Hacker Newsnew | past | comments | ask | show | jobs | submit | jwagenet's commentslogin

GTAV had a 60GB install size over a decade ago.

I've used a Dell Precision 5530 professionally and got a 5570 refurb this year from ebay for ~$800. The fit and finish of the Precision 5000 series is great as far as I'm concerned, though I'm happy the camera is back on top of the screen and would appreciate a 10 key. The work model I used for 3 years and basically the only issue I had was on the Windows side with sleep states (waking up from sleep while commuting). I rarely work long off ac power, but <40% is always kind of a danger zone, especially when doing intensive tasks like CAD modeling. Again, worked connected to Dell workstation dock 90% of the time, so ports are not an issue, but the state of unpowered usbc dongles/micro-docks with hdmi/usba/usbc/++ makes stationary use a non-issue. I also had a 2016 XPS13 I only stopped using as a primary due to lack of ram expansion.

I got a used Precision 5540 from my work. I prefer it to my husband's 2024 Dell. The Linux battery life's even quite good, ~10 hours.

What on earth, I've got a 5540 and Linux battery life is atrocious, maybe 3 hours under light usage.

What are the specs?


SV already wrecked HW engineering by paying far more for SW than market rate HW such that anyone with financial ambition made the switch long ago.

The hard part with 3d part creation isn’t the graphical interface or language, it’s actually describing and translating part requirements to a manufacturable design, weighing material, weight, fit, geometric, and cost tradeoffs. Openscad, opencascade, etc have been around for a long time and have specs for describing features in a way that llm should be able to handle, but if all the part constraints were available it’s far faster to make accurately in Solidworks.

This is my experience too. I took a course a long time ago in design for manufacturing, and it became abundantly clear that just because you can conceive of an idea doesn't mean that you can build it. That requires a lot more work and technical know-how that isn't always put into books or other "training data".

In MCAD, “parametric” does not mean a high level part or feature is driven by editable parameters or procedurally generated features. Parametric refers to the underlying storage format representing part features in a parametric way rather than as a mesh. Mesh formats like stl cannot represent a circle by its position and radius, while a parametric format like step can. This distinction is more akin to raster (bmp) vs vector (svg) graphics. Both can be generated procedurally by “parameters”, but only with svg can sub-features be faithfully extracted or transformed.


I have some understanding of "parametric" vs "mesh". I looked it up when I saw so many people going on about it.

Maybe it is the export or something. I run the 3D toolbox and often models are not manifold.

I see things like two circles in slightly different positions but both are connected in different ways to the surrounding "single" instance model. Things like this mean you end up with "infinitely small volumes". There is no fully enclosed "volume" and so mathematically there is "nothing to 3D print".

As a model this makes no sense to do, and so it irks me.

But clearly the slicer software doesn't care or autocorrects and people make their 3D print happen just fine.


Sorry, separate point:

>Mesh formats like stl cannot represent a circle by its position and radius, while a parametric format like step can.

This is where I think the Geometry nodes can help. A node (function) can be used to represent the circle with inputs and outputs set or changed as required.[0]

I have not fully explored this space though and so my "hopes and dreams" may well be as useful as thoughts and prayers...

[0] https://docs.blender.org/manual/en/latest/modeling/geometry_...


The problem with FreeCAD and every other free/open source MCAD project of note is the Open Cascade kernel they are built on. While Open Cascade is fairly mature, it has dealbreaker issues in a few key areas: fillets cannot consume connected faces and may fail for a number of other reasons, cylindrical and spherical faces require seams which often cause issues with boolean operations, and shapes like helixes are also often troublesome.


On a scale from "big chunk of work" to "complete rewrite", how much work would it take to fix those issues in Open Cascade?


It would probably fall somewhere between "substantial architectural overhaul" and "partial rewrite" because it’d require redesigning the topological representation to eliminate seam edges.

Some of these issues are long standing and really hard to solve. Someone could probably defend an entire PhD thesis on “redesigning the topological representation to eliminate seam edges” without making much practical progress


How do the other kernels e.g. Parasolid work without seam edges? Without a seam the 2D parametric boundary is not closed.


It’s not about seams in 2d but 3d curved surfaces.

OpenCascade’s kernel forces you to deal with periodicity in topology (the shape structure), while Parasolid handles it in geometry (the math). A cylinder is mathematically continuous because there's no actual "seam" where it starts and ends. But in OpenCascade there’s a seam from 0 to 2π and this seam edge becomes a real topological entity that every algorithm has to deal with.

In Parasolid the cylinder is periodic so when you query a point at U=2.1π, the kernel just knows that's equivalent to U=0.1π. The periodicity is a property of the surface math, not the shape structure. It’s not using polygons/edges/vertexes but a system of equations to calculate the surfaces.

This is why boolean ops fail so often in FreeCAD: it’s asking the kernel to intersect with an artificial edge that shouldn't exist. The seam creates edge cases in intersection calculations, makes filleting near seams a nightmare, and complicates things. Parasolid's implicit handling requires smarter surface evaluation code upfront, but then everything else just works.


Is there any canonical literature on this? I've been interested in what's inside the brep kernels recently.


Boundary Representation Modelling Techniques by Stroud is probably the most popular one. It's expensive but Anna has it in her archive.


Thanks!


> fillets cannot consume connected faces and may fail for a number of other reasons

I can't recall a single CAD system which did this differently. Has modern solidworks figured this out?


Sandia seems to have some form of kernel, but only Federal-associated entities can get access to it.

It would be interesting to see if they would license that out further for some amount of money.


If you're referring to Cubit, they license the ACIS kernel under the hood.


They’re (possibly) referring to “Scalable Geometric Modeler” (SGM)

https://github.com/sandialabs/sgm

Originally open-source, but since taken back in-house. As I understand, which should not be construed as an accurate accounting, Sandia wants to flesh out the basics further before (potentially) open-sourcing it again.


I was referring to Cubit. Phooey on the fact that it's ACIS.


I don't get the gripe. AirTrain gets you to A,E,J,Z, and LIRR, all of which get you to "Manhattan" or a significant number of intermediate destinations in about an hour. LGA is far worse.


Having to take AirTrain beyond the terminals at all is annoying. LIRR should just go to JFK directly. AirTrain is slow as molasses, and the fact that it costs money is absurd. It works and I'm glad it exists, but it's nothing like e.g. the Paris RER connecting CDG.

You generally never want to take A/E/J/Z because they're sooo much slower than LIRR, unless you live along them.

Yes, LGA is far worse.


> the fact that it costs money is absurd

Bart from SFO to downtown SF is about $11 due to a surcharge and the combined fare AirTrain + subway is also about $11.50. LIRR is a bit more expensive. The Paris RER is €13. I don’t see how the fare is objectionable.

I personally appreciate the subway connections exist. Taking LIRR would require a subway transfer to most destinations anyway.


You're comparing apples and oranges. The LIRR already is the train ticket. I'm complaining about the fact you have to pay two fares. Using two different systems.

And like I said, you don't want to take the subway unless you live along its route, it's so much slower.

If you need to pay for the construction cost of the AirTrain, it should just be funded as part of the airport generally, because that's what it is. Charging for it is as silly as if you charged to take the AirTrain between terminals.


The SAT and ACT plots indicate an accelerating downward trend beginning in 2018 though, later exacerbated by COVID.


That is exactly the bad analysis I was calling out.

If you take a data set and point out (while squinting) that it appears maybe to be turning down in the last two data points, any reasonable analyst should point out that those look like routine outliers and that if you want to project a trend you need more data.

Instead, you'd taking a very large (and well understood) signal in the next 5-ish data points and saying that it's proof of the trend. Which is silly.

No, that chart shows covid, period. If someone wants to show an uncorrelated effect across a signal that big, they need to come to the table with a lot more sophistication than a "WTF Happened?" blog post.


Fine, the dips 2018-2019 could very well be noise and we don't dont know if that trend would have borne out after 2020. However, the ACT and SAT composites gradually declined following peaks in 07 and 12, respectively, given the available data.


Come on. Extremely mildly down-sloping if at all, and without anything like an inflection point that would justify "WTF Happened in 2007?"

You can't do this like this. I mean, you sort of can if the effect is big enough (c.f. the 1971 site which inspired this one, and which makes a much better case). But if you can't eyeball a very clear angle in the chart without argument, you need to come to the table with some kind of fitted curve and real research.


> Maybe gaming

Only if you limit scope desktop gaming, sure. 75% of gaming market share is on mobile and consoles.


Oh true.


Yea the open cascade kernel underpinning freecad (and python libraries like build123d and cadquery) has a number of rough edges. Many can be avoided, but that unfortunately requires a bit of experience.

IMO, lofts are often not the best tool for many jobs, but planning around fillet issues can be a pain.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: