Network Render License Options Poll

[glimpse]

I might understand those who have multiple RIGs with 4 or even more GPUs that are unhappy with limitations, but why on earth hobbyist would want to network render on two rigs if they don't have good GPUs (maybe other parts are limited too, maybe even Your lan isn't the best..)

What I want to say, if don't have right gear Your network rendering will not scale linearly (take a look here: viewtopic.php?f=9&t=40400). The best would be not to buy extra licence, but invest into new GPU (if You look to save some money, maybe taking second hand 580 or so.. - for 300$ You can have GPU+ PSU if Yours is not enough =)

Even building from scratch, You can get 80$ CPU, 200$ Mobo, let's say 100$ PSU (that the cost of Octane Licence) & few cards for Your choose & budget..-there's no need to our force on developers about their pricing policy. It's not a charity..

It seems that whatever someone does (in this case OTOY) there will be a crowd that will start to talk how bad the progress is.. The program is cheap in todays standards, so it makes sense for OTOY to charge for extra work they have put through the years in order to bring these extra features..

if You can't afford the licence or hardware, maybe it's time to start thinking' 'bout cheaper hobby to play with (origami or such)..why You want other to work for free?

Look this way: You already have a possibility to render with 12GPUs that almost the same as having 12 PC running with single CPU - but You don't need to build 12 PC (as in CPU rendering field)..

What if Your client comes & say: Hey You've already charged me for First visualisation, it's "only to rotate & re-render", make three shots more for free..- What would be Your answer? (the same goes here..)

Don't want to insult anyone here, but I guess some of these requests are like insults for programers. They already giving out free version for students, have a demo to play..say thanks, & grab a licence at discounted price while You can or think of putting the same money to an upgrade.

[geo_n]

I might understand those who have multiple RIGs with 4 or even more GPUs that are unhappy with limitations, but why on earth hobbyist would want to network render on two rigs if they don't have good GPUs (maybe other parts are limited too, maybe even Your lan isn't the best..)

For me its productivity. I never render on the same machine while working. There's just too much drop in performance when something is rendering in the background. So I always do test renders on main workstation then send out render sequence to render nodes immediately.

Btw follow up question on my previous post.
With octane, with two gpus on one machine, does one gpu render one frame and the other gpu another frame? Or do they render the same frame at the same time? So if you have 4 gpu they are simultaneously rendering 4 frames. Seems a waste of allocated memory if they all rendered the same frame.
I remember when octane started 4 years ago it only supported one frame rendered by multiple gpu like vrayspawner. Is there a mode now that can render one frame per gpu? I rarely work on stills and rendering sequences is more useful in my case.

[glimpse]

For me its productivity. I never render on the same machine while working. There's just too much drop in performance when something is rendering in the background. So I always do test renders on main workstation then send out render sequence to render nodes immediately.

if You have multi GPU, let's say two, You wouldn't feel any performance drop if one would be rendering while other would be used for screen & other tasks (like accelerating PS, or powering viewport in max, etc.)

However, if You do network rendering Your system will be more in use as CPU needs to send data in out through LAN (I assume CPU is responsible at least partly for this task, but I might be wrong).

In other words. If You don't want to feel anything at all that something consumes system resources (or hang it by some unknown reason) it's the best to have a standalone app on the other rig & render there (as You said). In this case You need one extra licence. Even if You're running Max with plugin, simply export full scene & open it in standalone.

Btw follow up question on my previous post.
With octane, with two gpus on one machine, does one gpu render one frame and the other gpu another frame? Or do they render the same frame at a time? So if you have 4 gpu that are simultaneously rendering 4 frames. Seems a waste of allocated memory if they all rendered the same frame.

all GPUs are rendering the same frame. But as scene should fit in every GPU that is doing the rendering work..even if they would render diff frames there would not be any waste of resources..as You would need the same time & same memory to finish these let's say 4 frames (I don't see this coming anyway as there's no benefit..).

[geo_n]

Thanks Glimpse.
With regards to rendering a frame per gpu, maybe there's no advantage to octane because apparently gpu scales perfectly linear when adding multiple gpu. So 3 titans do actually perform 3 times the speed perfectly.
But with cpu rendering thats not the case, the swarm method rendering is not perfect. So if you used 3 I7 cpu to render a single frame its not perfectly linear increase compared to using single I7. Its still faster using 3 cpu than 1 cpu just not perfectly linear. This is the case with vrayspawner and modo renderer.
Anyway a developer suggested that to do frame per gpu just open another instance of the app and assign a single gpu to it and render the frame ranges per gpu. Neat idea since that way a machine with 3 gpu can each be loaded its maximum memory each and render a frame per gpu. That's sort of sending the scene to a cpu based renderfarm.

Back to normal programming.

[Paulski]

I don't really understand why the 2 Nodes a 6 GPUs option is so popular. There is barley any motherboard that is capable of holding that many GPUs.
There is actually no desktop board that is capable of holding them without a riser...

[Tutor]

Thanks Glimpse.
With regards to rendering a frame per gpu, maybe there's no advantage to octane because apparently gpu scales perfectly linear when adding multiple gpu. So 3 titans do actually perform 3 times the speed perfectly.
... .

It's been my experience that "... [like multiple] gpu(s) ... scale perfectly linear" in a sort of unusual way. If it takes 90 secs for one such GPU to render a specific scene such as the Benchmark scene, then adding a second copycat GPU will cut the render time for the same scene to 45 secs. If you want to then cut that 45 sec. render time in half (22.5 secs), then you'll need to add two more of the same GPU. Then to cut the render time (22.5 secs) of the four GPUs in half (11.25 secs), you'll have to add four more of the same GPU and to cut that time in half (5.625 secs) you'll have to use 16 of the same GPU, and so on and so on.

I don't really understand why the 2 Nodes a 6 GPUs option is so popular. There is barley any motherboard that is capable of holding that many GPUs.
There is actually no desktop board that is capable of holding them without a riser...

Some, like me, may be using dual video processor GPUs such as the double wide PCIe slotted GTX 590, GTX 690 or the triple wide slotted GTX 790 (aka the Titan Z). In the US, a Gigabyte Up4 motherboard can be had for under $250 (US) and sports four double wide PCIe 3.0 slots into which you can install four GTX 590s or four GTX 690s or four of what ever other double GPU suits the owner. Since Octane sees each video processor as a separate rendering device, it'll treat four GTX 590s, for example, as eight GPUs. In the US, GTX 590s can be had each for under $400 (US) on Ebay. So, I'm not talking about a particularly expensive system because you don't need to spend lot of money on low end Sandy or Ivy Bridge CPU if you're into GPU rendering. Different people make different choices about their systems and some, like me, build their own and they should not be penalized by being charged more for using their choice to load their systems with many GPUs when it costs Otoy no more to support such self-builds than it costs to support those with less expensive, equally or more expensive, but more limited other systems.

BTW- I'm a Cinema4d user and I've been network rendering with Octane since Team Render was introduced in C4d V. 15; moreover, I paid the full load for each of my current Octane rendering seats whose total number is insufficient for the number of systems that I have and as a consequence I also use and am exploring other less expensive, network capable, GPU render software. Also, I used Cinebench (CB) 11.5 and 15, along with Geekbench, to tweak my rendering systems' CPU performance since Cinebench is the closest metric for projecting CPU rendering performance in Cinema 4d.

[geo_n]

I don't really understand why the 2 Nodes a 6 GPUs option is so popular. There is barley any motherboard that is capable of holding that many GPUs.
There is actually no desktop board that is capable of holding them without a riser...

There is this board with 6 PCIE.
http://www.asrock.com/mb/Intel/H61%20Pr ... /index.asp

[riggles]

I don't really understand why the 2 Nodes a 6 GPUs option is so popular. There is barley any motherboard that is capable of holding that many GPUs.
There is actually no desktop board that is capable of holding them without a riser...

It's also perhaps because some are using GTX 590/690s which have dual GPUs in one card, and you can fit 3 of them into many desktop boards.

[riggles]

I'm a little confused, but not entirely surprised, by some of the clamor around the limitations of network rendering. I am not bothered by the 12 GPU limit. I'm not close to reaching that and I'd wager that neither are many who are complaining that it's too low. It's not about the number of GPUs, it's that each computer needs a full license and there isn't a less expensive render-only license available yet. Well, that option honestly should've been there at the launch of network rendering, it wasn't though, and OTOY are working on it. But to be extremely disappointed that one can't just use all the available resources on their network to render for free is a little silly and naïve. Features cost money. Are you a CG artist who makes a living that way? You should be familiar with this concept. Are you a hobbiest? You should be familiar with it, too, and this isn't Blender. I agree that there should be a cheaper render-only Octane license, we just need to wait for a OTOY to finish putting it together.

I'm not worried about having different level tiers much, either. One tier might be for enthusiasts or freelancers which could have less GPUs per node for a low price, and another could have higher GPUs per node (like 8 or something) at a higher price for companies looking to build a server-level render farm. No reason to think they're trying to price us out of building our own personal "cloud" so that we use theirs in the future. If OTOY does their cloud rendering service right, both price and ease-of-use, they have nothing to worry about - people will get addicted to that speed.

[geo_n]

The current trend with cg software is offering a lowend indie license and a pro license and a studio license.