Jump to content
HWBOT Community Forums

Gautam

Members
  • Posts

    382
  • Joined

  • Last visited

Posts posted by Gautam

  1. I'm afraid the only scores that don't seem to be bugged are the 118k and 123k. The 92k seems bugged low. The jumps are very unnatural from 92 to 118, 200 FPS more in the last 3 game tests which are gpu bound with no increase in gpu clocks. Then once again they're absurdly high from 123k to 141k, esp with such minimal increases in clock speed. The 141 is already bugged, then the 164 is even more badly bugged (GT3 extremely bugged high in that one)

     

    Proper score around your speeds should be like this one, but quite a bit lower since your card clocks are much lower.

     

    Could mean a part of your system is unstable, maybe the CPU. Switching to a newer OS (Vista/7) and setting affinities could alleviate it.

  2. Doesn't make any sense to invalidate a score just because a screenshot isn't there. hwbot is supposed to be a database and a valid score is a valid score. You can't say it never happened just because a screenshot isn't there; the fm link is proof.

  3. This is very strange...but anyone would have to admit that there is consistency here :S

     

    The 03 bug strikes in strange places and spits scores out almost randomly. This is one big thing that people need to keep in mind when watching for bugs or cheats. The consistency. Cheaters usually aren't even consistent with themselves when it comes to efficiencies and subscores/game tests etc. Bugs affect tests and runs randomly. You rarely see a sensible scaling in CPU/GPU speeds versus scores when dealing with bugs. For the 03 black-screen bug, I've seen it hit any or all of the 4 game tests, and in any combination. I'll be the first to say that bob(nz)'s scores look impossible to me as well based on what I know, but right now I've got 3 runs of his available for me to see. Two being the ones he's posted in this thread and one on hwbot right now, an SLi 285 score. It's certainly way out of the norm for a single 4870X2 to outscore two 285's in Aquamark, but other than that, they are mostly consistent. The new Catalysts work wonders that are almost unbelievable in all the other benches. Conventionally, for example, a pair of 285's killed a pair of 4870's or a single 4870X2 in 05 canyon flight, but now, the latter appears to be superior. The AM3 subscores are wonky. With whatever ATi did starting with Catalyst 9.4 that gave ridiculous boosts in 03, 05, 06 and even 01 it wouldn't completely surprise me if it had a similar affect on the GPU subscore of AM3. Some recent testing I've been doing is forcing me to keep an open mind on the "validity" of these scores, especially about Vista. Has anyone else even tried Nehalem+Vista+4870X2 in AM3?

     

    Most of you, I'd imagine, would be rather surprised to know how much better Vista is than XP for Futuremark benches on Nehalem. (and yes, I mean all of them, not just PCMark) Very very few benchers have picked up on this, and it's hurting their scores something fierce. It's possible that Vista might be as strong in AM3 as well, and might just be an oversight on the part of hundreds of benchers. I can't ever say I tried. When I went to run AM3, I loaded up XP by default and ignored ATi altogether. But conventional wisdom has stopped holding true in plenty of things with Nehalem benching...so I wouldn't be quick to dismiss these scores either.

     

    Course there's a high likelihood I'm overthinking things and these scores are simply bugged. It's definitely given me something to test though...

  4. At the end of the day we need to step back and think about what the rules are for. As far as I'm concerned, if someone put in effort for a score, they deserve points, and hardware sharing rules are in place to prevent people from taking points for someone else's work...no?

     

    I think you're going a bit overboard with this Paul. Rules and regulations are important but for what end if we lose sight of what truly matters?

     

    massman and Pt1t got a great 32M run off. Who cares about where the hardware came from or who it "belongs" to anyways? I think if we waste time arguing over such technicalities, we quickly forget about the true spirit of competitive overclocking, something that's happening all too often these days.

  5. The obvious thing would be an extension ;) Vince aint stupid, the compo will be fine

     

    Yep Vince is extending it. Keep benching, I already know that there will be some surprises when the contest resumes. ;)

  6. Okay, I just got off the phone with Fugger. I know jack about databases, so I don't have a firm grasp on the situation. First of all, it seems like the database of the forums in their current state are somehow erroring. The database isn't starting up. The database import takes 2 days, and Charles has done this several times over the past week already as you might have guessed, running into problems ever time. Normally that 2 days should've been the upper bound on the downtime. I think he also said that the original hard drive SAS array got busted somehow. He's still trying to work through all the problems, as clearly plenty of things went wrong. Right now the compromise solution he's thinking of is putting up a backup of the forums while he does this, but the backup is from about a month ago, so the data won't be current...

  7. Bro, I really do not see any difference between my point and what any of the other guys posting here except you and Gautam. It's not about me, I really don't like PCM05 so much (you have to have "best appearance" set, and Windows Media Player or converter or whatever installed and so on...bleah!!! that is not a bench OS :D ) and I really do not care about PCM2005 so much....in fact...I pretty much dislike it, just like I really don't like Aquamark. But there are guys that do and I think that there should be a more relaxed dialogue between HWB Crew and those guys. HWB is first of all a community made of benchers, so when you want to make that kind of decisions you should also take into account the opinion of the community. It's not about what you want or about what you think, heck...if I would be HWB Crew I would probably take out Aquamark instantly, but probably that would not be fair to those who like it. Be more relaxed, see their point of view, and find a solution together. That is all that I am saying.

     

    And yes, you guys have a strong argument with 9000MB/s, we all know that that is not real. But neither are those those >1000 fps that we see in Nature or Mother Nature. Sure, you can have that ammount of FPS in an old game using new hardware, but besides the fact that most of the games have a cap, anything over a constant, solid minimum of 60-75 fps really does not matter. So the situation is kindda similar. Benching is one thing, real life is another thing :)

    No, those high FPS are what the GPU is actually rendering. It's real FPS.

     

    The closest analogy here is using GPU PhysX for the CPU tests in Vantage. It's not an accurate gauge of CPU power there, but of the GPU.

     

    PCMark doesn't test the hard drive when it gives such crazy results; it's just testing a cache in memory.

  8. Bingo! Exactly my point. PCMark05 scales perfectly with the drive technology you use. This is a justification AGAINST your argument.

     

    To address Gautham, yes, the 9000mbps test was obviously using some sort of caching. So what's the problem? It's identifyable, define some limits and let people enjoy the benchmark. ...just like what has happened with the 220mbps limit, and now the 300 mbps limit.

     

    It's funny how people didn't talk about dropping everyone's points for 3DMark03 when AMD GPU's started producing bugged runs.

     

    I'm gonna step out of this thread now as I've said my bit and just want to get on with benching, including PCMark05.

     

    If it's using caching then it's NOT scaling perfectly with drive technology...;)

  9. Let me attempt to simplify this debate a little bit. I don't really mind the HDD bias myself much, IF that were what the problem were. ;)

     

    If the hardware were legitimately capable of 9000+ MB/sec bandwidth then sure...but let's just step back to reality here? Do you any of you really believe that ANY RAID array is capable of that? You could string up a million SSD's for all I care and RAID 0 them, but that still won't take away the inexorable fact that your controller isn't going to capable of delivering much more than maybe a tenth of the bandwidth that the virus scan is claiming. (At best)

     

    RAMDAC's own HD tune benches show 773 MB/sec peak, and consequently we're supposed to just believe that the 9100 MB/sec PCMark is showing is accurate??

     

    The point is not just that the "scores are too high" it's just that the benchmark is obviously not even testing the hardware. Even though MFT is off in this case, somehow it's caching something for that test for sure. It's just common sense...

     

    Let me reiterate this in case it's still not clear. The problem is not that the hardware is "too new" or "too good"; it's that PCMark isn't testing it accurately

  10. With 2 OCZ APEX SSD I had an average rate of ca. 350MB/s, which was good but with 4 OCZ in RAID 0, I had just about 460MB/s, then with 6 APEX SSDs- it was a bit more than 550MB/s.

     

    And yet your HDD Virus Scan was 8979 MB/s. :P:

     

    Yeah it's looking like time to can this benchmark...it's no fault of the community, it's just that the bench itself was poorly designed from the getgo and we've just all been working around it's shortcomings. =

×
×
  • Create New...