8800GTX - lockups but no indication of the problem

Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by Ray Proudfoot, May 25, 2008.

  1. The_Swead

    The_Swead Member

    Messages:
    32
    Likes Received:
    0
    GPU:
    asus gtx280/4870X2
    I havent read all post so i dont now if someone already suggested this but
    set frames to render ahead to 6 or 8 and see if it helps .
     
  2. mikeyakame

    mikeyakame Master Guru

    Messages:
    207
    Likes Received:
    0
    GPU:
    GTX280 729/1566/2700
    i'm a take a shot in the dark here....but any ram errors will effect everything, graphics, cpu, etc, since the fundamentals of running any software which requires memory for storage are making 'malloc' calls, whether it be directX or not, still requires allocating memory, copying memory, moving memory and freeing memory...if you have even one error in memtest you have a big problem....

    fix your ram problem before you go and waste your time rma'ing a working card....either you have really bad luck (same problem with your previous card?) or your memory dimms are simply bad....you cant troubleshoot a system with memory errors, any results you get from such an effort would be invalid due to ERRORS when reading or writing areas of memory...

    food for thought.

    edit:

    the comments you made about system responsiveness are definitely RAM related, what happens when data is either read from ram, or written to ram is, there is usual sanity and data checks, you malloc dynamic memory for storage object, you declare the physical object, you allocate the object's address a pointer which points to its storage data, then you dereference the pointer to write data to the memory location, and finally you do simple truth comparisons, and check the return data value with the value you stored....if this fails, memory is reallocated elsewhere by the os, and the process is repeated, incorrectly stored data can easily cause OS lag in response, cause it could mean having to re-evaluate large blocks of data storage and repeated execution of code blocks to attain that data.....

    ie. fsx is dx9 based, and dx9 maintains two identical memory allocations in both gpu ram and system ram, if the data stored in the gpu and system ram is not the same bit for bit, kernel may assume paging fault or corruption in data, and this leads to freeing for example a 500mb of texture data then writing that texture data back to both system ram and gpu ram then checking sanity once more, one corrupt bit of data can cause all sorts of mayhem on a system, from slow downs to complete lock ups/crashes. data in malloc must always be bit for bit true with the source from which the data was read.
     
    Last edited: May 29, 2008
  3. linyanti

    linyanti Master Guru

    Messages:
    372
    Likes Received:
    0
    GPU:
    Zotac 1080 GTX Founders C
    let me say this about that

    Memory issues should be fixed before proceeding any further.

    Be that as it may, here is an update to my situation:

    The two gpu's in sli are underclocked to the minimum using EVGA Precision Tune. It reduces the clock speeds about fifteen percent for these cards.

    It does not stop the error but it enables the system to keep going. When I join a round for the first time that day the screen flickers white and black vigorously for five seconds and then the monitor is off for five seconds and then the game starts and the performance is glorious.

    Here are the stats: Most of the time the FPS is one hundred. It will dip to 60 now and then. When the error happens (and this will only happen at the very beginning) the FPS drops down to 2 and the flickering starts. The temperature on the GPU is 55.

    I do not think that it is temperature because after a really intense hour long session the GPU was up to 74. Cooking, in spite of two 120 mm fans, a 5.25 bay fan, three 80 mm fans, V1 CPU fan and a fully vented (nine additional 80 mm cut ports) Coolermaster Ammo Case (great case) and the performance was fantastic. I have taken the side panel off my computer. This was done to prevent fire and not to improve performance.
     
  4. mikeyakame

    mikeyakame Master Guru

    Messages:
    207
    Likes Received:
    0
    GPU:
    GTX280 729/1566/2700
    Question mate. The toughpower you are running is a quad rail is it not? Keep an eye on the +12v and +3.3v. If either dip too much or sharply under sudden load, you will lose the display, if too suddenly it wont re-post and refresh.

    Rivatuner logging would be one approach to this, using Everest plugin for voltage readings. This way you can log to file, and even if the system crashes, you can open the logs on reboot and see what the system was doing just before it locked up. Might give you some insight on the problem.

    I've had problems with my old TT Toughpower 750W and single 8800GTX, their quad rail design struggles to keep the voltage stable and smooth for a brief moment in cases where the load rises too rapidly for it to correct and compensate the spike.

    There was a point in clocks where it would fail near immediately, ever since changing to a big single rail PSU I found the problem disappeared, and hasn't arisen there after. I happily run the card clocked at 675/1512/2052 on the standard leadtek cooler and Ultra bios... the core clock is the kicker as it's usually near impossible to get the card stable at those clocks without increased voltage.

    In my experience i've seen the strangest problems with next to no obvious relationship to insufficient power supply, be just that. I try and think outside the box these days when it comes to problems I can't pinpoint or come to a conclusion for.
    Advanced Gunning Transceiver Logic (A/GTL) is way too volatile when the signal noise or jitter is close to the switching threshold. Too little room for adequate noise compensation mixed with noisy voltage stepping circuitry in the power supply, makes for a problem with no obvious cause.

    Edit:

    The only constant you are changing by reducing the clocks on both cards below reference is the amount of voltage required to operate at those clocks. The card will only draw as much voltage as the chips and switching circuitry need to process the supplied operational workload.

    One thing you would be achieving in this experiment, is finding the point of where your power supply begins to rapidly show the inability to manage rapid spikes and rapid drops, where the given output vs peak demand where the value of Amax ( highest current draw ) causes the resistance circuitry to work outside the maximum variable correction of its design, needed to keep the voltage at the 12V side of the rails within the stable operating boundaries of the hardware it supplies.
     
    Last edited: May 29, 2008

  5. linyanti

    linyanti Master Guru

    Messages:
    372
    Likes Received:
    0
    GPU:
    Zotac 1080 GTX Founders C
    it does have four rails and runs at 11.7V

    My Thermaltake Toughpower 850 has four rails. Everest has it maintaining 11.7 V.

    I will log it and see what happens. It may be power supply related in that the issue occurs in the first few minutes of the game and that is when there is a large load on the gpu's.

    What I may do is change the power cables. I have the modular model which has two 6pin/8pin gpu power cables hardwired from the psu and two optional modular 6pin/8pin cables to plug in. They recommend using one hard wired cable and one modular cable for 2 gpu sli but I may switch it around some and see what happens.

    If it is still a problem then I could get a single rail psu or see if the new gpu cards coming out suck up less juice.

    Thank you for your idea, I will check it out. Fortunately, the temporary fix works and it gives me some time to tinker around with it.

    Update: I ran Everest with logging. I did stock clocks first and then ran the underclocked gpu and it actually crashed to BSOD with the stock gpu clocks. Usually it just crashes to desktop. After examining the logs there was no difference in CPU, memory, gpu or any difference at all in voltage. It was all rock steady. The only difference was that the temps got hotter with the underclock. That is because I was able to play hard for thirty minutes before taking a break versus only five minutes before it broke down.

    Thanks for the idea about logging. I had never done the HTML format that comes with Everest. It was easy to do and to read. It has a color coded bar everytime there was an event such as executing punkbuster or the crash.
     
    Last edited: May 30, 2008
  6. Ray Proudfoot

    Ray Proudfoot Guest

    Messages:
    101
    Likes Received:
    0
    GPU:
    Asus Extreme 8800GTX
    I'm in the UK but thanks anyay. The card was confirmed as faulty today and I swapped it for an Asus Extreme 8800GTX. Memory speed is 575 compared to the XFX 600. Doesn't seem to affect FSX.

    I ran a graphics stress test for 1.25hrs and the temp of the card never rose above 84C. I'm a happy bunny and I've learned a lot about useful software. Thanks to everyone who helped especially Tommy.

    I need to address the memory errors next. I'll raise that in General Hardware in he next few days.
     
  7. linyanti

    linyanti Master Guru

    Messages:
    372
    Likes Received:
    0
    GPU:
    Zotac 1080 GTX Founders C


    Fantastic! glad that was solved.
     
  8. Lex Luthor

    Lex Luthor Master Guru

    Messages:
    781
    Likes Received:
    356
    GPU:
    RTX 3060
    Gratz

    Lex
     
  9. opozoid25

    opozoid25 Member Guru

    Messages:
    118
    Likes Received:
    0
    GPU:
    2x BFG GTX 285 OC SLI
    I just noticed this thread and have some advice for the OP (if still needed). If your problems persist, keep the following 2 things in mind:
    (1) Many ppl including myself have had stability problems with the 170 series drivers and 8800GTX/GTS in Vista - most seem to report that 169.25 is stable, and has been for me too.
    (2) Your memtest errors could be northbridge related. Your P35 board should have a voltage in bios for "north bridge" or "MCH" or something like that, and the default is 1.25v. I had to increase mine to 1.33v to gain perfect stability. You may wish to try increasing it - but I would recommend not going above 1.45v.
    Hope this helps.
     
  10. Ray Proudfoot

    Ray Proudfoot Guest

    Messages:
    101
    Likes Received:
    0
    GPU:
    Asus Extreme 8800GTX
    Thank you for your helpful post. I haven't yet tried the 17x drivers. I'm currently using the Omega-tweaked 169.25 ones. With the new card things have been fine.

    I'll look at the voltage for Northbridge. I suspect it's set to Auto. I'm always cautious about changing these things. But the number of reported errors in Memtest does seem linked to voltage settings so this is worth investigating.

    Once I raise the topic in GH I look forward to solving these irritating supposed errors.
     

  11. ---TK---

    ---TK--- Guest

    Messages:
    22,104
    Likes Received:
    3
    GPU:
    2x 980Ti Gaming 1430/7296
    http://www.techpowerup.com/gpuz/wd8xs/ these 8800gtx`s are capable of much more than the standard 575/1350/900. mine is factory overclocked but wont go any higher
     
  12. Ray Proudfoot

    Ray Proudfoot Guest

    Messages:
    101
    Likes Received:
    0
    GPU:
    Asus Extreme 8800GTX
    Does o/clocking the GPU by 10% bring the same dividends as o/clocking the CPU by the same amount?

    I will certainly look at this once I'm ready to overclock. I need a good CPU cooler first. For ease of fitting I'm looking at the Arctic Cooler Freezer 7 Pro. The Zalman requires the mobo to be removed - no thanks.
     
  13. ---TK---

    ---TK--- Guest

    Messages:
    22,104
    Likes Received:
    3
    GPU:
    2x 980Ti Gaming 1430/7296
    i dont know about the cpu gpu comparison. but you have an asus motherboard? do you have by chance an option in the bios to DISABLE LEGACY USB SUPPORT? i had to disable legacy usb support because i was getting errors in memtest.
     
  14. Ray Proudfoot

    Ray Proudfoot Guest

    Messages:
    101
    Likes Received:
    0
    GPU:
    Asus Extreme 8800GTX
    Yes, the P5K-E Wi-Fi. I will have a look tomorrow and report back. I remember the setting but I'm not sure what it's set to.

    Thanks.
     
  15. mikeyakame

    mikeyakame Master Guru

    Messages:
    207
    Likes Received:
    0
    GPU:
    GTX280 729/1566/2700
    well if overclocking gpus were pointless. i wouldnt be running my 8800gtx at 100mhz core, 162mhz shader, 252mhz mem, above reference. as well as most of the others here. a card only overclocks as good as the chip scales, and the other components on the card.
     

  16. Ray Proudfoot

    Ray Proudfoot Guest

    Messages:
    101
    Likes Received:
    0
    GPU:
    Asus Extreme 8800GTX
    Perhaps the benefits depend on the game being played. I only 'play' Flight Simulator which is CPU-bound. Other games aren't so might benefit more than this one.

    All the talk on FS Forums is about o/clocking the CPU rather than the GPU.
     
  17. Ray Proudfoot

    Ray Proudfoot Guest

    Messages:
    101
    Likes Received:
    0
    GPU:
    Asus Extreme 8800GTX
    I'm opening a new thread in General Hardware on the MemTest problems. Seems more logical than discussing it here.
     
  18. ---TK---

    ---TK--- Guest

    Messages:
    22,104
    Likes Received:
    3
    GPU:
    2x 980Ti Gaming 1430/7296
    disable legacy usb support in bios and retest your memory dude!
     
  19. Ray Proudfoot

    Ray Proudfoot Guest

    Messages:
    101
    Likes Received:
    0
    GPU:
    Asus Extreme 8800GTX
    I did. It seems worse now. Anyway, please see my other message.
     
  20. danielmpr

    danielmpr New Member

    Messages:
    1
    Likes Received:
    0
    GPU:
    MSI 8400GS TD-256
    I also have random lockups, but only with graphics intensive games!, like BioSHock FSX, even FS2004!, other games don't lock up, even after hours!, my card is a MSI 8400GS TD256 wich I bought @ work, the idle temp reported by Everest is @ 51C, is that hot? I'm used to farenheit!, also, how the heck u turn off TurboCache? my vidcard reports as having 512MB!!!, when it's 256MB!, and I only have 1gb of ram!???, are these cards known to cause headaches?, is there a BIOS upgrade to fix this?, if I use the ATI onboard video of my m/b, have no problems!, but I like NVIDIA better, and as I said before, I only have 1gb, so if I use my internal video @ 256MB, i'm left with less then 768MB!, :bang:

    P.S. I DON'T OVERCLOCK!, NOT CRAZY ENOUGH TO DO THIS!
     

Share This Page