I’m using the subdivision modifier with my models to bake some textures and I noticed that I can’t go higher than about 4 million polygons with my i7 4790K, because Blender starts to lag. I was hoping that somebody with Haswell-E line of CPU’s can tell me how far they are able to subdivide their models without lag. I was hoping to go as far as 6 or 8 million polygons. Can somebody tell me if it’s possible and with what PC components? Thanks in advance.
I don’t have a Haswell E processor, but if you’re talking about the display itself, then what will have a far greater impact for performance would be if you upgraded your GPU instead (because it’s not the job of the CPU to draw the polygons).
That’s not to say a faster CPU wouldn’t be beneficial for a host of other things (such as calculating the modifiers themselves).
I just upgraded my GPU to GTX 970 about 6 months ago, but already have plans to get GTX 1070 when they come out. I hope to get some improvement with it.
What I want to do is have a lowpoly version and higpoly version of the model and bake the textures from the highpoly model over the lowpoly model. As I said, the PC starts to lag at about 4 million poligons, but that’s not enough to get my model detailed enough to bake the texture and was hoping to subdivide it to about 8 million polygons. Is it still the display what affetcs it and is a GPU rather than CPU what I need?
So, the GPU is what I need to upgrade rather than the CPU, as I said I plan to get a GTX 1070 when they come out. Maybe I’ll wait for the custom models rather than buy the reference models quickly.
Hi
System:
i7 5820K / 3.3 GHz Processor - LGA2011-v3
ASUS X99-D
4x8GB DDR4 2133MHz PC4-17000 CL15
Made a quick dirty test.
Subdivision.
8 M Face…No problem at all…Viewport is Smooth as butter…
32 M Face…Close to the same…Smooth.
Brushes work as normal.
Multiresolotion:
Same as Subdivison.
Mutires can go to 128 M Face…See Picture…But viewport LAG…Preview 8
Brush stroke work good.
Note:
Subdiv use close to 2 times the memory as Mutires…Subdiv 8 M Face = 3 Gb…32 M Face = 5 Gb
Multires…Use 6 Gb when it subdivide from 6 to 9…When the mesh is updated it go back to 2 Gb
See it in Picture…It is possible to sculpt with 128 M Face If You have Time…) Brush work as normal.
But viewport Damn…He he
Edit:
Forgot to say…I wrote Blender in Suzan’s right eye and Rock in the dot after…Puff Puff
…
Tai
Thanks for the info, it’s really useful. I’m not able to see the pics, can you fix it please?
I was wondering if there’s any settings for my GPU that will help me with the lag.
May I ask what GPU are you using?
I was wondering if there’s any settings for my GPU that will help me with the lag.
Sorry for the double post.
Hi
I upload the pic again…Hope you can get the Attached Thumbnails…
I use a 980 ti 6 GB…I di not wrote it because it’s in the picure…Don’t think You can do much with lag…
Tai
Attachments
So, there’s a change that the improvement comes from the GTX 980Ti instead than the CPU. Am I right?
I made a test on my PC:
i7 5820K 3.7GHz, 32GB DDDR4 RAM
GTX 970 4GB
Monkey with applied subdivission 6 - 8 million faces - still smooth
another subdiv - 32 million faces - it is already quite laggy - system memory 16gb used, gpu memory 3.2GB used
Thanks for such useful piece of information. This is exactly what I was looking for.
I’ll do a test also with monkey in my system:
i7 4790K 4.0 GHZ, 16GB DDR3 2400 MHZ RAM, GTX 970. I’ll look for highest polycount where it lags and post it here.
I think that I will wait for the Broadwell-E CPU’s that are supposed to be released in Q2 2016. They should give me even better performance than my i7 4790K and even better when I add the new GTX 1070. Thanks for your help!