Jump to content
HWBOT Community Forums
avalanche

Tweak book for cheats

Recommended Posts

Leaning how to OC many years ago I used to marvel @ the big scores done in Hwbot. Not anymore :/

So called tweaks " the secrets " that trick a benchmark to go faster. 9_9

yaf1h1d776eyaci7g.jpg

Your top #3 or #4 in the world banned for manipulating pc clock. Obvious cheat in x265

Then this guy H20vs LN2 shoots himself in the foot bragging his score is better xD Time measuring inaccurate therefor invalid score.

What a joke! Submitted & it's there in the yellow warning box.

_________

Cinebench windows not rendered fully. Then covered to mask the deceit.  

Members in here learning others of Aquamarks & early 3D marks how it can be cheated.

Pulling the monitor cord out ... & this was allowed. FuKKen hilarious that one

LoD on the modern 3D13 you can not even see the scenes first half of bench. This is just shit, make you feel all warm & fuzzy the powerful gpu :P

Haha, only a few days a go learned  ... less maxmem in a popular benchmark bugs a few hundred extra in score.

________

So how many of you TOP overclockers think it's OK to accept your ill gained scores until proven cheated.

I don't believe not anymore.

 

 

 

  • Like 1

Share this post


Link to post
Share on other sites

There is a thin line between a tweak and a cheat, depending on how you look at it then it could be one or the other.

As with most of the issues with this hobby, it is often hard (or impossible) to tell when someone is bending (or breaking) the rules.

Edit: it also doesn't help that HWBot has had quite a big staff turnover over the past while, so there are now only a handful of mods - and moderating loads of scores is harder than you think.

Edited by Noxinite
  • Like 1
  • Thanks 2

Share this post


Link to post
Share on other sites

Certain benchmarks R15 & x265 ... you can reasonably predict an output per multi.

I've spotted Cinebench 2003's that look dodgy downclocks. Just no way they could hit the scores with shit mems.

Endless bloody submission where the 3D bench cpu + mem clocks were down, yet massive scores that I could not get with superior hardware + speed.

Just gave up there, tweaks that were shared. Scores can never be beaten.

_________

There is no thin line. You know if you step over the line. These bandits I'm talking about, deliberate cheating.

  • Like 1

Share this post


Link to post
Share on other sites

Fire Strike Ultra rules state that changing LOD and Tessellation is allowed, which to me means that if the result on UL states that AA settings are modified then the result is invalid. Yet looking at many of the top scores this problem is present. Not to mention timer errors. Flagrant disregard for the rules like this just harms the integrity of the hwbot database. 

Now, if all these questionable scores were to be reported, would they be even taken down? 

  • Like 1

Share this post


Link to post
Share on other sites

People told you to report scores like the ones by H2ovsln2 - who has done this multiple times before even when I was still a mod, but got no time out for whatever reason. But talking about cheats when someone changes driver settings, or uses LOD where allowed is simply not leading anywhere. Benching is already much too much pay to play and soldering instead of optimzing skills. If you think that legal driver tweaking, or waza, or disabling services at os is cheating, bench prime 95 like others do and cheat on temps ;) - the gb3 desaster you talk about is constantly ignored by hwbot btw, it is not only that low maxmem bugs the score because of not enough drm-memory left to correctly run the benchmark, everyone who has eyes also sees that the results at settings like 4600-4800c14 bugs at multicore mem copy, but as long as majority rather plays this instead of reporting tons of scores, nothing will change, I am tired of messaging people on fb or talking ther about it, discussing this and see no result, It is the community who needs to change this, and if you don´t do it, don´t complain 😛

 

P.S. Lod was allowed long ago, and that aa changes are illegal to fm(UL) is not older than 2 years, while 3dm01 is now nearly 20 years old. So what sahll hwbot do, delete all scores made applying to rules?

  • Like 5

Share this post


Link to post
Share on other sites
22 minutes ago, Noxinite said:

Edit: it also doesn't help that HWBot has had quite a big staff turnover over the past while, so there are now only a handful of mods - and moderating loads of scores is harder than you think.

We are not talking about staff issues, now you edited.

It's the let slide a score that has been noticed here & there. We all expect it to be right / correct. Naive on my part till now.

Easily 100x scores that are dodgy incorrect I'd spot in a year. Some are just no cpu-z in the screenshots.

Drop you bastards in here a heap of points if I cared to report. Just did not give a fuKK or log it.

  • Like 1

Share this post


Link to post
Share on other sites
2 minutes ago, cbjaust said:

Fire Strike Ultra rules state that changing LOD and Tessellation is allowed, which to me means that if the result on UL states that AA settings are modified then the result is invalid. Yet looking at many of the top scores this problem is present. Not to mention timer errors. Flagrant disregard for the rules like this just harms the integrity of the hwbot database. 

Now, if all these questionable scores were to be reported, would they be even taken down? 

The simplistic way it was explained to me when the "replay mode all" tweak was found within Nvinspector was:

Tesselation = AA for AMD drivers

AA = Tesselation for Nvidia drivers

The truth is probably somewhere in the middle, but this tweak has been common practice by everyone who seriously benches 3D for quite some time.  Does that help clear it up?

  • Thanks 1
  • Confused 1

Share this post


Link to post
Share on other sites
11 minutes ago, avalanche said:

Certain benchmarks R15 & x265 ... you can reasonably predict an output per multi.

Lets say... there are ways to get another ~4% out of R15 in a way which doesn't involve touching the screenshot at all - a small enough difference to be considered "efficiency". Its really obscure and doesn't make any sense so I consider it likely nobody else has stumbled upon it yet. I've shared it with one of the mods on here and am not using it for results.

  • Like 1

Share this post


Link to post
Share on other sites

Geekbench isn't bugged IMO, nothing like the bugs you get in other benchmarks sometimes.

And if the memory subtest is purely testing bandwidth, then why shouldn't it scale with higher freq??

  • Like 3

Share this post


Link to post
Share on other sites

Because scaling on single and multi core is not linear with frequency but explodes, geekbench 3 was not designed for frequency like this and it simply looks as if the lack of maintenancemakes calculation not work properly under certain circumstances. Your safe benchmark btw also bugs at win 10, this was tested, and as said the maxmem settings also boost the score if too low bc of above mentioned reasons. At least some plausibility checks should be done, especially when you see that other benchmarks that profitate from high memory like spi32m which have a firm calculation base are faster at 4100c12 vs around 4800c14 at tests shown here on hwbot

Share this post


Link to post
Share on other sites
55 minutes ago, Gunslinger said:

The simplistic way it was explained to me when the "replay mode all" tweak was found within Nvinspector was:

Tesselation = AA for AMD drivers

AA = Tesselation for Nvidia drivers

The truth is probably somewhere in the middle, but this tweak has been common practice by everyone who seriously benches 3D for quite some time.  Does that help clear it up?

And yet no allowance was made in the hwbot rules for changing AA.

Still don't know if reporting non-conforming scores will be actioned. 

Share this post


Link to post
Share on other sites
1 hour ago, avalanche said:

These bandits I'm talking about

No one has any fucking idea what you're talking about. 

  • Thanks 2

Share this post


Link to post
Share on other sites
48 minutes ago, cbjaust said:

And yet no allowance was made in the hwbot rules for changing AA.

Still don't know if reporting non-conforming scores will be actioned. 

Why should hwbot make an allowance for AA? The fun fact is that all results before I think 2017 which had AA changed were valid with Futuremark if I remember correctly, so it was clear this was allowed. Suddenly FM changed this for hall of fame, implemented a check on this and excludes new results from ranking. If you look at programs like nvinspector there are 100+ driver settings you can change, so each time fm is bored and changes rules we delete all old scores and start at zero? I am not sure this makes sense, but agree that this might need clarification especially for new members like you?^^

  • Like 2
  • Thanks 4
  • Haha 1

Share this post


Link to post
Share on other sites

It was my mistake, I read the rules for the benchmark, didn't see AA explicitly allowed and so suggested it wasn't, as I'm not a big 3d bencher (for modern 3d at least).

I'll make a note to change the rules page to avoid situations like this in the future. If the score hasn't gone already I'll be removing slinky's score with the timer issue.

 

  • Like 1

Share this post


Link to post
Share on other sites

Newer members not understand the difference between a legit and an illegal cheat is as much of a fault on those experienced members for not making this clear before hand as it is on the new guys throwing the accusations. 

My advice for future threads would be to lead with a question instead of accusations if you wish to have dialogue on the issue. Otherwise you're just insulting members who have been here for some time. 

  • Like 6
  • Thanks 2

Share this post


Link to post
Share on other sites

My 2 cents is that we do from now all scores 100% valid on FM/UL,. not even driver non approved... VALID or out... 

We are looking into updating the rules, though this will take time.

If you spot a dodgy score report it, if you spot it and don't do jack, you are an accomplice lol

Edited by Leeghoofd
  • Like 2
  • Confused 1

Share this post


Link to post
Share on other sites
1 minute ago, Leeghoofd said:

My 2 cents is that we do from now all scores 100% valid on FM/UL,. not even driver non approved... VALID or out... 

We are looking into updating the rules, though this will take time

Why do we have to be dictated to by what FM says?

We didnt give a shit about LOD at least 15 years ago, why would we start now?

  • Like 5
  • Thanks 2

Share this post


Link to post
Share on other sites
3 minutes ago, Leeghoofd said:

My 2 cents is that we do from now all scores 100% valid on FM/UL,. not even driver non approved... VALID or out... 

We are looking into updating the rules, though this will take time.

If you spot a dodgy score report it, if you spot it and don't do jack, you are an accomplice lol

Makes sense for new benchmarks, but I don't envy whoever has to enact the purge if that's extended to the stuff with things like LOD and tess "grandfathered" in as tweaks.

  • Like 1

Share this post


Link to post
Share on other sites

First off, I'm fairly new compared to many of you and despite seeing the LoD tweak being allowed on the rules page it took me a while to actually understand how to implement it. Whenever I changed the LoD it didn't do anything to alter my score so I was like "why is this stated that it is an allowed tweak if it doesn't even work?" In my head it was the equivalent to saying "4 out of 5 dentists choose Colgate over Crest" - which may or may not be true but sure doesn't do anything to get me better 3dmark scores. Eventually I found a forum discussion where it was explained that the LoD tweak was enabled by also utilizing sparse grid supersampling. Sure enough, my scores started becoming competitive.

I never did figure out the tessellation tweak and just assumed it was an AMD only tweak. We don't have the option to modify tessellation with the nvidia driver. But now someone above mentioned that it is tied to AA? Not sure how I'd ever make that connection and maybe I'm already using that tweak ignorantly via supersampling? 

Now, just to be clear - whether Futuremark allows it or not - the hwbot rules allow for this, right? I personally don't care one way or the other - if no one is allowed to use these tweaks or everyone is allowed to use these tweaks all that matters is that I'm not on the wrong side either giving me an advantage or putting me at a disadvantage.

Regarding LN2 vs H20's submission - please realize I just learned about this as I read this thread - but maybe I'm missing some key information posted somewhere else and I couldn't find any comments on his recent results from anyone so excuse my ignorance - but the timing issue - what exactly is the problem?  Every so often after running a 3dmark benchmark I'd get an error message in my submission where it complained about timing inconsistencies. I forget the exact wording but I treated it the same way I treated the error about not having an approved driver. Since I wasn't manipulating the timer (and until Matt's recent informative XTU research was published I wouldn't even have had an idea on how to do that) I chalked it up to being an anomaly of pushing my system components too hard. If the score was better than my previous best score I would submit it. I tried to find one in my Futuremark saved history of results but couldn't but I know I have one or two somewhere. Is this the same thing that you are all talking about in regards to H20 vs LN2?

Maybe I've been cheating all this time and am about to find out that I just admitted it publicly - I certainly hope not and look forward to being corrected on any of my assumptions which aren't accurate.

Anyways, on a positive note - love the new changes to the site, love the new rules with the new seasonal ranking (haven't been this excited since my first few Rookie Rumbles! Feel like the Bot is new again!) and think richbastard and the other left over mods that are still keeping the ship afloat deserve some sort of trophy (that includes websmile and the other guys who may not be official any longer but still contribute on threads like these.

  • Like 2

Share this post


Link to post
Share on other sites

I don´t like playing the devils advocate, but what do we do when UL changes rules again? Delete all scores again? And what about the fact that UL has major problems to guarantee security and check for several new benchmarks (Timespyyy cough) and I do not even talk about legacies like vantage and older? We can of course stop tweaking, but will it help to stop cheaters and will it work in the future?^^

 

Marco , the timer issue is mainly a problem if you use win 8 and 10 and pre skylake hardware, becuase there it is often a sign that rtc clock was (or could be) manipulated - the time shift effect to boost scores illegally you can find at win 8 and 10 rtc bug article.

  • Like 6

Share this post


Link to post
Share on other sites

my comment was full of sarcasm lads...  there's no way to win as somebody will discover a workaround to glitch the detection system... thing is we need to be able to simplify things, sadly over and over again I'm repeating myself, things evolved this badly thanks to them "skills" of a few...

If you see a dodgy score report it!!!!

Edited by Leeghoofd
  • Like 3
  • Confused 1

Share this post


Link to post
Share on other sites

OK so who's gonna start sharing the supposes "tweaks" and who's gonna decide if it's a cheat?

No legislation will cause issues. 

Open game play game. Open benchmark, tweak a bunch of stuff and have a better than realistic score.

  • Like 1
  • Haha 1

Share this post


Link to post
Share on other sites
47 minutes ago, ShrimpBrime said:

OK so who's gonna start sharing the supposes "tweaks" and who's gonna decide if it's a cheat?

No legislation will cause issues. 

Open game play game. Open benchmark, tweak a bunch of stuff and have a better than realistic score.

Sometimes, I open 3dmark11 when I bench Catzilla

  • Haha 4

Share this post


Link to post
Share on other sites
1 hour ago, websmile said:

Marco , the timer issue is mainly a problem if you use win 8 and 10 and pre skylake hardware, becuase there it is often a sign that rtc clock was (or could be) manipulated - the time shift effect to boost scores illegally you can find at win 8 and 10 rtc bug article. 

Sir Webs,

Right - I got this  and know about it from all the times I've had to switch over to the HPET timer to run certain benchmarks so no worries. This is why I'm confused about this entry from H20 vs LN2 - since even though I don't actually know which submission is being discussed yet I can't imagine that my team captain made a recent submission using anything pre-Skylake. If anything as you all know probably better than I do he made a submission using some frakenzoid Xeon hooked up to his refrigerator or in a swimming pool full of mineral oil. Wouldn't that make things a little less suspicious? Please don't think I'm being purposely ignorant and am sincere in my question.

I guess it would help to at least to be familiar with the specific entry that was brought up. Anyone have a link? I see his last two submissions have the LOD/AA issues tagged to his submissions but nothing about timer issues. Maybe I'm reading too much into all of this and Avalanche's mention in the thread opener wasn't meant to be interpreted the way I am understanding it but it seemed like he was upset about cheating and then went on to associate a submission by Slinky as being an example of that despite the fact the rules allowed for them? Did I understand things correctly? If not I guess I'm asking questions about something nobody else is discussing. LOL!

  • Like 1

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...