at what point does supporting an older operating system like XP cost you more money then it would to spend the money upgrading to a later version of Windows?
Well... when expensive capital equipment runs XP, that point is essentially "never."
For example: Old versions of LabVIEW run on XP, but not 7. A license for one desktop runs $10-50k, depending on options. Lots of sciencey equipment requires software that only runs on XP. Want to upgrade to 7? Ok, here, buy a new $100k laser / $20k UV-vis-IR spectrometer / $10k multi-channel scaler / $10k oscilloscope / whatever. Another user mentioned his CNC machine (those aren't cheap).
I mean, you can do that, but $10k buys you a lot of extended support (which is curiously unavailable to everyone who isn't a big business). I'm sure most university labs would jump at the chance to pay a few hundred dollars per machine to support instruments incompatible with 7.
Seriously. A few years ago (2010ish) I was roped into helping out with a microprobe array that would have cost about $150k to replace, whose controller software would only run on Windows 95... not OSR2, meaning that USB wouldn't even work (we used ZIP disks to move sequence files on and off of it). The fellow in charge of the thing would grab systems being surplussed from elsewhere in the department that might run Win95 and clone to their drives... when I left he had 4-5 spares in a closet for the day the current controller died.
EDIT: And we did attempt to virtualize it, but at least at the time were unable to do so successfully (timing issues, I think).
It's not. But from what I've heard, it's much less than that perseat.
I'm saying that the PIs of most university labs would gladly pay several times the retail price of XP just to be able to continue having support for multi-thousand-dollar instruments.
But I think this nitpicking over a relatively minor point--the important part is that upgrading from XP to 7 costs tens or hundreds of thousands of dollars per computer for some people.
Edit:
In case you were referring to the sentence "$10k buys a lot of...," I can see how my intent was unclear (and I was sort of talking about both small and large labs at the same time). My understanding was that there is a minimum buy-in for extended support (something like $100k per year), but that such a contract would cover hundreds or thousands of computers. That's on the order of $100 to $1k per PC.
A medium-sized university lab might have dozens of instruments costing on the order of $10k to $100k. Now, paying $100k to support $100k worth of instruments for a year doesn't make too much sense, but paying a few hundred dollars per $10k instrument does. Even at $1k per PC per year, avoiding an instrument purchase buys a least a decade of extended support. It's just a question of getting a critical mass (or, by convincing Microsoft to open up the program to smaller organizations).
Comments
Well... when expensive capital equipment runs XP, that point is essentially "never."
For example: Old versions of LabVIEW run on XP, but not 7. A license for one desktop runs $10-50k, depending on options. Lots of sciencey equipment requires software that only runs on XP. Want to upgrade to 7? Ok, here, buy a new $100k laser / $20k UV-vis-IR spectrometer / $10k multi-channel scaler / $10k oscilloscope / whatever. Another user mentioned his CNC machine (those aren't cheap).
I mean, you can do that, but $10k buys you a lot of extended support (which is curiously unavailable to everyone who isn't a big business). I'm sure most university labs would jump at the chance to pay a few hundred dollars per machine to support instruments incompatible with 7.
Seriously. A few years ago (2010ish) I was roped into helping out with a microprobe array that would have cost about $150k to replace, whose controller software would only run on Windows 95... not OSR2, meaning that USB wouldn't even work (we used ZIP disks to move sequence files on and off of it). The fellow in charge of the thing would grab systems being surplussed from elsewhere in the department that might run Win95 and clone to their drives... when I left he had 4-5 spares in a closet for the day the current controller died.
EDIT: And we did attempt to virtualize it, but at least at the time were unable to do so successfully (timing issues, I think).
I don't think custom support is that cheap. There is typically a floor of at least $100k for Custom Support Standard.
It's not. But from what I've heard, it's much less than that per seat.
I'm saying that the PIs of most university labs would gladly pay several times the retail price of XP just to be able to continue having support for multi-thousand-dollar instruments.
But I think this nitpicking over a relatively minor point--the important part is that upgrading from XP to 7 costs tens or hundreds of thousands of dollars per computer for some people.
Edit: In case you were referring to the sentence "$10k buys a lot of...," I can see how my intent was unclear (and I was sort of talking about both small and large labs at the same time). My understanding was that there is a minimum buy-in for extended support (something like $100k per year), but that such a contract would cover hundreds or thousands of computers. That's on the order of $100 to $1k per PC.
A medium-sized university lab might have dozens of instruments costing on the order of $10k to $100k. Now, paying $100k to support $100k worth of instruments for a year doesn't make too much sense, but paying a few hundred dollars per $10k instrument does. Even at $1k per PC per year, avoiding an instrument purchase buys a least a decade of extended support. It's just a question of getting a critical mass (or, by convincing Microsoft to open up the program to smaller organizations).