"Useful" distributed project for stress testing?

[mmaestro]

As I've been playing with different OC settings for my system, of course I've been stress testing, mostly with Prime95, but leaving it on for hours on end burning electricity number crunching without doing anything useful seems incredibly wasteful to me. My experience with Prime is that if I have a bad OC, it'll bluescreen - I've not seen a single error thrown up yet - so I figured maybe stress testing with a more useful piece of software where it's actually doing some scientific research would be nice. Last night I downloaded BOINC, threw on Milkyway@Home (I chose that one as it's both multithreaded and will make use of the GPU), and let it run. Unfortunately, it's not maxing out my processor: I'm seeing only 90-95% utilization, and the difference between something like Prime and this in terms of heat and stress on the CPU seems obvious.
Are there any good distributed computing projects that'll actually max out my CPU (a 2600K) so I can feel like my stress testing is contributing to the greater good, or are things like Prime and Intel Burn Test really the only good stress testers out there?

 

[Edrick]

PrimeGrid will tax your system well, especially the GPU.

 

[IGemini]

My personal approach has been to load down the GPU with Folding@home (with one CPU core, at least on my setup, not sure how CUDA distributes the work), and load the rest of the CPU down with Rosetta@home. Prime95 is good for thermal testing but I'd usually find out if the OC was stable with this setup in under an hour.

 

[waffleironhead]

You can attach Boinc to primegrid.com and run any of the LLR tests. The Proth Prime Search (LLR) is fairly CPU intensive and will show an unstable overclock.

 

[totalnoob]

Folding@Home maxes quite well for me. I recommend you keep running it after your tests are done. It's for a great cause.

 

[Phynaz]

wcgrid

 

[AdamK47]

I run Folding@Home. 3x GPU clients on 3-way SLI GTX 580s overclocked with 11 threads of the CPU client on an overclocked 980x causes the lights on my street to dim.

 

[aigomorla]

F@H is the best..

Why? becuase it puts a ton of load.. and then another server error checks it with other results.

This means its the most solid proof if your system is stable.

VS. having the program run and pass on the same machine.. vs checking results with a pool of others.