Hi. I’d like to know what Pc I need to get 24 fps playback with my projects without having to render-viewport all the time.
So, what fps do you get for each file, and what’s your pc and monitor specs?
(With gpu subdivision off, since that seems to slow down low-poly scenes. Not to be confused with subdivision modifier. Please leave that and simplify at level 1. And leave line Art on.)
( I.e. Play the files as is. But with Gpu subdiv off.
I know these things slow things down but I want to see if a fast PC can play 24fps at high quality, instead of having to render viewport everytime I want to see if the animation looks as intended).
The more responses, the better. Even if you don’t have all the info.
2-Characters File (open link, and click the arrow-and-line symbol in top right corner if download button doesn’t work)
4-Characters File:
.
.
.
Here’s what I get:
Two Characters: 13-14 fps with gpu subdivision off
Four Characters: 8.5 fps with gpu subdivision off
My Pc:
CPU: i7-3770
GPU: Radeon Hd 5770
RAM: 24 GB
Screen Resolution: 1080p, 60Hz (since this can affect performance, i think)
Bottleneck: Cpu single thread
(only one thread is maxed, other threads ~70%)
Blender 3.6.5
.
.
.
Plz let me know. Black Friday is coming up!
Also, if you have any optimization tips, let me know. But I’ve tried everything. If you wanna know what I’ve tried, feel free to ask.
I played it first on my ‘light’ machine. A Mac Mini M2-Pro with 16GB RAM.
That gave 25-27 fps for the 2 Char file, and 16 fps with the 4 Char one.
Then I put it on my ‘proper’ machine. Intel i9-12900K 64GB RAM and RTX 4070ti.
That only did 11 fps with 2 Chars, and 6 fps with 4 Chars.
That’s not the way I expected it to go. The PC will normally destroy the Mac when it comes to proper rendering. Although the Mac actully does surprisingly well with less complex scenes.
You do have Simplify enabled. If you turn the viewport value down to 0 instead of 1, then that helps with numbers. I didn’t fiddle with any other settings.
Doing that, the PC gave me 22ish fps with the 4 Char file.
The Mac gave 32 fps with the 4 Char file with that Simplify setting.
For interactive stuff I’ve been surprised by the Mac. Doing EEVEE and Workbench renders, it compares well against the PC (for scenes that will fit in the GPU RAM). The PC totally rules with Cycles renders though.
Yeah that’s not what I was expecting either. It’s slower than my 3rd gen cpu. What resolution is your monitor for each computer? I think that can affect performance since 4k monitors need faster PC parts than an old 1080p monitor like mine.
Maybe I’ll consider a mac mini. I don’t use cycles. But I might wanna play PC games. I’m hoping more people will reply with more info.
The Mac was using a 4K screen at 60hz.
The PC was using 2560x1440 144hz with HDR (I think) - I’ve switched it off now
The PC was my main machine, but I use it mostly for more heavy-duty work and games now.
The Mac has become my main machine. I VPN to the office for actual work. I prefer it to the PC, as it’s silent, and doesn’t heat the whole house while I use it. It also works better for Zoom & Teams calls.
Thanks for the info!
Any idea why your 12900k pc is slower than my 3770 pc? I mean, sure, your screen is a bit higher res than mine. But someone else with a 10900k got 19-22fps and 13-16fps respectively (without touching subdiv, which is what I wanted to test). They even had an ultrawide monitor Edit: 5120x1440.
interesting that i get slightly better with sub on, but definitely worse with sub off than you, Safetyman. i wonder if it’s just that bit of a faster clock you’re running?
Just loading and playing the files as downloaded, I get the following:
2 char: 18fps GPU subdiv on
4 char: 10fps GPU subdiv on
2 char: 26fps GPU subdiv off
4 char: 17fps GPU subdiv off
This is using:
CPU: Ryzen 9 5900X
GPU Nvidia RTX 3080 Ti (note, this makes no difference at all, it’s all single core CPU bottlenecked)
RAM: 32GB
Resolution: 2560x1440 (also no impact, the GPU is doing stuff all)
Blender 3.6.5
As for some reasons why, well here’s a few.
You are basically doing low-poly models and animation, but have subdivision modifiers over everything. Maybe to smooth things out a little? Ideally just model directly and if things aren’t as smooth as you like, add an extra loop cut or two, it really won’t make any difference to performance.
In some cases those subdivision modifiers are before or after the armature. They should always be after the armature and in fact should ideally always be the last modifier. Reason being, if they are before, you are weight painting on the non-base mesh, this can cause a number of issues and second (and this is the reason the preference setting for GPU subdiv has an impact), is that for the objects that have the subdiv modifier BEFORE the armature, the GPU can’t speed them up and data is likely going back and forth between CPU and GPU, just slowing everything down.
All the mesh object have a mirror modifier before any rigging and weight painting… No, just no. Once all modelling is done, any mirror modifier is applied. The armature modifier is then the first at the top of the stack.
It’s possible there are other things that can be tested and refined to speed up performance, but that would require pulling apart each character rig, rebuilding and testing to see what impacts speed the most.
Plus on top of that, the Lineart of course isn’t helping, as that slows down playback speed by 10fps for me and it’s possible not much can be done about that.
This may help to provide some clues to what can impact viewport playback performance.
2 char: 22fps GPU subdiv on
4 char: 12fps GPU subdiv on
2 char: 32fps GPU subdiv off
4 char: 21fps GPU subdiv off
This is using:
CPU: 12th Gen 16 Core i9-12900K 3.20 GHz
GPU Nvidia RTX 3090 Ti
RAM: 128GB
Resolution: 3440x1440 (also no impact, the GPU is doing stuff all)
Blender 4
12th-gen+ Intel CPUs have some sort of Hybrid cores that excel at single-core tasks that are not multi-threaded. Maybe AMD has similar tech now? I’ve no idea.
Did you do comparisons with subdiv off altogether too? Simplify at 0?
How are you getting a lot faster fps than other people with only a 10900k?? Did you change any of the settings? There’s a reason I left subdiv modifier and simplify on level 1 and line-art on.
Or did you mean “subdiv off” when you said “GPU sub off”?
Is yours a laptop? Could that be why? Cuz I’m graphing these results, and your computer looks like an outlier that’s slower than expected (for gpu subdiv off).
Then again, there’s at least 1 other outlier that’s not a laptop that I can’t explain.
Ps. thank for replying.
it is a laptop, but it’s supposed to have a desktop cpu in it (with the multiple power supplies and extra inch of thickness for heat venting to match ) i might bump up the clock speed and see if that makes a difference, but can’t really do much until sunday, as laptop is mission critical until then. i .might. have had chrome open during one of the tests too… i know i shut it down for the first test (which would have been sub on), but not sure on the 2nd.
@Safetyman No worries. Thanks for replying. @KDLynch Don’t worry about it. Good luck on your project. Thanks for replying @thetony20 Thanks. I think I’ve seen your video before.
I tried applying mirror modifiers for everything, but didn’t get any fps improvement
My particular workflow is easier if I have mirror modifiers, so if there’s no benefit for me, I’ll leave them in. But thanks for the tip.
Yeah, I guess i forgot to put subdiv as last modifier for some objects.
) i might bump up the clock speed and see if that makes a difference, but can’t really do much until sunday, as laptop is mission critical until then. 