Jump to content
HWBOT Community Forums

RULE

Members
  • Posts

    561
  • Joined

  • Last visited

  • Days Won

    1

Posts posted by RULE

  1. There are a few issues with this... for example, you define efficiency where the only relevant measure is the CPU speed, which is wrong. Efficiency must be measured for the system as a whole, if not there's more or less how well the other parts perform - fast ram vs slow ram, some mobos are more efficient than others etc. The other issue is that efficiency improves with a lower CPU multiplier, so someone running 6000Mhz WILL have better efficiency than those running 6600MHz if the only difference is the CPU multiplier.

     

    I don't mind having a ranking like this for fun, but not for points.

     

    I'm talking about global efficiency.

    Basically run better ram and better mobo give you best efficiency.

    Exemple if someone run the ram 1100 mhz vs 1300 mhz, the efficiency is already reflected in the frequency*time.

     

    The problem is that cpu frequency and superpi time don't scale linearly. For each increase of frequency, your time improves less. As an effect, you will see the best efficiency at lower frequency. Similar in principle to the law of diminishing returns.

     

    One solution would be to improve the efficiency formula to account for the diminishing returns.

     

    I'll study the Diminishing returns later, interesting stuff.

     

    Also Rule you have shimizu CPU speed wrong,you have his CPU at 6600mhz his speed was 6660mhz,did you recalculated his effieincy at 6600 or 6660

     

    You're right mate, i'll fix that later. You are on top now and deserve it.

     

    *Perfect example - try to input the recent top 32m submission from MOA 'A' class into the above table... it seems there would be a new leader on the efficiency table.. and with quite staggering results/settings by contrast.... .....

     

    This Efficiency ranking for me have more sense apply at the qualifying contests, because at the final is more important have efficient overclockers instead of lucky overclockers. Imho there's the 32m in all the contests.

  2. Hi everyone,

     

    some people work hard for have good efficiency in our benchmark result, but they can't be always competitive, not all the hwbot member can't spend the same quantity of money.

     

    I wanna discus with you about this idea:

     

    5xSLoZY.png

     

    I'm sorry if some value on the table is incorrect, the example is based on 4770k spi32m rankings.

    Efficiency is = frequency(mhz)*time(s)

    Improvment rankings is = original rankings - efficiency rankings

     

    Of course the Efficiency rankings can only apply for hw point not global, because you can't compare the frequency of different architetture.

     

    Spi32m is the best candidate because without bugs.

     

    The only problem i see is easy cheating lowering the frequency before save the screen.

    Is also probably on this table someone have lowered the frequency for save the result.

     

    I wanna discuss this with the comunity because is important catch new member, but is also important work for keep the overclocker on hwbot.

     

    Is also important for some sponsor, sponsoring efficent guys is more probably some WR, if the WR is the point of sponsorship.

  3. Important Update: there will be no LN2 bios and no Afterburner Extreme for the GTX 750.

     

    The reason is that the procedure to make 40 LN2 BIOSes (that's how many models there are worldwide) is too much work and the LN2 bios only increases the power limit slightly anyway. It should be possible to max out the card even without the special LN2 bios. For the Afterburner, it comes down to the PWM IC used for the GTX 750 doesn't support voltage control. So even with a special version, there wouldn't be any gain voltage wise

     

    So yeah ... go crazy on the PWM modding! :D

     

    There's the nvidia gpu clock limit, we need homemade bios to bypass it....

  4. I realy don't understand, why is all bugged if is all inline with cpu frequency?

     

    http://hwbot.org/benchmark/ucbench_2011/rankings?start=5&hardwareTypeId=processor_1741&cores=4#start=0#interval=20

     

    If you post a 4400 mhz with an higher score than a 5200 mhz can be strange, but we are all inline with frequency/score.

    In Hwbot rules there's something about 'etic' and nobody here have broken that rule, unlike you that when you found out how to have bugged score you posted a lot of invalid results.

    so stop with this submission report...

  5. Guys I know that this question is a little "stupid" for many of you but...

     

    I already have the Z97 + G3258 + GTX 750 + SSD + RAM. Do you guys think than a good 850W PSU can be enough for run that hardware and do extreme overclock?

     

     

    which card do you have chose buddy? 1gb/2gb?

    You already have one on your hand? can tell us which vrm controller use?

  6. You at least know how to use those ePower, this would be my 1st time with GPU under sub-zero temps.

     

    Yes i know how to use the ePower, u need some skill to solder it on the board. Is not a matter of first experience sub-zero but some experience with soldering iron.

    If you are interest can read some thread on http://kingpincooling.com/forum/forumdisplay.php?f=31

     

    That's one of my jobs http://kingpincooling.com/forum/showthread.php?t=2671

     

    i have a feeling 2gb will be king, but its just a feeling. i hope not cause only 1gb in sa lol!

     

    the 2gb card have a poor alimentation on gpu, ePower need for sure.

    Msi release ln2 bios for all the cards? That's important.

  7. So we in Class B are looking at

     

    N750 1GD5/OCV1

    N750 2GD5/OC

    N750 TF 1GD5/OC

    N750 1GD5/OC

     

    is it unlikely that we could expect a "Hawk" or something to be released with a decent VRM?

    or more likely that this is it for the product range and thats it?

     

    I think the best choice can be N750 TF 1GD5/OC + 2 EVGA ePower. I don't know if 2GB can help in firestrike.

×
×
  • Create New...