knopflerbruce

Checksum is enough. You can easily see if 3 or 4 cores were used, the time difference is HUGE, way beyond any tweaks:p If two submissions get the same score, one cooled by LN2 and "push Ok and run"- style, and the other is tweaked to the limit on air, and get the same there is no way we can reward them differently. Skill may not be correctly rewarded in all cases. But by setting the limit to 50%, and making the rewards relatively small for the first boint giving submissions we reward those who can tweak their air cooled systems as well. However, I don't see why it's not possible to just put some ice cubes in a bucket with water, and make a water loop with no radiator, and then get an advantage that way instead of using a stock cooler:confused: Cooling skills should be rewarded, too. Sure people have benched after rev. 2 before, but it's not THAT different for grinders: buy cheap HW and max it out, getting as much boints as possible. The onyl big difference i can think of is that before you get automatically rewarded if less than 20 users benched (rewarding low skill, as it takes no effort to get boints), and now you have to beat a few guys to make it.

The problem with rare HW is that it's hard to make an algorithm that takes care of the cases where most people run on their daily rigs and at stock speeds more or less, and where the top bencher(s) have really maxed their chips and reached great scores. Lippokratis had a couple of sli/cf scores (not THAT many, but probably quite valuable in rev. 2 because of the bug that HAD to be fixed, as it was rewarding 3d-benchers with WAY too much boints than they really should have had). Example of why rev. 3 is better than rev, 2: http://www.hwbot.org/result.do?resultId=777212 CPUZ validation: 6/13, now 0.1 boints - before... maybe 2.0 or so? 5.3% OC... IMO 0.1 is way closer to the truth than ~2.0. "lots of time" is also hard to define. If you spent 6-7 hours straight trying to get that validation and perhaps a quick superpi run, and risk killing it several times in the process by trying different means to really max it out, then I agree - it's worth a fair share of boints. Is that what you meant by "alot of time"? Or do you refer to the overall time needed to get the complete collection of results Lippokratis has? The term "easy boints" is also quite weird... why should there be such a thing at all? Just run the bench, tweak nothing, do nothing, and get boints... not something to reward in a competition. You may get some achievements, though (if you have enough submissions). That's enough if you don't manage to sneak into the top 50%. Getting boints means you've done something that's not so bad - not that you've pressed the start button and ran the benchmark. About the slow boint degradation discussion I'd like to add that if you beat the second place by a mile, you certainly deserve a bit more boints than 10% of the 2nd place, and the same if you've spent some time tweaking to be able to beat whoever is first. Normally, the top spots are pretty good - and beating them should be rewarded properly. The best thing you "rare HW complainers" should do is to bench more - the more competition, the more boints for the best scores, and more available slots for boints.

There's no doubt this is a valid result. Rejections don't really mean alot, only if the score itself is weird in some way...

In some cases they are OK.

As long as the cards are actually comparable in terms of performance (both stock and overclocked), it's a good idea. But if the 512mb memory overclocks alot worse than 128mb, resulting in worse scores for the 512mb cards, there should be separate categories.

The whole top 5 is too much, being rewarded for being last is just stupid. 3/5 is good in some cases, others not. With 2/5 only it will be the top 40%, and 60% with 3/5. IMO it's better to reward 60% than just 40%, as there may be real competition between the top two, which would leave the rest without boints. Sure "bad" scores will be slightly more rewarded in some cases, but I think it's a better "sacrifice" than the other, which won't encourage benching those categories at all.

It's still visible for mods I think as it's not actualyl deleted. Post a link and we'll take a look.

Sounds strange that someone else than a moderator can block the score...

As posted a handful of times already: there are a couple of bugs that affect the hardware boint rankings, plus the complaints are all over the place. Some people think it's all about the money now, and some think $2000 rigs have no future here... it's obvious there are alot of misunderstandings - both of these statements can't be right!

I did it, after this thread was made:p

Lmao, that last one is PURE fail. We were on first spot in the teams ranking - by a MILE, and now we're second... We also had 5 out of the top 11 in the global rankings, now we have 5 of the top 14. I believe we had one more in the top 20 as well, before the change. We don't gain much from this, at all. And FYI there has been very little talk about "omg the system is so unfair, i wanna quit, i dont see the point anymore", and more in the direction of: "what do we have to do to take advantage of this new system?". I must say it's kinda weird if it's the lack of "giving-up" attitude that will lead us to the top...

I must say I don't understand all these complaints about rev. 3 being about $$$. For the millionth time: now you can bench older stuff and get lots of hardware boints in return (compared to before, where you would need a good chip and LN2 to compete in the popular categories), bench a cheaper single-core GPU (instead of a MUCH more expensive x2-GPU, or even worse - quad-GPU setups, which was the case in rev. 2). The only place where we've actually removed boints are for rare hardware, and there are just too many personally related comments that all can be about those (or those CPUs/GPUs would be more popular than they are ). What would be good to know is where you can extract "rev. 3 is for big bucks only" out of the changes, when this was what we focused most on when we made this new revision. As far as I can tell, you need less cash now than before to compete. Maybe more skills, but that's without a doubt a good thing. As long as the "common bencher" also posts scores, the engine will automatically reward those benchmarks with more "slots" that receive boints, too - which will benefit the air/water benchers alot. The increase in number of boint-receiving results in general is insane, so it will be easier to get a little bit of boints than before.

There was also posted a replicable superpi 32m score that was WAY out of line. So... we must remove anything except PiFast, CPUZ and AM3:p

There are a few bugs around... SOME of the low boint issues may be because of that. 50% of a hardware category will receive more than 0.1 boints, if the system works properly. Don't come here and tell me that half of the results are made on sub-zero cooling! In fact, air and water cooling benchers are the ones we thought about when we made the new rules... before only top 20 got boints, no matter if there were 10 or 1000 results, now that number is increased for most categories. There are MORE hardware boints to fight for than ever before, but of course you can't expect to be generously rewarded if you don't show some skill as well... Overclock the crap out of whatever you have, and see how high it goes.

Could be a bug, I found some similar stuff in the A64 rankings. Is it a category for an old CPU or GPU? And I mean more than 0.1, yes (like 0.2 and above )

If bugged scores are enough to remove a benchmark, we need to kick the whole FM series from HWBot:p If the bug is obvious, it's no problem.

If you look at the same rank, that's not true. The best 50% get HW boints, so 7th and 12th should get the same 0.1 boints, but 5th will give a bit more.

As long as people haven't been benching "rev 3 style" yet, I don't think it's a good thing to start complaining to soon. It's a different approach, and naturally the ranks may be a bit weird in the beginning, until people start benching in January;) Edit: 1000 posts before 2010:D

14 boints for running alot of stuff at stock was too much. And I know what I'm talking about, as I did get alot of cups by benching VERY unpopular stuff. IMO 1.5 for being first when there is no competition is far from being too little... Plus, there was nearly no difference between 1st and 3rd when you had few results in rev. 2. In all competitions where you're given points for your results, you always get much more for being 1st than for the others, even if you compete with 100 others. I can't imagine anyone been given 75% of the winner's boints when they ended up being 3rd out of 4 participants, or something like that. Edit: everyone had a chance to see the test server, and to make suggestions for this revision. Those who hate this revision and never posted their thoughts when the test server was online etc should remember that until rev. 4, and not make the same mistake again

Exactly. It takes an effort to be first, and that should be rewarded. Receiving 75% of the reward the best result gets when you're #5 out of 10 is far from fair. Perhaps a better approach would be to reward by comparing the results to the average, or something... if you're 20% better than 2nd place (and even more to the rest), then you should get a much higher reward than if you're just 2ms better in some test. That would be rev. 4 material, though...

http://www.hwbot.org/listResults.do?cpuModelId=256&applicationId=3&filterUser=true&filterBlocked=true&limit=100 ~55 scores from different users, still just 3 boints for the top spot.

http://www.hwbot.org/result.do?resultId=878680&tab=info - no verification links... My guess is that this score was done using the checksum validation, but that info isn't visible anywhere;) Also, there is a bug in the wPrime 1024m tricore rank - alot of unlocked CPUs are there, but when I try to moderate them I realize that they are already identified as quads by the engine, so they shouldn't really be there at all... This is not an issue with the 32m tricore ranking. btw.

If that's the case, maybe you can unblock my blocked GF6800 score? Always wondered how a score that's 9/10 can get blocked because of missing resolution:p link: http://hwbot.org/compare.do?resultId=690179 On topic: IMO that score shouldn't be blocked if it's not better than what you'd expect from an untweaked, similar system. However, someone once told me that mods should be expected to follow the rules much more strictly than other members... can't remember who that was, maybe Massman?

If the performance differs from the 512mb version, there will be made a new category, if not - just post in the 512mb one.

It's a tgood idea to add the ES categories, too - but they won't get much boints if there are no retails they can compete with:p