I need help with choosing the right cpu upgrade for my blender projects.
I run a ryzen 5 5600x and find that it struggles with geometry nodes (hair nodes specifically). It literally takes 5 seconds every time I change a value for it to be reflected in the viewport.
Since I need it to be quite responsive for the small tweaks I make here and there, I was wondering what CPU upgrade I should get?
I thought about the ryzen 9 5950x, but since it’s “old” now, I thought maybe it was a better idea to upgrade the whole thing for a 9000 series?
Pricewise I think the 5950x would be a better option (since I wouldn’t have to upgrade my motherboard), but I’m concerned it would still not be enough.
I can’t recommend something, but geometry nodes should be able to take advantage of multiple-core CPU…
That said it could be natural that some GN setups are long to compute, in case of hairs unless you specifically build your system to be responsive, ( like taking inspiration from video game techniques) , it’s very likely that it will take some time and you need one system for the viewport with only a few hairs, and a complete system for rendering…
with some quick maths (correct me if I’m wrong) …:
if your current CPU take 5s to compute :
a CPU twice as fast will take 2.5 s
a CPU 4 time faster will take 1.25 s
a CPU 8 time faster will take ~0.75s
a CPU 16time faster will take ~0.35s
That’s what I was thinking, but I asked the tutor I’m following (Smay on YT) what cpu he has since he gets near instant results in his tutorials.
He told me he has a 10core i9 (I don’t have the specific model, although he said it was not high end, so I assume a bit old).
So it seemed to me that the 5950x would be better since it has 16 cores.
When I look at the task manager whenever I make a change all 6 cores and 6 threads are utilized, so adding 10 more cores and threads would make the process faster, no?
Would you know what kind of operations I should look at on benchmarks to make a choice? Is a geekbench score reliable for this kind of application since it averages different kinds of tasks?
Also most benchmarks online look at render times but it’s not what interests me here.
Also considering the fact that I would need to spend 3 times as much to get all the hardware compared to just buying the cpu, would you say a 50% increase in single core perf is worth it or overkill for that kind of task?
well, maybe if you have access the exact same file your tutor made you can try a fair comparison…
it might be that some parameters make yours particularly slow…
TBH I don’t know about AMD Cpus so it’s hard for me to tell if you have a really slow CPU or if it’s useable but you need to optimize your scenes…
In any case, it’s generally always possible to make things faster, which is a nice skill to have even with powerful hardware !
So a couple of things. The 5600x is a 6 core 12 thread CPU, hence may need to expand/adjust task manager a little to show everything that is going on.
Second, how much are they being utilized, are all hitting 100% or only 1-2 cores. Unless the 5600X is being totally maxed out, then going to a 5950X would make little difference (outside of the fact that the 5950X clocks a bit higher, etc).
I’d be doing a bit more testing/benchmarking first to make sure exactly where and what the bottleneck is. Otherwise you could be spending money on a new CPU that only ends up 10% faster.
I did not even think about making a comparison with the provided material
Turns out the provided .blend files were much more responsive than my own node trees (not snappy but better).
I was not thinking about optimization since I’m going for realism, but I kind of forgot you can optimize the viewport performance, without affecting the final result. That’s a noob mistake I guess but I’m starting out.
A note for people dealing with hair nodes and struggling with performance: If you are using an interpolate hair curves node, try lowering the “Viewport Amount” slider to something your computer can manage.
My CPU still struggles so I bought the 5950x (300€ at the time of writing). It won’t set me back too much if it’s not enough but I’m sure it’s going to be better.
All the cores and threads are maxed out, so blender is using ryzen cpus quite well.
I bought the 5950x since it’s going to add more than twice the cores and threads. It’s not as good as the latest generation but it’s at least twice faster than the 5600x and is more cost efficient than upgrading most of my computer for the time being.
Optimizing the viewport + this purchase should solve the issue. I’ll update the post accordingly.
