GeForce2 GTS vs V5 6000 vs ATI Radeon
Home | Reviews and Features | Special Reports | Forums |

Results 1 to 7 of 7

Thread: GeForce2 GTS vs V5 6000 vs ATI Radeon

  1. #1
    Join Date
    Apr 2000
    Location
    Canada
    Posts
    203

    GeForce2 GTS vs V5 6000 vs ATI Radeon

    In the left corner, weighing in at 64 megs of DDram, wearing 1 GPU, the Geforce! In the right corner, weighing in at 128 megs of SDram, wearing 4 CPU's, the Voodoo5 6000!

    what about ATI? with all the voodoo/geforce rivalry, don't you think the Radeon MAXX will kik ***? when is it supposed to come out anyways?

  2. #2
    Join Date
    Jan 1999
    Posts
    1,424
    My money is on the Voodoo5 6000. Of course, very few HAVE the money to put down for this treasure. We will have to wait and see on Radeon, if ATI gets their act together and releases good drivers, we could have a GF2 competitor on our hands.

    ------------------
    This signature space FOR RENT

    This signature space FOR SALE

  3. #3
    Join Date
    Feb 2000
    Location
    Deutschland
    Posts
    936
    Ditto. Unless the GF2 gang manages to get 3ns DDR chips (as opposed to 6ns, which is the fastest DDR right now), the V6000 should have easily twice the memory bandwidth.

    What can I say... I'm still waiting. I hope 3DFX can deliver at least that one around the promised time. I must confess that while I admire their taking the time to fix all problems before release... I'm already VERY tempted to say "screw 3DFX", and go buy a GF2 instead.

    ------------------
    Moraelin -- the proud member of the Idiots' Guild
    Moraelin -- the proud member of the Idiots' Guild

  4. #4
    Join Date
    Dec 1998
    Location
    Grand Haven, Michigan, USA
    Posts
    11,332
    Yes the GeForrce2 GTS is so memory bandwidth restricted. DDR is nice, but does not solve everything. I can't see it getting much higher than 400Mhz this year. They either need QDR (Quad Data Rate) or a dual data bus. The core of these things can reach 250Mhz, but the memory needs to run at least twicve as fast.

    I think the Radeon MAXX will be the knig when it hits. This thing will be 200Mhz with 400Mhz DDR and a dual chip solution. But most of all will be some of the tweeks like the Hyper-Z and Charisma engine.

    The Voodoo5 6000 will be fast though. I just wish they could have used .18 micron. Even if they couldn't get enough 200Mhz SDRAM, they could have removed the fans then.

    256bit data pths will be the next thing. These cores are plenty fast enough, but the need the memory speed for 32bit color and higher resolutions (and all the other neat things like T&L and FSAA).

    Kyro (VR3) is the other posability. This hing performs about the same s the GeForce 256 DDR in 32bit color, but one Anand tested only had a 115Mhz core/memory and two texture units. It will ship at 125Mhz. The main thing though will be the VR4. This thing will be .18 micron and have DDR (333Mhz) memory, 166Mhz, and a Quad texture pipeline. That will make it almost 3 times faster.

    It managed 22.4 fps in Q3 HQ 1600x1200 with a 230Mpixel fill. The VR4 will have a 667Mpixel fill and the same memory bandwidth per pixel (125Mhz for 125Mhz 2 TU and 333Mhz for 166Mhz 4 TU). That means it would perform about 2.89 times faster than the one Anand reveiwed, resulting in 65fps in Q3 1600x1200x32 for something about as advanced as a GeFroce256 DDR... 128bit memory at 333mhz and a single chip with quad texture unit. The main diferance is that it will have about 1/2 the transisters and at .18 micron should run very cool and be cheap to manufacture.
    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

  5. #5
    Join Date
    May 2000
    Location
    Curitiba, PR, Brazil
    Posts
    261
    Don't forget that by the time Radeon is around NV20 will be almost hitting the streets, by the time a Radoen MAXX is out NV20 will probably be out kicking some buts...

    And what about if the NV20 has Kyro's technology to eliminate overdraw? Did any of you guys see Kyro's perfromance at www.anandtech.com, yeah, it's not the fastest thing around but look at it's specs!! It's specs are even worse then a TNT2 non-ultra!!! But it's perfromance is pretty superior to a TNT2, and it's still on a early beta stage. Now just imagine a gfx card with the power of NV20 (wich I suspect is going to be pretty damn powerfull), and probably a decent memory solution such as double-channels or Quad-date-rate, and with NO overdraw (wich will increase the bandwidth eficieny ALOT, since overdraw is coming close to 3 in the current games, we can assume a spec'd card that has 2.5Gtexels/s with overdraw will become a real 7.5Gtexels/s with no overdraw).

    For now the best solution is Voodoo 5 6000, wich has a good bandwidth, but it lacks on some features in my opinion...

  6. #6
    Join Date
    Feb 2000
    Location
    Deutschland
    Posts
    936
    Unless nVidia does something about memory bandwidth, the NV20 will offer no improvement at all over the GF2. See the overclocking situation with the current GF2's. Overclocking the core by a good 20% yields a measly 1% or so better fps, while overclocking memory goes almost linear.

    I don't know what will they do. Maybe tile based rendering, maybe a 256 bit bus (a la FireGL 2 and 3), maybe something else.

    Or maybe they'll just stick to the current strategy. Spew a load of hype and purely theoretical numbers, and watch people buy them anyway

    ------------------
    Moraelin -- the proud member of the Idiots' Guild
    Moraelin -- the proud member of the Idiots' Guild

  7. #7
    Join Date
    May 2000
    Location
    Curitiba, PR, Brazil
    Posts
    261
    If one thing will change on NV20 it HAS to be the memory bandwidth. Don't forget NV20 is a new cycle, so they will have to come up with something new anyway...

    How about this:

    Tile-rendering
    3ns 250Mhz QDR (Quad-Date-Rate) memory
    256-bit memory bus width
    Double-memory channel

    Let's just make calculations:

    250Mhz QDR would be efieciently 1Ghz (lol)
    256-bit double pumbed would be a 512-bit path

    Now, 1000x512 = 512000 bits/s
    Wich is: 64Gb/s. And let's not forget tile-rendering, since overddraw is coming close to 3, it's real bandwidth will be about 192Gb/s (64x3), I don't think nVidia would have to worry about memory bandwdith in this case, ahahahah...

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
  •