@andybot said:
@kokoriko17 said:
Titan is a waste of money. GTX 980 Ti offers the same performance for half the price.
You're right, you can get good value from the high-end nvidia cards. Main thing is CUDA seems to be what most render engines use, OpenCL is not getting much support these days (I found that out the hard way
)
Another reason why I didn't go for the new Macs as they dumped Nvidia and got back to ATI ( dumb move I must say )
@andybot said:
Seems the best thing for vray currently is adding as many cpu cores as possible. With a big budget, I'd suggest getting a dual cpu motherboard (Supermicro makes most of them) and putting two Xeon E5 chips (like the 2680 v3, 12 core) on there. Graphics card doesn't matter for vray currently, and it may be a while before it is up to the same level as cpu render. If you do get a high end graphics card, better stick with nvidia Titans.
Andy
Titan is a waste of money. GTX 980 Ti offers the same performance for half the price.
@jql said:
I just said that with my SSD AND without thumbnails, I don't get 2 secs saves. I don't even notice saving anymore.
@unknownuser said:
I understand that but if you need your render to look nice you need to import some complex models ( flowers, vase, figurines etc.. ) and these really hit sketchup hard.
You should use proxies for that sort of stuff.
@unknownuser said:
Check the remaining time for just merging faces :
Merging faces is not saving a model, it's a complex face operation like explode or import.
My problem is not the rendering, I can render relatively quickly. But the problem with Sketchup itself.
I have been trying to load a 118 MB skp file into sketchup but it took over 6 hours. What's funny is that 3DS Max could load it in 2 mins.
We can't deny that Sketchup is slow, really slow and I refuse to leave it for basic modeling as it has a lot of potential to compete with the big 5 but then we need a major update.. maybe a complete re-written Sketchup.
@jql said:
@kokoriko17 said:
2 seconds
That's too long!
Frame your model nicelly and save. Then turn off thumbnail generation on "Model Info".
I understand that but if you need your render to look nice you need to import some complex models ( flowers, vase, figurines etc.. ) and these really hit sketchup hard. Check the remaining time for just merging faces :
@pbacot said:
@kokoriko17 said:
Ok so bottom line, no multithread support in the next year or so ?
It's astonishing that even with a Core i7 3.9GHZ 4C/8T, PCIE SSD , 32GB of RAM and GTX 780, I have to wait 5 hours to clean up a 110 MB high details file because my CPU usage is 15% at max while rendering the scene at very high details takes 4 hours.
Also with complex models, I get a lot of hiccups and sluggish performance and it's definitely not my Nvidia as GPU-ID reports 10-20% GPU usage while sketchup is freezing or not responding most of the time with 15% CPU load ( the maximum it can use ).
I'd swear by sketchup if it were multithreading.
What process are you doing that you wait 5 hours? CleanUp plugin or ..?
What? So to be clear: You hit a scene and wait 4 hours to see it on the screen. Or is this for some sort of output like pdf? Until it gets faster, break up your files. If it's details, why put them all in one file?
Hiccups. Yeah the autosave for large files is a pain. As someone noted regular save is relatively fast, so why is autosave slow?
Hello and thanks for your reply; Yes the cleanup plugin takes too much time. Not that I am trying to compare sketchup to Autodesk products but Autosave usually doesn't lag me that much as I'm on a PCIe SSD ( 800+ MBps read/write ) I hit save as and it takes 2 seconds to save a 150MB sketchup file.
It's not only about the cleanup. For instance if I want to import and obj file using Simlab plugins sometime it takes 20-30 mins depending on the file size because only one thread is active out of 8.
I understand that OBJ files can be complex but after cleanup they shrink significantly.
Here's an example of a scene I'm rendering with sketchup ( needs 14 hours to finish ) :
Before cleanup file size was around 350MB then it went down to 74MB but it took literally 20 hours to finish the cleanup.
Ok so bottom line, no multithread support in the next year or so ?
It's astonishing that even with a Core i7 3.9GHZ 4C/8T, PCIE SSD , 32GB of RAM and GTX 780, I have to wait 5 hours to clean up a 110 MB high details file because my CPU usage is 15% at max while rendering the scene at very high details takes 4 hours.
Also with complex models, I get a lot of hiccups and sluggish performance and it's definitely not my Nvidia as GPU-ID reports 10-20% GPU usage while sketchup is freezing or not responding most of the time with 15% CPU load ( the maximum it can use ).
I'd swear by sketchup if it were multithreading.
http://sketchucation.com/forums/viewtopic.php?f=322%26amp;t=63373
If you really have a lot of cash to burn, Xeon
Processor E7-8890 is the fastest to date.
Cheers...
@sketchuh said:
wow thanks for the suggestion. my company's budget could be higher though.
- will Nvidia GTX 980 Ti render faster than using Intel
i7-5960X? but this will give me some reference.
Actually for rendering in VRAY you only use the CPU unless you're using VRAY RT which will be set to CUDA. It's when things get interesting, you can preview whatever you're working on really quickly with such a capable card.
Normally GPUs are much faster than CPUs when it comes to processing power. Not that they are more powerful but they have limited instruction sets. That's why they're used over CPUs when it comes to bruteforcing and cracking keys etc...
Intel Core i7-5960X
32GB RAM DDR4 2133
Samsung PCIe SSD ( 512GB is OK )
Vray doesn't use GPU unless using RT CUDA ( for this get Nvidia GTX 980 Ti - Titan X isn't worth the negligible performace )
CPU + RAM + SSD + GTX price all together around $2450
Add another $150 for a powersupply + Chasis
Connect it to your iMac screen using minidisplay adapter or get any 30" screen. ( i'm using my iMac and it's great for rendering - mine is Core i7 3.9 + SSD + 32GB RAM + GTX780m )
37 views and no reply which is a bit sad but I managed to fix it. For some weird reason I had to reverse faces, paint and revert. Now it's rendering correctly.
A note to developers, sphere lights are way too bright for an intensity of 30 if they are within a group. I had to put them outside the chandelier group for a proper brightness.
So I've followed literally every tutorial on earth to create a lampshade with translucency and after wasting 5 hours I realized that Vray RT does render the desired output but surprisingly not the final renderer.
I've used 2 sided material and set the color to RGB 104 with forced 1 side.
I hit R and below is the output.
I hit RT ( CUDA ) and below is surprisingly a different output
Everything else is rendered correctly in VRAY R
Machine :
iMac 27 Core i7 3.9 GHz
32GB RAM
PCIe SSD
Nvidia GTX 780
Appreciate some guidance.
Cheers.