Best GPU for Sketchup + Thea GPU Rendering?
-
@liverpudlian82 said:
Thanks for your opinions.
So understand that CPU is the bottle neck in SU. Problem is: it does not get much faster than what I already have
@liam: 50.000.000
@solo: Love your avatar
One thing I have used in SU to get around this problem is create a model in multiple documents and then assemble in the render engine. Or only compile the model when you want to output for a render. I have special tags in my geometry so the bits piece together like lego. I can then delete that geometry in Thea if that makes sense?
And if you cant be bothered to do that you can use the layers panel. I have managed some fairly big models in the multi million polys on my laptop.
-
Well, I just did it, placed an order on Amazon for a Titan, expedited to arrive Friday. Base camp may be out of reach as I spent $1000 on a video card (wife cannot understand these things, kids think my work machine is their next gaming rig...dream on)
Would like to see how a GTX 660ti and a Titan work together for rendering.
-
@liverpudlian82 said:
So understand that CPU is the bottle neck in SU. Problem is: it does not get much faster than what I already have...
the fastet i7-3770K@3.50GHz reaches 9.577 Passmark points, the i7-4930K@3.40GHz of the recent 'Haswell' series 13.624 Passmark points ~ 42% better number crunching.
combined with a nice SSD (-> Samsung 840 Pro) this may give a neat boost for SU.
[update]
oops, the 4930K is a six core CPU and thus won't help much for SU besides over-clocked... but maybe helpful for rendering though.
[/update] -
@sketch3d.de said:
This is a DirectX benchmark... how should this say anything about sketchup performance?
http://www.passmark.com/products/pt.htm
http://www.passmark.com/products/pt_adv3d.htm@sketch3d.de said:
Quadros typically do not show any advantage for SU display output speed, i.e. burning money for nothing.
I'm not sure about this... I'm waiting for a clarification for years now. Sketchup is still openGL, so theoretically there should be an advantage. I really would like to see some fundamental tests of this, but a problem is that there is no proper benchmarking method. Maybe by playing an automated sequence of a bigger scene, but then the adaptive degradation would have to be disabled for proper results...
But yes, i think the best way to speed up sketchup is still a high single core CPU performance.
@solo said:
kids think my work machine is their next gaming rig...dream on)
-
@liam887 said:
For example two GTX 680's in SLI will out perform the Titan as a whole but i dont know if Thea supports that? Pete?
Running two cards will help, but forget connecting those to SLI, it wont help or actually may cause issues, as SLI is only for "game" graphics... it does not work with CUDA. Sure Thea can use both (darkroom), but for IR rendering it will use one dedicated to it and the other one will be used for running display, so you get good response when working with UI.
-
@unknownuser said:
for IR rendering it will use one dedicated to it
How does one decide this, through the Nvidia panel or Thea studio? what about in the Thea4SU plugin?
-
@solo said:
@unknownuser said:
for IR rendering it will use one dedicated to it
How does one decide this, through the Nvidia panel or Thea studio? what about in the Thea4SU plugin?
From nvidia control panel you can just enable/disable cuda for certain program/globally.
Thea and Thea4SU has devices panel that can be used to set priority, interactive and so... that is the place where you select what is interactive GPU.
-
@numerobis said:
This is a DirectX benchmark... how should this say anything about sketchup performance?
benching the trianglation throughput speed of GPUs by processing facetted 3D meshes surely provides some valuable results to derive from for the display output performance of SU even if MS Direct3D is used. The relations of the values between the respective models will probably be very comparable if the OGL stack would be used.
@numerobis said:
I'm not sure about this... I'm waiting for a clarification for years now. Sketchup is still openGL, so theoretically there should be an advantage.
SU is currently using the capabilities of the ol' OGL version 1.5, therefore the advanced capabilities of the Quadro FX drivers certified for high-end modelers as e.g. Catia, NX or Creo etc. are simply not required.
@numerobis said:
But yes, i think the best way to speed up sketchup is still a high single core CPU performance.
definitely, but for a fast system the components combined also need to be on a par, a fast graphics accelerator won't help if the main processor is the bottle neck and vice versa.
-
I currently have a 680 GTX 4gb and it is really fast. In talking with the guys over at Eon, I will be getting a Titan next.
-
@unknownuser said:
I currently have a 680 GTX 4gb and it is really fast. In talking with the guys over at Eon, I will be getting a Titan next.
Is this because you got some inside info?
We are talking LumenRT or is Vue going GPU?
-
@solo said:
Well, I just did it, placed an order on Amazon for a Titan, expedited to arrive Friday. Base camp may be out of reach as I spent $1000 on a video card (wife cannot understand these things, kids think my work machine is their next gaming rig...dream on)
Would like to see how a GTX 660ti and a Titan work together for rendering.
@Solo: Did the card arrive yet? If so, please let us know about the rendering experience with Thea's GPU render engine
-
@unknownuser said:
@Solo: Did the card arrive yet? If so, please let us know about the rendering experience with Thea's GPU render engine
It did indeed and it's installed.
I decided to take Giannis's advice (Thea developer) and keep my existing GTX 660ti 3GB and add the Titan in next PCI express slot, I kept my dual monitors connected to GTX 660 and left it as main GPU.
Now in Lumion and Thea I set it's priority in the Geforce settings panel for the Titan as top card so now when I work (render) the Titan takes on all the work and my other card is pretty free to still surf net, work in Photoshop etc.
I did a test using a Thea bench mark scene.
Opening file "C;\ProgramData\Thea Render/Scenes/Benchmark/Presto/MeccanoPlane/MeccanoPlane.scn.thea" Finished in 0 seconds! Device #0; GeForce GTX TITAN Building Environment... done. (0.305 seconds) - Polys; 503556, Objects; 91, Parametric; 0 - Moving; 0, Displaced; 0, Clipped; 0 - Instances; 0, Portals; 0 - Nodes; 158062, Leaves; 158063, Cache Level; Normal Building Environment... done. (0.315 seconds) - Polys; 503556, Objects; 91, Parametric; 0 - Moving; 0, Displaced; 0, Clipped; 0 - Instances; 0, Portals; 0 - Nodes; 40851, Leaves; 122554, Cache Level; Normal Device #1; GeForce GTX 660 Ti Device #2; Intel(R) Core(TM) i7 CPU 940 @ 2.93GHz Device #2; memory 4189/12278 Mb Device #0; memory 503/6144 Mb Device #1; memory 476/3072 Mb Device #1; 205 s/p Device #0; 538 s/p Device #2; 140 s/p Finished in 5 minutes and 2 seconds!That's s/p 883.
When beta testing (without Titan) I got s/p 330.
As you can see Thea used the Titan the 660 and the CPU.
So as you can see the Titan made my rendering almost 3x faster.
-
@solo said:
@unknownuser said:
@Solo: Did the card arrive yet? If so, please let us know about the rendering experience with Thea's GPU render engine
It did indeed and it's installed.
I decided to take Giannis's advice (Thea developer) and keep my existing GTX 660ti 3GB and add the Titan in next PCI express slot, I kept my dual monitors connected to GTX 660 and left it as main GPU.
Now in Lumion and Thea I set it's priority in the Geforce settings panel for the Titan as top card so now when I work (render) the Titan takes on all the work and my other card is pretty free to still surf net, work in Photoshop etc.
I did a test using a Thea bench mark scene.
That's s/p 883.
When beta testing (without Titan) I got s/p 330.
As you can see Thea used the Titan the 660 and the CPU.
So as you can see the Titan made my rendering almost 3x faster.
Thx for the update. Appreciated!
-
Next purchase Pete:
-
@liam887 said:
Next purchase Pete:
Firstly can you imagine the price?
I was looking at a dual Xeon system recently but decided to rather stick with my very tired 1st generation i7 (never did get the funds to take advantage of the deal I arranged early this year for community) I instead gathered up all my sheckles, borrowed a few from wife and bought a Titan card instead.
My thinking...
What do I need a computer for, well it's modeling, rendering and animation, so with Thea now being GPU and as an insider i know it will only get better and support all features very soon, so I really needed GPU power as animation pretty much is all GPU, and now I only render with GPU and with my new 6GB card even heavy scenes are a breeze, so why need more CPU? In fact the Titan card extends my machine another year...at least that is what I hope.
-
Indeed. Rumours is will be priced around 1k.
-
@liam887 said:
Next purchase Pete:
Just wondering: A second Titan would cost as much and increase render speed to a greater extent, wouldn't it?
-
@liverpudlian82 said:
@liam887 said:
Next purchase Pete:
Just wondering: A second Titan would cost as much and increase render speed to a greater extent, wouldn't it?
A second Titan would increase render speed BUT not add to the Ram as Thea only will use the ram from primary card, so IMO instead of getting a second Titan get a very fast GTX 7XX at less than half the price.
Remember the Titan is king only because it has 6GB ram and high clock speed, still not the best, the best is this: http://www.newegg.com/Product/Product.aspx?Item=N82E16814133494
IF YOU GOT $$$$$
-
@solo said:
I was looking at a dual Xeon system recently but decided to rather stick with my very tired 1st generation i7 (never did get the funds to take advantage of the deal I arranged early this year for community) I instead gathered up all my sheckles, borrowed a few from wife and bought a Titan card instead.
I'm interested to hear this as I'm in the same boat. I have an early i7 2600K and I find it lagging on lots of projects. But in this year alone I bought a license for Rhino, a license for V-Ray for Rhino, upgraded my V-Ray for SketchUp and upgraded my Modo license. I managed to get good prices on all of them (it is worth it to wait for deals) but it adds up to a lot of dough. I figure a new computer is going to set me back at least $2500. If I could get a speed boost for another year by buying a Titan, that would be more realistic considering the carnage that all that software did to my bank account.
-
@arail1 said:
If I could get a speed boost for another year by buying a Titan, that would be more realistic considering the carnage that all that software did to my bank account.
Next generation of nvidia cards (Maxwell) is announced for feb/march which should feature unified memory architecture to use the system RAM. So i would wait at least for this.
For the rest of the system... if you can wait a bit longer, Q3 2013 will bring Haswell-E with 8 cores and DDR4.
Advertisement