For the last two years I’ve made extensive use of the nVidia Quadro FX 5800 GPU (Graphics Processing Unit) to crunch through trillions of calculations required in my antenna simulation and design work. I’m now in the process of upgrading one of my workstation computers accept an even more powerful GPU solution; the nVidia Tesla C2070. The compute power of these devices is simply mind-blowing. For example, the C2070 has 448 CUDA cores, 6 GB of RAM and can push through in excess of 1 Tera FLOP (Floating Point Operation) per second. The Quadro is an older model with only 240 cores and 4 GB of RAM but it’s still 10 to 80 times faster than my dual quad-core Intel Xeon machine on the antenna calculations! Simply amazing!
More amazing is that nVidia markets even higher performance GPU’s to gamers. Cards such as the nVidia GeForce GTX 590 are even faster than Tesla and Quadro for gaming applications. The difference is that Tesla and Quadro are aimed at scientists and engineers who demand ultra reliability, super efficient double precision calculations, and larger RAM memories to hold simulation data. The latest Tesla cards also use ECC (Error Correction Code) memory to help ensure that simulation results are not polluted by the occasional random memory error. None of these factors matter for gamers so the GeForce cards are substantially cheaper and sold to the masses at retail outlets such as Best Buy.
Now, what does any of this have to do with the security of your data? Well, there are several things to consider. Each somewhat tangentially affecting either your security or your computer’s reliability.
First, as I mentioned in this post, hackers are using these video card / GPU solutions to hack your passwords using brute force. Next time you’re walking down the aisle at Best Buy or browsing Tiger Direct perhaps you’ll take notice and see how cheap this incredible computing power has become and take to heart my message to use really good passwords!
Second, it occurred to me that many of you may want to consider buying one of these new video cards to improve your gaming experience or the responsiveness of your CAD programs at work. Who knows, some of you may even want to start experimenting with hacking passwords? Whatever your end use, I wanted to remind you that adding such a potent video card / GPU may require you to upgrade your power supply as well as your motherboard and cooling capabilities. As you can imagine, the faster we compute, the more power we draw. The newest GPU cards have substantial power requirements. Consider that the older Quadro FX 5800 is specified at 189 Watts, the Tesla C2070 at 235 Watts, and the GTX 590 an astonishing 365 Watts!
Clearly, without enough power your computer and new GPU won’t do anything very fast so you’ll want to run a power budget to decide if your current power supply will work or if you’ll need a new one. You can do the budget by hand or you can use one of the on-line calculators such as this one:
which has all the common CPU’s, video cards, and accessories in a database so you don’t have to look up specs on every device in your computer. Once you know you’ve got enough power, you should confirm that the power supply can provide sufficient current and has the proper connectors to support the graphics card. You’ll also need to determine if your motherboard has the proper slots to accept the card. Most high-end video / GPU cards need a PCIe X16 slot. If you don’t have one you’re probably looking at a motherboard / CPU / memory upgrade too which probably signals the need for a whole new computer.
Lastly, even if you’re not in the market for a new graphics cards, you are well advised to check the power budget of your current computer to make sure your power supply is adequate to handle whatever devices you’ve added to it since you bought it. An overloaded power supply is a potential failure point that can impact your data so I suggest you take a moment now to check it out!
Update February 23, 2012:
I’ve added an article on building a GPU Workstation that my also be of interest.