Splave Posted February 14, 2014 Posted February 14, 2014 Here is a screenshot not to make you angry but to show you how to bug UCBench to run stupid high. using -ri7 after the instruction set I would say 1 in 10 runs will be buggy like this In the screenshot you will see its not there because I erased it before taking it. I think the answer to this is to require a valid link for top 20 in each category. The valid link will show the actual string you used in the command line. Quote
GENiEBEN Posted February 14, 2014 Posted February 14, 2014 I agree. I will also contact the author to see what can be improved security-wise and maybe a different launcher. Quote
K404 Posted February 14, 2014 Posted February 14, 2014 Here is an observation. UCBench has been around for years, but it's only when HWB give big points for it that people A/find problems B/complain about them and/or C/anything is done about them. Quote
Splave Posted February 14, 2014 Author Posted February 14, 2014 I agree. I will also contact the author to see what can be improved security-wise and maybe a different launcher. sounds great mate @ Kenny: Only purpose of the thread was to show how to bug it so it will be fixed and how we can help make it more secure Sorry we broke it hehe Quote
K404 Posted February 14, 2014 Posted February 14, 2014 Oh, it's nothing against you, Allen, or the other guys that have been pushing the benchmark recently. If anything, the last few weeks have shown that letting a benchmark age and be worn in before awarding points means absolutely nothing given how points-focussed most of us are Quote
der8auer Posted February 14, 2014 Posted February 14, 2014 Yea we really have to do something about this benchmark. Don't wanna have a 2nd PCMark05 Quote
HiVizMan Posted February 14, 2014 Posted February 14, 2014 Guys may I suggest that until such time as there is a fail safe mechanism built into the verification process we suspend points for this benchmark. The testing that I have done for our team has shown way to huge a divergence of scores and methods of manipulating the scores. There is no harm in trying something. There is potential for harm in once trying and finding out it is not fit for purpose in its current form, of persisting doggedly with that experiment. Shelve or suspend for now, reinstate once sorted. Quote
Chi-Kui Lam Posted February 14, 2014 Posted February 14, 2014 I just report the UCBench issue to El , It is easy to clear out what is fault run and what is real, How to reproduce it and what cause the fault run. Hope Hwbot will give out some rules for UC Bench Quote
Moose83 Posted February 14, 2014 Posted February 14, 2014 Of course we need to do sonething on that bench...4670k at 4,9 4Core WR sounds to legit lol 1940p at 4,9....Bench is more "Tweak"-"Cheat-able" than PCM05 in total lol Quote
Moose83 Posted February 14, 2014 Posted February 14, 2014 So i read out of there not using -ri commmand i have no bugged score.... Quote
ObscureParadox Posted February 14, 2014 Posted February 14, 2014 If you can reproduce that and prove that the -ri command is the cause of the bugged runs then I say we just ban that command like we have banned tweaks from other benchmarks. Quote
dhenzjhen Posted February 14, 2014 Posted February 14, 2014 According to Genie -ri(x) parameter is ok just don't use odd numbers. Quote
Moose83 Posted February 15, 2014 Posted February 15, 2014 Splave also use 8,45,64 on his actually run, but no -ri. Genieben saying -ri allowed, but no odded Numbers^^ Can we please have clarified rules for this bench??? Quote
Splave Posted February 15, 2014 Author Posted February 15, 2014 sorry but odd numbers is not the issue, here is bugged run using -ri7 and even thread counts Quote
Moose83 Posted February 16, 2014 Posted February 16, 2014 The bench itself makes the bugged runs thought^^ Quote
Calathea Posted February 16, 2014 Posted February 16, 2014 So guys what's ok to use? -ssse3 -cpus=56, 60, 64? I Have a little ln2 left for a celeron 450 and don't want to be reported for bugged run Any more tweaks ok? Quote
der8auer Posted February 16, 2014 Posted February 16, 2014 Why? so this one would be bugged aswell? Or not bugged because I also used 2 and 4? http://hwbot.org/submission/2501341_der8auer_ucbench_2011_pentium_e2220_(2.4ghz)_447.3_mpt_score Quote
Calathea Posted February 16, 2014 Posted February 16, 2014 Then there's TONS of bugged runs. I need to use the LN2 today or it will evaporate, should I only bench with -ssse3? I will be beaten by results with 500 Mhz lower clock... Quote
GENiEBEN Posted February 16, 2014 Posted February 16, 2014 Why? so this one would be bugged aswell? Or not bugged because I also used 2 and 4? http://hwbot.org/submission/2501341_der8auer_ucbench_2011_pentium_e2220_(2.4ghz)_447.3_mpt_score Not bugged. Then there's TONS of bugged runs. I need to use the LN2 today or it will evaporate, should I only bench with -ssse3? I will be beaten by results with 500 Mhz lower clock... Yeah. pretty much. If single core cpu then select anything you want. Quote
Calathea Posted February 16, 2014 Posted February 16, 2014 I don't think I got any less confused by your answers Genieben Quote
Moose83 Posted February 16, 2014 Posted February 16, 2014 8,45,64 will only bug out in 1 of 10 runs... Problem isnt -ri or odd numbers... Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.