I.M.O.G.

All those places would be good. I was thinking on the profile mainly, and signature did cross my mind as it already can do "best submission". Team page would be good. There are already views in different places that show the newest or the best recent team submissions, but nothing for featured/favorite. If its a busy day for the team, lower scoring submissions are buried in both the newest/best recent views (cold OC'ers dominate a teams submissions list, harder to find other submissions that may be commendable but less point worthy). I agree about what would be ideal like you mentioned in the first part of your comment, and also about the limitations. I think letting users do a favorites/featured might be one way to reward difficult to reward submissions. Maybe not the best, or even worth doing, but just something that crossed my mind.

It would be cool if we could mark 1-3 of our own submissions featured, then they would be highlighted on our user pages - like maybe in a box vertically placed between where our points are and the tabs below. While some submissions are worth a lot and they get noticed, other submissions may be worth very little, but were still very hard and we may still want to show them off. If I were to think of my 3 favorite scores, I'd probably take the 8GHz result, the 7.74GHz sp32m (worthless, but highest frequency sp32m on record anywhere), and my global 4x UCBench record (also worthless). Would only add one query to the user page, and seems easy to setup. Just thought it would be cool, so mentioning it - I'm sure others also have scores they feel are cool and would want to show off, even though they aren't worth much.

You played with subtimings? That is what people have said about MA.

So it isn't a bug, it's just a performance boost that when applied, increases the score such that onboard raid outperforms $2500 hardware raid with 4GB dedicated ram cache. Understood. Try similar tests on Areca or IRST, and you cannot duplicate it. Only AMD, only with raidxpert. And only with disk benchmarks that don't error out when trying to test after making the changes. Doesn't make sense, but nothing with PCM05 does, so business as usual.

You are the only one crying, ironically, about other people crying. Everyone else is talking the way normal people do. Maybe you could join us? FWIW, I think sandbagging is fine, just not for me. I think the capture the flag/king of the mountain thing I suggested earlier would be interesting, but it would be more complicated to implement and it isn't really necessary. The current competitions are popular and fun.

I take a different implication. The current system is setup in such a way so as to encourage sandbagging - there is nothing gained by posting high and early. In fact, posting high and early is incredibly ill advised. In the current system, the only reason anyone would post high and early is to raise the bar, letting others know what your hand holds and encouraging them to push harder. Anyone posting high and early is either foolish, or dismissing strategy to encourage others to push harder. Not such a bad thing, and there are a lot of people on hwbot that say screw strategy, we're here in the spirit of furthering the hobby not just to win. This is only a function of the current hwbot system. It could change in the future. If competition points were won by "capturing the flag" for a specified amount of time, the new system would reward posting high and early, and encourage everyone to push harder as soon as possible... Rather than the current system, that encourages posting high at the last minute, to take the lead only as close as possible to the very end. "Better" is always a never-ending debate however.

Easier said than done! If we all benched to the limit every time, each CPU/GPU would effectively become single use disposable... The only way to be sure you've pushed it to its limit is to kill it pushing it harder. I push harder than I would otherwise when there are a couple people close to but ahead of me. If I'm in the lead, I try to ensure the hardware lives to run another day so maybe I don't increase that extra few decimal places on voltage... Few of us have endless buckets of hardware/money. Sofos just beat one of my gold cups by 23 points. I would guess he's sandbagging a backup. But I hadn't pushed that hard to take that gold cup, now that I see there's some competition, I'll push it a bit harder to try to hang onto it... And hopefully the GPU still lives to run another day. At the top of the rankings, half the battle is putting up the best score, and the other half is an endurance race to see who burns up their cherry gear first. Sandbagging is a good strategy to get others like me aiming lower than they need to, and maybe the next time they try to comeback on your score the gear will have degraded. I've always put up my best results though, and just waited for others to come after it, then rerun if needed. That said, I have more than a few GPUs that would score great on my current IBs, but have degraded after running them on SB.

Classifieds staff knows more about problems than members. Problems on ocf are rare. But I door knocked and repoed two items in the past year in Ohio. Not possible internationally. We have around 100 sales open any given day, and have a dozen or so disputes we mediate annually.

Good idea. It is hard to sell extreme cooling equipment in forums, there would be more interest here due to critical mass. If you do a marketplace here, you would save yourself a world of headaches by making it as premium of a service as possible... I have some experience in that. On Overclockers we only permit traders with at least 100 posts, and they must have ISP email. Why? Because that makes it quicker and easier to troll/con/rip someone off at any other forum but ours, and then we only are providing the service to the most loyal, trustworthy members. If I were going to launch a classifieds section on HWBot, I would restrict access to the classifieds here based on post count in the forums and/or a certain number of hwboints. I would set the limit at something like 100 HWBoints. Yes, that would cut out a lot of legitimate members, but it would also almost completely cut out the riff-raff to ensure having the section isn't a management headache... Then the most dedicated hwbot members can trade in relative confidence that they are trading with someone who has an investment in the site, and is less likely to rip you off than someone just passing through. Making it premium/restrictive makes the section manage itself much more than it would otherwise, and also rewards/motivates more loyal members.

Nice score and zombie work, but where's the screenshot? (I think you forgot to post the result)

I like the idea, but could see it being harder to finetune. I also think the current point system does something very interesting and important to the hwbot "magic". By artificially making the point reward very drastic between 1st, 2nd, 3rd, 4th, and 5th... It is VERY motivating if you are in the top 5. The way this plays out... if you are currently scoring 20th in a ranking, it isn't hard to tweak a couple settings and move up a few spots (low difficulty), but it doesn't get you many extra points. However if you are currently scoring 5th in a ranking, it can be very hard to tweak settings and move up one or two spots (high difficulty), but if you can move up those couple extra spots you get a lot of extra points. I'd be willing to bet that more hardware dies in the top 5 of the rankings due to the way the current point system works, and I kinda like that. (I'm not going to kill a card to get 15th place, I'll let it live and rerun when the next generation CPU comes out... But in the top 5, I'm going to kill that sucker and worry about finding another when I need it down the road)

That is correct. However you can use RAID expert and get "normal" onboard ram caching scores in general usage and virus scan if you don't modify cluster/sector size. With Areca, the RAID controller has RAM onboard, and it will produce high gen usage and virus scan scores consistently regardless of stripe size, sector size or formatting... It produces those big scores because running the storage tests multiple times it puts a lot of the storage tests in its RAM cache. (you must pre-run storage tests to initialize/preload the ram cache on an areca... Without preloading the ram cache on Areca, scores are only "good" not "great") The important difference, in my opinion, is that PCM05 whigs out with RAID expert and changing the cluster/sector size. Yes, Raid expert supports ram caching... But running the storage tests a second time does not increase the scores the way it does for Areca. It doesn't need to be preloaded for ram caching to boost the score. The cluster/sector size changes just produces really high scores every time, because PCM05 is detecting/reporting/calculating the speeds wrong. Certain SSD firmwares perform much better in virus scan and gen usage. The firmware's aren't tweaked for PCM05, just certain drivers with certain optimized firmware happen to work much better for PCM05.

I agree with that. Any 3770K I've tested (just under ten or so) can do about 6.2GHz on LN2 - take it out for a spin on 3D and you've got a gold mine for hardware points. Similar was true for any SB at 5.5GHz+, except that was doable on DICE/SS. In the popular 3D hardware categories those are a formula for low skill automatic/easy points. That is hardware points though mainly, and possibly a separate topic - there are lots of ways to grind hardware points depending on platform and cooling method... This topic isn't about hardware grinding. This topic is global points. The global points in question in this topic are almost exclusively exotic globals with relatively low competition. You can't take any random 3770K full pot and dominate in 3d03, vantage, 3d11, or heaven in multi-GPU. Global points in those categories are hard in that they require hard clocking on both CPU/GPU. If you look at multi-GPU benchmarks in categories other than those, there are a lot of scores from years ago where top spots are still held because no one has bothered running it again. 4x 7970 on IB would be easy to take those spots versus 4x 5870 ran just above stock, but pros/enthusiasts mostly don't even bother with those easy points because they'd rather compete in 3d11/vantage/heaven where the GPU scaling matters and people pay attention. My goal with bringing this topic up is to raise awareness about "gaming" global points, and perhaps lead to the rankings being more "valid" in the eyes of the community.

Nice sub. 23 points ahead, really? lol

I think global points could use a tune up. Just opening up a discussion. I don't really care for the outcome... It's just easier to make suggestions for hwbot than it is to make your own hwbot. Taking a quick look, at a glance the following is true for the current global point categories for CPU and GPU benchmarks... Covering the 3d tests divided by GPU count: Doesn't scale with anything beyond 1x GPUs (scores in 2x-4x just get worse): 3d01, 3d05, aquamark, 3d06 Doesn't scale with anything beyond 2x GPUs (scores in 3x-4x just get worse): 3d03 Doesn't scale with anything beyond 3x GPUs (scores in 4x just get worse): Unigine Heaven Scales with 4x: Vantage, 3d11 Covering the 2d tests divided by core count: Doesn't scale with anything beyond 4x cores (extra cores cannot be utilized): PCM05 Doesn't scale (really) with anything beyond 12x cores: wprime32m, wprime1024m Scales on everything (but doesn't get any globals): UCBench2011

They should be in my opinion. However I assume its more complicated than that, because multigpu categories should also be merged. Scores go down with more gpus in various 3d tests. EDIT: To be exact... Covering the 3d tests divided by GPU count: Doesn't scale beyond 1x GPUs and shouldn't have globals in 2x-4x: 3d01, 3d05, aquamark, 3d06 Doesn't scale with anything beyond 2x GPUs and shouldn't have globals in 3x-4x: 3d03 Doesn't scale with anything beyond 3x GPUs and shouldn't have globals in 4x: Unigine Heaven Scales with 4x and should have globals in 1x-4x: Vantage, 3d11 Covering the 3d tests divided by core count: Doesn't scale with anything beyond 4x cores and shouldn't have globals in anything higher: PCM05 Doesn't scale (really) with anything beyond 12x cores and shouldn't have globals in anything higher: wprime32m, wprime1024m Scales on everything: UCBench2011 (still doesn't award any globals)

No subtest can use more than 4 core. If you might clock higher or more stable, 4 cores are best.

Congrats to winners! Had fun!

No good. With the super tweaks artificial capping would be a joke. Transparent window, audio/video, memory latency, text edit, and I probably missed one in there... they can all be supertweaked.

Yes but its a bug that tricks the software into thinking it ran faster than it did. It doesn't actually produce higher results in ANYTING, except PCM05, because it bugs out on a setting which has nothing to do with actual performance. But then, nothing in PCM05 has anything to do with actual performance anymore, which is why only a handful of people bother running it anymore. I shouldn't have said anything about the settings I mentioned, because that same handful of people will use it, despite it being a bug, and everyone will clamor over how great the scores are. Certain drivers and firmware have the same result, they bug out PCM05.

Here's a screenshot, big gen usage and 9K virus scan by modifying those kinds of formatting settings on amd platform (3x crucial m4 raid0): [ATTACH]1337[/ATTACH]

You'll want to test different cluster sizes when formatting the drives, and for the RAID stripe sector size. I've seen a teammate do extra large scores in virus scan that way, same as the submisions in question... That was some time ago, my teammate didn't submit them because he didn't want to bother with the questions/criticism because it looks bugged. Changing cluster sizes shouldn't have that sort of effect, but when using it with the amd storage manager for some reason it can do that.

It's not that different than the crazy audio/video encoding scores and no one has complained about those yet (rules say no replacing codecs - are they replacig codecs?). A 6.1 ivy does 10k in each with basic codec setting tweaks. I hope no one gets in trouble for pcmark submissions. It is so hard to tell what is legal and what isn't. The whole "no tricking the benchmark to make it believe it ran faster" rule also doesnt apply to pcm05 - memory latency scores when tweaked disrupt the measurement.

Bugged, if not ramdrive. Only areca ramcache or hardware ramdrive do gen usage like that. Virus scan is high, but certain drive setups and firmware yield extra high virus scan. Dunno about that high. Gen usage is the one hard to fake. Virus scan can get crazy. I've done gen usage like that, but with $2500 worth of storage.

IRST/Matrix storage manager is ram caching, but legal - I haven't ever seen it go this high. I'm not sure anyone knows how to police PCM05 anymore with the mega-tweaks though. lol The gen usage does seem like a giveaway however, unless that was done on a hardware ram drive... Virus scan with 2 SSDs and matrix can go high, but not in general usage. There are a few different ways to do high general usage, the only legal ways I know of are hardware ramdrive (mediocre) and hardware raid with 1GB of RAM cache (that's why people use the PCIe raid cards).