Newbie looking into Folding@home

[Hi-Fi Man]

I've got some spare hardware lying around that I want to put to good use, I just don't know if it's worth it. I've got the following hardware on hand,

Phenom II X2 545
4GB DDR2
ASRock AM2+ nForce 730a mobo with PCIe 2.0
2U rack case with support for half height cards

Yes, the hardware is old as crap but I don't pay for electricity and don't mind running it 24/7. With that said does it make any sense? Or would a Phenom II X4/6 make it worth it?

What about GPU folding on this? Any decent half height cards worth using with this?

Sorry if this seems like an obvious question but I've never folded before and would like to get into without spending too much.

 

[biodoc]

I wouldn't bother folding on the 2 CPUs ( very low points per day (PPD) if you finish the Work Units on time) but you may want to consider buying a half-height GTX 1050 like this one.

According to this database, that GPU may produce 129K PPD. There are also low profile GTX 1050Ti available for nearly twice the cost but the PPD may go up to 193K PPD. Hopefully the cooling is good enough on these little cards for 24/7 folding.

 

[crashtech]

Folding is all about GPU now. The best half-height card for Folding will probably be a 1050ti. If you could mod the case to take a right-angle PCI-e adapter, then you could run a GPU like a 1060, which will give over twice the ppd (points per day) of a 1050ti.

Biodoc beat me to it.

 

[StefanR5R]

Right, what @biodoc and @crashtech said. Folding@Home really favors GPU computing. In terms of pure performance, as well as performance per Watt, very large = expensive 14 nm CPUs can just barely hang in with middle-sized 28 nm GPUs.

Though apart from using the Phenom as a GPU feeder, you could also look through the sticky DC projects list for projects which are not able to run on GPUs. If you are particularly looking for medical/ bio-science projects, then Rosetta@Home and the current sub-projects of World Community Grid are such projects whose applications rely on CPUs only. Of course, more cores or/ and newer cores means proportionally more results per day in pretty much all of those projects. (But if this CPU sits there unused, then why not put it to the task.)

 

[crashtech]

I don't know if there is a list of CPU projects that can use AVX or AVX2, avoiding those with older CPUs would be best, I'd think.

 

[StefanR5R]

Most prominent AVX users are PrimeGrid's LLR based subprojects of course. Optimized SETI@Home CPU applications make some use of AVX too. Asteroids@Home tries to use vector extensions too, but doesn't seem to get much out of it. Furthermore, a few less prominent math projects use AVX, e.g. SRBase.

 

[ao_ika_red]

Most prominent AVX users are PrimeGrid's LLR based subprojects of course. Optimized SETI@Home CPU applications make some use of AVX too. Asteroids@Home tries to use vector extensions too, but doesn't seem to get much out of it. Furthermore, a few less prominent math projects use AVX, e.g. SRBase.

TN-Grid has multiple default apps, one of which is using AVX and also Rakesearch has a very extensive optimized apps (SSE2, AVX, AVX2, and AVX512).

edit: answer to original question: As explained by all above me, you better use a GPU to fold (Maxwell and Pascal seem very good at it) because there's significant point difference between CPU and GPU folding. Even 750ti or gt1030 will do better than 12-threaded i7.

I've found an old thread about CPU PPD but it can be used as a very rough PPD estimation.
https://www.w7forums.com/threads/ppd-database.7947/ (look at post #3)

 

[Ken g6]

Since you're just getting started with F@H, you might consider doing your initial setup with the foldingcoin.net browser. You'll get a weird username, but you can earn folding coin with it and you can still select team 198.

 

[StefanR5R]

Before you look into foldingcoin, consider the possibility that cryptocoins which can be converted into cryptocoins and then into cryptocoins are just a pyramid scheme. (Somebody makes money out of those who lose money.)

 

[Ken g6]

Before you look into foldingcoin, consider the possibility that cryptocoins which can be converted into cryptocoins and then into cryptocoins are just a pyramid scheme. (Somebody makes money out of those who lose money.)

Possible. But you don't lose any money by setting up to mine FoldingCoin. You may not gain any money either if nobody wants to buy FoldingCoin, but it's only those who buy FoldingCoin who might lose money.

 

[crashtech]

It seems clear that Stefan finds the entire concept morally questionable, but I hope he won't be upset if some on the TeAm need a little monetary help to support their DC habit. I know I have been tempted, but there is something to be said about keeping the hobby purely altruistic.

 

[StefanR5R]

Who am I to judge in which ways people make (or try to make) their money? In fact, it doesn't concern me, and of course shouldn't.

However, when I see cryptocoins advertised to others here in the DC forum, I will feel compelled to post some fineprint beneath those advertisements, if they come without it. (Maybe I'll be able to resist to do so, as it may become boring really fast.)

 

[Hi-Fi Man]

Never mind cryptocurrency... I've had my fill of that.

I would definitely be interested in the CPU only medical folding projects if I can't manage to get a hold of a decent card. Speaking of which, I see that you guys mention nVIDIA cards but what about AMD cards? They have very good compute performance but how does that translate into folding performance?

I've also seen a few half height Radeon HD 7750s (DDR3) out there and was wondering how they compare to a GeForce GT 1030. Everything else seems insanely expensive due to miners right now so I would prefer to get a deal on a used card if possible.

 

[ao_ika_red]

I would definitely be interested in the CPU only medical folding projects if I can't manage to get a hold of a decent card.

Well, you should try BOINC-based World Community Grid. This is purely powered by CPU apps (still no GPU apps last time I used it in Dec 2017) and very diverse research fields like OpenZika, SmashChildhoodCancer, OutsmartEbolaTogether, etc (look at "My Projects" once you logged in there).

I see that you guys mention nVIDIA cards but what about AMD cards? They have very good compute performance but how does that translate into folding performance?

AMD cards indeed have very good computing capability but if the apps was already optimised for CUDA environment and preferred in single precision computing, it usually didn't work very well with AMD GPU, like in Folding and Seti@home for example. But AMD cards will excel in other DC projects that need double precision computing, like Einstein@Home, Enigma@Home, Milkyway@Home, and PrimeGrid's Genefer21 app to name a few.

I've also seen a few half height Radeon HD 7750s (DDR3) out there and was wondering how they compare to a GeForce GT 1030.

GT1030 should be as capable as my old R7 360 (which is the cut-down version of HD7790) about 60k+ PPD and based on database linked by biodoc, a GDDR5 7750 will only have 15k+ PPD, so very low.