PCA Forums

Is the gang from the old AGN board all here?

Check out my new editorial at:

http://hardware.planetcrap.com/story.php?id=86

Oh yeah, Corn, I kept it within one page (1600x1200 resolution, medium fonts, etc.). I believe that is a record for me.

im here...what else matter? nice read. but radeon is where the action is

Greg

Just took a quick peek at some of the posts. It seems that the video graphics board has turned into an anti-3dfx coalition. What happened to those "3dfx to the end" people?

U know alot about the 5500? if so Pidge has a pritty good question. and I have the same one. can U use a 5500PCI as a second videocard and use it 2 play games? I dont think it is anti 3dfx...dotn forget about s3 they suck 2. I had a 5500 and liked it more then the ddr I had just not as much as the radeon.

Greg

All the 3dfx to the end diehards are holded up with this loser awaiting the coming of the end. Or rampage whichever comes first.



Everyone that is except cobosdan. He somehow got away and has made it his mission to piss off Dirk.

Yes, you can play games using the Voodoo5 5500 PCI and another 3D capable card like the Radeon. However, you have to remember that you can only play 3D games on your PRIMARY card (either the PCI or AGP). That is, 3D acceleration is only present on the primary card. The other card will only be in 'software' or 2D. This is a limitation of the operating system. This means you will have to go to your BIOS setup and keep switching your primary card from AGP to PCI and vice versa.

The exception to the rule is glide. Glide games will run on the 5500 PCI even if it is set as your secondary card.

When I select "default OpenGL" in Quake II, the game runs on the TNT2 Ultra. When I select "3dfx miniGL" (which is an OpenGL-->glide wrapper), the game runs on the Voodoo3 2000. Glide will automatically go to your Voodoo5 5500 PCI. Now if you have two V5's, one an AGP and the other a PCI. Glide will run on the primary card.

I used to have a Voodoo3 2000 and TNT2 Ultra set-up, which is why I know these things. I haven't tried all possible games/applications out there so you may be able to do things I didn't try. (Oh, early drivers of 3dfx and NVIDIA cards, as in back in May of 1999, did not work well together so a lot of people with this setup may have unstable systems. Even 2D applications like Outlook Express would crash. By Fall of 1999, though, they work well together. Well, at least, the a Voodoo3 and GeForce256 card worked together as my brother had that set-up for months.)

The TNT2 was then replaced with a G400MAX. In theory, I had the capability of a three-monitor setup using only two cards but I never tried it to see if it works.

My current card is the Radeon (see .sig). My brother now has the G400MAX. All the 3dfx and NVIDIA cards I have bought the past two years are now collecting dust. But...I'm really tempted to put back the GeForce2 GTS 64MB that I also have. Driver updates from ATi are not very frequent. (Yes, video cards is a hobby of mine. I don't have a V5, waiting for the price to drop.)

I use the Windows 95/98 operating system, I assume you do as well.

Waiting for the price to drop on V5? Geez 3dfx just slashed prices on V5's to $199.

V5 5500 AGP $178.00 at mwave:
http://192.216.185.10/mwave/skusearch.h ... 11&bop=and

V% 5500 AGP $176.95 at buycomp:
http://www.us.buy.com/retail/product.as ... 18&loc=101

Considering that the TNT2 Ultra can beat the Voodoo5 5500 pretty badly in MDK2, then yeah I'm waiting for the price to drop---to drop even more.

800x600x32:
TNT2 Ultra--63.0 FPS, Voodoo5 5500--41.1 FPS

The Ultra (that is a TNT2 Ultra, not a GeForce2 GTS Ultra) is only about 50% faster. And remember, the TNT2 Ultra does not have any hardware T&L support but still pounds the Voodoo5. The 'real' ultra would be about three times faster because of the hardware T&L.

It is funny. Most reviewers wanted to compare the Voodoo5 5500 to the GeForce2 GTS/Radeon but neglected to compare the V5 to last year's cards to see the improvement. By constantly comparing to speed demons like the GTS/Radeon, people do not see the limitation of the V5. That is what you call 'bias', a prejudice. They weren't trying to make the V5 look good. On the contrary, by comparing to the GTS, they were trying to make it look bad. They ended up accomplishing the opposite. Well, to be more accurate, they ended up making it not look as bad as it could have been. A hyped up card being beaten by video card using NVIDIA's second best graphics chip of last year!

Naturally, we all know why the V5 got beat. It is the drivers. Their OGL drivers are finely tuned for Quake III so the V5 can do well in the Quake III benchmarks but not much else.

Do you know what is even more funny? 3dfx loves to hype that holy grail of performance--60 FPS. But it is the TNT2 Ultra that is capable of achieving this 60 FPS, and not the V5.

1024x768x32:
TNT2 Ultra--41.6 FPS, Voodoo5 5500--36.7 FPS

This is on an 800Mhz Athlon. The V5 still got beat.

For most 'bang for the buck', which card is the better buy? I'll wait for the price to drop.

But don't sweat, though, it does win at 1600x1200x32:
TNT2 Ultra--17.2 FPS, Voodoo5 5500--21.6 FPS

Considering that MDK2 uses up a lot of textures, the frame buffer sharing of the V5 must have helped out significantly to not only finally catch up to the TNT2 Ultra but surpass it.

See the scores for yourself here

Of course, the V5 is faster than the TNT2 Ultra in many other benchmarks. In those benchmarks, the V5's superior fill rate and bandwidth overwhelms the TNT2 Ultra finally.

Someone in my other thread posted that the V5 5500 had very bad multi-monitor drivers so I just bought a V3 3K from nexus. I just need it for Tribes and maybe Diablo II anyways, and as a second video card.

Its certainly interesting, perhaps even a science, how videocards are compared and benchmarked. Some of this false measurement and faulty conclusion making is done on purpose {selective comparison} and some of it is due to ignorance. No doubt about it reviewers need to get better informed about what exactly they are doing and looking for. In addition readers of these articles need to do the same for themselves. Certainly everything needs to be more clearly presented to readers also.

Any comment on the <A href="http://www.rivastation.com/index_e.htm" target="_new" title="Rivastation.com">Quake 3 results</a> the new 3dfx V5 beta drivers HSR features Roscoe? All the diehard 3dfx fanboys at <A href="http://www.3dfxgamers.com/boards.asp?BO ... oardpage=1" target="_new" title="3dfx fanboys going crazy thread at 3dfx gamers">3dfxgamers are going crazy about this wonderous new feature.</a> <img src="http://www.geocities.com/sbp7777/rolleyes.gif">

[Edited by sbp on 11-30-2000 at 11:18 AM]

ATi Radeon Default clock speed of 183MHz. Drivers off the CD. (i.e. 2X FSAA not working, 4X FSAA the only thing available). I upgraded my machine today. 256meg RAM and 950MHz Slot A Thunderbird Athlon. Anyway....

Kingpin Demo using 'tweaks'.

1024x768x32: 160 FPS
1600x1200x32: 117 FPS

1024x768x32 4X FSAA: 66 FPS.

There were visual anomalies that made the game unplayable but they make great benchmark scores.

The secret is to disable certain graphical functions. Kingpin allows you to do that for testing purposes. set r_drawworld, gl_reflections, gl_dynamic to "0".

On my radeon, these functions are disabled 100% of the time. But if I was a programming whiz and know how to incorporate these "tweaks" into the drivers, I would have the functions disabled PART of the time rather than all of the time. Some frames are not drawn properly while others are drawn well. The result is much faster performance. The more frames that are not drawn properly, the faster the frame rate. You can even have settings. zero for no cheating. 1 for for 5% of the frames not rendered properly. 2 for 10% of the frames not rendered properly. 3 for 35% of the frames not rendered properly. 4 for 50%.

You can incorporate "smart" cheating where frames that will require a lot of fill rate are the ones most likely to be not rendered properly while other less fill rate intensive frames are more likely to be rendered correctly.

You can even incorporate "smarter" cheats. Instead of having the entire frame being rendered wrong, only parts of the frame is rendered wrong while the rest is rendered correctly. The part that is rendered correctly would be the part that did not change a lot from the previous frame. This would reduce the occurrence of visual anomalies since it would make them harder to be seen.

I am not saying that Voodoo5 5500 is using these "tweaks" to improve performance and if they are cheating, I'm not saying they are using the same cheats. Time will tell if they really have HSR support in their drivers or if they are cheating. I am just saying that those ultra-fast frames that the V5 is reporting is not impossible. Need to know the "benchmark" game and take these "scores" with a mountain of salt.

http://forums.nv10.net/ubb/Forum1/HTML/002873.html

This thread over at 3DGPU shows a screenshot of what the V5 has to offer with the new HSR drivers. The author of the picture notes that there were "flashes" where the frame is not drawn properly. He had to take several screenshots in a row before he can get the one with the image glitches. Funny. What does that sound to you?

Of course, there were also reports of dropped frames completely which of course can greatly improve the frame rate measurement.