Jump to content
HWBOT Community Forums

I.M.O.G.

Members
  • Posts

    869
  • Joined

  • Last visited

Posts posted by I.M.O.G.

  1. 1 150L Dewar (tank) - *Tested* - Vacuum is good (it holds cold well, walls don't sweat/frost). This was my primary dewar for the past year or two, and it works great.

     

    Heatware Feedback is 35-0-0: click here

     

    This is the 3rd unit of this exact same model I have sold. One was sold to Janus67, the other was sold to icebob... This is my personal dewar and has been filled with over 2000L over the past year or so. It holds very well.

     

    Description and Price:

    1 150L Dewar (tank)

     

    • $400 + shipping

     

     

    ================================================

     

    Shipping:

    • Shipping Methods:
      • Freight (200lbs, 20"w x 58"h)

      [*]Local pickup encouraged, shipping cost could be $200-$500 (or more)

    Payment Methods (in order of preference):

    • Paypal
    • USPS Money Order
    • Cash

  2. Ya, people may be less likely to fill in all the fields... That is a downside. I often fill in more than I need to, to help myself later so I remember what I did and how.

     

    It also may not be ideal to sort of trick people into filling it all out. Maybe achievements or other motivators can be used to encourage filling in as much detail as possible, and bring more attention to why supplying the data is useful.

     

    I dunno. Would be cool if there are other ways to get that data. I think the pre-populate fields logic was a major step forward. Making new submissions easier for newbs probably isn't that big of a deal, but could be a good thing if it gets more people hooked.

  3. The most common challenge I see for new members is filling out the new submission form.

     

    It is overwhelming, there are dozens of fields that can be filled in and its not painfully clear which ones are optional and which are required.

     

    In reality, there are only a few required fields.

     

    Submitting would be more accessible to new people if they were presented with the required fields in its own section at the top of the new submission page, and then optional fields were below in their own section.

  4. I think some of you underestimate the skill needed to be 1st out of 10-15. It's just as hard as being ~10th in a very popular hardware ranking, even though the reward is much less. FOr no-competition stuff it's different, but you don't need much competition before it gets alot harder to be number 1.

     

    I agree sometimes, but probably not most times. I think if we could figure out how to also account for who you beat, this situation could be improved. If you are 1 out of 10, a lot of the time its a bunch of noobs who aren't sure what they are doing. But other times, there is contention for top ranks against one or two other skilled players in the same ranking. If you beat someone else with a lot of gold cups, maybe you should get some bonus points. (Goes back to my idea of not just how many you beat, but also who you beat)

     

    Mainly, where I agree with you for example, is that what you do very often by taking golds in low competition categories is often more challenging than taking 40th on an 8800 GPU. Yet, 40th on an 8800 only takes some cold and a modern CPU, and it awards almost the same number of points as you get for 22nd. It isn't hard to do that on an 8800, probably less hard than taking some gold cups. But I don't see a better solution for that, which is straightforward.

  5. by penalising folks for taking an alternative route just seems counter productive.

     

    1 the easiest way has been discussed previously on numerous occasions. reward folks on the percentage of oc achieved. much like a cpuz val apply the principle to other benches. one of my proudest moments was getting my old e66 past 100%oc. anyways, it's a nice idea that's never going to happen.

     

    I don't think we are penalizing alternative routes, just rewarding oddball setups with fewer points - there are a ton of leagues, rankings, and rewards that can be sought after... If the only reason people are benching certain things is to get the 60 globals, then motivations are in the wrong place. This only seems like a penalty, because by unexpected side effects of the old algorithm, it provided ways to exploit massive points which weren't intended - edge/odd cases hadn't been taken into consideration previously. This makes sure there is a reward, but also that the reward isn't totally out to lunch.

     

    Percentage overclock is no good. It would be great if benchmark security were at a higher level, and if monitoring/recording were integrated in the benches, but we don't have any of that. This isn't realistic due to the limitations of screenshots and cpuz being the fundamental method of verification... Easy to change frequency for screenshots. Also frequency doesn't account for efficiency, also an important element of skill.

  6. Just scores like this warrant extra verification. A lot of people have spent a lot of time looking at efficiency, and results like this break conventions of what is known to be possible.

     

    I have ran a lot of wprime on a few different platforms and know it to be very strongly tied to core frequency, and only minimally tweakable. OS selection makes an important difference, but less than 100ms. This score is 300ms faster than anyone else running 5300mhz on a 2600k, and only 60ms slower than my 5750mhz 2600k submission. From my experience, at a glance it makes me say something isn't right.

     

    If its win8, the pros would be on it and everyone would be talking about the efficiency change... Submissions would skyrocket as people rerun on win8. I don't see any of that happening.

  7. Sadly, points are mostly rewarded by the number of "noobs" you beat, rather than the actual skill required to beat your score. Not only that, inactive rankings get devalued simply because there are few recent subs, not because the submission skill level has decreased.

     

    2 things:

     

    1. Higher participants increase probability that one is competing against many non-noobs. Lower participants increase the probability that one is the only one playing who knows what he is doing. With the breadth of hardware and variety of rankings and rewards, I think this algo update is a step in the right direction... It isn't perfect, but it isn't too difficult to do, and its a more logical way to reward participants that scales both with high or low competition. With so much hardware and so many rankings, I can't think of a feasible solution that would be better at addressing actual skill.

     

    Well, I can think of another solution, but it probably wouldn't be feasible... Instead of only awarding points based on number of participants, the algorithm could also account for the quality of participants you beat. For example, if you beat 9 other participants, that would be worth 10 points. If 2 of those participants were ranked highly within their league, you get additional bonuses because you beat other skilled Overclockers. Probably too complicated to implement, however it would address both the amount and quality of competition.

     

    Ultimately though, I am the opposite of rasparthe on this issue. I don't believe awarding the same points for beating 2000 people or for running hardware no one else bothers to run is any good, especially when it rewards stock clock submissions that rank people higher than others legitimately trying to overclock.

     

    2. How do inactive rankings lose points? Is there a decay element in the algorithm? When does point decay start? I have never seen this personally - I've only seen points decrease when my submissions are beaten.

  8. Super efficient, how did you do that?

     

    Probably the same way he did the pifast result the other day that I reported.

     

    Couple possibilities:

    1. He found a BIG but legit tweak that is worth several hundred mhz extra core speed in cinebench/pifast

    2. He is using a cheat to submit super-efficient results that are run at several hundred mhz slower than everyone else

     

    The super-efficient pifast result was deleted so its probably possibility #2: http://hwbot.org/submission/2396554_

  9. I noticed something pretty strange yesterday. Boot into windows with 100 BCLK x 35 Multi 2000 Mhz Memory. Result was 360 Points. Increased clock with TurboV to 104 BCLK and reran the benchmark. Still 360 points. Closed benchmark, ran again and again 360 points. Tested several times even after a restart the result was the same.

     

    Boot into windows with 104 x 35 straight pushed it to 372 points. Now raising the BCLK to 107 the system still runs stable but the score stays the same. Pretty weird.

     

    Testing I've read so far has been similarly weird also. You aren't alone on this sort of behavior.

  10. $250 shipped USA, contact to inquire for international shipping.

     

    Selling my REX. It is in excellent working condition, prepped for subzero (petroleum jelly) - strap a pot on and its ready to go.

     

    The Northbridge is cooled with a copper heatsink and compound fan assembly, which screams - give it as much voltage as you like, temps will be happy. Additional copper heatsinks are present where needed, southwest of the CPU socket. Includes the LCD poster display as well. I haven't ever tested the mobo for maximum FSB, but 638MHz FSB was the farthest I have been able to push a chip on it: http://valid.canardpc.com/cache/screenshot/2333587.png

     

    Excellent, fun board. I am crying inside as I type this but I don't have time for it anymore. Please give it a loving home.

     

    [ATTACH]1536[/ATTACH]

     

    [ATTACH]1537[/ATTACH]

×
×
  • Create New...