Too many instances?
-
I've been having problems with SU dragging pretty badly with lots of instances of even really simple components. For example, 200+ instances of a solar panel. I tried ditching the render-ready one I was using and instead importing just a simple box textured with solid colors. It still lags 10 seconds or more trying to copy a row of them once I have more than about 50 in the model.
I'm assuming part of the problem is the old Dell Latitude E6400 I'm using for most of this, and it will be replaced soon, but what is the most important thing, hardware-wise, to eliminate delays like this?
-
Good CPU and GPU with adequate RAM.
If it's a laptop you should get a dedicated nVidia GPU.
-
@rich o brien said:
Good CPU and GPU with adequate RAM.
If it's a laptop you should get a dedicated nVidia GPU.
I'll be switching from the dual core 2.66GHz laptop with 4G RAM to a quad core 3.6GHz desktop with (initially) 8G RAM and a Radeon with 2G VRAM, so hopefully that will help. Need to talk the boss into adding a fair size SSD to it soon, too.
I know part of the problem is the other stuff I pretty much always have open, so maybe having a couple more cores and stuffing more memory in there will get it all reasonable.
-
If you can test SU with that AMD card prior to purchasing it would be worth doing so.
AMD cards are notoriously SU unfriendly. Plus for GPU rendering you're limited to OpenCL compatible render engines. Which is an incredibly small pool.
Given a choice I'd steer clear of AMD
-
@rich o brien said:
AMD cards are notoriously SU unfriendly.
Don't you mean SU is notoriously AMD card unfriendly?
Unfortunately, for the initial purchase I'm stuck with only a few package options. Going to a desktop, however, means that I'm not nearly as limited when it comes to upgrading. (And that I have an excuse not to haul stuff home unless they want to spring for another laptop and another SU Pro license.
) -
@airwindsolar said:
@rich o brien said:
Good CPU and GPU with adequate RAM.
If it's a laptop you should get a dedicated nVidia GPU.
I'll be switching from the dual core 2.66GHz laptop with 4G RAM to a quad core 3.6GHz desktop with (initially) 8G RAM and a Radeon with 2G VRAM, so hopefully that will help. Need to talk the boss into adding a fair size SSD to it soon, too.
I know part of the problem is the other stuff I pretty much always have open, so maybe having a couple more cores and stuffing more memory in there will get it all reasonable.
Since you have( or will have) 8GB of RAM I assume you have a 64 bit machine. Some things you should know. SU runs as a 32 bit application and uses only a single core, RAM in devices subtracts from that available. Since Su runs as 32 bits it can only address 4 GB of the RAM. If you have a 32 bit OS, which I doubt ref above, the kernel and the application split the ARM equally and you would only have 2GB for the application.
As you can see it is very important when you post a question you include all info. I have given up doing the search for folks , it is your problem so some effort is required.
It would also help if post a copy of your model if possible. It sounds like it is not very complex and should not be very big. Also have you done any model approach to help like using layers for visibility control and monochrome redndering during work to reduce load on the present graphics you have? -
@mac1 said:
Since you have( or will have) 8GB of RAM I assume you have a 64 bit machine.
Correct; it's Win8.1 64 bit, and the CPU is an AMD quad core at 3.8GHz. Since I usually have a couple of SU instances and few other things open at the same time (browser for CRM, Google Earth, and sometimes Kerkythea) it's not like the other cores are just sitting around bored and wishing they could play with SU too.
Performance is already noticeably better, though there's certainly still room for improvement. Hopefully the SSD will provide a fair bit when it comes in. Kerkythea is already dramatically better; even running only two threads while I'm editing a fairly large model at the same time, it's farther along a fairly complex MLT render after 36 minutes than it would have been in 2 hours on the old machine. -
I question if the SSD will do as much as you want. I use one in may machine and of course opening, saving should be quicker but normally the drive should not see a lots of use unless the model load is so large the paging file needs to be used , but it looks like you have more than enough memory. My working memory sets and private bites usually in low to mid MB range and CPU usage and memory in the 25% range. I do not render however.
Advertisement