Multiple reviewers, including myself, took note of problems with OCZ hardware when we covered the company. Tech Reports review of the vector 150 explicitly called out these problems,.

Ars covers a different spread of topics then it used to when I wrote there, that's true. But it's still an excellent site. The coverage mix has shifted, the quality of that mixture (in my personal opinion), has not.

Your tone implies you think differently, which is fine. It's still on my personal short list.

Full disclosure up front: I currently write for ExtremeTech and Hot Hardware. In the past I've written for Ars Technica (2007 - 2009) and briefly Tech Report (2H 2005). Before that, I wrote for a now-defunct site going back to 2001.

Obviously I could be biased and plug the sites I write for. I write for them for a reason, after all. But since no one is going to buy me telling you to read my own work, here's where I go, personally:

For in-depth, excellent analysis (in alphabetical order)

Anandtech (Anandtech.com)
Ars Technica (Arstechnica.com)
Tech Report (techreport.com)

For ultra low-level analysis:
Real World Tech (www.realworldtech.com)
Agner Fog's CPU blog (www.agner.org)
Lost Circuits (www.lostcircuits.com)

All three of these resources update only occasionally. But the information is second to none.

For spot-checking or specific issues:

TechSpot.com does great CPU/GPU scaling articles. LaptopMag or NotebookCheck are great for their particular areas. CPU-World has good general database information, VR-Zone often has interesting scoops, as does wccftech -- if you're willing to filter out a lot of rumor / speculation from the latter. Tom's Hardware has useful dynamic databases for product performance. So does Anandtech.

Don't be afraid to read a review on a site you haven't heard of, or with a layout from 1999. While established names and high-quality writers tend to go together, they are neither exclusively matched nor guaranteed. A good reviewer will document issues, give a thorough discussion of the topic, and won't come off sounding like a marketing employee.

Can't Mathematica be told to stick to an 80-bit precision output? If you can specify that in software, it shouldn't matter what code the underlying platform runs on.

Yes, absolutely an attempt to drum up false news about Tesla.

Alternately, an accurate report of new guidelines released by the NHTSA YESTERDAY. As opposed to the claims and NHTSA unhappiness over the summer.