Certifiably Insane

DSC_8997As the Computer Aided Design (CAD) world transforms for good or ill, under the mounting pressures of The Amazon, software services, subscription models, ozone depletion and virtualized streaming, the long-held institutional definition of what constitutes CAD hardware is changing.  Few these days are keenly aware of a hardware certification process – where specific workstation hardware is qualified by software vendors as a certifiable configuration ready for duty.  The process is slow and methodical – not to be confused with minimum or recommended software requirements – but instead testing and documenting precise machine configurations.  The practice is more common the higher up you go in the CAD marketplace, with the presumption that more expensive products are supposedly pushing the hardware performance envelope.  We know that presumption is certifiably insane in itself, but let’s not get ahead of ourselves.  As technology has moved forward (or sideways?), even the most entry level hardware is becoming quite capable from a 3D performance standpoint.  So an important question to ask: does certification really matter anymore?

The canary for hardware certification, if there ever was one, is the purposely bifurcated graphics card market.  A recent LinkedIn discussion drew attention to the differentiation that the NVIDIA wizards have wrought between the Geforce line and the Quadro.  It is thus written in all the available marketing that Geforce is for gamers, and Quadro is for professionals.  So your choice on graphics card depends on whether you’re blowing stuff up in your space battleship on a lazy Sunday, or you happen to be building your own in an unmarked hangar in the Mojave Desert.  The Quadro line is clocked more conservatively for improved stability and carries ECC memory (again for stability) available in larger capacities.   Aside from that, the core hardware is exactly the same.  The major difference, double precision floating performance, is defined by the software drivers – enhanced in the Quadro, and severely limited on the GeForce counterpart.  That one tweak makes all the difference when moving from the gaming universe, to the content authoring universe.  In the gaming space, simplified geometry tricks are covered up with tessellation and shaders to make all those explosions extremely pleasant to the eye.  Call it the Michael Bay approach.  It’s a stark contrast to the CAD world which is more about pushing heavily vertexed geometry with relatively boring shading or in –gasp– wireframe.  The price differential between Quadro and Geforce however is rather amazing, colossal.  So why pay space battleship prices, when a space rowboat might do nicely? It would seem that providing software that could support either scenario on both hardware variants (i.e. a truly unified driver) is easily feasible.  Except that won’t happen.  NVIDIA knows that the CAD market is willing to pay substantially more than the gamer market.  The former might spend thousands or tens of thousands on software, while the gamer market gets excited by $3.75 sales on Steam.  The practice is market segmentation, and it’s a tenant of capitalism.  They do it for the same reason AT&T can charge $40/month for a POTS dial tone.  Because they can.  What’s included in the software certification basis: of course, the Quadro.

Regardless of marketing shenanigans, the whole process of certification seems to be losing value fast.  For one, certification involves quite a substantial bit of testing, and as new hardware is introduced there is a notable time lag before certifications are published.  Of course, that’s been tolerated in the CAD industry up until this point when upgrades cycles may happen only once every time the polar ice caps melt or 10,000 years, whichever comes later.  Adding insult to injury, certification has never been a guarantee of anything – software bugs and incompatibilities still happen.  Certification only provides assurance that perhaps a monkey or highly intelligent border collie sat in front of a particular hardware configuration on a particular version of the software for some finite amount of time, and no one got killed or teleported into hyperspace.  In the past,  concrete distinctions between workstation and consumer class hardware existed, in the modern market much of that differentiation has largely faded.  Certification instead has become more of an artificial device to hold over users calling in for support: not using a certified setup?  Well, go pound some sand.  Final point – some of the hardware selection is well, rather obvious.   For example, do you think the next Quadro card will not be certified for the cognizant apps?  Bottom line is that choices have mattered less as the hardware technologies have matured.

As the availability of certain CAD technologies continues to migrate further down into smaller markets and the consumer space, expect a lot of people trying to do content creation on substantially less capable hardware.  Those users will not only be on run-of-the-mill desktops but also mobile devices, especially tablets.  The 3D capability now possible even in low cost devices is rather impressive – hardware is leaps and bounds better than what was available five or ten years back.  Expect that to translate into higher demand for more serious products as time moves on and users graduate to more advanced concepts.  However, that demand will not be in the context of workstation-class hardware, but perhaps more along the lines of an average gamer rig.  And perhaps not even that considering all the virtualization solutions beginning to emerge. Even in more traditional markets, as the penetration of Software as a Service (SaaS) solutions and subscription models become more pervasive, the cycle time between software version changes will no doubt decrease dramatically.  Do you think the certification methodology will be able to keep up with that pace of change?  Probably not.  And will anyone actually care?  Signs point to no.