Folding@Home

Creator: Stanford University

Folding@home is possibly the most successful GPGPU application at the moment, with ATI pioneering GPU-based folding years ago with its Radeon X1000 range of GPUs. Since then, folding has spread to the point where Nvidia has supported folding on all GPUs since its GeForce 8-series.

Folding is a client-based distributed computing application that simulates protein folding to aid our understanding of complex diseases such as Alzheimer's, Mad Cow Disease (BSE), CJD, ALS, Huntington's, Parkinson's disease, and many Cancers and cancer-related syndromes. To add some interest to the furthering of scientific and medical research doesn't hold any interest for you, Stanford also has team- and user-based leaderboards, so folding is also a competition.

To fold on your GPU, visit the High Performance Client page and download the latest version. The client downloads work units (aka 'projects') from Stanford, and each work unit is worth a certain amount of points.

You can monitor your folding progress via the FahMon tool, which will calculate how quickly your card can get through the current work unit and extrapolates a points per day (ppd) figure. You can join Custom PC's folding team (currently ranked fifth in the world) by using the team number 35947. Your contribution is valued.

Folding@home performance

Nvidia is keen to push folding performance, as the company is pushing the GPGPU capabilities of its cards and it runs folding via CUDA. As such, Nvidia cards tend to have roughly double the performance of ATI cards. The GTX 275 received the massive p5904 project worth 1,888 points and we couldn't get it to receive smaller work unit no matter how many times we re-installed the folding client. As bigger projects take ages to complete, they usually yield less ppd than smaller work units that are quicker to complete.

However, ppd output is also dependent on GPU speed and resources, so the higher clocks of the GTX 285 also explain why it folds a bit better than the GTX 275. The fact that it received project p5900 that was worth 472 points probably also helped. The GTX 260-216 worked on project p5769 which was worth 353 points. However, as this card has fewer shader units, it's no surprise that it lags behind the other two GeForce cards.

Meanwhile, the ATI cards received more sensible work units; the HD 4870 1GB received project p5746 which was worth 384 points while the HD 4890 received project p4750 which was worth 477 points. The extra clock speed of the HD 4890 gives it roughly 600 ppd over the HD 4870. We should mention that the HD 4890 makes an horrendous racket when folding though, which isn't great if you tend to run folding in the background as you work or browse the web. The GTX 275, like other GT 200 cards, isn't quiet but isn't especially loud either.

Folding@home power consumption

We tested the power consumption with a Watts Up? Pro power meter, using the device to record the total system power consumption at the wall socket, while we set the card folding. Using the data recorded by the meter we could determine the peak output.

As both ATI and Nvidia test benches are now identical Core i7 systems, we can accurately determine the apples to apples difference of what power both PCs take to run.


The HD 4890 is very power hungry when folding, but the other cards are of an equal level. This means that the HD 4890 is a very inefficient folder, while the GTX 285 and the GTX 260-216 are fairly efficient.