maxwell.

Looking closely at your scene... It looks like two different apertures are being used between the two renders (look at the frontmost edges of the front box)?

[suvakas]

@suvakas:

Yep, Fry looks like it has the same "softness" as Maxwell.

Looking closely at your scene... It looks like two different apertures are being used between the two renders (look at the frontmost edges of the front box)?

Yes, probably. I just hit render and nothing more. Wanted to check out noise levels and overall speed.

[OnoSendai]

So i'm playing with Fryrender now.

It looks like it changes its sampling strategy after each "sample level" is reached. For example, until the first "pass", the noise is very pre-Indigo 8, and to the next "pass", it looks like indigo's mutation strategy (i.e. RQMC), then for the next "pass", it changes again. Kind neat to watch.

The ability to just select an area that needs "help" (shadow areas, caustic areas) is a HUGE plus.

The default lens "vignetting" looks good. I like it.

No Blender plugin = for the lose. I won't be using fryrender seriously or considering buying it until there is such a plugin, or .obj import support, or something. I could at least play with Maxwell because of that - and AFAIK they still do not have a Blender plugin available either.

Interesting. Possibly region-based adaptive sampling densities based on running variance calculations?

Tell me.. How often does the image buffer update?
Because such adaptive sampling is a bit easier if you have less frequent updates.

[zsouthboy]

Image buffer updates about every 15 seconds, and that increases as the sampling level gets higher (max is probably 2 minutes between updates?). Similar manner to Maxwell, as the last time I used it.

EDIT: after watching it a few times, the sampling strategy changes in the same manner each time, in the same order - so it's not changing based on conditions, more than likely it just has a LUT for each "Pass" as to the strategy for that "pass"

EDIT2:
Neat, you can choose which lighting to render (under "Kernel Configuration").
Direct light (E[S|D]L), Emit light (EL), indirect light (E[S|D])

Direct and emit light both definitely look like just standard pathtracing.

EDIT3:
No offense to the fryrender guys, but why even bother with a "toon render" mode?

[OnoSendai]

ok

[zsouthboy]

Here's an example of what I was talking about, the differing strategies.



Heh, the "scanline" effect actually looks neat.

[matsta]

suvakus are u gna be able to render those ?