“But can it fold? “
This question confused me for years until I finally followed the links. It took me somewhere new: the Stanford University Folding@Home project. I poked around briefly, but moved on.
Later, after I joined the Icrontic community and had built my first quad-core system, I was encouraged to use said system for Folding. This time, I really delved into what it’s all about.
Folding@Home has taken off in many internet communities because it allows for friendly competition (via points assigned for each unit of work turned in) and benefits a great scientific cause. The largest increase in Folding performance has shown up recently with the new GPU2 clients for NVIDIA and ATI video cards. Previously the GPU client only allowed for ATI 1xxx series video cards, and the symmetric multiprocessor (SMP) client allowed for a single client to use multi-core processors which offered a huge productivity increase over running a single client per processor core.
The rise of the GPU
If you follow PC hardware, you’ve likely heard of the AMD spider rig, which allowed for the astounding realtime rendering of the first Spiderman movie in 30fps. This was where GPU power was really starting to rise beyond simple gaming and 3D rendering.
Then, NVIDIA announced that they were going to release a programmatic framework within their video drivers, CUDA, which would allow a folding client to harness the power of the GPU directly. This gave rise to the GPU2 client, the first time that both major video card manufacturers could run the GPU client, and it ran so very well. Folding@Home point values nearly doubled overnight.
Making the switch
Switching to GPU2 clients presented a challenge, because on Windows XP the GPU2 client requires 100% usage of one CPU core to feed the GPU information (Vista doesn’t require the entire core). With each GPU requiring a CPU core, many people either stopped CPU folding or had their CPU folding clients use only the remaining cores. Making sure that these two different clients didn’t compete with each other for resources was solved with a simple program called Process Lasso, which forces processes to use only the CPU cores you want it to.
Why run a GPU client instead of a CPU client? Anyone with an ATI card 2xxx series or NVidia card 8xxx series (or better) can run these clients, and are missing out on huge production potential if they’re not. The GPU client only uses one CPU to feed the GPU with the data it needs, so you force one processor core and one GPU core to full load. The production output by an Intel Core 2 Quad core running SMP is 2000-3000ppd (points per day), but it’s squashed easily by many of the video cards that are populating gamer rigs across the country and sitting around in custom PCs. For instance, a G92-core 8800GT can do 5,184 ppd at 675(core)/1728(shader)/972(memory) clock values. Currently, ATI cards do not produce as much as NVidia cards (of course, these are overclocked numbers, but honestly, if you’re Folding, I doubt there is anything stock about your computer except the color).
There are tables published at the Folding@Home forums that break down points per day by card, clock speeds, and even work unit.
Installing GPU2
The installer is fairly simple. It has only a few things to enter: team number (93 is Icrontic’s), your name (pick a username), some advanced tags, and a slider to select a priority level over other Folding clients. The application runs in the system tray and allows you change the settings while it is running (though you must restart the client to apply the changes). It places a shortcut on your desktop, which you can put in your Start->Programs-> Startup folder to have it start automatically when you login to Windows. The great thing about this client is that it is easy to set up multiple clients if you are running more than one video card, though they can not be in SLi or Crossfire mode.
To get started, download the GPU2 client and open the downloaded file to start installing; make sure to select ATI or NVidia depending on your video card manufacturer.
Click next.
Read the agreement (yeah sure), check I Agree, then click Next.
Click next again and the install will start.
Click Close and find the new shortcut on your desktop to run GPU2!
Once the program starts to run, a new icon will appear in your system tray. Right click that icon and choose Configuration from the menu. The first time a GPU2 client runs, this configuration window will display automatically.
Type in the user name you want to be known as while Folding. Add a team number (not name) to also have the points attributed to the team’s score; Use 93 to fold for Team Icrontic.
Go to the Advanced tab and make sure the Core Priority is set to “Slightly Higher” and the slider is all the way over. This gives GPU2 priority over other Folding clients.
There is a small trick to get two instances of this program running. You’ll need to generate a different folder for each instance; the easiest way is to install one instance and then copy the entire folder and rename it. If you name the folders after which card they are going to be used for, primary or secondary, it helps you keep track of pathways.
Click to enlarge
Here, I have created a second shortcut that points to the new folders I created for the second instance of GPU2.
The machine ID does not matter unless you are running multiple GPU2 clients or a GPU2 client along with a SMP client. If you are, simply change the machine ID to the next sequential number. For instance, if you’re lucky enough to have a quad core processor and two GPUs, you can set an SMP client to machine ID 1, a GPU client to 2, and a second GPU client to 3 and all will work together correctly.
Click to enlarge
Notice that the machine IDs are different and the bottom text box has what are called flags. The “–gpu #” flag is when more than one GPU2 client is running on a machine to tell the software which GPU to use; 0 is primary and it counts upwards.
Problems with your install? We’re here to help. Check our GPU2 folding thread and post your problem (just register first – it’s fast) along with as much information as you can gather, and our community will help you out.
Performance
The wonderful and truly intelligent thing about the GPU2 client is the built-in pause button that allows someone to turn it off quickly and resume just as fast. Since running the client impacts your video response, it’s useful, especially for gamers, to be able to pause and restart it on demand. Team Fortress 2 can be played on high settings with a G92 8800GT at stock settings while still Folding, but you can expect stutters. Watch out for websites that run very visual applications or have a large numbers of GIF images, because it can slow your computer for several minutes while it loads (since it’s also trying to deal with the GPU2 client).
No longer are Folding farms (lots of dedicated Folding computers) needed to rack up 5,000 points in a single day. Now, a single computer equipped with two NVIDIA video cards and a quad-core processor can produce 10,000 points per day. The numbers are staggering.
But remember, kids – this is more than showing that you can be the biggest and strongest little runt on the playground; this is about helping Stanford fight diseases and understand this intricate machine that is your body. Now is the time for all gamers and computer enthusiasts to show what their rigs are made of. Folding, Folding, Folding, get those computers a foldin’…
Articles RSS