Massman

The second issue to address was the sudden decrease in motivation to bench old, rare and exclusive hardware because of the tweaked HWBoints algorithm that attempts discriminate between huge effort and low effort results. Although this algorithm does seem to work pretty good in the categories where there's a bit of competition (~20 participants and more), in the extremely low competitive categories the algorithm makes the curve very steep. The key issue is that this creates a a spiral of non-participance in the category: the ranking has no competitors, so the points are low. Because the points are low, less people are motivated to submit into the category. Because less people submit, the points remain low. To solve this, we have will tweak the algorithm a bit for these categories. Not to reward those who are first per definition, but to motivate people to increase the competition in that category. How does the algorithm work? The concept is a very simple one and actually comes from rev2 of the HWBoints algorithm. For all categories having <20 participants, we apply one and the same lineair point distribution. From the moment the category hits 20 participants, we switch to the HWBoints algorithm. The first column represents the points currently awarded when 10 people are competing, the second column how the lineair algorithm will distribute points and the third column what happens when the HWBoints-curve takes over. Note that this change will affect the points in the categories with 11 to 19 participants: the current top scores will see a small decrease in points initially. Our main objective is to make each category competitive enough to fall under the HWBoints-algorithm. We've added a poll to see if everyone can live with this. If you have any comments, don't hesitate to throw a reply into the thread. But please, keep it civil and understandible. For complaints, go here.

From the feedback we received after launching the actual HWBoints Rev3 product, we noticed two big problems. The first was the low reward for breaking the overall world records. Since pushing the limits at the absolute top of the benchmark rankings does require skill and effort, two of the main ideas behind rev3, not awarding big numbers indeed doesn't sound right. So, we have whipped up a fix that is supposed to, well, fix this. What benchmark applications? As the name already indicates, this is only applied to the 3D benchmark applications because that's where the main issue is located. Running 5 pots under LN2 (1CPU, 4GPU) is a hell of a work and should be rewarded accordingly if the score turns out to be great. Benchmarks applicable: - Aquamark - 3DMark01 - 3DMark03 - 3DMark05 - 3DMark06 - 3DMark Vantage One might ask, why not do it for all benchmarks. Well, single-category benchmarks, such as SuperPI and Pifast already discriminate the world records from the non-world records as they are all in the same ranking. In other words, the top score will get the most points. For Wprime, this could've been a solution as well, although this benchmark doesn't scale with skill per definition. World records are set using a highly expensive server; rewarding financial input is not a key characteristic of the Overclockers League. PCMark05 might also apply for this rule, but as far as we can see, the top scores are already awarded the top points as well. Points distribution? As the name already suggests, this is a bonus to reward 3D WR attempts, not to reward buying expensive systems. So, these points are awarded based on the world record page, which contains the results of HWBoints-participants users, the non-participating users and the scores using unreleased hardware which receive no points. ALL of these scores have an affect on the 3D WR bonus; in other words, if someone breaks the world record using unreleased technology, then this will affect the 3D WR bonus points. Why? These points are WR points. World records are what they are: world records. They are not based on anything related to 'being fair' or anything related to this concept. The rankings that are fair fall under the HWBoints algorithm. Now, points are awarded to the best 20 scores of each benchmark application, as an added bonus to the existing hwboints. The points are awarded as follows: We've added a poll to see if everyone can live with this. If you have any comments, don't hesitate to throw a reply into the thread. But please, keep it civil and understandible. For complaints, go here.

How ... did I miss that. I suppose the difference will be quite noticeable?

Are you sure the Evga is supporting the Q3QP? There's a difference between Gulftown and Westmere-EP variants of the hexacore technology.

Top-20 requires verification links. You can make about 5 on one account without buying any licence.

Tomorrow VMOD

Very nice memory clock!

As I expected ... memory is not great. I'm running 285MHz stable max. Is everyone still running stock voltage?

I think VGA memory isn't so good. The card is the same like in these scores: http://hwbot.org/result.do?resultId=556169 Only 290MHz on memory :-/

I thought this was dead

Looks like backup games are in progress

Oh, must be a new bios update then. I had the PT880U based board for a while and a Wolfdale-based CPU was locked in multiplier x6. Couldn't change the MP, so ... :-/

Awaiting Vesuvius ...

Hey guys, don't forget to indicate the brand of the VGA card you're using ... always nice to see the stats on that

They support it, but not the 333FSB ones.

Damn :-/

You guys need me to bring out the AthlonXP?

We have our in-house designs ready for a league as you described. However, due to recent events in other, similar global competitions, we prefer to not make something that isn't supported by all sides of the overclocking scene: manufacturer, overclocking participant and overclocking spectator.

You can run at 1 core, but the hwbot engine will always classify that CPU as 4xCPU.

For BUG reporting, please post in the bug thread. I'll keep this one open for complains, questions etc BUG thread: http://hwbot.org/forum/showthread.php?t=4579

BUG: - http://hwbot.org/rankings/benchmark/wprime_1024m/rankings?cores=2 => 3oh6 is first - http://hwbot.org/result.do?resultId=939192 => Score is ranking #2 and points indicate #2 same issue here: http://hwbot.org/result.do?resultId=935049

Using the word "fake" can upset a user

Not faked; user switched numbers 4 and 1

Quadro4 != Geforce4

http://hwbot.org/browseHardwareVideoCards.do?gpuFamilyId=4 The engine allows cards visible on this page