Jump to content
HWBOT Community Forums

denvys5

Members
  • Posts

    52
  • Joined

  • Last visited

Posts posted by denvys5

  1. Just now, previousslayer said:

    ? I don't remember that convo, had this idea on my own for a while

    CountryCup 2022 stage discussion on hwbot's discord

     

    On 4/12/2023 at 1:23 AM, previousslayer said:

    Zen1/+ CPUs, APUs and Threadrippers (socket TR4)

    I'd suggest if we have either single core stages, or multicore with per core scores.

    Since for mem there are not a lot of options, valid and pyprime. Yeah, lets do the hell challenge, valid on TR4 ? 

     

    A thing to note, I want to hear a clear definition, is the compo intendent for "no rules, best score", or more focused on the efficiency of scores, with cpu clock or cooling limitations. Maybe something funky, like, fastest zen1 pyprime with tRCD=20 and max cpu freq 4000 :D 

    • Like 1
  2. 25 minutes ago, yosarianilives said:

    W8 they're not linked?

    They are linked. On most boards. But, there are exceptions:

    1. Crosshair X370
    2. Crosshair X470

    Those boards have a feature to unlink bclk from cpu ref. clock and clock them separately. Cpuz registers one tied to cpu. 

    It is easy to check manually in cpuz valid, attention to memory clock.

     

    25 minutes ago, yosarianilives said:

    And I think we'd do it by chipset so at least c8 would be out cause x570 sux

    120mhz should be doable on most boards, with unlocked specific bios options (to disable the chipset :D ), so the only other option become boards with external bclk that do not have 120mhz limit (no idea which would fit here, C8 maybe? giga x470?)

  3. Quote

    Am4 max ref clk

    true ref. clock, or cpu ref clock?

    If true ref., sounds bad, we have no way to distinguish it from cpu ref clock (cpuz shows cpu ref clock on AM4) and we will see a lot of 120mhz results, once everybody figures it out.
    If cpu ref clock, boards with external bclk AND ref. clock target switch (for example, crosshair 6 and 7) will dominate the stage, which also doesn't sound great.

  4. 6 minutes ago, mickulty said:
    • Currently the DDR4 AMD GPUPI 1B stage would include mobile-only "Zen 3+" as optimal hardware.  If that's banned, it'll require Bristol Ridge which afaik is locked to 2400 max mem mult and a bit of a compatibility minefield.  Might be a good idea to ban mobile and reduce to 4 subs (zen 1/+/2/3).  Although I supposed it would be funny to see people benching mem on bristol ridge, so I won't complain if it's kept at 5 ?

    .

    If restriction is done by core lineup (like in the description), not generation, then we can do Cezanne+Vermeer+Matisse+Renoir+ for example Summit Ridge.

    That way, nobody needs to use BristolRidge

    • Like 1
  5. 7 minutes ago, Leeghoofd said:

    Nothing is final bro, also the UAT is only recalculated manually. So its pretty hard to see how balanced it all is. 

    Tim programmed this within a few hours and its pretty fast on UAT (which is super slow), but we still need to work on the globals 2D vs 3D

    Yes, we understand that this is only "demo" and it is nowhere near final. We just give feedback on its current state.

    8 minutes ago, Leeghoofd said:

    Don't you think it is time that user start to bench something else than rebenching the same hardware over and over again?

    This new point system is there to reward anybody that wants to bench. with the current setup people don't even bother as they will get 1 point.

    Yes, that is good for health of the system. But it is important to control hardware grinding, which may impact competitiveness of some ranking systems (hw-masters for example). *User* starts benching new platform = good. *User* does good scores and gets rewarded for that = good. *User* subs popular platform in stock and gets rewarded by a lot = not good. *User* subbing 100500 systems in stock to gain hw-masters position = bad. Just an example of a problem, which didn't exist with current system, but may emerge with new one.

    Rewarding low scores is ok, but reward shouldnt be too much in relation to actual top-scores. 1155 is good example of that, but we, currently, cant propose solution that would be good for all parties, including performance impact on bot

    12 minutes ago, Leeghoofd said:

    Thirdly the gap closes between LN2 and the non extreme users as the points variance is no longer dependant on the top score. Also the discrepancy between eg the top 5 is way less, again making it more competitive or balanced if you want.

    Yes, that is what we like about this new system the most ?

  6. After short discussion in NotExactInRound, we have come to some conclusions:

    1) The main goal of boints is to encourage benching and competition, and current formula update is improving that aspect. It brings motivation to bench different platforms, that were LN2 only (like AM3+, for example)

    2) Current "demo" system on UAT is really unbalanced in regards to point curves. Unpopular and mid-popular hardware gives too much for golds, really bad/weak scores (e.g. stock) give too much points, golds in popular hardware give not enough points (to differentiate them from less competitive rankings), global rankings are somewhat broken (e.g. vantage extreme, there worst 20 scores get ~15 boints just for their existence)

    3) Broken (in regards to points) platforms became even worse. It is 1155, the worst case of this. Popular platform, and really low score ceiling brings ridiculous point rewards for subs in those categories.

    Judging by statements above, we expect that current hardware masters would get less motivation to improve and sub new hardware (even reaching a point there their ~top100 subs have maximum points), while there may come new hardware masters, who would just stockpile junk and make ~stock submissions = non-competitive hw-masters ranking. However, we have expectations that after proper point curve adjustments, and/or implementation of better formula, there will be higher quality of competition in both hardware, global and personal rankings

    • Like 2
  7. 24 minutes ago, mickulty said:

    likely getting pretty much the same score

    On the other hand, these scores may become more different over time. Currently, disabling e-cores allows for higher ring clocks and a bit better p-core clocks.

    In some time, we may see 1T runs with, e.g., 1+1 configurations, for the best score in P+E category. I dont think that result would be directly comparable to 1P core result. Although, thats just a worse result than 1+0 in 1T by definition.

    From another point of view, E cores maybe be used for score optimization. *Speculation* Like we already have 3-4 core runs in SPI/old 3dmarks, those cores may be E-cores with seperate frequency control, which may allow to clock higher single P-core (1P+4E)

    But, I dont have any particular opinion on the solution for this right now.

×
×
  • Create New...