Announcement

Collapse
No announcement yet.

Two identical G450s delivering 8-17% different 3Dmark 2000/2001 scores

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Two identical G450s delivering 8-17% different 3Dmark 2000/2001 scores

    Hello,

    I have a PC (P3-S Tualatin 1400/133/512K, Asus TUSL2-C, 2 x 256 MB PC133 CL2 SDRAM) with a Matrox G450 AGP in it.

    I recently acquired yet another, identical Matrox G450 card. So I decided to see if these two cards deliver the same 3D performance, just for the fun and sake of experiment.

    =================

    Both cards have these markings:

    model: G45+MDHA16D/OEM
    PCB: 971-0301 REV.: A
    RAM: SAMSUNG K4D62323HA QC60

    I flashed both cards to BIOS v2.1.37.

    I checked working parameters of both cards using Matrox MTSTU. Both were using AGP 4x with sideband addressing ON.

    Conclusion: everything about them seems to be identical.

    ==============

    I inserted a blank HDD and installed a fresh Windows 2000 SP4+ system (patched with AutoPatcher 2000), with only necessary drivers (Directx 9.0c, Matrox classic drivers v5.96, no sound or network drivers, etc.).

    I performed batch benchmarking using 3Dmark 2000 and 3DMark 2001. I have all results saved as separate TXT files (autimatically generated by 3DMark).

    After I finished benchmarking the first card (mine), I swapped the cards (without reinstalling the driver or anything - there's no need!), and repeated all benchmarks with the other G450 card.

    ==========

    I was ASTOUNDED when the results showed 8-17% differences (depends on resolution and color depth) in 3DMark scores between these two cards!!


    I can not explain this phenomenon. If anybody is interested in this case, I will send him all 3Dmark 2000 and 2001 TXT results for inspection.


    Here's an example result from 3DMark 2000:

    SETTING:
    Matrox Millennium G450 DualHead - English 1024x768x16 Z16 DB D3D Software T&L

    CARD 1: 2145 3D marks
    CARD 2: 2437 3D marks

  • #2
    Isn't there an option in the bios to have it re-assign IRQs, etc...?
    Maybe both cards are assigned different stuff....

    I know it is a long shot, but other than that I have no clue...
    pixar
    Dream as if you'll live forever. Live as if you'll die tomorrow. (James Dean)

    Comment


    • #3
      The main BIOS (on the motherboard) settings were:

      AGP 4x
      AGP aperture size 64 MB
      USWC video mode
      All SDRAM latencies at 2T
      All PCI IRQs are manually set (SCSI on IRQ 10, LAN on IRQ 4, sound on IRQ 5, TV on IRQ 7, AGP on IRQ 11, etc.).
      Windows 2000 is installed in non-ACPI mode (standard PC HAL), so it did not mess up IRQ settings, like it would have if it was in ACPI mode.

      Anyway, the BIOS settings remained the same with both cards, so it should not be a cause. And as I said, I checked the working parameters of both cards when they were installed using Matrox MTSTU. both were running at AGP 4x with sideband addressing ON and fast writes OFF.

      Since I flashed the BIOS-es of both cards prior to testing, that should reset any possible hacks or fiddlings in their BIOS-es which somebody might had done in the past.


      Wierd, very weird :S

      Comment


      • #4
        So, the G450 was a good performer after all??
        Join MURCs Distributed Computing effort for Rosetta@Home and help fight Alzheimers, Cancer, Mad Cow disease and rising oil prices.
        [...]the pervading principle and abiding test of good breeding is the requirement of a substantial and patent waste of time. - Veblen

        Comment


        • #5
          Would GPU-z detect them to see if they are running the exact same speeds ?
          Not at work so can't check with the 450 in my work pc...
          PC-1 Fractal Design Arc Mini R2, 3800X, Asus B450M-PRO mATX, 2x8GB B-die@3800C16, AMD Vega64, Seasonic 850W Gold, Black Ice Nemesis/Laing DDC/EKWB 240 Loop (VRM>CPU>GPU), Noctua Fans.
          Nas : i3/itx/2x4GB/8x4TB BTRFS/Raid6 (7 + Hotspare) Xpenology
          +++ : FSP Nano 800VA (Pi's+switch) + 1600VA (PC-1+Nas)

          Comment


          • #6
            I checked with GPU-Z, and both cards are running at identical clock speeds.

            Comment


            • #7
              Just checked here, and it doesn't detect my G450, it detects the deactivated i915 driver.....
              I cannot help you any further, sorry....
              i could try 3DMark on this PC, but its already bogged down with crappy software....
              PC-1 Fractal Design Arc Mini R2, 3800X, Asus B450M-PRO mATX, 2x8GB B-die@3800C16, AMD Vega64, Seasonic 850W Gold, Black Ice Nemesis/Laing DDC/EKWB 240 Loop (VRM>CPU>GPU), Noctua Fans.
              Nas : i3/itx/2x4GB/8x4TB BTRFS/Raid6 (7 + Hotspare) Xpenology
              +++ : FSP Nano 800VA (Pi's+switch) + 1600VA (PC-1+Nas)

              Comment


              • #8
                Seems from your post you only performed the test once...3D mark always had some fluctutations, perhaps you've just stumbled upon results which are close to sensible minimum/maximum for G450.

                Do numerous tests.

                Comment


                • #9
                  have you checked that it's physically identical memory chips on the cards?
                  If there's artificial intelligence, there's bound to be some artificial stupidity.

                  Jeremy Clarkson "806 brake horsepower..and that on that limp wrist faerie liquid the Americans call petrol, if you run it on the more explosive jungle juice we have in Europe you'd be getting 850 brake horsepower..."

                  Comment


                  • #10
                    Memory latency...good one...
                    PC-1 Fractal Design Arc Mini R2, 3800X, Asus B450M-PRO mATX, 2x8GB B-die@3800C16, AMD Vega64, Seasonic 850W Gold, Black Ice Nemesis/Laing DDC/EKWB 240 Loop (VRM>CPU>GPU), Noctua Fans.
                    Nas : i3/itx/2x4GB/8x4TB BTRFS/Raid6 (7 + Hotspare) Xpenology
                    +++ : FSP Nano 800VA (Pi's+switch) + 1600VA (PC-1+Nas)

                    Comment


                    • #11
                      I seem to recall that matrox did vary clock speeds for oem g450's. When called on it they more less just said so what.

                      Comment


                      • #12
                        @Nowhere: I performed batch tests on each card THREE times, and fluctuations on each card were less than 2% (usually less than 1%).

                        @Technoid: The memory chips are physically identical, and have these same markings: SAMSUNG K4D62323HA QC60


                        Anyway, I sold the slower card yesterday to an old-school photoshoper for 10 EUR (never miss a good opportunity!), so I can't perform any additional comparisions and verifications.

                        Comment


                        • #13
                          Good price, I bought G400 DH AGP4x for SIT 2500 (10.50 EUR) 4 years ago. I then sold my G400 16MB SH. I don't have the heart to sell my G400 and Parhelia, the Millennium PCI is still chugging in my server.

                          Comment

                          Working...
                          X