Folding@home is a project started by the Pande group at Stanford University. Intended initially for helping accelerate protein folding simulations for biomedical research, the project today spans numerous laboratories from around the world with the ultimate goal of finding cures for diseases such as cancer, Alzheimer's, Parkinson's, mad cow and diabetes. As long as your notebook meets a few key requirements, it's a relatively simple affair to set up Folding@home, letting you join the fight against these debilitating and deadly illnesses.
BACKGROUND:
My first introduction to F@H was around 2004 when I had just purchased my HP zd7000 notebook. At that time, F@H ran with just a console client, although a GUI version was also available. Performance was sluggish at best, despite having the fastest single-core HT processor, taking often days to complete a work unit. Back then, running F@H was entirely the realm of desktops (and the more boxes you had the better). Considering also the load it placed on my system, it was fast becoming apparent that I would not be able to run F@H 24/7, requiring much beefier hardware.
 |
| Have you joined the F@H race? |
Should you make the jump and begin using your own notebook for F@H? There are a few key hardware considerations to keep in mind before you commit to the task, including your notebook's processing power, cooling capability and acoustics.
USING THE RIGHT CPU:
F@H is all about wringing the most performance out of your hardware, and is great at exploiting high-end processors, especially when using the SMP client. The minimum you should be looking at if thinking about folding on your notebook is a quad core CPU. 4C/8T mobile Sandy Bridge Core i7 models perform great here, as do newer Ivy Bridge variants. Older machines should also work with a QX9300 or similar, but keep in mind the lower you go with core count and frequency, the slower you will fold and complete work units. It's a very simple equation - the faster your CPU is, the better it will fold.
For those lucky few rocking a 6-core desktop processor inside their notebook - I envy you!
As for how (poorly) the low-end performs, I've been able to get modest results running a Core i3-2367M. ULV Sandy Bridge should be the absolute minimum if you are inclined to tackle F@H with a 2C/4T processor. Forget anything lower, as a 2C/4T Atom N570 won't be able to complete a work unit in the allotted time, and consume so much CPU cycles that it will cripple the UI.
AMD processors will also work, but again performance is dependent on cores/frequency - the more cores and faster frequency you have, the better your performance will be. Quad-cores are desirable, dual-cores should be avoided.
USING THE RIGHT GPU:
Got an nVidia GTX8800M GPU or newer inside? You'll be able to run the F@H CUDA client (drivers permitting) and fold with your graphics card in addition to your CPU. The CUDA client runs different work units than the SMP client, and depending on your hardware, can also complete work units a lot faster than your CPU. The ability to fold with CUDA is a big argument in favor of notebooks with nVidia hardware, even older models such as my 9800M GTS.
The new mobile Kepler graphics cards are the only exception, however, as Stanford hasn't quite optimized F@H to maximize folding performance with the latest GPU architecture. You can still fold, but obtaining an optimized client may take a while, and your scores (PPD) will be mediocre.
Adding confusion is nVidia's naming scheme that puts some last-gen mobile Fermi cards under the new 6xxM model lineup. GT620M, GT630M, GT635M, GTX670M and GTX675M are all Fermi-based, while GT640M, GT650M, GTX660M and GTX680M use the improved Kepler architecture. Picking one right now will mean either great F@H performance using the Fermi-client or great gaming performance with Kepler graphics.
The choice? I'd say go with Kepler - the improvements to gaming alone compared to Fermi will be tangible across many 3D titles, while it's likely just a matter of when (and not if) Stanford throws out a Kepler-specific beta client. Choosing next-gen is always the better buy because it extends support and prolongs the life of your investment.
Got ATI Mobility on board? Stanford's version 7 client recognizes Mobility Radeon cards, so anything better than a HD4000 series should work. The big drawback is the inability to run workloads designed for the Fermi client, meaning your performance relative to nVidia hardware will be less, producing less PPD.
Don't have a dedicated GPU inside your notebook? Itching to use HD3000 or HD4000 from your Intel Sandy Bridge or Ivy Bridge processor? Sorry - no support! Other than the benefits Quick Sync provides for transcoding work and HD video playback, you won't be able to use Intel's integrated graphics for doing GPU folding. Certainly not for the foreseeable future.
WATCH THOSE TEMPERATURES:
If you haven't figured out by now that F@H is a killer when it comes to stressing your hardware, then your notebook's temperatures will surely tell you so. Running your GPU and CPU at 100% for extended periods will raise temperatures to levels you don't see during normal use, or even while gaming/transcoding. Assuming that you have the hardware to run F@H, it's imperative that your notebook's fan(s) provide adequate cooling.
 |
| Keeping tabs on your temps is vital |
 |
| A must-have item for notebook folders |
For those notebooks with fans and heatsinks that simply do not provide adequate heat extraction, your safest choice is to not run F@H. Many newer ultra-thin models fall into this category, as the small/thin profile of the fan blades fails to move enough air, or it's diminutive heatsink is incapable of moving the maximum thermal load (TDP) from the CPU over an extended period. In a similar fashion, notebooks that lack underside cooling (no bottom fan vents) will also stress under load, as they will be unable to benefit from the direct airflow of a notebook cooler placed underneath.
A good first test is to run your CPU and GPU clients together and gauge the results after the first 30 minutes. If temperatures haven't hit critical levels, keep going and look again after running for another 30 minutes. If temps after an hour are no higher than 2-3 degrees of your first measurement, all should be OK. Now keep running for a full 24 hours. If your system hasn't shut down, shown any graphical glitches (usually caused by running F@H on a GPU that is overclocked or on bad drivers) and is still within comfortable temps (fans quiet), I would say your system is stable enough for running F@H.
One thing I would advise against doing is running F@H with the lid of your notebook closed - keep your notebook open. The keyboard and palmrest actually help dissipate surface heat, and closing the lid will prevent the system from staying as cool as it should. Newer chicklet-style keyboards in particular use the gap between keys to help ventilate the system, either by letting the CPU fan pull in airflow from above or having heat expel via passive venting.
CONSIDERATIONS ON NOISE:
For those of us who, by virtue, sleep with our computers, noise will become a major concern as well. This is especially true when running F@H as it will want to keep fans running at full speed in order to deal with the heat. Add to that notebooks that come with an already obnoxiously noisy fan to begin with, and you begin to "hear the problem".
Fortunately, adding a notebook cooler underneath can help a lot, as the additional cooling can in fact lower fan noise, which is regulated by temperature. My X305 runs decidedly quieter because the additional airflow from underneath keeps the CPU cooler, and thus prevents the fan from spinning up. And most, if not all, aftermarket notebook coolers run far quieter than a notebook would, making the combined acoustic load actually lower than running the notebook alone.
Then there's the issue of those fans that have an annoying high-pitched whine when running at full speed. There's not a lot that you can do here, so if adding a notebook cooler doesn't bring temperatures down (and quiet your fan), then you'll either need to put that notebook some place far from your bed, or enjoy sleeping with earplugs.
Finally, if all else fails, you can try lowering your thermostat! Ambient temperatures will directly impact how hot your gear runs, and a 3-5C degree drop in your room's temperature can be a big-of-enough difference to make your notebook run quieter than in more warmer surroundings. You may even save a buck or two on energy costs.
RUNNING OVERNIGHT VS. 24/7:
We've determined we've got the hardware to run F@H, our temperatures are in check and we're happy with the noise coming from our system. The question now is how do we run F@H - do we just keep it on overnight or run it 24/7?
Obviously, running 24/7 will net you with the most points and get you up in the rankings fastest. It will also put the heaviest load on your hardware, meaning fans spinning constantly with no breaks in heat output. Of course, how long your system can take it is entirely another question, but such use will require more frequent maintenance, and possibly lead to failures in hardware due to 100% up-time earlier than expected.
Running overnight allows you to retain use during the daytime for multitasking and other CPU/GPU intensive workloads. It also allows the system to take a break from a constantly high thermal load, prolonging fan bearings and making dust cleaning less frequent. You even have the option to turn your unit off and let it cool down completely.
How you ultimately choose to run F@H will depend mostly on what you want to do with your notebook. As a desktop replacement, I can see F@H running 24/7 for many enthusiasts that have high-end hardware. Casual users may feel happier keeping their machine on 24/7 but only running F@H overnight, making use of a notebook cooler. Likewise folks who work on-the-go may find that using F@H overnight while they plug-in and charge their battery for the next day is a good fit.
For users that fold on both CPU and GPU, you can choose to run both or just one as well depending on your needs. Leaving my machine on overnight to work through an encoding queue, for example, can still allow me to use the GPU for folding. I can even get away with 3D gaming and still run F@H on my CPU, and actually have lower temperatures than folding flat out. It all depends on your hardware's capabilities, and how fast you want to rack up points.
DEALING WITH FAN FAILURES:
I can honestly admit that the single biggest negative I have had to contend with over my time using notebooks deals around maintenance, and with that fan replacements. Unlike desktops that have adopted standard fan sizes, notebooks entirely use proprietary fans to fit the very tight internal spacing, making finding any replacement very difficult. Also, unlike desktops, getting access to notebook fans often requires complete disassembly, a task that can take multiple hours and requires step-by-step disassembly instructions. My X305 falls into this category, and making it more painful is the use of a 3rd-party brand/model of fan that has at best only a 12 month lifespan. That means I go through a complete notebook disassembly procedure every year in order to replace a dying fan - no fun!
Many other notebooks, fortunately, have a less-complicated chassis, allowing you to get to the fan just by removing the bottom panel - such is my HP zd7000 and 210 Mini. If you are using your notebook to fold, it's wise to have a spare fan on hand in case your machine does shut down, allowing you to swap out and not lose complete functionality of your notebook. What's the best indication that your fan is about to go out? CPU temperatures become higher that normal under load, you hear more pronounced fan noise, or you come home to a machine that has shut down due to a thermal hardware trip. Early detection is key here, since you want to avoid damaging your processor due to extreme heat or having to reapply thermal paste that's become hard-baked, further complicating the situation.
CONCLUSION:
Folding@home is a great way to take advantage of the idle time your notebook offers, so long as your hardware meets the requirements and can handle the load. Knowing that you're helping fight humanity's deadliest diseases can bring a strong sense of empowerment. This is especially the case for dedicated folders who have lost a loved one or know someone who is impacted in some way/shape/form by illness. The competitive nature of F@H further spurs participation thanks to the ability to generate points for completed work units. Now that it is no longer the exclusive realm of desktops, F@H can be adopted by notebook users ready and willing to join in the crusade for a better life.
Even if you are not in the race for points/rank, participation alone can be a great contribution, and as part of a team, genuinely appreciated. In my case, I'll never reach the top 100 or even top 500 - my current notebook hardware simply doesn't cut it. But I know that if I stick around long enough, I'll eventually make it into the top 1000, and that's good enough of an achievement for me. Anything better will be a milestone I can certainly enjoy owning, and feel good knowing my effort has helped someone, somewhere.
No comments:
Post a Comment