Any work arounds to Nvidias RAM monopoly?

[studiocampus]

but..... in a multiple gpu configuration, the ram of the two or more cards is summed? I always thought so! octane shows in the bar just the amount of ram of one card (ex. 1.5 giga) but loading the same scene in a two card setup the amount of ram used goes 50%, I assumed that the load was splitted for the two cards.
or not?

e

[Rhodesy]

Sadly no

[studiocampus]

rhodesy are you answering to me or to the topic subject?

I just made a test and the same scene with 1 card loads 400megs and with 2 cards 200megs. If is correct, I cannot understand the gtx400s vs tesla post...just take 4 gtx and you will have the same ram but 4gpu (four times faster), maybe saving some money too.

by the way I think that today a system with less than two gtx480 cards (a card goes for little more than 500 euros here) is not for people who intend to do unbiased octane renderings for WORK, or simply just occasionally do some renderings for work. Simply couse if doubling the speed of your work does not value 500 euros, maybe you are not working.
I consider double speed becouse realtime speed up the entire process of creation of a scene, not just the time required for final rendering.

e.

[GeorgoSK]

rhodesy are you answering to me or to the topic subject?

I just made a test and the same scene with 1 card loads 400megs and with 2 cards 200megs. If is correct, I cannot understand the gtx400s vs tesla post...just take 4 gtx and you will have the same ram but 4gpu (four times faster), maybe saving some money too.

by the way I think that today a system with less than two gtx480 cards (a card goes for little more than 500 euros here) is not for people who intend to do unbiased octane renderings for WORK, or simply just occasionally do some renderings for work. Simply couse if doubling the speed of your work does not value 500 euros, maybe you are not working.
I consider double speed becouse realtime speed up the entire process of creation of a scene, not just the time required for final rendering.

e.

No, he answered to you. The scene has to be allocated on each of multiple GPUs you have, so you only have the amount of ram as the smallest of them has.

[Rhodesy]

I mean as far as im aware the whole scene has to fit into each graphics card RAM so if you have a GTX 480 1.5GB and a GTX 260 900mb a scene that was say 1GB wouldnt render on the 260 just on the 480. If all RAM was combined then all our problems would indeed be solved and buying extra cards would be a no brainer double the speed, double the RAM capacity. But im afraid I dont think that is the case. Indeed if you were to be using Octane as your principle renderer for your profession then you should indeed be investing, but unless your company has money to burn (£3,500 on a single card and then multiples there of) then your stuck with the current 1.5gb offerring from nvidia which is maybe the equivilent of going back to a 2GB traditional system

Im an Arch vizzer working for an architectural firm and I use Vray for all my production stuff. Ive bought a GTX 480 and Octane for my own machine just to test it out so not really intending on using it commercially at this stage just seing how it develops and when I think I can use it commercially get my boss to fork out for some more hardware but I kind of feel that may be somewhere off not due to the quality of the software but more the affordability of the hardware.

EDIT: Beat me to it!!

[studiocampus]

mmm....you right, as well my considerations are personal and not in contrast with yours....
Going back on ram usage, If every card have to load the entire amount, why on my setup I see 200 megs used if I use two cards and 400 if I use just one? (of course the file is the same and also the resolution, just loaded the file than set one or two cards).....
help RADIANCE!!



ps: long time ago, when 4megs of ram on a video card was huge and ram was far more expensive than now, I remember cards with expansion slots for ram. Maybe if we wanna push for something with a subscription we can ask for this option in next cards (so just one card have to be engineered and produced)....

[GeorgoSK]

Guys, if you're watching some tech news, do you think future might lie in solutions like Intel's Knight Ferry (crippled larrabee) or AMD's Fusion... ?

[n1k]

We will just need to wait and see about fusion and larabee. I expect intel and amd to jump on high power computing "train" very soon and shake a bit NV dominance. Recently Manju Hegde, nVidia’s VP for CUDA and PhysX, moved to amd to work on fusion.
Very interesting times are ahead of us:)

Cheers,
n1k

[Rhodesy]

Yeah especially if Octane moves onto openCL from CUDA then that should benefit everyone and make it a more competative market. Im pretty sure Vray RT GPU is going to be OpenCL based. You guys are also right I suspect with Intel and AMD joining the party, its all got to be good news for the likes of us - just wish it would all hurry up!