TechSpot

Insanely horrible 3dMarks

By SubKamran
Jan 27, 2004
Post New Reply
  1. Here's my word document pasted:

    Specs:
    • Pentium 4 2.4CGhz (800FSB) Northwood
    • OCZ EL 256MBx2 (512) Platinum PC3200 DDR RAM
    • Asus P4P800-Deluxe Motherboard
    • Sapphire Radeon 9800 Pro 128MB

    Test 1

    • CPU FSB @ 240Mhz = 2.8Ghz
    • DRAM Timing: 320Mhz (5:4 divider)
    • DRAM @ 2-2-3-5
    • Catalyst v4.1

    3dMark03 Standard Score: 2084 Marks

    OK. THAT REALLY IS HORRIBLE. Something's wrong, no? Is it my drivers? ARGH!
     
  2. MrGaribaldi

    MrGaribaldi TechSpot Ambassador Posts: 2,512

    Are you perhaps running with FSAA & AF enabled in the control panel? Or (for some reason) with TruForm?

    You're not using Seti@Home or any other DC program in the background?

    Considering I'm gettin way more marks than that, and I'm only on a 9700pro, you have to have goofed in one of your settings...
     
  3. SubKamran

    SubKamran TS Rookie Topic Starter Posts: 166

    Yes, it's maxed out with TruForm.

    Is that bad? :D I'll try it with TruForm disabled and Application Preference.
     
  4. SubKamran

    SubKamran TS Rookie Topic Starter Posts: 166

    OK, test 2, much better :D

     
  5. farmer

    farmer TS Rookie Posts: 48

    You ought to try 3DMark 01 SubKamran. It does a better job of showing what the whole system is doing. 03 is largely video card dependent...

    You've a 20k 3DMark 01 rig there, waiting to bust out...

    Later....
     
  6. ---agissi---

    ---agissi--- TechSpot Paladin Posts: 1,977   +15

    Yeah I dont use '03, I recon '01 was much better.
     
  7. SubKamran

    SubKamran TS Rookie Topic Starter Posts: 166

    Ok, thanks! Yes, see, I KNEW it. It's mostly video card dependant, that's why I got only 5598 on my last test even if the rig is at 2.8Ghz :D

    I'll run some 3dMark01 tests today.
     
  8. Arris

    Arris TS Evangelist Posts: 4,606   +287

    Athlon XP @ 2Ghz with 9700 Pro 128mb.

    4790 in 3dmark03.

    Just make sure you have the card set without AA or AF (or true form). Set all the sliders in the catalyst drivers to the lowest setting and you should get the true score.
     
  9. MrGaribaldi

    MrGaribaldi TechSpot Ambassador Posts: 2,512

    Hehehe...

    When you're running a benchmark and want to get max results, allways use Aplication Preference... Otherwise your system will be doing more than asked of it (seen from the benchmark side)...

    WRT 3dmark 01SE vs 03...

    It depends on what you want to see...
    3dmark 03 is, as has been said, gpu limited so you'll see how your performance is in different aspects (PS2.0, Fillrate etc) and can use this to judge how future games with heavy usage of said features will perform...

    It won't tell you exactly how many FPS' you'll get, but will give you an idea if there are some features you should avoid/decrease, and which you can increase to get the max from your system...

    01SE is, now, system dependent and will tell you, when compared to other similar systems, if you're very unoptimized or if you have a bottleneck in your system...
     
  10. SubKamran

    SubKamran TS Rookie Topic Starter Posts: 166

    Yup, I downloaded 01 SE so I'll run that once I get home! :)
     
  11. SubKamran

    SubKamran TS Rookie Topic Starter Posts: 166

    I get:

    Settings:

    http://service.futuremark.com/compare?2k1=7499349
     
  12. Mictlantecuhtli

    Mictlantecuhtli TS Evangelist Posts: 4,345   +11

    That's what I don't understand in benchmarking: If the scores are low, let's just decrease image quality and try again..

    It's like fooling yourselves just to get those extra points, at the cost of what? Do the same people who minimize quality in benchmarks do the same in games?

    By the way, has anyone ever ran any OpenGL tests? All benchmarks I see being used are D3D, for some reason. Yet a lot of games use OpenGL.
     
  13. MrGaribaldi

    MrGaribaldi TechSpot Ambassador Posts: 2,512

    There aren't as many OpenGL benchmarks out there as DX...

    So most people then use an OpenGL game with benchmark abilities to test their systems...

    As for why decrease IQ, it's the easiest way to get a comparable result...
    My settings doesn't neccessarily correspond with yours, but running it as Application Preference will allways be the same... (Also helps when comparing between different cards, as one IHV's implementation of AA often differs from the others and it's the same with AF)

    And then there is that point about letting the benchmark control what it's doing, instead of overriding it...
    If you want to run 3dMark with AA&AF, buy it, and you can choose whatever settings you want....
     
  14. SubKamran

    SubKamran TS Rookie Topic Starter Posts: 166

    Thanks guys! I guess 16277 is pretty good without even OC'ing the memory (in fact, it's underclocked, 196Mhz) :D

    I can't reach the 250Mhz FSB because I can't boot up. 249FSB and I get artifacts and it's unstable. So, 245 it stays until I get better memory :D

    And it's fine for me! I'm going to install my video card's overclocking software to see if I can OC it a bit (like 10Mhz).
     
  15. farmer

    farmer TS Rookie Posts: 48

    It's not the memory that's holding you back. That is as good a memory for the P4c as there is. Try raising your vdimm to 2.85v and see if it will boot into Windows. Also might relax the timings a bit, 2.5-2-3-6, and see what you get. For your machine, 250fsb is doable, just gotta find the right settings. With my PC3500, and the 12x multi, I'm at 270/216, running 24/7, so you can do better.

    The best OC'ing software for your Radeon, is the Rage3D tweak utility. Get the build for the 4.1 drivers, then OC away. I can run mine at 440/370 to benchmark.

    You also might need to let your new hardware do some burning in before you get maximum OC. Just some ideas here....

    Good luck....
     
  16. SubKamran

    SubKamran TS Rookie Topic Starter Posts: 166

    I'll try that today, I forgot I can increase the memory voltage! :D The RAM is rated for 2-2-3-5 :p, but I'll touch them down a little. What's REALLY funny is that it won't boot if the timings are on SPD mode :p
     
  17. SubKamran

    SubKamran TS Rookie Topic Starter Posts: 166

    Farmer, 2 questions:

    What does '2-2-3-5' actually mean? I mean, I know what they mean but what order are they in? (precharge, delay, etc.)

    And second, my RAM says "Supported voltage: 2.6v" on NewEgg but you have my PC35(7?)00 version of mine so... :D
     
  18. SubKamran

    SubKamran TS Rookie Topic Starter Posts: 166

    I think I'm at maximum. I got to 250 and I set it at 252 because 260 won't boot.

    My RAM is set at 2.5-2-3-6 so we'll see if it will complete a 3dMark01.
     
  19. SubKamran

    SubKamran TS Rookie Topic Starter Posts: 166

    It runs stable at 250Mhz. 252 crashed in 3dMark01.

    If you have any other suggestions, please say but right now I'm satisfied. It gave me an extra 400 points in 01. So now I'm OC'ing the video card to see.
     
  20. BrownPaper

    BrownPaper TS Rookie Posts: 407

    i would run something like prime95 to really test how stable is your processor. 3dmark 2001SE does not put as much load (and heat) on the processor as prime95.

    there are people who think they run fine until their prime95 gets messed up after a few minutes. then games like CoD and Halo reveal instabilities with their OC setup.

    you have a 2.4c so your cpu should be able to be stable at 3.0.
     
  21. farmer

    farmer TS Rookie Posts: 48

    Ok, first question:

    CAS# Latency 2.5 (kinda important)
    RAS# Precharge 2
    RAS# to CAS# Delay 3 (very important)
    DRAM Precharge Delay 6
    DRAM Burst Length 8
    This is specific to your P4P800 mobo

    Second question:

    OCZ rates their memory at a certain voltage, but they do as good a job with their PCB's as anyone, and the ram can handle the higher voltage, no problems. I only with this mobo had a higher vdimm....

    Here is a link for some serious Bios tweaking you might check out. The guy really seemed to do his homework, so you might see something that interests you...

    Later....
     
  22. SubKamran

    SubKamran TS Rookie Topic Starter Posts: 166

    What's the DRAM burst length do and is 8 or 4 a good option?
     
  23. farmer

    farmer TS Rookie Posts: 48

    Man, you'll need someone smarter than me to explain it to you! All I know is it is douvle the speed of 4, and it does it stable for me, so I thought it might be a setting you could use.

    Also, when you run 3DMark01, disable HyperThreading. HT enabled lowers your score for some strange reason. It's the only benchmark I've found that is affected. Just some tips that I use here...

    Later....
     
  24. SubKamran

    SubKamran TS Rookie Topic Starter Posts: 166

    Thanks.

    Unfortunately, when my FSB is at 250, I found out it IS unstable because WarCraft III: Frozen Throne dies about 15 minutes into the game.

    I found this out while trying to play Network :D Lowering it to 247 did the trick though.
     
  25. Arris

    Arris TS Evangelist Posts: 4,606   +287

    Mict,

    Aquamark and Vupline are 2 openGL benchmarks I've seen in the past.

    Plus the drivers should be set to application preference or off, as the test parameters in the 3dmark benchmark program actually specify that it is a test being run at X resolution with no AA, no trueform etc....

    Its not cheating. its just not overriding the settings that should be used for the benchmark. I don't agree with extreme changes to your system from its standard configuration to benchmark and get the best score you can but you have to have a standard test setup to be able to do any comparison between systems.

    3dmark often allows users to compare their results with that of other users running the same system. Having exactly the same components and having your system running extremely badly and hence getting a low score and performing under par in comparison to someone else with exactly the same hardware would annoy me. If I'm not at least getting the performance I paid for without overclocking I would be even more annoying. In this way 3dmark and other benchmarks make a valid contribution to the computer enthusiast world by providing references to let them know their system is running as it should be without problem.
     

Similar Topics

Add New Comment

You need to be a member to leave a comment. Join thousands of tech enthusiasts and participate.
TechSpot Account You may also...