Folding Forum

Hello!

I have a few older CPUs that have been laboriously crunching on Boinc for a few months now.

One, a dual socket AMD Opteron6128 (16 cores total) and 2 Intel Xeon w3530s. I have a few other smaller fellas - an i5-2300 and 3570. What should I be expecting for any / all of this in terms of PPD?
I know the power draw is amazing on the AMD in particular and I won't be using it in the summer. I have a Ryzen 3600 in mind for later perhaps paired with a 1660 or 1660ti.

Are any of these older machines even worth folding, with all of the amazing GPU folding going on here? I feel like strictly CPU folding I'm such a small fish in a huge pond with my tiny little RX570.

Thanks to the amazing effort being put into all of this. You guys are doing fantastic - scientists, administrators, moderators and all! Keep it up!

AMD Opteron6128 8 threads SSE2
Xeon w3530 8 threads SSE2
i5-2300 4 threads AVX-256
1i-3570 4 thread AVX-256

So all of them are faster than what i fold with.

CPU folding excels at different molecules than graphics cards. Accurate simulation of small molecules is ideal for the CPU.
GPUs prefer to simulate very large molecules, as it is hard to make use of more than one shader per atom. CPUs don't have that issue as few Proteins have ewer than 8 atoms.

JimboPalmer wrote:AMD Opteron6128 8 threads SSE2
Xeon w3530 8 threads SSE2
i5-2300 4 threads AVX-256
1i-3570 4 thread AVX-256

So all of them are faster than what i fold with.

CPU folding excels at different molecules than graphics cards. Accurate simulation of small molecules is ideal for the CPU.
GPUs prefer to simulate very large molecules, as it is hard to make use of more than one shader per atom. CPUs don't have that issue as few Proteins have ewer than 8 atoms.

Out of curiosity, what do you fold with?
Thank you for the explanation. It is appreciated.
I think AVX256 came out with the introduction of Sandy bridge processors if I am not mistaken? And directly after Phenom II processors e.g. the generation of the opteron.
Re: GPUs vs CPUs - that is a very interesting observation and I guess it doesn't matter if my PPD is not in the millions as long as I am contributing. As long as science is progressing that is the important thing.
How far along does a folding WU have to be before the PPD shown in the web interface is accurate? or does the machine in question have to complete a certain number?
I do have a Ryzen 1800x going at the moment, and the CPU portion is showing 75000 PPD, which seems like an awful lot but it also could very much be true.
I think I have an old geforce 750ti here somewhere. Should add a little help.

[6 laptops: Pentium to Core i5, 2 Core 13 desktops: one with a GTX1050ti, one with a GTX 1060. I use 4 laptops as POS in the lobby of a local theater, my wife uses 2 to judge BBQ]

I know that by Haswell F@H is using avx-256. It is faster than sse2, but not that much.

I have been Folding since a Semperon and a HD 2600 Pro. I tend to use trailing edge hardware.

I don't use F@Hs site to track folding It takes a while before this guy starts tracking, but then it is pretty good
https://folding.extremeoverclocking.com ... =&u=429597

As a user of older equipment, some times it would be cheaper to buy new and use less electricity.

A GTX 750 should help a good deal.

JimboPalmer wrote:[6 laptops: Pentium to Core i5, 2 Core 13 desktops: one with a GTX1050ti, one with a GTX 1060. I use 4 laptops as POS in the lobby of a local theater, my wife uses 2 to judge BBQ]

I know that by Haswell F@H is using avx-256. It is faster than sse2, but not that much.

I have been Folding since a Semperon and a HD 2600 Pro. I tend to use trailing edge hardware.

I don't use F@Hs site to track folding It takes a while before this guy starts tracking, but then it is pretty good
https://folding.extremeoverclocking.com ... =&u=429597

As a user of older equipment, some times it would be cheaper to buy new and use less electricity.

A GTX 750 should help a good deal.

I just found vast.ai. Definitely not going to be buying anything rtx related, and even the higher end gtx series are pretty high up there for me. But mostly it's due to a lack of room.
Renting a GPU, or multiple GPUs, for $3-$8 a day is pretty decent. I'm struggling to imagine the costs of powering 8x 1080ti. This is definitely not something that would pay off long term, but for the shorter term it will definitely help.
Hopefully we see these new servers come online and workunits will resume coming in rather than playing the waiting game.
I've got a Lenovo x1 extreme, but would hesitate to fold with it simply due to heat. It's got a 1050ti in it.