1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Bringing in the big guns.

Discussion in 'bit-tech Folding Team' started by Petaflops, 19 Sep 2018.

  1. Petaflops

    Petaflops Minimodder

    Joined:
    27 Aug 2013
    Posts:
    179
    Likes Received:
    1
    Haven't been on the forum for ages. I recently sold a pair of R290X's for reasonable money on Ebay. They produced about 250K PPD but ran very hot and were high watts to run. That has prompted a flutter of money on a Titan X Pascal and a Titan Xp. These are dropping to reasonable prices on Ebay now that the new Nvidia 2000 series has been announced. Wow do they produce some points but my study is rather hot and noisy now. I'm getting about 1.2 million PPD on the overclocked X pascal, and a bit too soon for the Xp. I only installed it yesterday so I'll wait to find out. I need to get my second PC going again as this had a problem with the enermax PSU. They were totally rubbish on the returns. I've got a pair of gtx 970's to go into that one though it might be time to sell those as well and upgrade to something a bit better. What's the latest best GPU in the mid range? The central heating is sorted for the winter, but the wife will be pissed!
     
  2. DocJonz

    DocJonz Another CPC refugee .....

    Joined:
    24 Apr 2009
    Posts:
    1,202
    Likes Received:
    97
    The team could do with a boost :thumb:
    Forum is a bit quiet these days ... looks like there are only about 100 active Folders on the team these days.
    The best mid-range GPU has been the GTX 1070 - mine do 650-700k PPD on 140W versions of the cards, so juice/heat not too bad.
    It will be interesting to see what the RTX cards do Folding-wise - I've seen that a load of Folding folk on one of the overclocking teams have pre-orders in, so we'll find out when their cards land ...
    (I had issues a number of years ago with several Enermax power suppliers - never again!)
     
  3. Petaflops

    Petaflops Minimodder

    Joined:
    27 Aug 2013
    Posts:
    179
    Likes Received:
    1
    The Xp is looking like it might do 1.3 - 1.4 Million PPD. Not too shabby. Yes I've rather been put off Enermax with my experience. Scan were good but Enermax were rubbish. I've still got to check that the repair has been done properly. I think the 970's were about 500K from memory but I'll see what I can get out of them and then decide what to do. I've heard good things about the 1070TI in terms of it's productivity and there are some increasingly cheap ones on ebay.
     
  4. Petaflops

    Petaflops Minimodder

    Joined:
    27 Aug 2013
    Posts:
    179
    Likes Received:
    1
    Tested the 970's and they did about 300k ppd, so a bit disappointing. I'm going to sell those on ebay. Bought a very cheap 1070ti on ebay and that put in 800k ppd overclocked a bit. Very respectable. I've moved the Titan X and Xp up to my mobile home to keep it warm this winter and to keep the damp out. These are together doing 2.2m - 2.5m ppd so I'm shooting up the leader board with results like that. The Xp isn't performing much better than the X (Pascal) really, so that's a bit of a disappointment, but then it was only 50 quid more. I've got another PC to build and I think I'll have that up the mobile home too, and leave the study whisper quiet for the wife's blessing. She says she is very sensitive to noise, so she says! I think I must be deaf as I can't hear anything. Probably ought to be mining with this kit but it all seems a bit complicated and mysterious to me. Think I've probably missed the boat on that one.
     
  5. DocJonz

    DocJonz Another CPC refugee .....

    Joined:
    24 Apr 2009
    Posts:
    1,202
    Likes Received:
    97
    That lot should give you a good boost.
    I'm with you on the cryto stuff - I've read numerous articles, but still don't fully understand it.
    PS I think my wife would be happy if I could re-locate my Folding rigs some where else as well!
     
  6. javaman

    javaman May irritate Eyes

    Joined:
    10 May 2009
    Posts:
    3,987
    Likes Received:
    191
    You should look into Curecoin....best of both worlds.

    Interesting the gtx970 doesn't do much more than my RX480. It feels like it is getting harder to find PPD and PPW measurements for cards these days
     
  7. Petaflops

    Petaflops Minimodder

    Joined:
    27 Aug 2013
    Posts:
    179
    Likes Received:
    1
    I've not heard of curecoin. What's that then? My results have been:-
    R290X - 250K - Sold for £175 about 2 months ago
    GTX970 - 300K - Sold for £160 this week
    GTX 1070TI - 800K O/C - now getting cheaper second hand on ebay - £270 - £300 - Great performer for the money.
    Titan X - up to 1.2 million O/C - getting much cheaper - £450 - £550
    Titan Xp - up to 1.3 Million O/C - getting much cheaper - £550 - £600
    I don't measure the electricity really, just the temperature!
     
  8. GeorgeStorm

    GeorgeStorm Aggressive PC Builder

    Joined:
    16 Dec 2008
    Posts:
    7,019
    Likes Received:
    559
    This has encouraged me to get some folding going again at least for a bit.

    Currently struggling to get fahcontrol to see my 970 I just put in my centos box, funfunfun!
     
  9. MLyons

    MLyons 70% Dev, 30% Doge. DevDoge. Software Dev @ Corsair Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    3 Mar 2017
    Posts:
    4,196
    Likes Received:
    2,777
    What would you consider a good way of measuring? If it doesn't take long then there's a possibility @Dogbert666 could implement it or I could test them.
     
  10. javaman

    javaman May irritate Eyes

    Joined:
    10 May 2009
    Posts:
    3,987
    Likes Received:
    191
    I think the difficulty has always been different projects running on each card giving different PPD results.
    Running a card for a week and taking the PPD average was a simple enough way that gave a ballpark figure for multiple units and PPW could then be worked out from power draw results.

    The biggest difficulty I've found is that most reviews simply don't run folding@home anymore.
    Some places use to have a spreadsheet of results but many cards where missing or the sheet hasn't been updated in years.
    Some other people on here may be able to give better suggestions
     
  11. MLyons

    MLyons 70% Dev, 30% Doge. DevDoge. Software Dev @ Corsair Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    3 Mar 2017
    Posts:
    4,196
    Likes Received:
    2,777
    If there's a demand for it I'm sure I can get permission to use some kit over a few nights and do a writeup. Just need to be told what info is needed and how to get it.
     
  12. DocJonz

    DocJonz Another CPC refugee .....

    Joined:
    24 Apr 2009
    Posts:
    1,202
    Likes Received:
    97
    I'd contact the guys at CustomPC, as they did used to do this as part of their hardware reviews a number of years ago.
     
    MLyons likes this.
  13. MLyons

    MLyons 70% Dev, 30% Doge. DevDoge. Software Dev @ Corsair Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    3 Mar 2017
    Posts:
    4,196
    Likes Received:
    2,777
    Been given permission to spend some nights at the office to get results. Should have the results done in the next 2 weeks or so assuming nothing comes up. Planning on testing 90/10/20 4/5/vega series and anything else I can find around the office. Any specific cards you guys want?
     
  14. DocJonz

    DocJonz Another CPC refugee .....

    Joined:
    24 Apr 2009
    Posts:
    1,202
    Likes Received:
    97
    Excellent.
    For a number of folks, the relative high-end would be of interest, so; RTX 2070, 2080 and 2080Ti and Vega 56 and 64. That way folks can gauge the PPD/Watt or PPD/£ etc., before forking out on the hardware.
     
  15. MLyons

    MLyons 70% Dev, 30% Doge. DevDoge. Software Dev @ Corsair Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    3 Mar 2017
    Posts:
    4,196
    Likes Received:
    2,777
    I'll be in the lab tonight benchmarking. I'll be running this and getting the power consumption of the rig. Hope to have the mainstream stuff finished tonight then will do the older stuff I've got coming in sometime later this week.
     
  16. DocJonz

    DocJonz Another CPC refugee .....

    Joined:
    24 Apr 2009
    Posts:
    1,202
    Likes Received:
    97
    Have you got the beer and pizza in :hehe:
    (Or isn't there a thing to do with 'chessecake' on the bit-tech site??)
     
  17. MLyons

    MLyons 70% Dev, 30% Doge. DevDoge. Software Dev @ Corsair Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    3 Mar 2017
    Posts:
    4,196
    Likes Received:
    2,777
    Seems that benchmark doesn't give me PPD hmm. Need to rethink how to get the results
     
  18. MLyons

    MLyons 70% Dev, 30% Doge. DevDoge. Software Dev @ Corsair Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    3 Mar 2017
    Posts:
    4,196
    Likes Received:
    2,777
    I'm going to need to work out a better way if doing this as after 5-10 mins f@h says estimated PPD of 94k which is wrong
     
  19. Gareth Halfacree

    Gareth Halfacree WIIGII! Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    4 Dec 2007
    Posts:
    17,129
    Likes Received:
    6,717
    Alternatively, don't worry about the validity of the results as they pertain to external benchmarks - because there are so many variables in Folding@Home that you'll never be able to do a one-to-one comparison anyway; instead, just use the results from the synthetic benchmark as relative.

    So, if you have cards A, B, and C, pick the slowest card as the baseline and assign it a value of 1.0. Let's say that was card A, and it scored 94,000 PPD. Card B scores 125,000 PPD, and card C 101,000 PPD. That gives you a final score of:

    A: 1.0
    B: 1.33
    C: 1.07

    Thus we can see that, for a F@H-style workload, card B is the fastest. Doesn't actually matter whether it was the fastest 'cos it got 125,000 PPD or 125 PPD - it's just faster.

    Then you can play around with performance per watt. Let's say card A draws 150W, B draws 300W, and C draws 200W. Play around with those figures and you get:

    A: 1.0
    B: 0.66
    C: 0.81

    Thus card A is actually *better* than B or C if you're going multi-GPU or just worried about your 'leccy bill.

    Ah, benchmarking!
     
  20. MLyons

    MLyons 70% Dev, 30% Doge. DevDoge. Software Dev @ Corsair Lover of bit-tech Administrator Super Moderator Moderator

    Joined:
    3 Mar 2017
    Posts:
    4,196
    Likes Received:
    2,777
    I was planning on doing something similar but then the benchmarking program just failed to open. As i don't want to bork the GPU rig I'm going to try setup a dedicated rig for it. Thanks for the help. Also my home rig is now folding for the team.
     

Share This Page