Massman

This is what I posted: How am I closing my eyes? This is a typical case of arguing for the sake of arguing. We both say that MaxxMem has some flaws, we both say it's better not to use it for points ... and still you're trying to accuse me of 'not listening' and now apparently also 'choosing money over community' ... huh? In the end WE provide the platform for competitions. In the end, that's ONLY what we do. Of course, we can make suggestions to organizers how to set up a competition, give insight on behavior of applications and benchmark rules ... but in the end, whoever sets up the competition has the final call. If you work at a bank, you can also suggest the client to invest the money in fund A, but you can't force him to do so. And I think we're now reaching an end of this discussion. Everyone agrees, let's stop arguing just to have fun. I accept one last reply from you, if you feel the need to have the last word

MaxxMem likes low latency ... you are comparing 4.5GHz NB + 2250CL8 to 4.7GHz NB + 2250CL7. The score looks perfectly normal for those frequencies. Two things: 1) At hwbot, we don't impulsively make decisions based on emotional reactions. It's normal that you guys suggest to do everything very quickly. 2) The benchmarks were NOT chosen by HWBOT. We have the platform to host competitions, apart from the hwbot challenges we don't make the rules ...

? I'm not following what you guys are trying to prove here? I think we all agree that the benchmark is bugging in some instances. Also, this is something we can't forsee when adding a benchmark ... it's only through extensive testing that issues can surface. Everest benchmark is not possible because there's not a single-score. If you want to rank benchmarks, you need a single score ... not 4.

Is that all I get? A score breakdown. If you want to make a point, please add a bit more information. FYI, the benchmark doesn't like it when you have programs opened, or are loading things in the background.

There aren't that much bugged scores ... no need to become overly paranoid. From the June OC Challenge, I've learned that MaxxMem is pretty consistent at stable settings (within 10p ~ 2% variance). When NB/IMC is pushed to the edge, latency gets bugged ... Our deal with MaxxMem is just that we include the bench if proven stable. Nothing else. That's because MaxxMem cannot work well with altered bios microcode ...

This issue has been popping up a few times in the last couple of months, so maybe it's time to work on solving it. Ideas and concepts have been floating around for months (maybe even years) and here's a first worked-out version of a competition that should make everyone happy (). Note: # This document is a WORK-IN-PROGRESS and certainly not a fact. We're releasing this to the public to get feedback on vital points we've missed or things we've overseen! # Name is work-in-progress as well # Biggest change for household overclockers: no ES in amateur ranking # We do NOT force anyone to join any category LINK TO THE FULL DOCUMENT: http://hwbot.org/blog/wp-content//ocproleage.pdf

Whoah, lightning still works

Not really. The benchmark is in fact pretty consistent as long as your system is perfectly stable. At the very edge of stability, the benchmark is producing out of the ordinary latency scores (bandwidth figures are mostly ok).

In comparison to the other scores already in the database (so not from the competition), the only reason why this score could've been possible was if the CPU speed would give a huge boost (others have 5.1G, here 5.5GHz). His first score was with the same CPU frequency and was not much faster than the others with same mem/unc frequency. Subtimings is not really an explanation. Everyone plays with subtimings. Problem is that this benchmark's producing bugged runs instead of crashing

The score is bugged. Previous run at approx. the same settings (http://hwbot.org/community/submission/1016877_jale00_maxxmem_ddr3_2109.9_marks) indicates that scaling does not come from CPU Frequency. A difference of 7ns with +/_ same settings is too much.

Not bad! How are the subtimings set for this run?

To be honest ... I'm always using this Fat pot, but I dont' own it. Within the team, we are just sharing LN2 pots

IMC running 1:1 with CPU, maybe that's where the confusion is?

12 is a team moderator ... this is a disgrace!!

Bugged latency score ...

I think the key for getting PI stable is running high Vdimm. Really? You just need to look for the performance shops. There's one shop in Belgium that has this performance minded boss and they have 5 980x in stock all the time

To sum up. What you need to mod your GTX 480's is: Hipro5 bios #1 Kinc's hardmod modifications And a software overvoltage tool such as Afterburner

And that's when exactly? Initially, I was thinking about only doing this for discontinued platforms, but it's not a bad idea to also do this for very new platforms (it's maybe easier to be king, but if it's easier more people will try to get it, so it becomes difficult again?) If we revise caps every 6 months, we can add this kind of achievement to every new platform no problem.

... "Achievement King", "Achievement Guru", "Achievement Junkie"

Just tried the query - the problem is that some impossible scores (that are not blocked) mess up the ranking.

It's not a right measurement since you run one benchmark multiple times for one score. If you see 6.7s in 1M, it means you spend 10 hours getting that score

Just noticed I forgot to remove 'King' from the name field. Just an example, so no biggie.

Good idea!

Any ideas for the icon. Underneath looks kinda lame, I think

about 80p iirc