Discussion | Cloud Rendering Experience

I am curious to know how users find this. The idea of faster rendering is (of course!) appealing, but presumably the files need to be uploaded to the cloud first, and then the final file has to be downloaded. My home internet isn’t very fast, so am I going to find this any faster?

Also, does anyone have a sense of the carbon cost? I run my little home PC on a green energy (all solar/wind electricity); I presume that’s not the case with Topaz’s servers, but are there economies of scale? Any thoughts would be appreciated.

I’m very surprised nobody has replied to your post.
There are quite a few factors involved and it gets complicated but I will have a go !
I’ve used Topaz software for 7~8 years right back to the beginning.
Would not claim to be an absolute guru or a platinum expert but maybe “experienced long time user” would be fair.
In the last 7 months I’ve done 200+ Cloud renders and more than 1500 Local renders on a Windows PC with i7 7700K 64GB and RTX4070 Super 12GB GPU.
I follow the Community Forums for Photo AI and Gigapixel daily.
My images vary from small online jpgs to 40MB RAW files.
My internet in Thailand is OK - a full version release of Gigapixel takes about 2~3 minutes to download a msi file and about 5~6 minutes to install.
So far I have ONLY used free complimentary Cloud Credits offered to Beta Testers who complete the beta test surveys.

You say “My home internet isn’t very fast …”.
Topaz gets a few user complaints in the Community Forums about the speed of its upgrade releases and of its Cloud servers.
My experience is that upload and download data rates are 25%~50% of what I get from Adobe, Capture One and DxO.

Specifically for Cloud rendering you are not guaranteed that a Cloud render or an initial image upload will start immediately you press “Enter” - there can be a wait in a queue.
In my experience I get queued about 10%~15% of the time - often for less than 30~60 seconds but a couple of times for 2~3 minutes.

8+ hour queue wait
This post today complains of an 8+ hour queue but my guess is that was due to some error or glitch in the image or the render settings because I don’t recall seeing other similar cases.

General Cloud variables

The idea of faster rendering is (of course!) appealing …
On my PC system a rough rule of thumb seems to be that my Local renders will take about 3x~4x longer than the Cloud render time estimate - that’s assuming no queue.

With the image upload - processing - download sequence my Local rendering of small and medium images can be faster or at least equal to the total Cloud times. If I try 6x upscaling of an original 1200x1024px jpg the faster Cloud processing is significant ie probably 3~4 minutes cf 9~16 minutes Local. BUT of course you pay for a lot of Cloud Credits - perhaps 14 - I think currently a monthly sub of $9.99 gets 80 credits - so 14 = $2 …
Cloud Credit Subscription Costs

Important to Understand that Cloud renders are DIFFERENT

A Cloud render does not simply produce a faster render than Local on your computer …
The Cloud AI models and AI database are completely different (as far as I understand) to those installed on your computer.
So your Cloud render will almost certainly be noticeably different to your Local render.
As an example -

An online original Turner litho print of Plymouth


.
.
Local Redefine Artistic Creativity Max Texture 5 x6 upscale - 7min 48sec

.
.
Same Max and Texture settings - Cloud Render 3min 13sec

.
.
Another example both using the same model settings with beta version v8.4.0.beta4 -

Local


.
.
Cloud

.
.
Overall my experience with Cloud rendering is -

  • almost always better IQ, that is more detail, better focus, fewer sky artifacts
  • frequently with higher Creativity 4 & 5 & 6 lots more fantasy (as one would expect)
  • not faster cf MY PC SYSTEM for small and medium input images
  • perceptibly more responsive to detailed text prompts than Local renders
  • often greatly improves smaller images without introducing fantasy with None, Subtle, Low and even Med Creativity - this tends to be very dependant on the quality of the input image
  • nearly always better as the upscale factor increases - more pixels for more details ?
  • potentially costly if you get carried away ???

In the future if/when my complimentary cloud credits run out I’m not sure that I can justify an additional monthly subscription cost of $9.99 - that would be $120 and more for a year than the cost of the software upgrades.
Even paying $78 once for 400 credits would be double the cost per photo compared with the subscription route. So a difficult decision.

I think that’s enough for today !
You don’t specify your existing computer. If you want to post details or to PM me I would be happy to answer any follow up questions and maybe more advice based on your hardware.

Of course I’ve not mentioned one of the outright advantages of Cloud rendering which is that if you have a lower spec computer that cannot process AI you may be able to send your processing to the Cloud to get the best possible results - all be it at the cost of paying for credits.

Good luck with whatever you decide to do.

3 Likes

Many thanks for that detailed and helpful reply. I appreciate you taking the time

1 Like