Will generative design ever get a local compute feature?
@Anonymous
Very well said, thank you.
I will accept this as a solution in that you have adequately shut down @mavigogun 
Quite frankly, the early adopter got jack squat.
I've had a Fusion 360 subscription for the last 5 years. The deal was that we are going to enjoy a $300 subscription for life (product life that is) if we keep the subscription.
The advantage was that we had access to the Ultimate version. The fact is that the ultimate version never gained much traction, because there were very few usable things that were actually ultimate. Now were all in the same boat feature wise. Every one who subscribes new gets to pay $495 (or so) and we continue to pay $300 and get 1000 cloud credits. Same as every other subscriber.
I agree with @Anonymous that this is a business model and not a limitation of computers and software.
However, at this price level creative use and exploration are out. However that exactly is necessary to develop an intuitive understanding of the software and methods.
@I_Forge_KC I am not exactly sure what you disagree on and where I am comparing apples with oranges.
If you are talking about industrial use then you are correct. You left out one package that has offered topology optimization for years and that is Altair Inspire and AFAIK is more affordable than any of the solutions you have mentioned. The also offer a generative design solution.
But I am NOT talking about industrial use. I am talking about a lot of creatives, tinkerers, makers etc. that in my opinion run circles around the traditional cubicle-engineers of today but are individuals that cannot afford shelling out that much money just to experiment with a tool.
Everyone forgive me... this is long and winding...
SolidThinking (with Structures) is $3195/yr
Inventor (standalone) is $1935/yr
Similar SIMP method in both.
What I disagree with is that there isn't a need for the cloud compute solutions. It's far more than a business model - it's a practical way to leverage high-performance computing without self-investment (currently the only way).
--------
There is also a huge problem here in that there is double meaning to Generative Design.
Generative design (little d) is a paradigm where we use heavy computation to achieve optimized geometry - this includes topology optimization, lattices, DfAM, topology synthesis, etc. When anyone in the industry at large says "GENERATIVE DESIGN" they are referring to this paradigm and 9 times out of 10 they are really referring to SIMP topology optimization. This is true of Altair, Ansys, Siemens, PTC, etc.
Generative Design (big D) is a marketing name Autodesk gave to Dreamcatcher - a level set method for topology optimization (aka topology synthesis). Very long story short, the guy behind Frustum (the other level set method out there) was an Autodesk research scientist before he founded Frustum (I'm still unclear if his departure was voluntary... though I tend to doubt it). That's the reason why Frustum and Generative Design are the closest matches in terms of comparison. In time, this name choice will make more sense because the system will grow beyond the current synthesis methods to encompass more of the paradigm.
------
@rkalak If your current design tool is Inventor... USE INVENTOR! Shape Generator inside Inventor is a SIMP method for topology optimization. The only sucktastic part about it is that the results are strictly mesh output (which is par for the course, really). Since you're a student, you also qualify for Nastran for free... that's what powers Shape Generator in Inventor but you can throw a ton more constraints into the full system. In both cases, the solver is local (and CPU based).
As far as a GPU SIMP solver, they do exist... though I'm not sure anyone in the marketplace is actively pushing them out (yet). There are several research papers released within the last couple of years that show their development - so I imagine they are still immature for production systems.
Autodesk is trying to position Generative Design to be a design exploration tool and let SIMP tools be the design validation tool. The inner workings of SIMP methods are the same as normal FEA methods. Since the gold standard for validation is FEA, it makes sense to use SIMP as the validation method for topology optimization. With that said, the only reason why AGD is good at design exploration is that it works on multiple studies in parallel - so that you can explore multiple designs in short order. The actual topology that comes out is perfectly fine from a functionality standpoint, but in the end, you're going to use a normal FEA program to validate it anyway.
(FEA = finite element analysis... the virtual testing tools)
@Anonymous The Enterprise Token system is basically unlimited (for organizations of size).
Very early on I had discussions with the managing team over Generative on pricing. The basics of the pricing scheme are that you spend roughly the same amount of money whether you have a design engineer cook up those results or you have the computer do it. It's not necessarily meant to be more cost-effective but rather be an on-demand resource. Basically, you're looking at spending $1,500 - $5,000 to solve a topology challenge but without actually outsourcing a contractor. It's certainly fuzzy math, but the point gets across.
I also would say that Nastran is the desktop solution from Autodesk and that the cloud stuff is where they are pointing people for on-demand resources... the same as Ansys or Altair does. Where the "rub" comes from is that Fusion is the long-term replacement for Inventor (yep, I said that out loud) and so there are features that will get folded into Fusion that won't make it into the old guard. With Fusion being cloud-centric, we'll be seeing more and more cloud-based features.
My wish is that they would turn people like me loose with an unlimited amount of CCs to go and teach the RIGHT way to use AGD. I literally wrote a LinkedIn article about this very topic on Friday and how there isn't a good guide yet. I don't claim to be the patron saint of AGD, but I've put in more hours than any other user outside of Autodesk and I know the best practices to make it sing. My business thrives on knowledge transfer and I don't like being in an ivory tower like I am on AGD. It's just hard for anyone to pony up several thousand dollars just to burn on inconsequential projects to learn from - and I totally get the heartburn there. Truthfully, over the last 18 months, I created hundreds of projects with thousands of runs to brute force my way through learning... I can't imagine the cost of it by today's pricing (but this was pre-release).
Now if only they could shoehorn the old Blue Ridge Numerics solver into Linux so it could be cloud compatible........
@I_Forge_KC wrote:
What I disagree with is that there isn't a need for the cloud compute solutions.
I agree that there is a need for cloud compute solutions, but that does not mean that it isn't a business model 😉
There is also a huge problem here in that there is double meaning to Generative Design.
Triple 😉
Because what we are talking about is really generative engineering, not design.
Or maybe generative mechanical design but sure as heck not generative Industrial Design.