Massman

Doesn't the Areca have onboard cache? So it's something like MFT or ramcache?

By the way, one of the biggest complaints in our previous HWBoint systems was that it was very difficult for the 'regular' guys to compete for the top ranks. Creating a system that tolerates a wider range of categories to generate +/_ the same amount of points will allow more people to be able to compete, but will also reduce the direct competition aspect of it. Direct competition as in "I beat you, so I'm ranked higher". It's hard to find a balance between making sure there's intense competition, which is an exclusive format as it requires lots of time and resource, and that everyone can have a shot at the top ranking.

Does the firmware/driver trick the benchmark or is it just boosting the one thing the benchmark is designed to test for?

Small update on consequence of this thread. - 32M result is removed (obviously) - John Lam's account suspended for one month Remember people, you are responsible for your HWBOT account and the scores that are linked to it. At all times, you should be aware of what scores you're uploading to it.

I opened it up for another month .

Heads-up: still going over the competition scores. Personal note: don't travel at the start or end of competitions.

5 out of 8 will show up at the live final and do a "final battle" there, as far as I understood.

Wait ... two Z77X-UP7 boards? :D:D

Rampage V Extreme? During pre-testing MVE and 4x 680 Sam and me noticed something similar. In 3DMark11, GT2 and GT3 would throttle and physics would hang the first second making us lose 1000pts in that subtest.

Sorry for the delay in response, I posted this one at the staff forums first for the other staff members to proof-read and comment on. So, as I've mentioned before, for the past two days (due to Campus Party event) the staff has been going over all the information posted in this thread, provided to us by email or found by ourselves. Of course, we also spoke with both Mad222 and John Lam about this to hear what they have to say. As far as we can see, there are two different accusations here: 1) The 4m47 and 4m49 SuperPI-32M originate for the same system and is 'score sharing' 2) 1M, Wprime32M and Wprime1024M are also shared + JL is lying about mainboards About the SuperPI 32M It's been confirmed with Mad222 and John Lam that these two results do in fact originate from the same system. Whether it was a deliberate act of result sharing or a simple mistake is up for debate as it's hard to talk factually about motives here. As far as I understand the situation, Hicookie went over to the HKEPC office to test the upcoming Z77X-UP7 mainboard together with Mad222 and John Lam. There are pictures of that session on Facebook as well as on the forum. John Lam states that he mistook the 4min 49s SuperPI run as his own because he had obtained very similar results with his ES CPU and a USB key was used to store the results instead of just keeping them on the SSD drive (as usual). Here is a 4min 50s SuperPI-32M run with the UP7 and the 3770K ES CPU. As said before, whether this was a deliberate act of result sharing or just a mistake is debatable. Personally, I find it hard to believe that someone would risk everything just to have a SuperPI-32M result in their profile that is 1 second faster than what they achieved. Even though, every mistake comes at a price and has a consequence. I'm currently waiting for the other staff members to check in for their view on the level of consequence. About the others So, later in the thread people came with accusations of grand scheme accusations of 'fooling the community' on a large scale. So, we went again looking at the pure facts and not let our mind trick us into jumping to easy conclusions. In essence, the accusations can be brought down to these 3x2 results: When reading through the thread, we've found these arguments were used to 'prove' this definitely was hardware sharing: - Screenshots look similar - CPU frequency/voltage is too similar - Hard to believe one team would have two great CPUs like that - Memory frequency of 1206.8MHz isn't possible, so result wasn't done on Gigabyte (and thus John Lam lied in HWBOT submission and used Mad222's results from the M5G) - John Lam is not known and therefore it's not possible for him to do great results. Of course, when seeing these arguments lined up I can understand it's logical to jump to the conclusion that this is a clear case of result sharing. So, we went over the arguments one by one. #1. Screenshots look similar Arguably a solid foundation to ban someone for. Definitely a reason to look closer at the results. #2. CPU frequency/voltage is too similar The argument here would be that no two CPUs respond the same to voltage and temperature, so it would be highly unlikely that there would be two CPUs doing within 20MHz of each other at the same voltage level. Der8auer has indicated that both of his CPUs run within that same range of frequency at the same voltage level. So, it is in fact possible to have two of these similarly acting CPUs. The likelihood of that happening might seems slim at first sight, but is quite hard to determine exactly. There are two other results in the Wprime1024M ranking that were done at same voltage within 20MHz from two different users, so Der8auer is not an exception. #3. Hard to believe two such great CPUs can be found in one team HWBOT staff member Christian Ney confirmed having seen at least 5 7GHz CPUs over livestream. Besides, we all know that finding a good CPU is mostly about testing in great volume (HKEPC has tested over 100) rather than skill. We don't have a clear view on how much chips were tested exactly, but we don't have that view on anyone's testing for that matter. #4. 1206,8MHz cannot be done on a Gigabyte mainboard One of the more complex arguments was that John Lam would have submitted a result using an Asus mainboard, without mainboard tab, but indicated Gigabyte in the result submission. The fundamental argument here was that the memory frequency of 1206,8MHz was not possible on a Gigabyte mainboard and that it proved that he used a result from Mad222 (with Asus board) to submit as his own. The claim was that John Lam is, and I quote, "a liar. He is deceives to vendors and thus all true overclocking community". We've established that this memory frequency is indeed possible on a Gigabyte mainboard here and here. The technical explanation was provided as well (SSC Spectrum) and we've also shown that you don't need any special software to alter this %, but it's just BIOS dependant. #5. It's all because of marketing and $$ This one's tied to #4 and basically insinuates that the sharing and (false accusation of) lying was done on purpose for marketing and/or financial benefit. Now, I have no clue whether or not any company pays HKEPC (doubt it), but there has been no PR on any of the SuperPI 1M, Wprime 32M or Wprime 1024M results. Actually, if it would have been done for marketing reasons, essentially John Lam would've shown that the UD3H is a tad worse in all benchmarks ... not really good PR I'd say. Highly unlikely, in other words. #6. John Lam is not known and therefore it's not possible for him to do great results. This one's a bit weird and not really objective to say the least. John Lam's results don't really come out of nowhere either. He did the Antec Computex gig which showed decent skill/creativity. We all know with the right CPU it's "easy" to set records; Ivy Bridge is pretty straight-forward overclocking. Another finding One thing we picked up on while trying to figure out what went on is that the file extentions of Mad222's and JL's scores were not the same. Mad222 always uses .pjpeg whereas JL uses .png (or .jpeg for the 1M). Under the assumption that this is a case of result sharing, the file extentions should in fact always match! Both sets of Wprime results have different file extentions, which makes it highly unlikely that they originate from the system. Unless you'd of course believe that they would pay attention to the file extention but then not give a damn about the similar voltage and memory settings. By the way, before you argue that the .png file would have been done after editing out the mainboard tab from the .jpeg file, let me remind you that the IQ decrease from saving as .jpeg will not be resolved by re-saving it as .png. Here's a test: 1) original .png 2) saved .png as .jpg (notice IQ degradation) 3) saved .jpg as .png (notice same IQ degradation) If John Lam was editing the .jpg results and then saving them as .png, there should be bad IQ in his screenshots. But they are very clean (like a real .png should be). Example: http://d1ebmxcfh8bf9c.cloudfront.net/u41156/image_id_820307.png In conclusion As far as this thread goes, there were 4 sets of results under scrutiny: SuperPI 1M, Wprime 32M, Wprime 1024M and SuperPI 32M. 1) SuperPI 1M: different #core, different memory timings, different cpu ratio 2) Wprime 32M/1024M: different file extention, 1206.8MHz possible on gigabyte, similarly clocking CPUs possible, multiple 7G cpus verified. 3) SuperPI 32M: mistake (?) while submitting result. I know Slamms and a lot of people-who-don't-know-him-but-still-agree have already made up their mind based on speculation and false facts, but here's a list of facts. We tried to be as neutral as possible, assumed all possible scenarios and acquired this list of findings.

No political stuff here, we're all awesome in our own messed up way

Yes. That one time you had to prove over livestream was because a third party, MSI, was involved too. Besides, HKEPC already showed over livestream that they have multiple 7G CPUs. The fact that they would run them simultaneously would not provide any more information other than how good my new internet connection is.

Same treatment. In fact, for all the previous 'cases' mentioned here we always backed up any decision with factual proof. That includes 12/diabolo, Andre's LOC and Zzolio/riska. But we don't openly advertise who gets banned and for what.

FYI, that 32M is actually confirmed (by Mad/JL) from the same system. I'll wrap up a nice text tomorrow

If anyone should be on HKEPC's side, it should be you. You've been wrongfuly accused before and were always incredibly angry about it. But now that it happens to someone else, for some reason you prefer to switch sides and jump on the "very strong evidence"-bandwagon. We requested to show prove your similar results only once. That was for a big competition like MOA and after several complaints of other overclockers from the Americas. Those complaints were not sent directly to HWBOT, but were in fact addressed to the MSI USA office. Not sure if me trying to figure out a way to get the problem solved was good or bad now.

Judging by the amount of reports we've received over the past couple of years, fire is everywhere and still spreading. But, judging by the stories we've heard from those who got accused, smoke doesn't even exist in real life. Just trying to say that smoke on the internet is a fairly vague concept and that it's better to just stick to real facts. John Lam's results don't really come out of nowhere either. He did the Antec Computex gig which showed decent skill/creativity. We all know with the right CPU it's "easy" to set records and we also know you can find the right CPU by just testing as much as possible.

What if there's factual proof there no sharing of results?

Ah, so no need for the staff to actually dig deeper then, right? In any case where enough "community members" make a choice, we can just issue a ban "by popular demand" now . Makes things pretty easy for us actually . Almost all of my recent scores are group sessions too. I just enjoy benching with fellow enthusiasts more than on my own. Well, guess I'm just a shit overclocker too then . That's why I specifically wrote "best I've seen" and not "perfection all around". We know why things went wrong in that stage too; we learned.

In case you guys are wondering, we're currently going over the actual facts in the staff forum as well as asking Mad/JL for feedback. But one thing that continues to bother me is that, when I re-read this thread, it seems that a lot of people have jumped on the "they are frauds"-bandwagon without real evidence and just relying on very basic information. People seem to find it normal to use the word cheater at random and even dismiss actual evidence as not valid because "i don't believe". In any case, could someone who has the ability to make a post without using the words "cheater", "fucker" or any other aggressive language please provide any other seemingly incriminating evidence of false play? From this thread, so far I've seen - superPI 32M sharing - "impossible" memory frequency - "never seen" SuperPI bug - similar 1M and wprimes Just one more thing. Unlike some of the folks I've seen posting in this thread, the staff evaluates facts and not just opinion. Anyone who just enjoys to speak their mind and figures it's okay to just throw out just anything that comes up can pretty much go do their thing elsewhere. There's a fine but clear line between raising questions and public slander. I'll move this topic to the HWBOT OC-Crime Center as that subforum was meant for stuff like this.

I had a similar issue like Gorillakos on the UP4 before, but it got fixed magically. What I did was: - boot up with pcie gpu - set igp to default in bios - save, exit shutdown - remove pcie gpu - boot with igp - clear cmos (long hold) - boot with igp always After that, I never had issues with booting from the IGP again.

Link? One of the two scores might be 2xGPU?

Lol. Royalties, here I come!

Guys, please resize your pics or at least use thumbnails. Now the thread is far too cluttered/heavy to really see anything.

It's very normal that rules/punishment evolve along with the system they active in. Hardware sharing had a much greater effect in all HWBOT revisions before rev4, so it's logical that the punishment was far greater back then too. Nowadays, the practical effect it has on other users/teams is far less, but of course it still is very annoying to see. HW/result sharing is an inherent problem to any system that gives reward based on benchmark result, by the way. It's not a problem exclusively to HWBOT, but to a type of system. Just think about the LOC 2010 incident: in an OC competition outside HWBOT and without anything like points, there was still result sharing between participants to win something. And, yes, there's definitely ways to improve the situation so hw/result sharing is even less of a problem (eg: get rid of user leagues, user league based on TPP, etc), but it doesn't mean it's going to be better. So far, the 'best' system I've seen that removes the benefits of hw/result sharing are all the team-based competitions (like Country Cup, Team Cup, ...) we've organised here at HWBOT. Anyway, good luck with the new project!

It got fixed ?