When you associate the state of CAD graphics to an alien conspiracy it just might incite a cheer, conversely perhaps a bit of rage, or in the very best case thoughtful commentary from a lagomorph well versed in computational geometry. The crux of my original article was not to convict or condemn, but rather to warn that the continually evolving CAD universe, and by association professional CAD graphics, is on the precipice of something here unto unforeseen and unimagined. It might seem like the end of all things. Something disconcerting enough that a forlorn and distraught Samwise Gamgee might longingly implore, "Don't professional graphics matter, Mr. Frodo?" For which our brave hobbit might reply: "They do Sam, they do. But not for me."
But how in the name of Middle Earth could professional graphics not matter? What of the vast armies that support development, optimization and maintenance of the drivers and hardware certifications? Are they merely coffee-sipping orcs overly preoccupied with esoteric machine precision and beating 19th degree Non-Uniform Rational B-splines into submission? Calling attention to the purposeful bifurcation of the graphics market might seem like an attack on such an army, an outright marginalization even. It's enough to get your kernels in a bind and make a break for Minas-Tirith.
But that's the wrong conclusion. They are not orcs. Although it's possible some of them might occasionally LARP as orcs, we're talking about some titanic efforts by some awfully smart engineers. Such efforts are indeed worthwhile, are indeed value creating and, most importantly, are absolutely critical for companies and even entire industries who work with serious CAD datasets.
So why isn't this just an open-and-shut case, nothing to see here, move along? In the prior article I mentioned the world of CAD graphics is complicated; I wasn't kidding. Let's continue the exploration.
Master model and the graphical arms race
Since the days of the Eldar, CAD modeling has continually pushed the need for faster, higher fidelity graphics capability pushing ever increasing amounts of data. The need to build, interrogate, and present a product's complete physical definition as a master model, a deep product structure containing hundreds of thousands or even millions of components, drives the relentless pursuit of ultimate computational performance.
Industries like aerospace, automotive, and shipbuilding especially have been at the forefront of this decades long graphical arms race. Here's the rub. With growing product complexity, professional graphics or not, we've long surpassed the capability to open the entirety of top level models on a single workstation.
Don't you dare load the whole plane
Here's one thing that doesn't happen at Large Aerospace Corporation: engineers wander in daily, bleary-eyed and coffee-infused, log-on to their workstations and promptly load the entire plane. If they manage to somehow attempt it, they peg the bandwidth monitor just before they run out of memory, CAD managers appear scowling from their dark towers and promptly wizard them into the ceiling.
Design in context doesn't mean design in all of the context. This is a critically important distinction - most day-to-day design is focused on editing individual parts in a relevant context, which involves loading far, far less geometry into memory. When you write a novel do you write it with every page simultaneously viewable in intimate detail? That's what I'm Tolkein about. Don't load the whole plane, mmkay?
What's good enough?
When you take professional graphics optimization over consumer graphics, for example hidden line display, it's usually demonstrated on a model that is typical of an irrelevant context; it's great for a tech demo to emphasize the performance differential. It's even legitimately necessary for those who do indeed have to work in an unusually large context. But I assure you, that is not most designers.
In the past, this has generally been a non-issue, the gap in performance between professional cards and consumer cards on a relevant context was still too wide to ignore. However, successive generations of graphics technology and a plateau in CPU compute performance have dramatically narrowed this gap. Many designers will open up a relevant context, and not necessarily notice the performance difference, and if they do it's in such a way that doesn't prevent them from getting their work done. You might think people will get upset about LOD dropout or frame rate irregularities, but you'd be surprised as long as UI latency is low. Which is why freelancers or engineers at smaller companies can jam a Geforce in a PC and be happily oblivious to an unoptimized (yet sufficiently productive) experience.
Case in point, ask yourself how SolidEdge can be certified on a MS Surface Pro running Intel 5500 graphics or that AutoCAD actually has Radeon consumer cards certified on the Mac side? These two certs exist for one reason, people want to get their work done on tablets and Macs, and the graphics, while probably rather awful, are quite simply good enough. This is one avenue of design democratization, the rising demand to use any device.
To be fair, in most cases, enterprises will stick with professional graphics because they aren't interested in the risks and maintenance costs of uncertified setups, especially if they are using the higher-tier tools like CATIA, NX, or Creo, and spending tens of thousands per seat.
Shift in graphics from workstation to infrastructure
But in even in formerly mentioned Large Aerospace Corporation, all the certified hardware in the world, still can't open the whole plane. This limitation is precisely why we have a variety of methods for large model management including selective loading schemes like volumetric filtering, lightweight formats, independent mockup tools for clearance analysis and offloaded virtualized graphics solutions like Nvidia's GRID. The emphasis of solutions like the latter will gain prominence in the future, because we're reaching the limits of practicality for stacking expensive cards in a box for these specific ultra-high-end uses.
It's reasonable to anticipate that the current army working drivers/certification will further specialize into this wholly enterprise-centric, physically-abstracted market, drifting away from enterprise-centric cards in your average PC, workstation or otherwise. In other words, moving from something individual users have to worry about (certified graphics) to something backstage that only enterprises have to worry about (certified infrastructure). So don't worry about the army or their jobs, they'll be replaced by robots well after most of us. That's avenue two for democratization, workstations, especially certified ones with purpose-built professional graphics cards, will matter less - at least for CAD.
But what if everyone does CAD?
Most of the people who will enter the CAD market in the next decade will be intolerant to the status quo. We'll be moving from just a few million engineers using hidden line display modes to tens and perhaps even hundreds of millions of makers creating and manipulating 3D design like it was a game. They will have trouble understanding what the point of an orthographic projection might be, and are totally satisfied with the 3D on their phone or tablet. Incidentally, when you download an app on your phone, do you make sure your phone is certified for it? This is the coming reality. The engineers and the enterprise scale design tools aren't disappearing, but they will be a voice in a much larger chorus, an edge case for very large design.
This ongoing democratization of design has instigated and will continue to catalyze the commoditization of CAD graphics. As CAD tools become more economical and accessible, CAD graphics will have no choice but to follow suit. In the meantime, I'll be on a boat. I'm headed west.
We're making professional 3D printing easy and accessible
There's a new solution coming your way this summer. Click here or on the graphic to the left to request updates and beta access.