1. Build a beowulf cluster of hot grit pouring Portmans.
2. ??
3. Profit!

This article isn't about how much RAM Firefox needs when *running*, it's talking about how much RAM Microsoft Visual Studio uses when *compiling* (actually, linking) Firefox.

We aren't talking about Microsoft Certified Whatever People here.. the Civil Engineering PE licensing process is actually pretty good at testing for knowledge and expertise in the subject area. Civil Engineering is a relatively straightforward discipline to test knowledge on with agreed upon best practices, etc, PE is a professional license which requires a degree in the field, years of practical experience and then a pretty extensive exam process.

Did I mention it's a license? You don't just lose your piece of paper if you fuck up, you are held personally liable. You can go to jail for it.

Western District of Wisconsin's court is in Madison (right down town) and there's plenty of tech here - if you count biotech.

Still don't know why they'd file in Wisconsin though. Motorola is in Chicago.. perhaps their law firm has an office up here?

That's just cheating.

Please, please, good people, I am in haste. Who lives in that castle?

Nope. It is now official. Netcraft has confirmed: Linux is dying.

Ah, ok. In the future if you're in need of an NLE that can do S3D natively, Sony Vegas 10 was just announced and has a full S3D workflow.

What are you using to produce this?

The 3 ghz clock applies only to parts of the processor die. And not the entire thing at that.

The PCIe clock rate is by default 100mhz. Engineering motherboard traces to do much more than that is a real pain because noise becomes a problem. Incidentally, dealing with noise is a large part of why PCIe uses multiple serial connections rather than one large parallel connection to increase bandwidth. Still, there's a physical limit to how many traces you can put on the motherboard.

PCIe is pretty damn fast, but there's a lot of overhead and lag to getting data from main memory out to the card and back. Once you set up the transfer though, it's fast. The problem for a lot of software developers is redesigning your software to deal with that lag.

Very true. There are other limitations to GPUs as well. They don't handle branching well at all, some scatter/gather operations are slow, etc.

But don't get me wrong; you'd have to pry my dual ATI (AMD) Radeon 5870's from my cold, dead fingers. And hopefully soon I'll get my hands on a GTX 485... :)

GPUs are highly parallel processors, but most of our computing algorithms were developed for fast single core processors. As we figure out how to implement new solutions to old problems to take advantage of these highly parallel processors, you'll continue to see stories like this one. But, there's a limit to how good they can be at certain types or problems. Read up on Amdahl's law.

Basically, traditional x86 processors are good at lots of stuff. Modern GPUs are great at a few things.

While we're beating dead horses, this is the type of post that would have made an excellent story on Kuro5hin!

Yea, because NOBODY has tried that before...