[deadalvs]

https://caustic.com/series2/index.html

Wondering...

Back in 2004, I burnt my fingers/wallet by buying 3 ArtVPS PURE cards. So since then, I'm very sceptical with custom hardware to accelerate stuff.

What do you guys think of this ? NL ?

Cheers !

[polynurb]

imho the most interesting thing about this is:

1
on board ram for geometry only... textures etc. are accessed via system ram

2
dynamic geometry manipulation (even cvs and such)
here i really wonder if NL has any plan to handle voxelization differently to make this happen too.

i wouldn't mind at all if MXW could run on that thing.. 1.500$ for the big card is ok too.. assuming the speedup is right.

[deadalvs]

There's a post on speed found here in the comments :


http://www.youtube.com/watch?v=TZFM4p3l2is

[deadalvs]

btw. would it be possible to add multiple cards in a system ? did I miss this somehow ?

[polynurb]

i doubt it is possible.. at least atm.

they are selling the dual card for dual cpu workstation. so there seems to be a very close cpu-rtu relation.

[Half Life]

This is designed for a different render engine (Brazil which is now owned by Caustic) so I don't believe there is any potential for Maxwell with this company.

Proprietary and hardware/software match-ups are a bad solution for the end-user IMO -- it locks the user into the engine and keeps them captive of their financial investment... good business strategy for the manufacturer, but not a good deal for the user, long-term.

Best,
Jason.

[numerobis]

They have an OpenRL SDK too. So to me it looks like the Causics Visualizer is using Brazil, but OpenRL can be used for other apps too.
Don't know what this means for the future and if this really can become a standart like OpenGL... But the system RAM usage for textures looks very nice

https://caustic.com/dev_intro.php

[polynurb]

btw. would it be possible to add multiple cards in a system ? did I miss this somehow ?

answer:

http://forums.cgsociety.org/showpost.ph ... ostcount=7

[mtripoli]

At $1495 couldn't one build a "bare box" with a fast processor and gobs of memory, network the thing and not have to deal with proprietary hardware...? Just asking...

[dmeyer]

This is designed for a different render engine (Brazil which is now owned by Caustic) so I don't believe there is any potential for Maxwell with this company.

Proprietary and hardware/software match-ups are a bad solution for the end-user IMO -- it locks the user into the engine and keeps them captive of their financial investment... good business strategy for the manufacturer, but not a good deal for the user, long-term.

Best,
Jason.

OpenRL sits between the hardware and the renderer.

So in theory the renderer sends commands to OpenRL which then sends it to CPU, GPU, and the Caustic card.

[hatts]

At $1495 couldn't one build a "bare box" with a fast processor and gobs of memory, network the thing and not have to deal with proprietary hardware...? Just asking...

Yes, one could do that, and it would be infinitely wiser...

IMHO buying dedicated hardware like this is like begging to waste one's cash on something that's destined for obsolescence. What if the rendering companies suddenly change their software architecture? What if Intel's next batch of CPUs is exponentially faster? What if cloud solutions actually become realistic and everyone ends up feeling relieved that they didn't spend another dollar on hardware?

[numerobis]

At $1495 couldn't one build a "bare box" with a fast processor and gobs of memory, network the thing and not have to deal with proprietary hardware...? Just asking...

Yes, one could do that, and it would be infinitely wiser...

if it really has the speed of 2-5 dual xeon systems where only one processor costs more than this whole card?

[dmeyer]

At $1495 couldn't one build a "bare box" with a fast processor and gobs of memory, network the thing and not have to deal with proprietary hardware...? Just asking...

Yes, one could do that, and it would be infinitely wiser...

IMHO buying dedicated hardware like this is like begging to waste one's cash on something that's destined for obsolescence. What if the rendering companies suddenly change their software architecture? What if Intel's next batch of CPUs is exponentially faster? What if cloud solutions actually become realistic and everyone ends up feeling relieved that they didn't spend another dollar on hardware?

If i could spend $1500 to get a 2-6x speedup in maxwell i'd do it in a heartbeat, even if knew it'd be obsolete in 2 years.

Even that is unlikely though. Intel's next gen Xeons are going from 8 to 10 core, overall, what, maybe 30% faster?

[mtripoli]

Years ago there was a similar system for doing rendering (may have been the same company, I'm sketchy on the details). I *almost* bought into it with the same thinking as dmeyer. At the last second I couldn't pull the trigger; as I looked around my office I could see stacks of hardware that was either obsolete or didn't live up to expectations. One of the beautiful things about Maxwell is the ability to render on any machine, regardless of speed or vintage. At one point in the early days I had 4 or 5 various machines rendering. When I moved I gave them up. Point is, with Linux being *free* one could go to the local mom and pop computer store and buy 5 "boxes" for a small render farm. I'd spend my money on this rather then a dedicated card.

As for the cloud... I think the possibilities are astounding. However, pointing out the obvious, if you have a crappy internet connection then the whole thing is moot. I've been using Dropbox for exchanging files. I save "locally" and the file "shows up locally" on collaborators computers; no emailing, no FTP. It's really very cool. However, I have a friend that lives in a relatively small town in Massachusetts where they barely have DSL speeds. No matter how "cool" or convenient the cloud might be, he'll never be able to take advantage of it until the local ISP (and everyone else involved) does something about the network. The biggest proponents of cloud based services are also the same ones with T1 (or better) connections.

[hatts]

If i could spend $1500 to get a 2-6x speedup in maxwell i'd do it in a heartbeat, even if knew it'd be obsolete in 2 years.

I wouldn't. More/faster processors also speeds up other tasks in your work, unlike a one-trick-pony. Furthermore for the price of two or three of these cards you could have a whole workstation, ready for several years of use.

@mtripoli, yes the cloud stuff is much less appealing to those with shitty internet connections. However there's a major difference between the current realities of cloud computing - where you're constantly uploading and downloading your working files to and from the cloud - vs. the planned future use of cloud computing, where the company hosts the actual software on their servers and all computing tasks are performed on beastly servers. Even if the workflow still requires UL/DL your files, the computational advantage of having 20+ nodes compute a process vs. your one computer could be revolutionary. Imagine using NX on a netbook...

Additionally (I'm not sure how familiar you are so forgive me if this is elementary), Dropbox doesn't actually upload your entire file if you save a change to it; it only uploads the new parts of that file. Pretty intelligent really. My point is, if using cloud-hosted software, the advantage in computation speed may more than make up for a sluggish internet speed.