In operating Fusion, how much speed is attributed to CPU Vs. GPU? Will a faster Gfx card make a bigger difference performance-wise than faster CPU.
Solved! Go to Solution.
Solved by smeggy. Go to Solution.
Hi, Thanks for the question!
This is something we are spending more time profiling and coming to deeper understanding of ourselves.
My belief is that CPU throughput (ie. not just clock speed) is the main driver of modeling workflow performance so long as you have sufficient System RAM and Video Memory (based on what resolution monitor you are driving and what "effect" options you have enabled - among other things).
What has your experience been?
We are happy to bring in our performance & graphics experts to discuss more if you'd like.
Thanks!
-Nathan
Hi Nathan, the new Fusion360 does not work with the Graphic card of my one year old Macbook Air. The computer is fast enough with its I5 chip. But when doing Rendering using the new Rendering feature the screen becomes black. The computer is not stopped. It simply does not work. Click disable will cure this symptom. That is the reason I just bought a MacBook Pro for that purposes. Fusion works great on MacBook Pro.
So it does need a matching graphic card. The one in the AIR sure does not work with Fusion360. It will be nice if it could do that. So if I am using the AIR I just use it the design but not rendering. But I can use the Autodesk RaaS. Maybe the latest generation AIR can work with Fusion. Unfortunately someone spilled water on my AIR and trashed it. So now I cannot discuss graphical performance of the AIR anymore.
Hi Kingson - Yes, I should have been clear that my comment was geared assuming you have sufficient system RAM and GPU memory capacity (Macbook Air has plenty of the former, none of the later since it lacks a dedicated GPU).
That said, Fusion is a unique CAD experience and we do strive to have it function well on a wide breadth of platforms and hardware capabilities and are continuing to look for ways to optimize on systems with more modest hardware and certainly apprechiate feedback!
-Nathan
Hi Nathan, Thanks for your fast reply. Now I had found a Graphic card unrelated problem with Fusion. I had been experimenting with WEB and RIB commands. I had found that using the MacBook Pro or other MAC, a slight sketching mistake will result in a report dialog box, and then program closes. Only a perfect sketch can prevent that. In other words, the WEB and RIB (both same in behavior ) cannot tolerate small sketching errors. But I understand those two commands more now. But the fast quiting of the program on slight sketching error might be something to improve on. I did not send the reports since it happened so many times while I was trying the two commands.
Well one thing I'm thinking of is getting a dual quad Xeon (older model but still much faster than what I have now) so can Fusion make use of all those cores/threads? Also I'm thinking about better a gfx card. So I guess from what I already know is that the better cpu's will get all the calculations done much faster but that won't necessarily translate into better screen performance all else being equal. Better Gfx cards will display the results faster but not help with core calculation. In the end it's all balance, I was just wondering if the program itself had a preference for CPU or GPU. If there's one thing I hate is slow screen updates, slow zoom/pan/rotation.
I don't know enough of the technical ins and outs to really make an educated guess as the best way to go so any guidence would help.
Hi Smeggy,
There are a number of operations in Fusion where additional processor cores are used/benefitial (RayTracing), that said - generally a faster serial processing pipeline will make the most difference.. ie for a given processor architecture (ie such as Romley-EP or if a bit older Nehalem-EP Xeon) a higher processor and memory bus speed are likely larger contributors to performance than the number of cores.
By "dual quad Xeon" I take that to mean a dual socket, two physical processors with quad cores each (so a 8 core machine)? My home system is a IBM with a single-socket quad core Xeon 5500 (Quad Core) and a rather anemic Nividia Quadro FX 570 (only 256MB!). In my case my best improvement would be to move to a 1GB GPU..after that GPU improvements would have rapidly diminishing returns (in terms of performance) and my next best bet would be to move to a faster processor.
So in the case of a Nehalem Xeon (ie. like x5550 or 55xx) you'd probably be better off turning off hyper-threading and turning on Turbo mode (which will shut down one or two cores and overclock the remaining) in the UEFI/BIOS configuration utility.
The shorter answer is that a Xeon EP (midrange) quad core system, even a few years old, is plenty up to the task if you have a appropriate GPU (I think 1GB is plenty good enough) should be fine.
-Nathan
Hi Kingson,
I'm sorry to hear that! Do you get the opportuntiy to file a CER or does it close without a crash report? If you can, please consider fileing one, this is something we want to understand better. I'll ask our QA folks to take a look at this and see if we can get a good reproduction for development to address.
-Nathan
Hi I did not file the CER. I think you guys will catch all these eventually since this WEB command is new. I found that RIB and WEB are working on different environment.
I'm using an Hackintosh i7-4770 16gb ram ssd and a low range ATI graphic card (HD7750 1gb). The hardware acceleration and openCL of the GC works perfectly as my Luxmark score is ok.
I've noticed that when you are sculpting you just use 1 core, but rendering use them all (which makes sense...).
The grafic card is used (i can hear the fan spinning faster when i turn my design in all direction) when i'm in the "model" mode. When you "select" a lot of design and make it turn in all directions, then it start to slow down, and the fan of the graphic card is spinning full speed. So i suppose the graphic card is also important...
For exemple when you select like this and turn it everywhere then it starts to slow down: