Rasparthe Posted November 26, 2011 Posted November 26, 2011 Don't you always have to run default? We allowed to run less resolution on 3dMark01 or other 3D now? Changing settings on the benchmark to give better score should be disallowed. Quote
xoqolatl Posted November 26, 2011 Posted November 26, 2011 Submission page says: Please run UCBench 2011 now at default settings. Quote
GENiEBEN Posted November 26, 2011 Posted November 26, 2011 Don't you always have to run default? We allowed to run less resolution on 3dMark01 or other 3D now? Â Default 2K1 implies DEFAULT ORDER for GT's too... Prove me wrong. Â Changing settings on the benchmark to give better score should be disallowed. Â Like I was sayin. Quote
Christian Ney Posted November 26, 2011 Posted November 26, 2011 Default 2K1 implies DEFAULT ORDER for GT's too... Prove me wrong. Â Well, for 2k1, you run all GTs right ? Maybe not in the default order but all GTs right? Â Here only SSE3 has been run, not all Tests Quote
Massman Posted November 26, 2011 Author Posted November 26, 2011 The analogy with 3DMark01 is not correct, because all those GTs matter in the score and with UCBench that's not the case. Â This is more like the CPU tests in 3DMark03/05 or the Feature tests in 3DMark Vantage. If you don't have a key, the CPU tests will be enabled by default in 03/05, but since they don't have any effect on the final score you just press escape before they start. In Vantage, everyone disables the feature tests in the paid version for the same reason: no effect on the score and might cause instability crash. Â The final score of UCBench only depends on the best method to crack a password and for most CPUs that currently seems to be with the SSSE3 instructions. Disabling all the other tests that don't have an effect on the score anyways is completely legit. The method of calculating (= the benchmark) is the same; you just skip all the parts that don't need to run for the score. Â For those who want to do it too, just add the following line after the command line in the benchmark application: "-cf=ssse3 test.rar" Quote
GENiEBEN Posted November 26, 2011 Posted November 26, 2011 @Ney: I can run all instructions and still get same/+ score: http://anrieff.net/ucbench/viewitem.php?id=2590 vs http://anrieff.net/ucbench/viewitem.php?id=2589 Dont even get me to start wdasm. Â @Massman: thanks for having common sense! Â Quote
Rasparthe Posted November 26, 2011 Posted November 26, 2011 (edited) The analogy with 3DMark01 is not correct, because all those GTs matter in the score and with UCBench that's not the case. This is more like the CPU tests in 3DMark03/05 or the Feature tests in 3DMark Vantage. If you don't have a key, the CPU tests will be enabled by default in 03/05, but since they don't have any effect on the final score you just press escape before they start. In Vantage, everyone disables the feature tests in the paid version for the same reason: no effect on the score and might cause instability crash.  The final score of UCBench only depends on the best method to crack a password and for most CPUs that currently seems to be with the SSSE3 instructions. Disabling all the other tests that don't have an effect on the score anyways is completely legit. The method of calculating (= the benchmark) is the same; you just skip all the parts that don't need to run for the score.  For those who want to do it too, just add the following line after the command line in the benchmark application: "-cf=ssse3 test.rar"  With all respect Massman I think you are wrong on this. This is not default settings. Then by rights you should be able to run 3d runs with lower resolutions or less than 32bit or whatever. The bench still runs, the frames still get calculated and the method of calculating is the same. Why run at default res when it gives a lower score?  The other UCbench tests do have an effect on the score, they lower it in some cases, and they should have to be calculated. I would agree with your reasoning if they had no effect on the score but they do. If every chip comes up with the same score with it enabled and disabled then sure leave it out but that is not the case.   Default 2K1 implies DEFAULT ORDER for GT's too... Prove me wrong.  In my opinion it implies nothing of the sort. Default is that you run each. Whatever order has nothing to do with default. If UCBench allowed you to run the SSE tests out of order and that improved your score...have at it. The point being that you ran every SSE test. Edited November 26, 2011 by Rasparthe Quote
Bobnova Posted November 26, 2011 Posted November 26, 2011 I'd say it's more like LOD for 3d tests, personally. Quote
Rasparthe Posted November 26, 2011 Posted November 26, 2011 I'd say it's more like LOD for 3d tests, personally. Â How does removing tests compare to LOD? Quote
Massman Posted November 26, 2011 Author Posted November 26, 2011 I don't understand the logic behind the comparison this tweak/trick with lowering the resolution on a 3D benchmark. The SSSE3 benchmark test does not change when you run it separately. When you lower the resolution, you do change the benchmark. Â As for the performance 'gain' you achieve: you can compare this with the behavior in 3DMark01. When you run, for instance, Nature as first test, the score will be a little bit higher than when you run it as third. The performance increase is not related to the benchmark itself, but due to OS/cache variables. Quote
Johni5 Posted November 26, 2011 Posted November 26, 2011 Hmm i runing Unbench Phenom II x4 945 4200MHz.Only ssse3 teszt OK.But default settings restart computer the test 80% Lol. Quote
Bobnova Posted November 26, 2011 Posted November 26, 2011 I see it as akin to LOD as it is a minor change in score (maybe) that is getting blown all out of proportion. Â Seriously though, the score actually used is always going to be the sse3 score anyway, so why run the other stuff? Quote
xoqolatl Posted November 26, 2011 Posted November 26, 2011 What other rules are going to get changed halfway through the stage? Quote
Rasparthe Posted November 26, 2011 Posted November 26, 2011 Bah, I see that I am misunderstanding how the benchmark works. I thought the score was based on some algorithm of each of the tests, not just which ever one is the best of the bunch. Then I have no problem with removing the other tests and your reference to Vantage and the like makes much more sense. Â My apologies for the misunderstanding. Quote
Massman Posted November 26, 2011 Author Posted November 26, 2011 I'm not sure if SSSE3 is always the fastest one, tbh. SSSE4 seems just a tad slower than SSSE3 ... maybe it gets a nicer boost when running that one separately ? Quote
Crew Leeghoofd Posted November 27, 2011 Crew Posted November 27, 2011 C'mon Belgium we need some more scores before wednesday !!! Â [hidden mode] " You know what to do Christian. Plz PM me your bank account for the money transfer. I'll transfer the cash after the score has been submitted. "[/end of hidden mode] Â Was fun guys yesterday, hope to do some more... though my place is a pile of dead/sort of alive hardware... and I feel pretty lousy and tired lol Quote
GunGod Posted November 27, 2011 Posted November 27, 2011 [hidden mode] " You know what to do Christian. Plz PM me your bank account for the money transfer. I'll transfer the cash after the score has been submitted. "[/end of hidden mode] Â wtf i got error "recepint account is already full, plese sent to another one" ?!? c'mon mate buy something cuz Stage IV is almost over Quote
Massman Posted November 27, 2011 Author Posted November 27, 2011 Lol. Â You guys are paying off the wrong person. He doesn't have REAL power LOL! Quote
ilusindo Posted November 27, 2011 Posted November 27, 2011 What's wrong with my ucbench submission, why can not enter Indonesia's score? http://hwbot.org/submission/2224989_ilusindo_ucbench_2011_pentium_e5200_2.5ghz_434.4_mpt_score Quote
Massman Posted November 27, 2011 Author Posted November 27, 2011 Because there's already someone else who used a Wolfdale CPU for the Country Cup: http://hwbot.org/submission/2224123_sky_coolzzz_ucbench_2011_pentium_e6500k_2.93ghz_437_mpt_score 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.