Folding Forum

I am considering my next gaming system, and would like to fold when not gaming, but I would like to know which CPU may be better for folding. I tried a search, but came up empty, I would like to know, is the new Phenom II x6 better than a Intel i7? or even the i3? Perhaps some ppd numbers would be beneficial to show?

I run the console client on my Quad Phenom II @ 3ghz and it puts out about 220 ppd per core/client as I run 4 clients that's 880 ppd total. I couldn't figure out the smp client and so I run separate single core clients, that also lets me run 1-3 clients when I want to shut down one for playing a game or something. I know the new x6 will do better since it has 2 extra cores, but let's see what 1 core does and then we can multiply from there by how many cores you have.

It's the CPU optimizations that I'm wondering about, as the newer the processor, the better the code should run right?

Also does the i7 do better than the Phenom II? If so, how much better does the i7 do? or does it any?

When choosing a CPU, you have to consider:
Number of cores.
Hyperthreading, which doubles the effective number of cores.
Speed of each core.
Memory type and speed.

With the Classic client I don't think you will see a significant difference in folding between AMD and Intel CPU's at any particular price point.
Except that building an Intel i7 based system will be a lot more expensive.
So your choice should be determined by gaming performance and budget.

However, there is a new client due for release "soon": See News

This will make it much easier to run multiple clients and SMP, which may alter how you fold.

If you are going to fold classic uniprocessor clients, then at stock my AMD Phenom ii 955 is faster than my i7 920, but with SMPs the i7 is much faster.
I think this is the wrong question. What processor can you afford, and is it to be a dedicated folder if not what else will you use it for.
Basically though more cores and higher clocks are fastest, yes surprise surprise, faster processors are faster.
NB the release of v 7 should change all of this and make it easier to run smps

You also forgot to mention price... Whats your budget.

AMDs are very well priced per performance, Intels not so much, but perform very well.

guest3412 wrote:I am considering my next gaming system, and would like to fold when not gaming, but I would like to know which CPU may be better for folding. I tried a search, but came up empty, I would like to know, is the new Phenom II x6 better than a Intel i7? or even the i3? Perhaps some ppd numbers would be beneficial to show?

Can't give you values for AMD as they are not marketed here (only one shop that is ridiculously expensive) but I can tell you that my i7-860 @ 3.6 gives me ~8K PPD (6701) while the remaining gives me ~12K PPD. (using -smp 7 as the system isn't dedicated)

guest3412 wrote:I run the console client on my Quad Phenom II @ 3ghz and it puts out about 220 ppd per core/client as I run 4 clients that's 880 ppd total.

My previous build has Core 2 Duo E8190 @ 3.0 where I would run 2 Classic Clients and each would give me about ~450 PPD (Not sure as it was last year Dec and I was running with -advmethod flag and got the ProtoMol Core)

guest3412 wrote:I couldn't figure out the smp client and so I run separate single core clients, that also lets me run 1-3 clients when I want to shut down one for playing a game or something.

FYI, it is highly recommended that you exit/pause all the F@H Clients when running CPU/GPU intensive applications like gaming. When I used to play games, I would exit all the F@H Clients so that there won't be any possibility of corrupting the WU. Remember that exiting the F@H Client may cause errors, very small possibility but it exists.

guest3412 wrote:I know the new x6 will do better since it has 2 extra cores, but let's see what 1 core does and then we can multiply from there by how many cores you have.

It's the CPU optimizations that I'm wondering about, as the newer the processor, the better the code should run right?

Also does the i7 do better than the Phenom II? If so, how much better does the i7 do? or does it any?

In theory, performance should scale perfectly but practically, it depends on several key factors that should all play together nicely. For F@H, it is the Core which determines how efficiently it can process the WU.

On a side note, if you plan that your system is primarily for gaming and secondary for F@H, I would recommend that you chose a Quad Core as majority of games will not be using the full power of the 4 Cores. With a Quad, get a third party cooler and OC the CPU which would increase the game performance and would yield more F@H points. Do remember that with the current GPU2 and GPU3 BETA Clients, Nvidia GPUs outperform ATI GPUs by a significant margin so keep that in mind when buying your GPU .

PantherX wrote:Also does the i7 do better than the Phenom II? If so, how much better does the i7 do? or does it any?

In theory, performance should scale perfectly but practically, it depends on several key factors that should all play together nicely. For F@H, it is the Core which determines how efficiently it can process the WU.[/quote]

Actually performance CANNOT scale perfectly. There are all sorts of features that make small deviations from perfect scaling and cumulatively, they can add up to quite a bit. It's really difficult to give an accurate answer, though.

For instance, take cache size. Everybody understands that a small cache will run slower than a big cache, but there is a certain size which is "right" for a particular project, and increasing only cache size beyond that point will not make that project run faster. Another project will have a different optimium cache size, so even if there were an easy way to find that optimum, nobody can give you a general answer about what's right for all of FAH. Moreover, you don't get to adjust cache size, so whatever size comes with your system is going to be what you use.

Cache size is only one of perhaps a dozen features and as I said, collectively they add up to enough to notice. The only machine that scales perfectly is whatever machine that the Pande Group uses for a benchmark machine. Everything else will do better/worse on some projects compared to other projects.

If your primary concern is gaming then I would go AMD 1050T / 1090T and ATI 5850 / 5870 that will give you a top of the line gaming and a 10000 to 15000+ PPD folding system for the least amount of $$$$. I fold and game with both AMD / ATI and Intel / Nvidia and ATI you do not really need anything above a mid range quad for gaming and allot of people say that is a waste. But for folding the Higher end the better and the I7's socket 1366 are the best for folding due to there large cash and ability to OC.

guest3412 wrote:I am considering my next gaming system, and would like to fold when not gaming, but I would like to know which CPU may be better for folding. I tried a search, but came up empty, I would like to know, is the new Phenom II x6 better than a Intel i7? or even the i3? Perhaps some ppd numbers would be beneficial to show?

I run the console client on my Quad Phenom II @ 3ghz and it puts out about 220 ppd per core/client as I run 4 clients that's 880 ppd total. I couldn't figure out the smp client and so I run separate single core clients, that also lets me run 1-3 clients when I want to shut down one for playing a game or something. I know the new x6 will do better since it has 2 extra cores, but let's see what 1 core does and then we can multiply from there by how many cores you have.

It's the CPU optimizations that I'm wondering about, as the newer the processor, the better the code should run right?

Also does the i7 do better than the Phenom II? If so, how much better does the i7 do? or does it any?

In my own personal experience, the CPU (regardless of being AMD or Intel, or quad- or hex-core) is best used to 'feed' a good GPU, if you want to get the 'best' Folding@Home scores.

In fact, I have found that buying 'higher-end' CPUs and Motherboards is actually counter-productive, if you want to get the 'best' (= highest) Folding@Home scores for the amount of money that you spend on a system.

I do realize that you need to have a 'high-end' CPU/Motherboard/RAM combination in order to get good Gaming performance. I am, however, restricting my following statements to getting the 'best' (= highest) F@H scores for the amount of money spent, as a general principle.

To get the 'best' F@H scores for the amount of money spent, I now buy an 'ultra-cheap' CPU/Motherboard/RAM combination, and put in a 'kick-a**' GPU, like a GTX 260, GTX 275, GTX 285, or GTX 480.

That is how I get over 100,000 points-per-day (ppd) from my 19 Folding@Home computers, with 18 GPUs.

For a specific example, my #1 system has an AMD Phenom II X4 Quad-Core 965 3.4GHz CPU (not overclocked), and two nVidia GeForce GTX 285 GPUs (also not overclocked), running a single CPU systray client (on one of the 4 cores) and two GPU2 systray clients (one on each GPU). The GPU clients contribute about 25,000 ppd (combined), while the CPU client contributes a few hundred ppd (it's so small, that I haven't bothered to actually measure it).
Folding Machine #1:
OS: Windows 7 Professional 32-bit
CPU: AMD Phenom II Black Edition X4 Quad-Core 965 3.4 GHz (AM3)
MB: ASUS M4A79T Deluxe
RAM: 4GB DDR3, 1333MHz
HDD: 2.5TB (1 x 500GB plus 1 x 2.0TB)
GPU: 1 x MSI N285GTX SuperPipe OC, 680MHz Core, 1 GB GDDR3 2500MHz Memory (factory overclocked)
plus 1 x EVGA GeForce GTX 285 648MHZ 1GB 2.484GHZ DDR3
PSU: Silverstone OP Series 1000 Watt, OP1000-Evolution

After gaining more experience on "what works and what doesn't" on getting 'high' F@H scores, this is the type of system that I build now:
Folding Machine #19:
OS: Windows 7 Ultimate 64-bit
CPU: AMD Athlon II X2 250 Dual Core Processor Socket AM3 3.0GHZ 2MB Cache 60W
MB: Biostar MCP6P-M2+ AM2+ AMD mATX NVIDIA GeForce 6150 NFORCE430 PCIe-x16 2PCI RAID Video Sound
RAM: 2GB DDR2, 800MHz
HDD: 0.5TB (1 x 500GB)
GPU: 1 x EVGA GeForce GTX 480 Superclocked Fermi 725MHZ 1536MB 3.8GHZ GDDR5
PSU: OCZ ModXStream-Pro 700 Watt, OCZ700MXSP

In my #19 system, the CPU/Motherboard/RAM combo cost me about $160 CAD total (= about $154.50 US, at today's exchange rate). The GPU (GTX 480) cost me about $570 CAD (= about $550.38 US, at today's exchange rate).

In my opinion, it is actually a waste of time and money, to buy a 'higher-end' CPU/Motherboard/RAM combo to run the F@H clients, if getting a 'high' score is your goal. 'You' (in the general sense) may need a 'higher-end' CPU/Motherboard/RAM combo for another reason (like Gaming), but it is not necessary (or even desirable, in my opinion) for running the F@H clients.

I can build two systems like my #19 system, for what 'you' (in the general sense) would pay for a single 'higher-end' system, regardless of the CPU make or number of cores.

On the basis of getting 'the best bang for my buck', so-to-speak, I buy AMD processors exclusively. Sure, Intel's top CPU beats AMD's top CPU, but that fact is completely irrelevant to getting the 'highest' Folding@Home scores (based on the total amount of money spent versus the results that you get), in my experience.

I hope this helps,
Art

@art_l_j_PlanetAMD64 But a high end cpu (running full time) can do pretty well folding Big Advanced is probably better output for around the same prices.
However as the OP wants to game and fold, the GPU route is probably cheaper/better. As the OP already has a good quad (a 940 by the look of it) adding the best GPU or 2 will give the best fold/game performance. Best would probably be a GTX 480 at present.

Yes that is alot of information, thanks guys for your input, I guess my main question answered is that the more mhz (or ghz) the faster the uniprocessor client will output.

Thanks kiore, that was what I was really looking to hear - and yes you nailed it my cpu is a 940!

My main use of the system is gaming during the weekends and a few hours a day on weekdays, I already have a GTX480 and will transplant it - running the gpu3 client.

I was wondering if an equivalent i7 (980x) would beat out the much cheaper AMD 1090T, since I do not plan to run smp and I will shut down the core when I comeback to my PC to game, I have my answer - AMD is cheaper, performance difference is marginal, if any difference, and my folding/gaming will not be affected by choosing either. Budget? What budget, when it comes to my gaming, there is no Budget, it's pick what you want, then worry about paying for it ;)

If I am wrong in this assumption, please let me know, thanks :)

I tried to GPU fold, but I realized it was a huge power draw. The hardware outlay expenses might be lower than for a CPU system, but over the year following, electric costs far outweigh that. Of course, if you pay a very different amount for electricity, the equation could change.

I was able to sustain 76,000 PPD using 800W of power - 2 LGA1366 systems with HD5450 running bigadv and an undervolted LGA1156 with a GT240 on A3/GPU. Core hardware costs (had my i7 965 been a more practical i7 920, and ignoring SSDs) came out to about $2500. Because I pay 16 cents a kW-hr, folding was costing me an additional $92 a month... about $1100 over one year. I would draw the same amount in electric costs that I paid for hardware in 27 months! Had I gone the GPU route, I calculated I'd spend the same amount in electricity in just 14-15 months.

Worksheet summary:
CPU route: 3 i7's with 2 on bigadv and 1 on SMP2 + light GPU
Hardware cost: $2500
Power: 800W
Yearly cost in electricity, dedicated: $1100
Points per day: 76k (~30 + 30 + 12 + 4k)

GPU route: 3 Celerons/Semprons with 3 GTX275's each
Hardware cost: approx. $2000
Power: 1500W minimum
Yearly cost in electricity, dedicated: $2070
Points per day: approx. 76k (8400 per card)

That's a classic apples vs. oranges question.
Another thing to consider is the type of games you'll be playing. Many newer games take advantage of multiple cores while others do not.
I've been quite happy with the performance and low power consumption of the Opteron processors. The 4100 series has one model with 6 cores and a TDP of 35 watts; less than 6 watts / core. They range in price from 100 to 400. Check in server processors on newegg if that's where you look.
Like another poster mentioned earlier, the CPU and chipset combination is important as well. Some play better and more efficiently together than others.

kiore wrote:@art_l_j_PlanetAMD64 But a high end cpu (running full time) can do pretty well folding Big Advanced is probably better output for around the same prices.
However as the OP wants to game and fold, the GPU route is probably cheaper/better. As the OP already has a good quad (a 940 by the look of it) adding the best GPU or 2 will give the best fold/game performance. Best would probably be a GTX 480 at present.

Hi, and thanks for your feedback.

In my experience, with 19 computers (of which several are 'high-end' CPU systems), the CPU contributes at most 5% of my F@H score (assuming that 'you' have high-end GPUs like I have in my #1 system (dual GTX 285s)).

Regarding the 'high-end CPU', it most certainly does not give 'you' "probably better output for around the same prices". In fact, the exact opposite is the case, again in my own personal experience.

I have spent quite a bit of time (and money) researching this very issue, as (although I officially retired 7 years ago at age 50) I do have my own company where I design and build custom-built computers for people in my local area (Vancouver, BC, Canada).

On a cost-per-FAH-point basis, the very best option is to build an 'ultra-low-end' system with a high-end GPU, like my #19 system. This is, beyond question, the most cost-effective way to go.

Building a 'high-end CPU' system, like my #1 system, is (also beyond question, in my experience) the least cost-effective way to get FAH points.

If you would like to discuss this matter further, please send me a PM, so we can continue our discussion elsewhere, without 'cluttering up' this Forum with our different opinions on this issue.

Thank You, and Best Regards,
Art

art_l_j_PlanetAMD64 wrote: Regarding the 'high-end CPU', it most certainly does not give 'you' "probably better output for around the same prices". In fact, the exact opposite is the case, again in my own personal experience.

Please elaborate, because I have yet to find a 300$ GPU that can output at LEAST 20K PPD & with greater PPD/W than any CPU.

art_l_j_PlanetAMD64 wrote:On a cost-per-FAH-point basis, the very best option is to build an 'ultra-low-end' system with a high-end GPU, like my #19 system. This is, beyond question, the most cost-effective way to go.

Again, what does cost-effective even mean? The term is thrown around so often that it is often meaningless. Perhaps for GPU's, they are cost effective, but only in the short run. Even then it is questionable. It will cost you more to run a GTX*80 GPU than any Core i7 CPU.

art_l_j_PlanetAMD64 wrote:Regarding the 'high-end CPU', it most certainly does not give 'you' "probably better output for around the same prices". In fact, the exact opposite is the case, again in my own personal experience.

Hi, and this time last year I might have agreed but I'm afraid your informative is no longer up to date. The purchase price is not the only price and high end cpu's that are able to do Big Advanced units are at the least very competitive and on 24/7 will easily outpace any gpu on the market currently. For example my relatively cheap to run i7 920 easily gets 23kppd while my GTX 275's barely do 10k ppd each and the cpu has more economical power consumption. Note this is talking about 24/7 dedicated machines.
Although for ease of use and for part time machines a good gpu is the best option currently in my opinion.

I know you asked that responses to your post be by PM but I don't think you're going to get that after posting your personal opinion bolded above, as although it is from your experience it is not a currently accurate statement.