Jump to content
HWBOT Community Forums

SteveRo

Members
 View Profile  See their activity

  • Posts

    1084

  • Joined

  • Last visited

 Content Type 

Posts posted by SteveRo

    Clarification in PCMark'05 about Tweaks // Tricks allowed

    in Benchmark software

    Posted

    I vote for no more PCMark05 submissions till Geniebens wrapper pops up...

     

    Latest submitted scores are again a joke to me, the benchmark is not regaining it's credibility if you still achieve a 10-20K difference between the top 3...this with just two SSD's in RAID...

     

    Not all are out of line, Moose's intel submits all look good to me as does Denis's latest 4x.

    When the decision to not allow non-512's is made - Moose will do the right thing. :)

    Clarification in PCMark'05 about Tweaks // Tricks allowed

    in Benchmark software

    Posted · Edited by SteveRo

    Thats is the problem, Amd raid eXpert is more like ramcaching that Rst, or i'm worng ?

     

    FYI. I never used amd raid expert.

     

     

    I haven't used amd raid expert that much either.

    The remaining question regarding amd raid expert is can you adjust the amount of system memory used by the amd disk controller and if you can how much?

    I think the main issue with raid expert (that we know of) is the use of amd raid expert to set a HDD sector size greater than 512.

    Clarification in PCMark'05 about Tweaks // Tricks allowed

    in Benchmark software

    Posted

    Fully agree, Intel allow Driver based no change of Cluster Size and no Caching options. AMD Raid expert allowes both in Driver, so there is for me an huge difference. With RST you cant score 650 gen usage and 9G VS, on AMD you can change these settings to trick PCM to believe it runs faster...

     

    Mr. Moose, I have not used amd raid expert that much, can you adjust the amount of system memory used by the controller? If it is adjustable, what is the range?

    Clarification in PCMark'05 about Tweaks // Tricks allowed

    in Benchmark software

    Posted · Edited by SteveRo

    Hmm, now I gotta find out how to check cluster size....

     

    Note - when you create a partition (non-bootable) you have the option to set the cluster size for the partition - I'm pretty sure this is different than the HDD sector size.

    On rotating storage I think the 512 sector size is hard coded - not so now for many controller/SSD combinations, pci-e drives.

    Can a check for this be added to the wrapper? - again not partition cluster but drive sector needs to be 512.

    Clarification in PCMark'05 about Tweaks // Tricks allowed

    in Benchmark software

    Posted

    Regarding amd raidexpert and intel rst - they are not the problem, they do allow some amount of raid caching but its not huge.

     

    The real problem with amd raidexpert or with any controller/ssd that allows it is the changing to a non-standard HDD sector size.

     

    This is how I think it works - pcm05 assumes the SSD sector size to be 512 all the time (hard coded?), if you use

     

    1k sector size, pcm05 reports 2x the actual datarate for virus scan

    2k sector size, pcm05 reports 4x the actual datarate for virus

    4k sector size, pcm05 reports 8x the actual datarate

     

    See below,

     

     

    I agree that we should NOT try to change who won what in previous competitions.

     

    However if you agree that we do not want to obviate storage then increasing disk sector size beyond the default 512 should not be allowed.

     

    Again, this trick requires only one SSD and a utility to increase the sector size from the default 512, this can be done on just about any motherboard with iodrive, probably several others.

     

    It could be - the true virus scan score for 4k sector score of 9500MB/s is simply –

     

    9500MB/S * (512/4096) = 1187.5MB/s

     

    Again, I agree that we should NOT try to change who won what in previous competitions but I believe that tricking the benchmark in this way should not be allowed and points (including previous points) should not be awarded for using it.

    Clarification in PCMark'05 about Tweaks // Tricks allowed

    in Benchmark software

    Posted · Edited by SteveRo

    I'm here without spot :D

     

     

     

    Beyond what any guy thinks, Pro has judged that the submission of Gluvocio not in line with PCMark05., Imho ( my opinion) if you look Hdd gen and virus scan is high in this sub., and noone ask the important question , what sotrage has been used ?

    But in this crucial time for the PCM. legality matters more than anything else.

     

     

     

    Good idea Karl. for all submission, in this benchmark ? all guys , or those in which we have doubts ?

     

    @Moose

    Please Moose, add picture of your rig in this submission,especially the storage used, thanks so much

    Opsss. and send me one Ssd :P

     

    Including pictures should probably be mandatory.

    Its a shame that we need to require it.

    Probably best to include pictures of both storage and cooling used.

    Clarification in PCMark'05 about Tweaks // Tricks allowed

    in Benchmark software

    Posted · Edited by SteveRo

    @Moose, why Virus Scan is so high, in Intel ? with 10 Ssd thats is correct ?(2094.77 MB/s) You have 10 Ssd ? :eek:, I never use many ssd., only 3 plextor

     

    http://hwbot.org/submission/2323631_moose83_pcmark_2005_core_i7_extreme_975_53547_marksP

     

    So, my friend please, explain this score, thanks so much :)

     

    Provided Master Moose used his areca 4gb + 10SSD for this - I believe 2000+MB/s virus might be achievable.

    Clarification in PCMark'05 about Tweaks // Tricks allowed

    in Benchmark software

    Posted · Edited by SteveRo

    I agree that we should NOT try to change who won what in previous competitions.

     

    However if you agree that we do not want to obviate storage then increasing disk sector size beyond the default 512 should not be allowed.

     

    Again, this trick requires only one SSD and a utility to increase the sector size from the default 512, this can be done on just about any motherboard with iodrive, probably several others.

     

    It could be - the true virus scan score for 4k sector score of 9500MB/s is simply –

     

    9500MB/S * (512/4096) = 1187.5MB/s

     

    Again, I agree that we should NOT try to change who won what in previous competitions but I believe that tricking the benchmark in this way should not be allowed and points (including previous points) should not be awarded for using it.

    Clarification in PCMark'05 about Tweaks // Tricks allowed

    in Benchmark software

    Posted

    Steve, check original post in thread. It has been updated.

     

    Also, regarding 'normal' scores, check Gluc's recent Core2Duo sub:

     

    http://hwbot.org/submission/2323429_gluvocio_pcmark_2005_core_2_e8400_(3.0ghz)_44665_marks

     

    33k Transparent Windows (with Mouse and Bob80's on 3960X @ 7970 I only get 25k)

    50 Web pages/s

    ~6k Audio/Video (about what I get with a 3960X OC + reg tweak, but he gets on C2D)

    345 pages/s Text Edit

     

    That's about right, though I'm not sure what he's doing Transparent Windows wise. Some other tweak I do not know probably

     

    Denis - If you would, please pm Mr. Pro with you tweaks for TW, thanks.

    Clarification in PCMark'05 about Tweaks // Tricks allowed

    in Benchmark software

    Posted

    Blocked most top scores. I will continue tomorrow, got 4.5h sleep tonight, so kinda tired :P I wonder if there will be any complaints... THanks fr the help on these numbers.

     

    WPR would be cool to know as well. And Audio compression.

     

    WPR is tougher to call, there are many legal tweaks for this subtest.

    On 6ghz ivy's - probably WPR of 75, (maybe even 80) and lower is good?

    For 1x semprons - WPR of 30 maybe 35 is probably good?

    Clarification in PCMark'05 about Tweaks // Tricks allowed

    in Benchmark software

    Posted · Edited by SteveRo

    Anything with TW over say maybe 40k, for sure anything over 50k was probably using (what is now) banned tweaks?

    Maybe less then that for 1x core?

     

    edit - if an individual thinks his tweak is legit - pm Master Pro and explain it to him but don't get your hopes up :)

    #2 edit - for older cpus (E84, E86, Q66, E66 ...) make that probably anything over 30k is probably using (what is now) banned tweaks?

    For 1x core probably anything over 15k is probably using (what is now) banned tweaks?

    Clarification in PCMark'05 about Tweaks // Tricks allowed

    in Benchmark software

    Posted

    Yes Steve Please Can you write correctly this in order that I edit the first post ? or any Guy with good writing in English ?

     

    In spanish few ones are going to understand it :D

     

    Thanks so much :)

     

     

     

    Better , much better, Meanwhile, the previous stays in functions from this night, night for me, until the wrapper are ready.

     

    Thanks by your clarification GEN, so shortly we will have unless problem

     

    Edit:

     

    Updated First post,

     

    ok I'll take a crack at verbiage for the additional screen shot requirement, note - might not be needed if the wrapper includes this. -

     

    1. "Futuremark link required for all top 20 submissions either global or hardware.

     

    2. For all non-top 20 submissions either Futuremark link or Pcm05 Results.txt required via screen shot."

     

    (This is a 1st cut, I'm sure others can improve on this).

×
×
  • Create New...