something creepy about 3dmark2001 - Page 2
Home | Reviews and Features | Special Reports | Forums |

Page 2 of 3 FirstFirst 123 LastLast
Results 16 to 30 of 34

Thread: something creepy about 3dmark2001

  1. #16
    Join Date
    Apr 1999
    Location
    Canada
    Posts
    879
    posted by Notorious AGD:
    Well I think that we all need new CPUs more then vid cards for this benchmark. The score only increases by 300 Marks when going from 1024X768 32bit to 640x480 32bit. This is a 2 times lower resolution with 2 times less of a fillrate need meaning that it would double in performance, but it does not. The score look better with faster CPUs.
    i think this is because most of the things in DX8 are software driven (unless you have a card that can do DX8 stuff, ie. Radeon or GF3) so increasing the CPU without changing the video card, it will increase it, because it can do the DX8 stuff faster, but a DX8 compatable video card can do it even faster, and relieve the CPU of these tasks. so really you do want to get a new video card, but not a GF3, well not right now. (Radeon would be a good intermediate choice)

    ------------------
    people that are slow are easy to pass, it's people who drive fast that are hard.
    people that drive slow are easy to pass, it's people who drive fast that provide a challange.

  2. #17
    Join Date
    Mar 2000
    Location
    Auckland, New Zealand
    Posts
    2,331
    Dissapointed. 3DMark2001 runs like a pig

    Anyone notice that as soon as there is a light source framerates 1/4

    If a 1000MHz PIII with a GTS card can't maintain 30 FPS at decent quality, well.

    As for the results, they are totally screwed up unless you can run all tests, its no where near a benchmark.

    And definitly, whats with a demo nature scene but 'can't do it' in the benchmark.

    Someone please run 3DMark2001 on a supercomputer and record it to an AVI file
    AMD Mobile 2500+
    A7N8X Deluxe rev2.0
    1GB Corsair XMS PC2700
    GeCube Radeon 9800 Pro
    Seagate SATA 80GB
    2xSegate 160GB in RAID0

  3. #18
    Join Date
    Dec 1998
    Location
    Grand Haven, Michigan, USA
    Posts
    11,332
    Mad Onion based it around the GeForce3 which he has had for a quite a while for testing. The program is also very CPU intensive. Performance will suck without a 1.33Ghz Athlon and GeForce3. I wonder if he made it SSE2 enhanced for the P4?

    ------------------
    The COMPUTER is your FRIEND!
    Happiness is manditory.
    AMD Phenom II x4 945 3Ghz | ASUS M4A77TD | 2X WD 1TB SATA 2 hard drive | 2x2GB Corsair XMS3 | nVidia GeForce 8800 GTS | ATI TV Wonder Theater Pro 550 | Antec P-160 case | Antec 650w Earth Watts | LG Blu-ray Super Drive | LG DVD RW | Windows 7 Pro

  4. #19
    Join Date
    Nov 1998
    Location
    Winston-Salem,NC, US
    Posts
    289
    I've another riddle, why are the fillrate scores 1/2 or less than 3DMark2000 scores and why have the polygon scores increased over 3DMark2000? Shouldn't they be the same since they attempt to find the theoretical maximum for each system? Or are we seeing the actual fillrate of the card rather than the theoretical fillrate? Or am I the only one who had low fillrate scores?

    Also on my system, 866T-Bird with GTS@218/378, I get 500 points going from the default 3DM2001 bench to 640x480x16bit with 16bit textures. 3100 to 3600, pretty shabby when some are getting over 4500 with Radeon and Ultras.

    BTW, it looks like the Radeon rules, almost 5000 for one system in the comparison list. I did see a P4 system with a Radeon card (I think, it may have been the GF comparison) but it was well down from all the Athlons so SSE2 optimation may not exist in 3DMark2001 or it can't match the Athlon FPU. Athlon was on top in all the comparisons I did but I could never get a complete list, MadOnion's servers must be swamped. Very interesting stuff benchmark, however.

    [This message has been edited by 23skidoo (edited 03-18-2001).]

  5. #20
    Join Date
    Aug 1999
    Location
    Norwich, UK
    Posts
    3,790
    If that is the case then it will be even more meaningless than the previous Mad Onion creations as a current D3D 3D Graphics benchmark. If it is overly CPU intensive it won't mirror real world game performance unless the GF3 really will be significantly faster when used with a high clock P4.

    Whatever it would appear pretty pointless at the moment and only good for you own tweaking or comparisons to systems of exactly the same spec. There was me thinking they might have learned something from 3D MK 2000's Geforce and P3 bias


  6. #21
    Join Date
    Nov 1998
    Location
    Winston-Salem,NC, US
    Posts
    289
    Finally got through, a P4 with a GeForce Ultra gets 5273. That's a P4 at 1875mhz, not sure where the Ultra is clocked at. This appears to be the default benchmark, it's by DigitalJesus so the scores are suspect, he either cheats or is the Greatest Tweaker ever (LOL) because he's always on top first. Next highest P4 is 5096 at what appears to be the default bench.

    The highest Athlon is 5351 but at 640x480x16, next highest is 5322 at 1024x768x16 and the highest 1024x768x32 is 4877. 1423mhz, 1550mhz and 1430mhz respectively on the Athlon clock speeds.

    Highest P4/Radeon is 4002 at 1024x768x16 and the highest Athlon/Radeon is, WTF can't get through now but if I remember right it's almost 5000, in the 4900's. MadOnion needs to increase their server capacity by an order of magnitude, this is ridiculous.

    Anyway seems to me that it's all good, P4 may have an advantage with higher bit depth but Athlon is ruling the roost. I was wrong about the Radeon but on the average I believe they are doing better because of the DX8 features.

    If anyone else can get through please double check me. I'm also curious to see if you see the same results I did or if MadOnion really has something screwy going on. I cannot get a comparison chart from the ORB because not enough data is available, LOL. I don't think they can collect and collate it fast enough and they can't get all the data because it's so hard to get though.


    Update, got though and the highest Radeon/Athlon I see is 4848 at what appears to be the default bench and Athlon clocked at 1152 mhz. Heres a good one the highest score in 3DMark2001 is 10421 with an Athlon at 1088 and a Quadro2/Pro. Cough <bull****>, cough, cough. Looks to be a modified 3DMark2000 score submitted as a 2001 score. This was submitted by The Snarf, thesnarf@noknok.nl, who I think is trying to tell us something. I think we all need to wait for some comparisons by all the reputable websites before we make any conclusions.


    [This message has been edited by 23skidoo (edited 03-18-2001).]

  7. #22
    Join Date
    Apr 1999
    Location
    Canada
    Posts
    879
    by 23skidoo:
    I've another riddle, why are the fillrate scores 1/2 or less than 3DMark2000 scores and why have the polygon scores increased over 3DMark2000? Shouldn't they be the same since they attempt to find the theoretical maximum for each system? Or are we seeing the actual fillrate of the card rather than the theoretical fillrate? Or am I the only one who had low fillrate scores?
    the reason for this, as far as i can see, is that the default benchmark is now being done in 32bit with a 24bit z-buffer, whereas in 3dmark2000 it was only 16bit with a 16bit z-buffer. it's just more bandwidth is needed to fill 32bit textures then 16bit ones. so most video cards take about <50% when switching from 16-32bit. unless your running a Radeon or Kyro based cards, in which case you'll see about <15% decrease in fillrate.

    ------------------
    people that are slow are easy to pass, it's people who drive fast that are hard.

    [This message has been edited by drzaius (edited 03-18-2001).]
    people that drive slow are easy to pass, it's people who drive fast that provide a challange.

  8. #23
    Join Date
    Nov 1999
    Location
    Oregon
    Posts
    1,990
    "Only the first 3 games metter in the Mark score."

    I found out how the result is determined :

    score = 10 x ( Game1_Low + Game2_Low + Game3_Low ) + 20 x ( Game1_High + Game2_High + Game3_High )

    As you can see, the high detail tests are given twice the weight of the low detail tests.

    ------------------
    Friends don't let friends buy P4s.

  9. #24
    Join Date
    Mar 2000
    Location
    Auckland, New Zealand
    Posts
    2,331
    It seems a bit to 'high end' to me.

    Did anyone else notice the max texture memory used exceeds 32MB often enough, how would this effect 32MB card users???

    A P4 with a 64MB GF3... yeah right, I'll take 3 <cheque's fine? )

    Also, I do think game 4 is counted in the final score... read the help files

    Its all bollocks

    There are other vid benchmarks out-there <of which marketing people havn't wrecked>
    AMD Mobile 2500+
    A7N8X Deluxe rev2.0
    1GB Corsair XMS PC2700
    GeCube Radeon 9800 Pro
    Seagate SATA 80GB
    2xSegate 160GB in RAID0

  10. #25
    Join Date
    Mar 2000
    Location
    Auckland, New Zealand
    Posts
    2,331
    Also, minimum spec 32MB video card with 500MHz CPU.....

    hahahahahaahahahahahahahaahahahah


    Sure, the program will boot
    AMD Mobile 2500+
    A7N8X Deluxe rev2.0
    1GB Corsair XMS PC2700
    GeCube Radeon 9800 Pro
    Seagate SATA 80GB
    2xSegate 160GB in RAID0

  11. #26
    Join Date
    Apr 1999
    Location
    Canada
    Posts
    879
    32meg video? then how are voodoo3 people running it?
    also i know that you need atleast 128meg RAM to start the program, it tells you to either upgrade your system or use an older version of 3dmark

    ------------------
    people that are slow are easy to pass, it's people who drive fast that are hard.
    people that drive slow are easy to pass, it's people who drive fast that provide a challange.

  12. #27
    Join Date
    Nov 1998
    Location
    Winston-Salem,NC, US
    Posts
    289
    Thanks, drzaius, I hadn't thought about that so I double checked it. 3DM2001 fillrate is ~900 and ~600 Mtexels/s at 1024x768x16bit with 16bit textures, triple buffering and 16bit z-buffer. 3DMark2000 was ~1400 MTexels/s and 600MTexels/s. So it really is down from 3DMark2000 in multitexturing. The effective fillrate vs. the theoretical fillrate of the GeForce has always been questioned because of the memory bandwidth problem and that may be what we're seeing.

    3DMark2001 may actually prove to be a more usefull benchmark and not just a testosterone power trip to get higher numbers. I'll bet the Ulta's and Pro's with the larger faster memory arrays don't see as much difference in the multi-texturing tests. Has anyone checked the fillrate numbers between 3DM2000 and 3DM2001 on an Ultra or Pro?

    Forgot to mention, the high polygon test went up another 1 million triangles/s with the reduced texturing requirements so polygon throughput is being limited by the rendering process. 14MTringles/s vs. 18MTriangles/s between 3DMark2000 and 3DMark2001. Still not at the 25 MTriangles/s claimed but much better.

    Sorry about the edits but I want to get it right.

    [This message has been edited by 23skidoo (edited 03-18-2001).]

    [This message has been edited by 23skidoo (edited 03-18-2001).]

  13. #28
    Join Date
    Apr 2000
    Location
    Morelia, Mexico
    Posts
    1,717
    drzaius is right.


    Most vid cards out there doesn't fully support DX8, and many instructions are executed using software engines. That's why lights and other stuff in 3dmark2001 decreases the frame rate a lot. Is not 3dmark related, is directx8 related. A faster CPU, will do the software instructions faster.


    Madonion and nVidia?, there must be some agreement of technology, just like the one with id software and nVidia, developers need to have knowledge of the future technologies, and therefore develope games and applications to use it.


    If the agreement wouldn't exist, only 3dmark2002 will use the gf3 technology, that would not be good, right?. 3dmark purpose is to test your computer's abilities to handle THIS year games.


    ------------------
    Keyboard not detected, press F1 to continue...

    [This message has been edited by Luis G (edited 03-19-2001).]
    If you were in a world full of crazy people, who would be the crazy person?
    OTCentral

  14. #29
    Join Date
    Nov 1999
    Location
    Oregon
    Posts
    1,990
    I mean no disrespect but apparently you and Madonion have a difference of opinion. 3DMark2001 is a benchmark of Dx8 performance therefore it is a measure of future games since none currently utilize the DX8 features that I am aware of. Actually, 3DMark2001 measures several features no currently available board supports.

    Personally, I don't consider this to be a bad thing. It gives us all something to grow into


    ------------------
    Friends don't let friends buy P4s.

  15. #30
    Join Date
    Nov 1999
    Location
    Oregon
    Posts
    1,990
    First off, I should correct myself. The fourth test (nature) will factor in the results if it is run. Unfortunately, MO has chosen not to allow software rendering for this.

    One other thing, I have read that everything in 3DMark2001 is standard DX8 stuff so all future board claiming DX8 compatability should be able to run all of the tests.

    I believe that the fact that Nvidia is about to release the first DX8 compatable board should NOT be held against them. There are several other companies that had the opportunity and did not pull it off. Actually, this is almost beside the point since most cards out today can run the first three tests in high and low detail.

    One more issue : these issues are with the cards of today. The scenario can change daily and is about to very soon. At least, I hope so


    ------------------
    Friends don't let friends buy P4s.

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Posting Permissions

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