Jump to content
HWBOT Community Forums

Strunkenbold

Crew
  • Content Count

    1736
  • Joined

  • Last visited

  • Days Won

    11

Strunkenbold last won the day on September 2

Strunkenbold had the most liked content!

Community Reputation

180 Excellent

About Strunkenbold

  • Rank
    hardware overlord

Converted

  • Location
    GERMANY
  1. hwbot "classic" rev8

    Well the problem of global and wr points is that they are really big. While you just can get maximum 50 hardware points, it is possible to receive over 160 points for globals and 170 points for worldrecord points. In other words an active bencher which posts some high quality scores with big points can be #1 in career ranking and in seasonal ranking at the same time and he can achieve this in some months (hypothetical). I think the career ranking should reflect the effort made over some longer time and not just some months. Currently those 30 slots for hardware / global / WR fill up too quickly with current made results because global / WR points have a so much bigger influence. Regarding comp points: I think its sad that the effort made in previous years diminishes. Seasonal ranking gives the chance to see the current level of a bencher while career shows what he did over the time. Its 10 slots for those points. If we shift up hardware point slots it should not have a too big impact anymore.
  2. hwbot "classic" rev8

    Well the point is that I didnt see too much constructive criticism in this thread. Just complaints. If you want that Frederik does change something in your direction you have to give him something he can work with. I mean it should be clear if Frederik just receives negative feedback of his work, he probably doesnt see the point to spend time on hwbot. Which ultimately means that hwbot close his doors forever.
  3. What is a Vega 56 Nano?

    Well I think AMD had actually the intention to release it. At least it was rumored and I really thought it gets released. However, I think AMD did not introduce another marketing name for it when the powercolor nano got released, so I guess it can be deleted from our db.
  4. hwbot "classic" rev8

    So the culprit is just the seasonal team ranking? I think this can be easily changed back to an all time ranking again. I dont see why this needs to be seasonal. Actually first I thought the seasonal ranking just add more complexity but the more I think about it the more I like the idea. The seasonal ranking idea has the advantage to attract new benchers to easily climb up. I mean hwbot is over 10 years old, if the ranking would be static for all times, newcomers see those thousand and thousands of points and give up early. Cause it makes a difference if you had 10 years time to bench those gazillion hardware / benchmark combinations or just some months. However the current approach to calculate the career points needs to get tweaked. As currently, if you have benched a lot of current gen hardware, you are way too good in the career ranking which pisses people who are here for a long time. As we all know, there is a new hardware generation coming each year, and each year world records gets beaten by new hardware. If I take a look at the profile of splave http://uat.hwbot.org/user/splave/#Points I see that his seasonal ranking doesnt differ much from his career ranking which leaves the question why having two rankings if it acutally doesnt look much different. So my proposal is to remove global points from career ranking and increase the submissions taken into account for the ranking, maybe to 50 hardware and 30 competition. Just to make a difference between one year and 10 years contribution to the bot. The seasonal ranking is a nice idea. But you also have to see that points get zeroed every year. So it might be also a good idea to create new achievements like best overclocker 2017 or something. And introduce rankings or leader boards for the last years. So we all could see who was the best of 2010 or something.
  5. Please add old IIT 3c87 FPU processor

    Well that is something unique for sure. However if I understand correctly, those Co processors can just work with a main CPU. Is any program able to detect those combos? I mean, I dont know if it makes sense to add co processors at all to the db, probably need another opinion from Antinomy here.
  6. hwbot "classic" rev8

    Dead links on uat. Is anywhere else a description how those points come together (for the users)?
  7. hwbot "classic" rev8

    Okay misunderstood the focus then. As no one replied after your RFC, I should say that it was just my personal impression that we should make the point algorithm easier. As I thought that the many broken rankings are because of this too complex algorithm. My personal motivation has been always the points I got in a ranking. I never cared too much about my ranking in a league. So thats the reason my focus seems to be different from the other guys. If rankings could get fixed and work more reliably, plus the socket ranking pages, I would be very satisfied.
  8. hwbot "classic" rev8

    hmm and I thought changing the point algorithm to something better understandable and simpler is all about this new rev... I mean, I understand that you have to keep those percentages for lower server load but I think the biggest problem of the current rev are strange and broken rankings. Youre right, I was wrong about the percentages, mixing hardware and global, but IMO the percentages should be really reverse. 50% for the globals, means like 4Ghz scores get global points where the 75% cliff for hardware points makes Air and Water benchers unhappy...
  9. hwbot "classic" rev8

    It would be very nice if this longstanding bug could be finally fixed. I think this is the reason for some serious problems on the website.
  10. hwbot "classic" rev8

    This all looks very promising, great job! Some small additions: Im not a big fan of the 50% cliff for the hardware points. I often see very weird rankings because of this. And its not good to build up rankings because user on air have no motivation to post results once a LN2 score got uploaded. If we want too keep the cliff I have the following ideas (not sure how hard it would be to implement): 1. To feed the ranking with results, award the first 25 results always with points regardless of 50%. If possible visualize those 25 places with another color or something else so people get the idea. 2. Draw a red line or something in the ranking page to illustrate where the cliff begins (So people know why place 36 get points while place 37 doesnt). 3. Calculate 50% of the current first place and show the number in the ranking page so people know which score they have to beat to get points.
  11. hwbot "classic" rev8

    +1 Would be very very cool to have this on the regular page instead hidden in the forums. I think this should be possible without too much problems, I mean I can search for SuperPi results for Socket 478 too and those get sorted by the best result. Why shouldnt it be possible to create a page for this?
  12. Please add old cpu ng80386sx-33

    http://hwbot.org/hardware/processor/am386_sxsxl_33/ Dont know if I should go with this name scheme though.
  13. The Official Team CUP 2018 DDR3 stage thread:

    Just want to say that we mostly rely on data from CPU-World and TPU GPUDB. I know that CPU-Z as GPU-Z have problems showing the correct code name for recycled or castrated parts. That often happens if the vendor doesnt change the IDs of those parts. Thats the reason why CPU-Z /GPU-Z doesnt match our db in some cases. On that FM1 performance issue: I put an answer already in the _mat's thread where he examined the issue. But in case you didnt read yet... The reason why older bios versions are faster is that AMD turned of a feature of those CPUs because they found a bug which can cause the CPUs to hang. GPUPI makes heavy use of division operations. Llano CPUs actually have a unit that can do those division operations very fast. But because of a bug, AMD decided to turn that unit off. Those operations need to go through other units of the CPU now which are a lot slower. In case you ask now why nobody in the last 7 years bothered, the reason is that division operation seem to be just not that important under normal workloads, yet GPU-PI makes heavy use of it. If you want to re-enable the performance, without downgrading your Bios or your OS: https://www.passmark.com/forum/performancetest/3705-amd-llano-a-series-benchmark-and-cpu-bug?t=3656 Cbjaust already confirmed that it works.
  14. FM1 and the unexplainable GPUPI performance

    Thx for testing this. Maybe _mat_ can confirm. I think this is getting us to a comfortable position where we can safely allow this "tweak". Its just like we did with The Stilts work optimizing Superpi performance on Bulldozer, afaik he was also messing around with CPU registers.
  15. FM1 and the unexplainable GPUPI performance

    Yes I remember there was something similar with those unlocked Haswell mobile CPUs. You have to prevent that Windows loads the intel firmware on boot. I dont think that you need to implement the fix in GPUPI. All needed is the patch exe from the passmark guys. To quote their readme: This is reminds me of The Stilts Bulldozer Conditioner... All we need is someone to test if it really works. I think I have somewhere a board and a CPU both from scrap and not known to work maybe its time to test this bundle now.
×