Scoring more on older hardware ?

[r0k0]

GTX 780Ti PPD Est. 590300
GTX 1060 6GB PPD Est. 369800
GTX 1070 SC PPD Est. 737700
RX5700XT PPD Est. 1027000

damn didn't know my 5700XT was so strong ! maybe i shouldn't have sold that 2nd for a profit from what i originally bought it for

i have a lot of cards at home that i bought for collection and testing and some for building pc for resell, some guy selling me EVGA GTX 1070 SC for 225$ CAD, currently bought 2, im thinking of picking up 2 or 3 more ! lol

The GTX 780 Ti is rated at 250 watts and the GTX 1060 6 gb at 120 watts, so before you buy more, consider your cost of electricity.

i don't pay for my hydro where i live ( all included in rent ) so it doesn't really bother me for the moment

[TMStech]

Honestly, "score" and or/ the definition of "work unit" appear to be pretty meaningless. I've been slowly chipping away for a total of 136 units, for a score of 101,131. Someone who's apparently been around and jumps around teams a bit just joined my team days ago, and with 10 units they blew past me with a score of 168,802. XD

Before, I thought it might have been "in the past they gave huge scores" - but obviously that's not the only factor. I don't know to what degree tenure and "what project you're working on" factor in, but it's not exactly encouraging to see someone get 23x the score per WU. Thus I remind myself that score means nothing, only giving my tiny push toward understanding COVID-19.

[MeeLee]

Your 780 Ti has 2880 shader cores, vs only 1280 on the 1060.
Your 780 Ti does run at just under 1Ghz, vs the 1060 running at under 2Ghz.
The 780 Ti has 2,25x increase in # of cores, running at half the frequency.
That means it's about 12% higher speeds.

Also that GPU has a Ryzen 1600 6 core 12 thread at 3,6Ghz helping PPD.
The Intel Xeon has only 4 cores 8 threads running at 3,8Ghz.

both cpu
Ryzen - credit : 5758
Xeon - credit : 4947

not really much a difference in cpu folding power, im surprised anyways that the 780ti is still a strong card to fold, i guess it was worth the 100$ i paid for it vs the 200$ gtx 1060 6gb lol

It really depends.
If you doubled your amount (to $200), you'd have more than twice the PPD.
The money you've saved, you'll have to pay back in electric cost.
The 780Ti at 250W uses more than double the power of a 1060 (120W)

[HaloJones]

good attitude. in the past points were 1 pt for one work unit. it was really hard to get thousands of points. now a good gpu can get millions in a day.

[JimboPalmer]

Tenure often means your Points per WU is very low.
On the team I fold for:
I am in 3rd place on the team, in 1,339 place over all, my computers have earned 552,348,567 points in 161,284 WUs
The 4th place on the team, in 1,435 place over all, Jafo, has PCs that earned 515,247,209 points in only 24,942 WUs. Jafo uses faster hardware than I do and did.

Which Project can be important if you own a expensive GPU/CPU.

You can buy a Graphics card with about 5000 shaders (threads) and if the Protein you are folding is small enough, it won't utilize all of them. My GPUs are mid range parts that do not 'suffer' from this.

In a similar fashion, the newest CPU may have 256 threads, And Windows has issues over 32 threads. Volunteers who own some Xeons, Treadrippers, or Epycs often have to divide them into multiple slots to get work.
I own bottom of the barrel equipment, often from a decade ago, so I do not 'suffer' from this. None of my CPUs have over 4 threads.

[r0k0]

in my understanding i thought PPD was just helping more, i know its more the work units, but i was under the impression the more points meant you donated more, pushing out more needed WU.

myself i dont own the most recent hardware for folding

HP Z800 2x Xeon X5675 / 24GB DDR3-1333 ECC / GTX 670 until i get my 1070ti back from RMA
HP Z220 Xeon E3 1245 V2 / 16GB DDR#-1600 / GTX 1060
Ryzen 1600 / 16GB DDR4-2400 / GTX 1070 SC ( previously housed the GTX 780ti )
Intel i7 6850K / 32GB DDR4-3000 / RX 5700XT

currently i have 3 towers running 24/7, my 5700XT rig is my gaming computer so it doesnt run all the time because i am gaming

Honestly, "score" and or/ the definition of "work unit" appear to be pretty meaningless. I've been slowly chipping away for a total of 136 units, for a score of 101,131. Someone who's apparently been around and jumps around teams a bit just joined my team days ago, and with 10 units they blew past me with a score of 168,802. XD

Before, I thought it might have been "in the past they gave huge scores" - but obviously that's not the only factor. I don't know to what degree tenure and "what project you're working on" factor in, but it's not exactly encouraging to see someone get 23x the score per WU. Thus I remind myself that score means nothing, only giving my tiny push toward understanding COVID-19.

if you get a passkey, the points are much much higher, i've folded 1 month and did a few hundred WU and only managed around 1millions points, now after i got my passkey, with 271WU im currently sitting at 6.8 millions points

[r0k0]

so I do not 'suffer' from this. None of my CPUs have over 4 threads.

so if my HP Z800 with 2x 6 core xeons, i should make them run 3 cpus instance so that i dont have more than 4 cores / 8 threads working for more WU production ?

[Neil-B]

Your xeons would be best put to use as a single 24core slot (if not GPU folding) … A lower number of higher core count slots turns the WUs around quicker which helps the science.

Your total thread count does not exceed 32 so you don't have to worry about the issue JimboPalmer was alluding to … I have 56 cores (actually threads but use the core terminology) so run a 24core CPU slot and a 32core CPU slot … I never have any issues getting WUs for those, hence the reason I advise a 24thread

[JimboPalmer]

I said this:

In a similar fashion, the newest CPU may have 256 threads, And Windows has issues over 32 threads. Volunteers who own some Xeons, Treadrippers, or Epycs often have to divide them into multiple slots to get work.

But it is possible you only read this:

so I do not 'suffer' from this. None of my CPUs have over 4 threads.

so if my HP Z800 with 2x 6 core xeons, i should make them run 3 cpus instance so that i dont have more than 4 cores / 8 threads working for more WU production ?

No. The Windows limit is 32 threads. If your PC has more than 32 threads, and runs Windows, you should consider dividing them up. You should only have 24 threads, your CPUs are from 2011.

If you also fold with a GPU on the same box the math gets more complicated, as F@H dislikes large Prime numbers of threads, or multiples of large Prime numbers.

_r2w_ben is doing research on which numbers are safe and which crash F@H. The FAHclient gets better at avoiding bad CPU counts as well.

2, 3, 4, 6, 8, 9, 12, 16, 18, 20, 21, 24 are good numbers of CPUs to choose.
5. 10. 15, 20 may work most of the time. Other numbers will bite you.

You want to avoid 7, 11, 13, 14, 17, 19, 22, and 23 threads; and understand that sometimes 5, 10, 15, and 20 may fail.

With a graphics card, I would choose 21 threads in the CPU slot.

[r0k0]

well crap, i think 98% of the time its a7 and 22 that are always running, i dont see other numbers then those working

[JimboPalmer]

well crap, i think 98% of the time its a7 and 22 that are always running, i dont see other numbers then those working

I am confused. What other Cores did you hope for?

Core_a7 is the only current CPU Core, It runs on CPUs back to 2001 (Pentium 4 Willamete) and improves the science on newer CPUs.

Your Xeons will use SSE2, newer CPUs use AVX.

Core_22 is the latest GPU Core and all COVID-19 GPU WUs use it. It needs OpenCL 1.2 and Double Precision floating point math.

Core_21 is an older GPU Core that will have some older Projects to complete. It also needs OpenCL 1.2 and Double Precision floating point math, but does not support AMD Navi (RDNA) graphics.
Because Core_21 is older, it uses a less capable version of OpenMM. It will do less PPD than Core_22. I expect much gnashing of teeth when those projects come back.

So 99% of the time the latest and best cores are used. When the older Core_21 comes back, it is to finish projects that were already started.

In the future:
If the AVX-512 'standard' ever become standardized, I would expect a Core_a9 to use it.
https://en.wikipedia.org/wiki/Advanced_ ... th_AVX-512
There is a new OpenCL 3 standard coming, it is possible Core_23 will use it.
https://en.wikipedia.org/wiki/OpenCL#OpenCL_3.0

[Neil-B]

Think there may be some confusion here ... a cpu slot uses the core_a7 processing unit but the number of threads/cores/cpus that can be assigned to that slot is where the choice comes in - the FAH advanced control can be used to set these ... gpu slots use either core_22 or core_21 processing units.

[r0k0]

You want to avoid 7, 11, 13, 14, 17, 19, 22, and 23 threads; and understand that sometimes 5, 10, 15, and 20 may fail.

that's what you said avoid 7 and 22, which are the ones that always runs on my pc

[PantherX]

...that's what you said avoid 7 and 22, which are the ones that always runs on my pc

Please note that if you review the log file, you will notice that while you have set it for 7 CPUs, it would automatically be rounding it down to 6. I am not sure if 22 does that too or not.

Why not post a sample from each system so we can have a look

[r0k0]

...that's what you said avoid 7 and 22, which are the ones that always runs on my pc

Please note that if you review the log file, you will notice that while you have set it for 7 CPUs, it would automatically be rounding it down to 6. I am not sure if 22 does that too or not.

Why not post a sample from each system so we can have a look

i dont even know where the logs files are, im still very recent and running F@H at a minimum basic hahaha, oops