My Computer

[relion64]

I have a pretty decent gaming computer, not new by and standards,(luckily its not that outdated!) that i built, AMD 2.8 ghz dual core processor(Windsor), ATI Radeon hd 4300/4500 series(for watever reason, all tho i cant figure it out, 4350 is sticking out), 4gb ram ddr2, 579 gb sata hardrive space, x86 windows 7 ultimate... pretty good, in my opinion to be running high performance versions, but i have no idea how to setup f@h for multicore/GPU usage.

[k1wi]

Hi relion64,

For FAH it is actually probably a little underpowered to fold with high performance versions! (FAH moves ahead FAST).

I am led to believe that F@H has stated that they don't recommend dual-core machines folding SMP (multiprocessor work units) and the HD 4xxx doesn't have much time left until it is depreciated by F@H because it cannot efficiently handle the new calculations being rolled out by the project and running the old software they can consume considerable CPU time unless your edit the environmental settings on your computer.

But that said, you can certainly fold multiple 'regular' uni-processor work units, either using the Beta v7 client which allows you to manage them all from the one control, or using multiple v6 clients (one for each core)...

k1wi

[Baowoulf]

Depending on the speed of your dual core you may or may not be able to do SMP. Depends if it's one of the newer dual cores or an older dual core. Your graphic card should be good until September I believe. That's when the deadline is for ATI cards 4xxx and lower because they don't have a sufficent CAL? level.

I'd recommend trying SMP and seeing if you can turn in a WU in time and if not do CPU/GPU clients or just GPU depending on the performance you get.

[Napoleon]

Since it is also a gaming computer, I think it unlikely there will be enough spare processing power to run the high-performance SMP client in a beneficial way. Two uniprocessor clients (or two uniprocessor slots in Beta v7 client) are just fine, though. V6 installation guide also tells you how to set up two or more uniprocessor clients (one per CPU core). I'd suggest you try them first, then reconsider adding a HD4xxx client/slot. See also FAH Support for ATI GPUs announcement.

[relion64]

huh.... thats weird... im not going to even bother with SMP/GPU clients now, but i can play newer games like Portal 2, Black Ops, and other not really new games but need good computers like TF2, MW2, World of Tanks, Audiosurf, and Grand Ages of Rome, just to name a few of my games... and every single one of them, except Black Ops, i can play with ALL graphics settings turned up to extra... black ops, i can turn a few up to extra... so thats why your responses have surprised me. I know that its not new, or state of the art, or anything like that... it was a Dell Dimension E521 that i decided to modify with spare parts from a computer that, yes is at least 4 years old, but when my dads friend first built it, EVERYTHING in that machine was top of the line... you couldnt beat it. As a matter of fact, the only thing that i havent switched out on this computer is the mobo and the things attached to the case... and i couldnt change out the graphics card because of a power problem with a NVIDIA GeForce 7950 GT that was installed in the other computer. But to make a long story short, im actually really, really surprised... believe it or not, this computer is the most powerful computer in my house... my computer blazes past my brothers 1 1/2 year old ACER gaming computer!!!!! so again, surprised!!

[Baowoulf]

There is more that goes into the requirements then what it takes to run the latest video games. The GPU clients require a high enough CAL(ATI) or CUDA(Nividia) level to run the GPU client. And SMP the thing is it's large WU's that have a quick return deadline. And with the SMP the number of cores your cpu has factors in highly to what you need to run it. But this is a very simplfied explanation about it. Someone like Panther or Bruce could give you a much more informed answer.

Also don't forget sometimes card makers can change how they decided to code/make their cards and that can lead to needing newer hardware which is what I think happened with ATI.

[bruce]

I'm not sure how much extra information Bruce can offer. The answers already provided cover it pretty well.

Welcome to foldingforum.org, relion64.

I do successfully run the SMP client on a Core2Duo. I can do it for two reasons:

First, it runs FAH 24x7. The deadlines for SMP work are much more stringent than the deadlines for uniprocessor clients, so 24x7 is essential.

Second, I don't use it for gaming. Two uniprocessor clients (or slots, if you choose to run the V7 beta) do a much better job of yielding resources to your game than SMP, so you're less likely to see a performance impact. If I did, I'd probably choose to shut down FAH while gaming which means that occasionally I'd forget to restart it. (All that time counts against the deadline, too.)

Some people choose to run FAH as a donation to science and others choose to get deeply involved in the competitive aspects of points and teams, etc. (and others do both). Here on the forum we take no position that one is better than the other. In either case, any donation of processing time or any points contributed to a team is important, no matter how small, so start now

[PantherX]

Everything important is covered above.

However in case you're wondering about your GPU, the reason that PG can't support it is because ATI/AMD has dropped support for Brook hence PG has no other option but to follow. Moreover, the new FahCore_16 uses hardware features that aren't present in HD 4000 Series. They are only present in HD 5000 Series and above.

Regarding your CPU, there are many new CPUs released each year and upgrading is taken for granted by PC enthusiasts. However for those that don't upgrade, you can always run the Classic Clients *if your CPU can't fold SMP WUs) which will have scientifically valid data that SMP or GPU can't do. Thus every single folded WU counts

[relion64]

well, i ive done some things and found out a few things. one, i can run v6 GPU client wiondows 7 x86 at 100 iter/sec, IF i dont have the viewer open. viewer open, it drops to 25 iter/sec. then after a few seconds, the client start to only update the screen every second, so its like major lag... BUT then i get 75 iter/sec with viewer open. also, like right now, i can run f@h, and boinc, and on boinc, right now, along with f@h, im running seti@home, enigma@home, malariacontrol.net,EDGeS@HOME, eon2, cosmology@home, and milkyway@home. of course boinc is only doing it 2 at a time, but i put my computer to the test. i was able to have the viewer on seti@home open, plus the viewer on f@h, plus internet browing on chrome, and listening to music on wmp. no lag, period. i was also able to play audiosurf while using seti@home and F@H. so, im kinda confused. everything you guys have said i just pretty much smashed. im not going to start an argument over it, but this is kinda weird. (im not a hardware expert, im software, btw.)

[relion64]

i also forgot one thing... isnt the cuda version of ati video cards called APP(Accelerated Parallel Processing)? secondly, my video card has APP... just throwing that out there. Thirdly, the docs, that i remember, call for a ATI 2xxx-5xxx video card, and thats going to be going out, right? well, my card is technically an AMD Radeon HD 4350. (emphasis on the HD). it is a 4xxx series card, but its the HD version, plus AMD just started to slap there logo, instead of ATI, onto their video cards, and my card is an AMD. (sorry if i sound condescending in any of my posts... i have problem[ill leave it at that] and thats the best way i can explain it).

[k1wi]

Hi relion64, below is a list of points that may help clear up the situation. I hope it will provide some more information into the world of folding

Unfortunately there is no version of CUDA for ATI/AMD - CUDA is a NVidia programming language, it would be like the Germans speaking French! (well, there would probably be more money involved between ATI/AMD and NVidia) ATI previously developed Brooks/CAL, but have since abandoned development of that in favour of a new open standard called OpenCL/GL (the English of the GPU world if we are to extend the previous analogy too far!) of which the new OpenMM core utilises.

Second of all: The folding viewer - don't use it! Why? Because as you noticed it uses the same resources you're using to fold to make the pretty images, resulting in a drop in performance. Plus it's also known to be pretty unstable. Also iter/sec isn't the best way to measure 'speed' as it can vary from one project to another. PPD (points per day) is a better measure.

Third, I don't doubt that you can fold and do all the other things on your computer - that's how F@H works. On the CPU it has the lowest priority, so while you're able to do everything else, F@H is making away for them, so you're not actually folding much. I cannot tell from your stress test whether you were just stressing the GPU with F@H, or whether you were also running the CPU client, or whether you were just using the CPU version of BOINC...

Fifth, a wide range of CPUs can fold, but many of these are limited just to folding uni-processor clients. In terms of GPUs, there are a number of GPUs that currently can fold, but GPUs such as the ATI/AMD 4xxx series and below are about to be depreciated for the reasons given by Panther above: the 4xxx series GPU might have introduced new hardware features, but the new features required by F@H to fold efficiently weren't introduced until the 5xxx series. There is no difference between an ATI 4350 card and an AMD Radeon HD 4350, infact you cannot get a non "HD" 4350 as all since the 2xxx generation of cards have the HD monkier attached to them. The other thing to note is that the ATI GPU client consumes a lot of CPU resources - try running just the GPU client and see how much your CPU is being used by it. If one of your cores is being used to feed the GPU then that will substantially slow down your SMP client, because it will be constantly 'waiting' on the core feeding your GPU.

The best way to monitor 'speed' in folding is in terms of ppd - that is your points per day. That is the amount of science being done by a given computer per day. Installing a monitoring app such as HFM.net or FAHMon will let you see that the points earned by your ATI/AMD 4350 is much much lower than the equivalent NVidia GPU of the same era - probably in the realm of 2000ppd vs. 5000ppd. Compare that with the lastest NVidia clients which are producing over 10,000 points a day. Unfortunately, your ATI card is not using the latest version of the GPU core, which after November will be the only one supported.

As I mentioned, PG recommend against dual processors running the multiprocessor version of F@H (SMP), but there is nothing stopping you from trying it out and seeing whether it is earning you a higher ppd than running single processor two clients. There is a great install guide next to the download link that gives step by step instructions as to how to install either two single processor clients or one SMP client.

In any event, my recommendation would be to install HFM.net or FAHmon, that will give you a much clearer picture of your folding performance, you can also use it to see whether you are able to meet the deadlines by running the SMP client!

k1wi

[John_Weatherman]

The thing to remember is that Stanford/the Pande group want the results back as quicly as possible (see http://fahwiki.net/index.php/How_Import ... adlines%3F). Using your machine for gaming, browsing or anything else means that folding will grind to a halt, or work very slowly. That's why it sounds best in your situation to try the CPU single core/uniprocessor client first. If you're going to have your machine on 24/7 then ok, try the other clients.

[bruce]

(im not a hardware expert, im software, btw.)

No problem, but let me explain something about the hardware that may be missed by software people.

Computers are good at computing information that's in real RAM. Modern operating systems have virtual RAM which allows lots of software to appear to run simultaneously. In fact, only one is being processed (per CPU Core) at any one time. That's fine for a browser or a Word Processor because they spend a high percentage of their time either waiting for somebody to move a mouse or type something or for data to be obtained from disk or from the internet and actually use the CPU in very short bursts.

That is NOT fine for things like BOINC or FAH. They will use 100% of the CPU for hours/days at a time. That means that you're not really running all those applications at the same time. You have a dual processor, which means that as soon as you start the third application, the first two slow down to 2/3 speed. As soon as you start the fourth application, they all run at 1/2 speed.

We would very much like to have you run Folding@home, but we also recommend that you make a choice. Dedicate one or two CPU-cores to a specific project. We'd much prefer that it be FAH, but if that's BOINC, so be it. Splitting the resources by running lots of DC clients is not a good idea. BOINC only runs two at at time for a reason, and just because FAH doesn't use BOINC doesn't mean that you can run two BOINC projects PLUS FAH and get any more processing done.

[relion64]

ok... one, wat i mean by im not a hardware expert is the fact like: let me explain it this way. My dad has worked on many computers over the years... he worked at dell for 5 years reparing laptops, and all this other stuff. my dad was working on computers in the 80's. so he has 20+ years experience on computers. he can build a computer with his eyes closed. I, on the other hand, dont have that experience. yes, i did take a Dell Dimension E521(OEM, bought from store, that is at least 2-3 years old), and i pretty much replaced everything on it: processor, ram, hardrives, etc, and turned it into a gaming machine. original parts, that i did not replace/modify is the mobo and stuff attached to the board. BUT, just kinda like my java programming skills, i dont know how it works. i know that typing certain code will allow me to do this, but its probably not the efficient way, and i dont know how it works the way it does. that's why im also stuck on Swing. command line stuff(mostly BlueJ), im great... swing, not so much. Two, i know alot about computers tho, even with that last paragraph. ive read many books on this subject, i do kinda know how a processor works, and ram and all that... i do know that "multitasking" in a computer is not actually multitasking(just to clarify, i know how to install, but with like overclocking, i have no clue what the new voltages and what not need to be). secondly, running F@H and Boinc, believe it or not, it was actually doing something. i set F@H to higher priority, and i was able to figure out how many iteration a second its doing by looking at the step number, immediately closing the viewer... counting to thirty, just to kinda get an average, then opening it back up, reading that number, adding one second to account for the time the window opened, subtracting the two numbers and dividing by 31. i have a quick eye, and reflexes, so thats possible. so i am getting 100 iter/sec even with Boinc open. I only have Boinc because i was only using my GPU's power... completely leaving the processor open. thus i can run two Boinc tasks, and F@H at the same time... and even with everything maxed out, i can still play games. now there, is when maybe Boinc and F@H will relinquish some resources, that would be the logical thing, but believe it or not... im actually being productive with these settings. Also, im kinda confused on one other aspect: according to what ive been told by my sources, the ATI Radeon HD 4300 Series and the ATI Radeon 4300 Series are actually two different things, with the HD series being released afterwords. so that would technically make it a newer GPU right? and sorry, i was wrong, my card is not an AMD... apparently my friend bought it before AMD started slapping their logo onto the video cards and started phasing out ATI.

[Baowoulf]

I think I get where you are getting confused. On one hand you have the card which is a ATI Radeon HD 4300 and on the other they are probably talking about the 4300 series like (4300 4350 4360). Note: I just picked some likely card numbers for example purposes I have no idea what the other 4300 series of cards numbers are if there is any.

Also how graphic cards are numbered can be confusing at first. Such as a 5300 might not be as strong as a say a 4960. But a 5960 would kick the 4960's butt. I think it's better to compare similar cards like compare the closest you can get to a say 3300 or a 5300 with your 4300. At least that's what I've gotten when reading about graphic cards(Nvivdia ones but I can't see why it couldn't be the same for ATI as well) since I'm going to need to build a computer soon.

Note: I used the card numbers just as an example. I could have went to ATI's website for exact card numbers but I don't think it was necessarily.