Author Topic: Rebuse render farm  (Read 3506 times)

2020-02-01, 13:43:18

minimaldesign - Blaž

  • Active Users
  • **
  • Posts: 158
    • View Profile
    • minimaldesign
Hi all,

Is there something wrong with the Rebuse render farm or with Corona?
Is this normal for test render of basic interior scene without displ, dof, etc...just a basic 200mb scen file and 3800pix render size (basically the same project scene with the same settings)???
To me these render times are extremely abnormal not to mention the price!

And specification from top to bottom:
Total cost in RP
6.71 RP
Ø cost / frame
6.71 RP
Sent
01.02.2020 08:43:42
Started
01.02.2020 08:45:55
Finished
01.02.2020 12:49:57
Queued time
0.53 min
Render time
244.47 min
Frames
1 of 1
Width
3800
Height
2850
Output
.exr
Size
630.00
Renderer
corona 5.0
Cam
Final

Total cost in RP
11.92 RP
Ø cost / frame
11.92 RP
Sent
31.01.2020 18:46:14
Started
31.01.2020 19:58:09
Finished
01.02.2020 03:11:26
Queued time
63.18 min
Render time
496.40 min
Frames
1 of 1
Width
2480
Height
3507
Output
.exr
Size
633.91
Renderer
corona 5.0
Cam
A4

Total cost in RP
20.54 RP
Ø cost / frame
20.54 RP
Sent
30.01.2020 23:27:22
Started
31.01.2020 00:09:10
Finished
31.01.2020 00:45:23
Queued time
32.05 min
Render time
68.25 min
Frames
Single Distributed
Width
3800
Height
2850
Output
.exr
Size
802.58
Renderer
corona 5.0
Cam
1 FRONT SIDE.1
« Last Edit: 2020-02-03, 11:05:12 by minimaldesign - Blaž »

2020-02-03, 19:09:12
Reply #1

johnnyswedish

  • Active Users
  • **
  • Posts: 237
    • View Profile
Hi,

I had loads of problems with Rebus. I have now switched to Ranchcomputing. Works much better with Corona :-)

2020-02-03, 19:10:29
Reply #2

TomG

  • Administrator
  • Active Users
  • *****
  • Posts: 5434
    • View Profile
Sounds like questions for Rebus farm, have you contacted them?
Tom Grimes | chaos-corona.com
Product Manager | contact us

2020-02-03, 19:48:12
Reply #3

minimaldesign - Blaž

  • Active Users
  • **
  • Posts: 158
    • View Profile
    • minimaldesign
Sounds like questions for Rebus farm, have you contacted them?

Hi Tom,

full of nonsense and logic for me...this is total rip off ... nothing else I can say, sorry :(
On one of our office single computer 2x Xeon with render time 135min for this scene and on Rebuse render farm this basic interior (200mb file size) render time was 68min + distributed on 25 cores and for comparison and they say everything is normal "The costs got that high because you distributed the rendering to 25 machines."...total nonsense! Cost calculator on rebuse site shows me 5€ and 5 min render time for that project "in my dreams i guess". I don't know if those "25 machines" are laptops from year 1990 or what?!
Then some basic 1min animation with that or similar scene comes around 45.000€ with those numbers!?!?!!
I hope that the other render farms you advertises on corona web site are not like Rebuse.

and the answer form rebuse:

"Hello Blaž,

we have checked it and found no errors. the price corresponds to the render time.
unlike local, this image was rendered distributed. 25 rendernodes were active each of them has to load the scene and assets.

Best regards / Beste grüsse,
Ilya
Rebusfarm GmbH
rebusfarm.net"

sorry because i'm a "little angry" pissed off but it's not clear to me where's the problem or is it in corona or with them and with the "ripoff party"!
« Last Edit: 2020-02-03, 20:08:28 by minimaldesign - Blaž »

2020-02-03, 20:15:14
Reply #4

TomG

  • Administrator
  • Active Users
  • *****
  • Posts: 5434
    • View Profile
Well, if it's rendering without errors on your local machine, it wouldn't be Corona I would say. How long does the scene take to set up on your machine (ie how much of the time is Scene Parsing and Geometry, in the VFB stats), that would give you an idea of the part of rendering that can't be done faster when distributed to multiple nodes, each machine would have to do that for itself. Then it contributes to the actual passes part of the renders, and that part of the time should be, say, 25 times faster (though maybe less, maybe more, depending on what nodes they distributed it to vs. your local computer).

As a note, TR is not the world's most robust system, and in general for large resolutions (which resolution was this?), you can need to adjust the TR settings - they should be set to Arbitrated, and then the packet size should be raised (e.g. 128MB vs 2), and the interval to send should be raised (eg to 30s from 5s).

EDIT Looks like 3800 x 2850, could be worth sharing what your TR settings are under the Corona render settings. If they are default, that might be the cause (what can happen, I believe, is too much data is sent too frequently, which leads to network congestion and lost data, which slows things down)
Tom Grimes | chaos-corona.com
Product Manager | contact us

2020-02-04, 08:50:41
Reply #5

minimaldesign - Blaž

  • Active Users
  • **
  • Posts: 158
    • View Profile
    • minimaldesign
Well, if it's rendering without errors on your local machine, it wouldn't be Corona I would say. How long does the scene take to set up on your machine (ie how much of the time is Scene Parsing and Geometry, in the VFB stats), that would give you an idea of the part of rendering that can't be done faster when distributed to multiple nodes, each machine would have to do that for itself. Then it contributes to the actual passes part of the renders, and that part of the time should be, say, 25 times faster (though maybe less, maybe more, depending on what nodes they distributed it to vs. your local computer).

As a note, TR is not the world's most robust system, and in general for large resolutions (which resolution was this?), you can need to adjust the TR settings - they should be set to Arbitrated, and then the packet size should be raised (e.g. 128MB vs 2), and the interval to send should be raised (eg to 30s from 5s).

EDIT Looks like 3800 x 2850, could be worth sharing what your TR settings are under the Corona render settings. If they are default, that might be the cause (what can happen, I believe, is too much data is sent too frequently, which leads to network congestion and lost data, which slows things down)

Hi Tom,

I try the same scene with the same settings on Ranch render farm! Guess what 2.2€ and 2min 18sec of render time on low priority!!! Wellcome Ranch!! :)
As said and meant ... crazy ripoff with sick/senseless excuses full of bull s... (rendering on 25 machines, render time 68min for 25€ per frame) by Rabus!
« Last Edit: 2020-02-04, 09:06:46 by minimaldesign - Blaž »

2020-02-04, 14:40:52
Reply #6

TomG

  • Administrator
  • Active Users
  • *****
  • Posts: 5434
    • View Profile
I'd definitely raise that with Rebus support, now that you have results from another farm to compare with - unfortunately we don't have any say in or control over how render farms set things up, that is totally up to them. At least now you can show them proof that this isn't a "Corona thing" and is dependent on the farm set up in some way.
Tom Grimes | chaos-corona.com
Product Manager | contact us

2020-02-14, 09:13:26
Reply #7

minimaldesign - Blaž

  • Active Users
  • **
  • Posts: 158
    • View Profile
    • minimaldesign
I'd definitely raise that with Rebus support, now that you have results from another farm to compare with - unfortunately we don't have any say in or control over how render farms set things up, that is totally up to them. At least now you can show them proof that this isn't a "Corona thing" and is dependent on the farm set up in some way.

Hi Tom,

I think something is also wrong with the corona on another farm because there are weird times for the same scene with the same settings?
it's a little too weird for this to be repeated on both farms

Regards, B

2020-02-14, 14:54:40
Reply #8

TomG

  • Administrator
  • Active Users
  • *****
  • Posts: 5434
    • View Profile
Sorry, not sure what those images represent? The filenames are different, so it's different versions of the scene? Or the same version but just different job name, so you mean the same scene with zero changes submitted at different times gives different results? Which farm is it? There's not a lot we can do with render farm performance unfortunately, since we're not in the loop on their set up, both hardware and any custom software they are using - things rendered on a local farm we can diagnose more, but on an external farm, it really comes down to the user talking with them (then perhaps they can contact us if they have something unexplained, and they can share more info with us for us to look into things).
Tom Grimes | chaos-corona.com
Product Manager | contact us

2020-02-18, 08:22:12
Reply #9

minimaldesign - Blaž

  • Active Users
  • **
  • Posts: 158
    • View Profile
    • minimaldesign
Sorry, not sure what those images represent? The filenames are different, so it's different versions of the scene? Or the same version but just different job name, so you mean the same scene with zero changes submitted at different times gives different results? Which farm is it? There's not a lot we can do with render farm performance unfortunately, since we're not in the loop on their set up, both hardware and any custom software they are using - things rendered on a local farm we can diagnose more, but on an external farm, it really comes down to the user talking with them (then perhaps they can contact us if they have something unexplained, and they can share more info with us for us to look into things).

Hi Tom,

In a previous post was a test of the identical/same scene with the same settings as seen below (in a previous post) on another farm. There are two different render times for the same scene also on this render farm.
So i'm asking if maybe there really isn't any issue in the corona render.
Because on two different farms is happening the same for the same scene with the same settings, it's weird to me or. is this possible to have such a differences in render time for the same scene/same settings?
I can also send you a scene and you take a look

Regards, B

« Last Edit: 2020-02-18, 09:36:59 by minimaldesign - Blaž »

2020-02-18, 14:02:25
Reply #10

TomG

  • Administrator
  • Active Users
  • *****
  • Posts: 5434
    • View Profile
Again, unfortunately it's impossible for us to diagnose, as a render farm introduces all sorts of things unique to them (their scheduling software etc.) It would have to start by talking with their support, who can talk to our support as then the two support teams can ask and answer questions back and forth. We could only diagnose things that happen on a local network of computers, using the standard DR, Backburner, or Team Render tools, as we can then try those for ourselves and see directly what's going on - but render farms introduce processes that we are unaware of, which may (or may not) play a part in what is happening.

By any chance are you able to test on a local network and see if you get anything unusual? Failing that, I would start by sending the issue to their support team, as they could then communicate with us directly if they believe there is a Corona issue involved.
Tom Grimes | chaos-corona.com
Product Manager | contact us