Jump to content
HWBOT Community Forums

FM_Jarnis

Members
  • Posts

    188
  • Joined

  • Last visited

Posts posted by FM_Jarnis

  1. One small note: 3DMark 2001 and PCMark 2002 uploads to 3dmark.com are not supported. Haven't been supported for some time now, so this is nothing new. Sorry, no way to support them without digging up 10 year old backup tapes for code, then disassembling and reverse-engineering it as no source code for the receiving server for these benchmarks exists any more :( - not going to happen

     

    Also for the rest of these legacy benchmarks (03 and onwards, in essence benchmarks that use "modern" SystemInfo): The result upload works for now, but since the products are no longer officially supported, we can't guarantee this will stay so forever. We will not intentionally disable existing support, but if at some point keeping them working would require too large additional development investment, we may stop supporting them on 3DMark.com.

     

    Now realistically this would happen only if we embark on a major rewrite of the whole online result system and no such thing is planned in the foreseeable future.

  2. "Time measurement data not available"

     

    Means that SystemInfo process that is tracking the timers crashed during the run and could not record the timers at the end of the benchmark. Usually this means your overclock is not fully stable. Note that if you get this error on Windows 8, it means "we can't tell if your result is affected by the downclock bug or not". On Windows 7 it means "we can't tell if your result is affected by some fairly obscure BIOS level "tweaks" that mess the timers". It does not automatically mean that the result is invalid, but it also means that the built-in timer checks are not present for this result, so it MAY be invalid, can't tell.

     

    "Time measurement data inaccurate"

     

    ..means that timers were properly recorded and they are off - something is causing different timers in your system to be out of sync. Most commonly this is due to the Windows 8 downclock bug. Result with this error is definitely invalid.

  3. Im playing with 780 Ti now for some days. When i run newer Benchmarks, ala Catzilla or Heaven everyrting is fine an card stays in 3D state. But when i run 3DM01-3DM06 with 780 TI, cards sometimes lost 3d state and turned back to low 3D or 2d clocks :(

    How can i fix this issue? LN2 Bios is flashed on TI, but same issue on normal Bios, how can i force 3d clocks on 780 Ti all the time?

     

    Not unexpected considering older 3DMarks (06 and older) are effectively CPU marks at this point...

     

    Same probably will happen also with Ice Storm which is too lightweight of a test for high end dedicated GPUs.

  4. Just a heads up, if someone doesn't have a 3DMark Advanced Edition yet (I know, most at HWbot do and some probably have press licenses and so on...), there is a good opportunity of getting it on the cheap for the next ~20 hours; it is available at 90% off on Steam in the Steam holiday sale.

     

    (key from steam version can be used with standalone version)

  5. Ours are on overclocked systems. I keep a overclock list thread. Currently allowing "not available", but not "not accurate" with Win7.

     

    Systems may not be fully stable, but stability is not a requirement, just a result. As long as it's stable enough to complete the bench. Quite common with Win7 to get "not available" and that's with the latest sysinfo..

     

    Up to you what rules you follow.

     

    Futuremark Hall of Fame requires systems that are actually stable enough to keep SystemInfo process from crashing during the run (which would lead to "not available" error with latest SystemInfo). This is because without it, there is no way to tell if RTC was being messed with.

     

    Note that allowing "not available" but disallowing "not accurate" means by your rules it is possible to cheat with the RTC. Would be trivially easy to set up so that SI process is killed during the benchmark run to "cover" any RTC mungling... And yes, it is confirmed possible to cheat the RTC also on Win 7. Just not as trivially (or by accident) as on Windows 8.

     

    If you have a reproducible "not available" error on confirmed stable system, please email us the details (hardware config, any result files you have etc.) and we'll investigate.

  6. Was running my 770 today, and noticed that the vast majority of Firestrike Extreme subs are illegal.

     

    Reported a bunch, only one gone so far.

     

    What do you mean? PhysX hasn't been used since 3DMark Vantage.

     

    3DMark Fire Strike uses DirectCompute for GPU physics.

     

    In which way you think the submits are "illegal"? Are you talking about HWBot rules or 3DMark Hall of Fame?

     

    Send any suspect score links and state your case against them to info [at] futuremark.com and we'll take a look.

  7. This message means "SystemInfo was disabled or your SystemInfo was older than 4.19, so there is no data about the RTC clocks in the result file. We can't say if your result had funky RTC business or not".

     

    And yes, the RTC verification is done on all OSes and on all benchmarks as there are other ways to trick the same effect that Win 8 caused unexpectedly (nope, not going to detail on that...)

     

    If you are getting this on SI 4.23, it would imply that the system is not fully stable (the SI process crashed during the benchmark run so it could not get the required RTC data)

     

    If you see this on stock clock system, we would like to hear all the details.

  8. No scores with SI older than 4.22 will appear in HoF.

    Existing scores with SI older than 4.22 will be gone from HoF.

    Nothing will change with scores on user accounts (we'd never delete any user scores) and they will not receive any kind of extra flags.

     

    So this is purely a change to Hall of Fame eligibility.

     

    We will also at some point in the near future bump up the minimum SI version for submitting at all (probably to some early SI 4.xx version) during the process of cleaning up the receiver code. I can't recall what the current minimum is but I do know that truly ancient (3.xx something) already cannot submit. Again, this will not influence any scores already on your accounts.

  9. Just an early heads up; within the next two weeks, Futuremark Hall of Fame will raise the minimum SystemInfo version requirement to version 4.22 or later.

     

    So if you are doing Serious Benching, make sure SI is updated or your record might be gone from FM Hall of Fame in a week or two.

     

    We are also in final stages of getting 3DMark v1.2 update ready. Major bits include fixed hardware monitoring and greatly improved GPU detection (featuring licensed GPU-Z dll - why reinvent the wheel and all that), and the inclusion of Ice Storm Unlimited tests (to keep Ice Storm in parity with the mobile versions).

     

    Should be available in a week or two.

  10. ID = 19411383

    CREATED = 2013-09-30 05:32:52.0

    STATUS_CODE = REJECTED_SYSTEM_INFO_VERSION

    PRODUCT_VERSION = 3.6.0

    SYSTEM_INFO_VERSION = 3.6

    ERRORS = (Unsupported SystemInfo version: 3.6)

     

    Simple as that. Install a modern SystemInfo. If you cannot update it (03 still sees the old one) then you are using the old version of 3DMark03 that has embedded SI. In such case you have to uninstall 3DMark03 and SystemInfo, then install 03 and latest SystemInfo standalone installer.

     

    You can download 03 from this page; http://www.futuremark.com/support/downloads

     

    ...and latest SystemInfo here: http://www.futuremark.com/downloads/Futuremark_SystemInfo/latest

  11. Investigating those at our end. If the 3dmark.com results turn into "invalid" we found them to be bogus. Stand by.

     

    Edit; Second of those showing completely bogus GT2 score, result is invalidated.

     

    Edit2; Yes, also the first result is conclusively checked to be bogus. Sorry.

     

    Dunno if the odd (7970 + 7950 in mixed crossfire) GPU config plays a part here. We really haven't tested on such a config.

  12. Recommend trying to dig up older drivers. Seen some driver "rot" in our testing when doing stuff with the "museum collection" in FM test area - later drivers not properly supporting very old cards in all applications even if technically the card is listed in the driver.

  13. Also you might want to take into account the fact that by the time Windows XP is officially retired by Microsoft (XP being the target platform of PCMark05), it is likely that Futuremark will also stop supporting it. It is, after all, 8 year old benchmark already superseded by two more modern PCMarks (with third on the way).

     

    I believe Microsoft's official "XP game over" day is in spring 2014.

  14. If you change the timer to one that actually works, and then change the bclk, I guess those results won't be marked as invalid?

     

    They are marked as invalid only if RTC is skewed (ie. actual time taken during benchmark does not match the time taken as per RTC). It doesn't care one bit what your BCLK is, it just validates the time from multiple sources and makes sure they agree with each other.

×
×
  • Create New...