Attention Overclockers and Benchmarkers: new features at HWBOT !

[ozzimark]

# results today: 149

Outdated you say ?

then how come my 3400+ sempron is grouped with the s754s, as a palermo with 2ghz stock speed

and imo, this new layout is a good way to push the little sites out of sight. not conductive to proper conpetition if you ask me

[massman]

then how come my 3400+ sempron is grouped with the s754s, as a palermo with 2ghz stock speed

If you run into troubles finding the correct cpu, please give our forums a visit and let us know.

[jmke]

and imo, this new layout is a good way to push the little sites out of sight. not conductive to proper conpetition if you ask me

I don't agree, a small group of OC'ers can get 1st spot in the ranking and climb very high; what we need now is a member to HWBOINT ratio for a third ranking

hey RIchbastard! HWBOINTS/MEMBERS = Average OC-bility of Team

can you throw together a page with such a ranking, that would be swell

[RichBa5tard]

Nah, that would discourage teams from accepting new, non extreme, overclockers.

Imho being a big team is a great achievement on itself. We shouldn't punish them for having a lot of members, which the hwboints / members score would certainly do. Small teams can still get a high rank in a benchmark specific ranking (which is top 10 avg).

I could add the hwboints / member ratio to the ranking, but purely informational.

[Dani]

(Thx to Mtzki for the brilliant idea)

Yeah, our big boy from Finland

[ozzimark]

If you run into troubles finding the correct cpu, please give our forums a visit and let us know.

i just did let you know. no need to register again to tell you the same thing (edit: just realized that probably comes off negatively.. didn't mean it to! )
WeStSiDePLaYa is sorta right though. while there are AM2 results, you're missing a lot of the chips at the moment.. kinda surprising considering the platform has been around for almost 6 months now

I don't agree, a small group of OC'ers can get 1st spot in the ranking and climb very high; what we need now is a member to HWBOINT ratio for a third ranking

hey RIchbastard! HWBOINTS/MEMBERS = Average OC-bility of Team

can you throw together a page with such a ranking, that would be swell

i though of that last night, but with how the scores are set up now, it skews really heavily against the big teams. what needs to be done is just less emphasis on the global points. while making it so that members with lower scores can contribute is fine and dandy, i see it as a lowering of standards. a dilution of the "extremeness" of it if you will. after all, i shouldn't be able to help out my team by submitting all the random scores i have laying around on my computer from a year or more ago

[RichBa5tard]

Phase at OCX suggested a very interesting alternative. The member ranking based on hwboints seems to work very well, the team ranking however does indeed favour huge teams a tad too hard.

The global points are not a very good system of scoring. Just show individual hardware rankings like you have been.

Only the top ten scores should count in the overall Team Ranking and only your top scores should count for individual rankings otherwise peeps will just keep submitting scores to increase their points. And those teams with more members like XS will always be on top. It must be a level competition or it becomes arbitrary in the ranking. Giving value according to how many members have submitted scores should not reflect the overall value. What does that have to do with top scores?

If there is not as many submissions then peeps should just download the application and run it, they would have a better chance to get a higher rank than some other benchmarks. The system you have been running is great why change it now? Unless you start over with a new competition it would not be fair and any new teams would not have a chance at all. The points should be awarded for 1st through 10th with a overall weighted score accordingly ie> 10,9,8,7,... points for each place in a different bench with all benchmarks having an equal weight in scoring regardless of how many scores are submitted. Then averaged overall with the top 10 scores of each team. Then you can break it down into hardware specific scoring. Just like you have already done for the last year.

Also in individual rankings it should be the highest score "only" of your top 1st-10th place scores in each benchmark ie>Super Pi 1m,32m,3DM01-06 that reflects your overall ranking, not scores from a year ago that are arbitrary compared to the new top scores. Those older scores should just be used for a lifetime ranking of some kind. Otherwise whats the point? You may as well use an old out of date CPU and become #1 in the world, even though your scores are not even top 1000 of the true world record. Old out of date hardware, should just be a separate competition all together or just a nostalgia competition of some kind. You do not run last years Ferrari at this years F1, right?


Every bench should be scored accordingly for every position you get. The points for that position should be awarded to each team (Or Individual) according to rank.

1st place = 10pts
2nd place = 9pts
3rd place = 8pts
4th place = 7pts
5th place = 6pts
6th place = 5pts
7th place = 4pts
8th place = 3pts
9th place = 2pts
10th place = 1pt

So the current standings would read like this>

XS
Benchmark ranking (current rank - weekly & monthly change in ranking)
* 3Dmark 2001 rank: 1 - - [ list scores ] - [ ranking stats ]
* 3Dmark 2003 rank: 1 - - [ list scores ] - [ ranking stats ]
* 3Dmark 2005 rank: 2 - -1 [ list scores ] - [ ranking stats ]
* 3Dmark 2006 rank: 1 - - [ list scores ] - [ ranking stats ]
* Aquamark rank: 1 - - [ list scores ] - [ ranking stats ]
* CPU-Z rank: 4 +3 +4 [ list scores ] - [ ranking stats ]
* PCMark 2005 rank: 16 -1 -1 [ list scores ] - [ ranking stats ]
* PiFast rank: 8 -1 -2 [ list scores ] - [ ranking stats ]
* SisoftSandra rank: 11 - -1 [ list scores ] - [ ranking stats ]
* SuperPi rank: 1 +1 - [ list scores ] - [ ranking stats ]
* SuperPi 32m rank: 14 - -1

Overall score = 69


OCX
Benchmark ranking (current rank - weekly & monthly change in ranking)
* 3Dmark 2001 rank: 2 - - [ list scores ] - [ ranking stats ]
* 3Dmark 2003 rank: 2 - - [ list scores ] - [ ranking stats ]
* 3Dmark 2005 rank: 1 - +1 [ list scores ] - [ ranking stats ]
* 3Dmark 2006 rank: 3 +2 +1 [ list scores ] - [ ranking stats ]
* Aquamark rank: 7 -1 -3 [ list scores ] - [ ranking stats ]
* CPU-Z rank: 3 +8 +12 [ list scores ] - [ ranking stats ]
* PCMark 2005 rank: 9 +21 +18 [ list scores ] - [ ranking stats ]
* PiFast rank: 21 -1 -2 [ list scores ] - [ ranking stats ]
* SuperPi rank: 2 -1 - [ list scores ] - [ ranking stats ]
* SuperPi 32m rank: 1 - -

Overall score = 68

That is the only fair way to do it.

[mtzki]

The suggestion for the team points by Phase above is pretty much the same as my original plan. Only that all the teams getting a score in a ranking should get some points, not just the top10. The decision to go with the kind of team hof we have now wasn´t exactly unanimous.

The big team bias could also be diminished by summing only the top n overall personal scores for the team (instead of all like now), with n e.g. 20.

[wittekakker]

Great stuff.
I hope to see some nice team fights.

It would be nice to have a program that is able to extract benchmark scores out of the avaible tools. I remember guys who developed something familiar for Need For Speed Underground 2. You had to put on the extract software, then run your game, set your new track record, alt-tab back to the extract sofware and submit it online. They even build in anti-cheat! You could add a possibility to safe your record to a .hwb file or something like that for people who might not have a internet connection at that time
I know it's allready on the to do list, but I think this way might be the best (most liked by tweakers) way to do it, allthough it aint going to be that simple to do

[massman]

We would need support from SiSoft, FutureMark, Aquamark holders and XS to do it decent enough. Without a anti-cheat build in, such a program is worthless imo

[wittekakker]

I wouldn't count on that one. FutureMark for example will not share their knowledge with any possible competitors (that's what I think). Aquamark should be possible. When you download Aquamark you will find a readme in the instal directory wich explains a few points. Bundle forces with the CPU-Z programmer would be cool, and off course having your own MultiPi benchmark program.

[massman]

1. Futuremark:

I wouldn't count on that one. FutureMark for example will not share their knowledge with any possible competitors (that's what I think).

Exactly. They won't be very happy if we made a program to submit their results to our database.

2. Aquamark:

Aquamark should be possible. When you download Aquamark you will find a readme in the instal directory wich explains a few points.

Believe it or not, we have already discussed a way to attach the AQ3 program to our database in order to let benchmarkers submit their results fast and simple to hwbot. We didn't find any (decent) way to do that.

The total score is based on the average frames per second which is the total number of frames divided by the time the benchmark needs to complete. The AvgFPS value is multiplied by 1000 and
the result is the AquaMark3 total score. The value of 1000 has been explicitly chosen to scale the result on high-end desktop PC at the time of release to a value of approximately 50 000. The CPU and graphics scores are based on the AvgFPS CPU and AvgFPS GFX described in the previous chapter. These values are scaled by 50 and 100 in order to be both between 6 000 and 8 000 on a high-end desktop PC.

I still have no idea of how we can build in the anticheat protection, athough there is a hidden checksum in the file you get after the bench ends:

click here for <a href="http://arc.aquamark3.com/forum/showthread.php?s=&threadid=46">AquaMark Frequently Asked Questions</a>
<br>
<br>
<input type="hidden" name="SubmitID" value="{D63C970F-9CAA-486F-973F-6B9DE1979529}">
<input type="hidden" name="CheckSum" value="Yx8oRpUaTsv5TrujtwNa8w==">
<input type="submit" value="Submit">

3. MultiPi / Cpu-z:

Bundle forces with the CPU-Z programmer would be cool, and off course having your own MultiPi benchmark program.

MultiPi is not our n°1 thing to do. A Cpu-z screenshot should be easy to attach to our database, but I don't think Frank really wants that. He has his own database .

4. SiSoft:
It seems that they have included a way to upload scores to an online database in their program, so maybe ...

[wittekakker]

Hwbot mods mark my scores as being legal. But what if I change them after you guys have mark it? I don't see it reapear on the main page as a new score being posted... so do you guys get a notify that I changed my score?
It doesn't get the 'unchecked' status either.
So basicly, if users change their benchmark score it would be cool to put them back unchecked.

[RichBa5tard]

Hwbot mods mark my scores as being legal. But what if I change them after you guys have mark it? I don't see it reapear on the main page as a new score being posted... so do you guys get a notify that I changed my score?
It doesn't get the 'unchecked' status either.
So basicly, if users change their benchmark score it would be cool to put them back unchecked.

A very good point. Added to the todo list.

[jmke]

we don't necessarily check all scores as they are submitted, rather the top 20-30 in each ranking to see if nobody is trying to cheat their way to the top.

We should be able to differentiate between accidental errors and scores entered wrongly on purpose; for the last ones we should be able to sanction or "tag" the offending user to help prevent any future problems.

[Turtle 1]

How do you register @ hwbot. Link please. Never mind I got registered. I will ask wife if I can register her scores when she gets home.


So I still don't understand the scoring. Does XS get points for me . with a P4c 3.2@3.4 in 3Dmark 05 . My browser pc scores in the number 1 position and in pcmark 04 I am second. Now does this give points to XS or hows it work.

[massman]

Your score will add about 2 points to the XS totals

EDIT: It's more like 6 points

[wittekakker]

I've found a 7800GTX score today wich seem to be fake. The thing is that I cannot report it because one of hwbot mods has put it as valid. How can I report that it is fake?

Onto the fake result: it has a Futuremark ORB link, but the 3D Mark 05 GT3 score is just way off compared to other 7800GTX. It even beets a 650MHz 7950GT with him only having +500MHz on the core. Where he got 12k, he should get only 10k considering his system specs. I also tried to contact Futuremark about this, I hope they delete the score, but I don't know if they will hear my, they don't have a "report" button on their site so I just picked one of their available e-mail adresses

Here it is: http://www.hwbot.org/compare.do?resultId=550240

[jmke]

that entry doesn't seem to be fake, or am I wrong?

check my new sexy signature image, it's available to all HWbot.org users, click the "bench profile" link after you've logged in.

[jmke]

here's yours:

[wittekakker]

Damm, you passed me, need to check my HD, maybe there is more :p


--------
About that entry. Look at his score on Futuremark ORB:
http://service.futuremark.com/compare?3dm05=2150955

He has Atlon X2 @ 2,65GHz and 7800GTX @ 560MHz.
My setup: Allendale @ +3,3GHz and 7950GT @ 650MHz.
I should 'own' him on one leg

Main Test Results
3DMark Score
(3DMarks)
12135 <-> 12756


Detailed Test Results

Game Tests
GT1 - Return To Proxycon
(fps) 54.5 <-> 46.1

GT2 - Firefly Forest
(fps) 36.9 <-> 37.0

GT3 - Canyon Flight
(fps) 56.9 <-> 77.9


The big differance on GT3 tels us he has bigger gpu performance then me.
Now look at this hwbot entry:
http://www.hwbot.org/compare.do?resultId=506615
Although it's in the single card competition, he descibes his setup to have two 7800GTX's. Now look to the results above, we see a patron

I didn't saw it before, but it's more clear now. It's an SLI score, not single card.

[massman]

You're absolutely right Geoffrey, sorry for the mistake.

From now on, this score will be known as SLI

[rjw]

How do I correct problem with the hwbot?

I posted a 3dm03 score, which ended up being put in the siungle card area as opposed to the dual card area.

Now it has beeen removed altogether?

[mtzki]

You can edit your submissions at the hwbot website:
http://www.hwbot.org/user.do?userId=6154&filter=false

I think your 03 score is still there and fixed now.

[rjw]

You can edit your submissions at the hwbot website:
http://www.hwbot.org/user.do?userId=6154&filter=false

I think your 03 score is still there and fixed now.

It is not listed except for in my proile with an n/a next to it.

Ok, I edited it. I don't know why the bot picked it up ioncorrectly though?

Question...Will my signature update automatically, or do I have to go back and change the url once in a while?