Look it up. You can begin just with CPUs supported. I have no way of knowing what individual hardware comprised your company's systems, but I'm betting that much of that hardware is supported by the Linux kernel. Certainly there exists hardware that Linux doesn't support and that is supported by some other OpSys but my statement stands. No single OpSys in history supports or has supported more hardware than Linux. I'm also betting you now understand what I mean just in CPUs alone.

It may be your opinion and I have to respect that but you'd need to show me some facts to sway my decades of experience. I have been in technology since the 70's. I've seen all sorts of companies and systems and uses. It may be that at some point based on total devices that have any sort of OS one might soon discover some change to the accepted notion that the world runs on Windows. It won't be Linux directly either that wins.

@jefro - I think you are misunderstanding me. I'm well aware that MS enjoys a 90++% market share on the desktop and nearly equals Linux in enterprise servers, with the only currently Linux dominated niche being supercomputers. That is not my point.

I'm not saying what hardware has, at one time, on some version, been supported over time, cumulatively since new versions of many systems commonly drop support for older or less common hardware. CURRENTLY, the Linux 3X kernel supports more different kinds of hardware than any other single kernel of any single OpSys, ever has. Have you compiled a linux kernel lately from source? or even just run "make menuconfig" or "make xconfig"?

This is a start for documentation
http://en.wikipedia.org/wiki/Compari...system_kernels

Here is a more hardware specific page
http://www.linux-drivers.org/

Incidentally, the point could be made that supporting more and different hardware is just as much a curse as a blessing. I make no value judgments about this but merely stated a fact. It is not a win/lose kind of thing.

http://en.wikipedia.org/wiki/Compari...system_kernels This page has little to do with supported hardware in any quantitative metric.

You should be aware that it isn't that difficult to get a kernel or microkernel from any OS to be modified to run. MS has created one of a kind OS's for various reasons before and not released it to general public. Just because a kernel from any OS can be made to run on a cpu there are maybe a billion other components in various service levels out there.

I admire you optimism but I still believe it's unfounded.

Let me reverse the issue. Show me thousands of devices that ONLY run linux. Why you say? Because most are not created to run Linux.

OK. I won't mention Android Phones, Chromebooks, or SuperComputers or the soon to be released SteamBox..... ARM CPUs, PowerPC CPUS, XBox, or PS/2. Let me turn this around. What hardware or even class of hardware does Linux NOT support?

The original target platform for hardware is immaterial. So is it immaterial if they can be modified (well, immaterial other than legalities) to run what they don't out-of-the-box. The bottom line, for whatever reasons, is that others have never supported whole classes of hardware as well as dropped support for some they did. Despite your polite (but nevertheless condescending) dismissal as "optimistic" it is not optimism. I have no stake in this and don't understand why you seem to.

Linux, at least as of Kernel v3.x.x has not dropped support and has for a very long time sought to run on anything. The very same kernel downloaded from kernel.org that runs on my PC will run on a Mac, a PS/2, or SuperMUC (150,000 processing cores), a router, an ATM, a SmartTV, a phone or an embedded system in my car.

So let's bring it to a head. I challenge you to name one single kernel that can do all that, that can support all that hardware, other than Linux... and you still haven't answered the question of have you actually looked to see what is supported in a generic 3.x.x kernel or simply relying on out of date anecdotal experience? I build custom kernels so I do see what is supported. Anyone, including you Jefro, can look to see for yourself. It's not like it's hidden or secret... well, not in Linux anyway.

Its called a rant and everything in it was intended to convey my frustration with the idiots at Matrox and by idiots, I mean their management not their fine engineers or support staff. Even their sales staff are quite knowledgeable about technical issues.
A little more background to answer some of the questions posed:
1) We did not find (Mr dugan) that open-source drivers can support the multiple heads that our SCADA application requires, so we were forced to use proprietary drivers. We found that Matrox had just issued (at that time) a new driver for the M9148 for Linux in December of 2013.
2) For jefro among others: we very recently purchased the M9148s (these are $600 cards), on Matrox salesman recommendations, about a year ago (late 2013). We bought 6 cards, by the way, so we are talking about a significant amount of money. The message we get on driver installation looks like something a programmer inserted to check the X version and exit, without any knowledge of what the future might hold. In other words, it's arbitrary. And my experience with X backward compatibility tells me the driver has a good chance of working without that one "if" statement.
3) Thank you enorbet, there was significant research and reason (and a new Linux driver) to expect Matrox robustness and we have read of issues with all cards. Now we will be evaluating Nvidia and ATI.

If anyone else has multi-head experience under Linux, please weigh in. Thanks.

I've run multi-head configurations before, but not with Matrox hardware. I've always used multiple Nvidia cards, and FOR THE MOST PART, their proprietary drivers simply work. Even now, the free drivers seem to work much better (although I admit I haven't attempted to use a multi-head configuration with the open source nvidia drivers).

@paz9 - Hello again
I hope I wasn't overly harsh in my first response so maybe I should explain that it likely reflected my disappointment that Matrox apparently dropped the ball. It's almost 20 years ago and I still recall my first boot to desktop with a newly purchased Matrox Millennium which literally took my breath away for a moment because I was so "knocked out" by the depth of color, sharpness of edges and clarity of fonts. This was on OS/2, btw, as they had a driver even for it. When the G400 was released I thought "These boys are serious!" and then..... crickets. I don't know exactly what happened that they chose what seems like such a radical game change, but they did change. Maybe they took a gamble and lost and decided to pare down to a comfortable niche of the largest potential market, but that niche seems to exclude Linux.

I think you fortunately did see my main thrust that "betting with the house" is safest and currently that house is nVidia. FTR I have no affiliation with nVidia or any manufacturer for that matter. I just appreciate how much effort they have consistently put out for decades for alternative systems. I was a bit surprised that Linus scolded nVidia but the very fact that instead of stonewalling they stepped up should tell us that they pay attention, have lofty goals and are willing to spend what it takes to be Top Dog.

Do you actually need dual-head or just multi-monitor? I know that 3 and even 4 monitors on single ATi and nVidia cards in Linux is fairly common.

Thanks enorbet, and no I actually find your posts the most useful.
We need multiple independent heads (so "no" to Xinerama, etc.), up to 4 under Solaris, but we are looking to increase that under Linux, perhaps with multiple cards. Yes, I remember the old Matrox. The new Matrox is not bad, but they are trying to churn their customer base. The M9148 quad card is exceptional: 4 micro DP ports, recent driver updates. Now, in order to use RHEL 7, we have to scrap all 6 and buy C-series cards, which admittedly can support 6 heads, and we'd buy them if we could also still use the M-series too. Good thing we are in early stages of our port to Linux.
We have ordered an Nvidia card and an ATI and will be evaluating these for our desktop systems (we don't use laptops Timothy Miller). I will post back with results as we generate them.
I'm disappointed in all manufacturers in that none of them support SCADA-style 2-D multiple-head except in their high-end products. This means you have to pay for 3-D graphics and bit-coin-mining-capable GPUs that you really don't need. The SCADA market is quite large, as I told the Matrox folks. I thought they were in my corner. There are power grid operations and expanding mass-transit throughout the US & the world. These kinds of app require a lot of screen real-estate, but not much demand for throughput. Our app runs on 60 or so Unix desktop workstations communicating with fault-tolerate 5-9s servers and front-end PCs. We also have call for several screen wall displays which can use up another 4 cards per wall. We are expanding all the time. I'm busy as hell (oops, those damn censors and don't have time for the nonsense.
I appreciate all comments.

@paz9 Ah! Now the lights go on - Instrumentation. I have been interested in the field a very long time and even bought stock in a company developing Field Emission Displays way back in 1999 that won a contract for instrumentation on the Abrams Tank. Unfortunately it was the company that "tanked". I'm still certain this a huge emergent industry and often imagined wild scenarios of a single computer sending multiple applications' graphics data through some sort of networking interface to some kind of remote, embedded and networked display module. I recently have wondered if this is an idea buried within Apple's Thunderbolt project.