Results 1 to 7 of 7
  1. #1
    New Member blinkieleblind's Avatar
    Join Date
    Jun 2010
    Location
    London, UK
    Posts
    5

    round off error checking?!?!

    hi,

    so i am have been running the large ffts torture test on p95 for about 11 hours now but i forgot to tag round off error checking on the options after installing new OS. i realise that p95 wont inform me of errors but usually when a core fails on me, realtemp etc tell me that the cpu load has dropped to 75% ish.
    This hasn't happened, realtemp still says 100% so i assume that no core has failed??? or does the round of error checking prevent a core from restarting in some way?
    i.e can i assume i have had no core failures for 11hrs p95 and i am still stable or do i throw away the past 11 hrs of test?
    Q8200 @ 3.15 Ghz | Biostar TP45 HP | Titan Fenrir | HD5770 @ 1 Ghz | Corsair Dominator 4GB 1066 |
    Akasa 850W powermax | Xigmatek Utgard

  2. #2
    Member Seanie's Show's Avatar
    Join Date
    Mar 2010
    Location
    Dudley, West Midlands
    Posts
    227
    Large FFT's test ram, so I would say your RAM failed the test, hence why all your cores are still running, Small FFT's test your CPU and Cache, and blend is a bit of both, thats why its best to test blend, although rather than testing in blend select custom and make sure you enter all your free/available RAM into the little box before hitting start, leave everything else in that window set to default.
    Attached Images Attached Images  
    Intel Core i7 920 @ 4.3ghz 1.328v
    6gb (3x2gb) Patriot Viper Extreme @ 1640mhz 8-8-8-24-2t 1.66v
    Gigabyte GA-X58a-UD3r 1.38v NB, 1.34Vtt
    2 x Sapphire ATI Radeon HD5830 1gb GDDR5 Crossfire
    1 x OCZ Vertex 96gb SSD AHCI
    1 x Samsung Spinpoint F1 320gb Hard Drive
    Creative Soundblaster Titanium PCI-e x1
    Sony Optiarc DRU-870S 24x DVD±RW SATA ReWriter
    Swiftech MCP350 10W, 360mm XFlow Rad, DD Stelth 240mm Rad, Swiftech MCR220 240mm Rad
    EK Supreme HF i7 CPU Block, OCuk 3/8" UV blue Tube
    Coolermaster CM690 II Advanced, Windows Se7en x64 Pro

  3. #3
    New Member blinkieleblind's Avatar
    Join Date
    Jun 2010
    Location
    London, UK
    Posts
    5
    yeah thats the way i thought it worked. I am a noob to OC'ing and i have been using greyskys guide ( http://forums.bit-tech.net/showthread.php?t=149050 ) as a walkthrough, he suggests using the large ffts as a good stability test and i have heard this from others too. afterall, if a core fails a core fails although i see your point that it is more of a ram test.
    good thing is that after i stopped the torture test last night, it reported no errors so im hoping i can get stability.
    so whats the best method to test stability? i have read that some use small ffts for 2 hours and then blend for 8-12hrs and some just use OCCT which i had never heard of till last night.
    Q8200 @ 3.15 Ghz | Biostar TP45 HP | Titan Fenrir | HD5770 @ 1 Ghz | Corsair Dominator 4GB 1066 |
    Akasa 850W powermax | Xigmatek Utgard

  4. #4
    Younglin's Avatar
    Join Date
    Jan 2010
    Location
    Calgary, Alberta, CANADA!
    Posts
    852
    JonSimonzi already has a ziodberg Jesus avatar. This could become confusing. XD

    As for a stability test intel burn test works well. Might have to run 3 or 4 simultaneously to get the entire cpu to 100% but let it run for 3-4 hours and it is a good indicator of cpu stability. MEMtest works well for stability testing memory.

    Honestly there are tons of programs for stability testing. If you google it ( I.E. " cpu stability test") you'll find a lot of good programs. Some are a bit dated and some of poor quality but generally any test that keeps 100% load will do.
    Imperfection is beauty, madness is genius and it is better to be absolutely ridiculous than absolutely boring. - Marilyn Monroe.

  5. #5
    New Member blinkieleblind's Avatar
    Join Date
    Jun 2010
    Location
    London, UK
    Posts
    5
    Damnit, i guess i will have to change my avatar

    on an initial note, after reading through the p95 documentation i conclude (and this is my opinion) that using the large ffts on p95 is the best way of determining general stability for long periods using p95.
    small ffts is a good way of testing the cpu this is true as everything fits into the cache and is probably the best way of testing an overclock initially.
    blend i do not see the point of really as it basically saturates your ram and tests it by storing results from the ffts for easy access whilst stressing your cpu to perform the calculations; this really looks to me like an overclock memory test whereas memtest is much better at this as it is designed to stress the memory using complex bits of code specifically written for this purpose.
    as it states large ffts is designed for maximum heat output which, correct me if im wrong, is one of the main things oc'ers want to minimize. it stresses the cpu evenly using longer fourier transforms than small ffts and for this reason needs to store some of the information in the ram (wont fit in cache). so to test endurance surely large ffts is the way forward when using p95 as it tests the cpu, ram and NB(fsb) to some extent, which to have a stable system you will want all these to perform without failing for the coveted >24hrs le mans torture test.
    this is my understanding after looking over the documentation but i suppose that experience counts for a lot more, so if anyone wants to correct me........ as my old head chef used to say "everyday is a learning day"..... god i hated him!!
    Q8200 @ 3.15 Ghz | Biostar TP45 HP | Titan Fenrir | HD5770 @ 1 Ghz | Corsair Dominator 4GB 1066 |
    Akasa 850W powermax | Xigmatek Utgard

  6. #6
    Younglin's Avatar
    Join Date
    Jan 2010
    Location
    Calgary, Alberta, CANADA!
    Posts
    852
    Makes sense to me. =]
    Imperfection is beauty, madness is genius and it is better to be absolutely ridiculous than absolutely boring. - Marilyn Monroe.

  7. #7
    Researches Meritless
    LIES for the Front
    Page and Super Mutterator

    Overclockers.com Editor
    First Responders

    EarthDog's Avatar
    Join Date
    Dec 2008
    Location
    Stuck in Maryland...
    Posts
    45,434
    Large FFT tests SOME ram....so its not a shoe in thats its the ram if it fails.

    Small FFT = CPU, Large FFT = CPU and some of ram, Blend = CPU + Ram....(Just like the screen says!!!!) Blend has hsitorically been the best for system stability.

    "We have more information and more ways of accessing it than ever, yet seem increasingly less inclined to do so."- Michael Wilbon

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •