Performance wise, the HD 6670 is not exactly a high end GPU. I have one myself and it is OK for most games, but then I don't expect to run my games at maximum graphics settings. If you expect high quality graphics with decent FPS, a 6850 would be the minimum. Or maybe a HD 7770.

Second, I can confirm some funny artifacts under Windows (windows switched to the background are not fully overwritten). This happens both with the HD 6670 @home and the HD 5770 @work, so I suspect a driver problem that is common to the 5xxx and the 6xxx series.

Right now, there is no perfect solution, only tradeoffs:
- If you have plenty of money and are willing to live with a closed source (but having a good rep) driver under Linux, the Nvidia GTX 670 looks good. But their midrange Kepler stuff is not released yet. The older Fermi "Thermi" models get clobbered on performance per watt by AMD.
- If you insist on open source drivers under Linux, you are stuck with AMD for serious graphics cards. But the open source AMD drivers under Linux suck at performance.
- For those who don't have big GPU performance expectations, Intel is becoming interesting with the HD 4000 integrated graphics. But it still gets clobbered by the HD 5570, see http://www.anandtech.com/show/5771/the-intel-ivy-bridge-core-i7-3770k-review. So even Intel's best integrated GPU still loses to a discrete lowish-end card ;-)

As NotSoHeavy3D wrote, there is the GT640 (Desktop, OEM). See http://www.anandtech.com/show/5784/nvidia-updates-geforce-600-oem-desktop-lineup-adds-gt-645-gt-640-gt-630.

Three versions under the same model number, and with significant differences in power consumption and performance. Good luck getting the one you wanted with your OEM PC. I guess the catalog will just say "Nvidia GT 640" and you'll get whatever the assembly guys have lying around at the moment ;-)

So I guess that's a reason not to buy OEM with a Nvidia GT 640 aboard.

Not the drivers, the security patches. If you go online without installing them first, it may become a race between malware infections and Autoupdate. A firewall might help, but installing the patches before you connect to the internet is even better.

This said, reinstalling the drivers and applications is something that WSUS Offline Update will not do for you. Using a recovery DVD from the OEM might just put the crapware back. So I think that part is the time-consuming part that justifies the $65 plus tax.

I think you have some good points, but there is more to it. It seems that some outsourcers have an attitude of "get the contract by bidding low, do the minimum work to fulfill the spec, then bill the customer extra for anything that was not spelled out".

A nice example is the following anecdote I heard from a semi-reliable source:
1) Company orders a custom application from an outsourcer. Specification says "must have a print button".
2) Application is delivered. Print button is there, but does nothing.
3) Reviewing the spec reveals it only says there must be a print button. There is no specification about what should be printed.

The tale as I heard it ended with 3), but does anyone bet against
4) Outsourcer makes a follow-up offer for actual printing???

Option 2. is about to collapse for some EU countries right now (Greece in the first place, but Spain and Italy might not be too far behind). Because at some point it becomes obvious that you can't pay them back, and no one will want to loan you money anymore.

Now I have no idea where China stands in that regard, as I don't follow the news about their economy much (if the numbers are public in the first place). But it is something to keep in mind when borrowing lots of money...

Good point, and the "break it up, and explain it exceptionally well" translates to detailed and clear specifications.

In my experience, not every organization is able to create those. Actually, my experience as a developer is more along the lines of being given a vague goal, producing a prototype, and then people would play with said prototype and start producing change requests. Which tends to developing the project piecemeal and with plenty of feature creep.

If your company is capable of writing good specifications, outsourcing may work for you. If it is of the "vague goal" persuation as described above, stay far, far away from outsourcing ;-)

While I've not tried it myself, it is widely reported that Vista SP1 is, in fact, much better that Vista at first release. There are your patches. BTW, the "new kernel" seems to be a minor upgrade, unless Microsoft has wildly exotic ideas about version numbering:

In my world, going from version 6.0 build 6000 to version 6.0 build 6001 might be a bugfix release, but major changes would get at least a version 6.1.

And under the hood, Windows 7 differs very little from Vista. One could say it is Vista SP3 with a new look and slighly relaxed UAC. I tend to believe this is true and the release of Windows 7 was mostly a clever marketing ploy to get rid of the tainted name "Vista".

By the way, this is not meant as an apology for Microsoft. While I don't buy every rumour about collusion with big entertainment, you might have noticed that I do think they are guilty of sloppy software engineering. One particular embarrassing example:
In Vista pre-SP1, various sources reported much diminished network throughput while playing back multimedia. See Mark Russinovich's blog: http://blogs.technet.com/b/markrussinovich/archive/2007/08/27/1833290.aspx.
To me, that does not smell of conspiracies, but of incompetence in project management. Obviously, Microsoft with all its ressources was not able to develop a mature new OS within the six years between XP and Vista. It took another year until SP 1 appeared and fixed the worst bugs. In other words, Vista was both "long overdue" and still rushed in terms of being not mature at release.

Wrong for two reasons:

1) Your latest version would still be worth more $$$ than the five year old one which can be freely copied => not all profit is lost. This said, I concede that 5 years are a bit too short. Maybe ten years from public release?

2) In many hardware projects, there is a need for specialized software that is
      a) delivered in five months, not copied from elsewhere 5 years from now when it becomes legal.
      b) may not exist elsewhere at all in the form that is needed.
This leaves a pretty large field in which software developers could still earn money.

5 years after release of the version in question, like with all Closed Source IP. And I think it would hurt GPL'ed software a lot less than Closed Source. Because unless the original project has stagnated badly, you could get a much more advanced version from the usual sources, usually for free. An example:

Assume that forking some five years old Linux version and selling it as Closed Source was allowed. Try doing that. Good luck competing with the latest version of Ubuntu ;-)

Sources? I've googled for that "kernel level encryption" or any difference in DRM from Vista to Windows 7. Did not find anything that looked loke a reliable report.

BTW, my point was that Vista was just released before it was ready, with some stupid hacks in place that killed performance. Not because of DRM, but because the release was rushed.