(background)
first i will upgrade the video card to either a voodoo5 5500 (gotta love the price 趦) or a geforce2 256. soon afterwards i will have a 550e or 600e on my BX6-2. i will be overclocking the cpu like crazy, so the AGP bus speeds will be 90+ mhz.

(now for the question)
does anyone know how well the voodoo5 5500 tolerates out of spec agp speeds?
how about the geforce2 gts?
or even the geforce1 DDR?

any help or comments on my upgrade path are appreciated!

------------------
aventari
"witty one-liner"

on the topic. will a new 7200 rpm hard drive up my performance noticeably? i have a maxtor 5400 8.4 gig. I'm thinking a new seagate 10.2 gig 7200 would be a good upgrade too

------------------
aventari
"witty one-liner"

I've got two IBM 9GB/7200RPM/2MB Buffer/8.5ms seek drives running RAID 0 and it is SWEET.

------------------
"Everyone's favorite implement for any task"
------------------

i think i'll do that, then save up for a gf2 gts or a voodoo5 5500

------------------
aventari
"witty one-liner"

------------------
Just your generic meaningless signature. Mix with 2 quarts water and stir till evenly coated.

------------------
Just your generic meaningless signature. Mix with 2 quarts water and stir till evenly coated.

------------------
-- incubus
As I chase the leaves like the words I never find ...

Does anyone have a link to the Promise ata66 modification for the new BIOS?

------------------
aventari
"witty one-liner"

quote:

---

Originally posted by aventari:
does anyone know how well the voodoo5 5500 tolerates out of spec agp speeds?
how about the geforce2 gts?
or even the geforce1 DDR?

---

No one can predict how well a video card will work when it's out of spec. It may work, it may not. Just because one GF2 or Voodoo5 went to some ass high bus speed doesn't mean the one you get will. The best bet is to get whichever card you like the best, if it doesn't o/c well, return it. If you can't return it at least you have a card you will like.

------------------
< www.slapmahfro.net >
ya know ya wanna slap it ...

------------------
~Anztac - The guy who had the really long sig (formerly Kriegman)

"I am easily driven into a flying rage by blithering idiots"
-cr0bar [The god of this domain]

[This message has been edited by Anztac (edited May 24, 2000).]

------------------
"Everyone's favorite implement for any task"
------------------

------------------
~Anztac - The guy who had the really long sig (formerly Kriegman)

"I am easily driven into a flying rage by blithering idiots"
-cr0bar [The god of this domain]

------------------
"Everyone's favorite implement for any task"
------------------

------------------
~Anztac - The guy who had the really long sig (formerly Kriegman)

"I am easily driven into a flying rage by blithering idiots"
-cr0bar [The god of this domain]

yes you can. you can say 'on average'. on average the celeron 300a will hit 450 mhz. on average a TNT2 can pull 89 mhz agp bus speeds. thats all im asking for..
i know how overclocking works, dude!

------------------
aventari
"witty one-liner"

geforce 2 gts all the way, 趤 (until may 26, with prmotional offers from the buy.com sites)

so it's an ultraata66 raid hack w/ 2 seagate 7200 10 gig'rs for me i think. Will i still be able to run my 8.4 gig, and 2 cdroms at the same time? (a total of 5 IDE devices)

------------------
aventari
"witty one-liner"

so, anyone think i need a new video card? (and what should i get?) (under 赨 preferably)

------------------
No signature required.

[This message has been edited by TeKniC (edited May 25, 2000).]

Direct3D is different; the T&L funcs are there, but the developer has to actually put in the T&L func calls himself. Since most cards did not have T&L before the GeForce, many D3D apps elect to do T&L in software without even checking for hardware T&L support.

And another issue that I have yet to hear a conclusion on...what if your CPU is fast enough that software T&L actually gets better FPS than hardware T&L? A good OpenGL driver would hopefully be able to predict which will be faster upon driver initialization...there are probably several other ways of doing this.

As for GeForce2 vs. Voodoo5, I'd go with GeForce2. I've lost my respect for 3Dfx, and I've seen their Voodoo5 fall behind the GeForce DDR in a lot of benchmarks.

As for Unreal and UT, I trust neither as a benchmark. Both are far too Glide-biased, and there's a lot of sloppy technique that nobody's worked out of the engine yet. I'm betting ten to one that OpenGL and D3D support in Unreal is actually just a layer that Unreal sticks in to translate standard Glide calls to the respective chosen API, much like a D3D wrapper. Yech.

As for IDE RAID, I heard a while back that IBM is making a 10,000RPM ATA-66 drive...wonder how that performs in a RAID?

And as for RAID 0 (i.e. just AID , I might warn some of you that the chances of losing all your data to a disk crash increase by a factor of the number of drives in your RAID. So unless you do regular full backups, my advice is not to choose RAID 0!

------------------
aventari
"witty one-liner"

Now in order to draw the image the vertex data has to first be transformed into the frustrum shaped jobby.

Because full openGL compliance in still a bit sketchy on most drivers, and whether the software openGL implementations included a transform engine for fitting orthagonal projected scenes to the frustrum was not certain, most games, wisely, chose to transform the vertices to the perspective projection themselves, using software.

This means that the GPUs T&L engine doesn't get to flex its muscles to transform the vertex positions.

Only games which have the capability to describe positions in orthographic projection will be able to take advantage of onboard T&L on graphics cards.

DirectX, i believe, has only added support for orthographic projection at version 7.

At least, this is what i have gleaned from a quick look at the OpenGL specs.

DuSTman

[This message has been edited by DuSTman (edited May 26, 2000).]

Maybe there's a way of enquiring if there is hardware T&L on an OpenGL , in which case developers could take advantage of it, but also have to duplicate this functionality in software to cater for drivers without transform capability..

And while I don't program in Direct3D (I'm a *nix guy), I remember hearing about hardware T&L functions before DX7, and maybe even before 6. It hasn't been around long in D3D, though, which is why OpenGL is such a popular CAD API.

------------------
~Anztac - The guy who had the really long sig (formerly Kriegman)

"I am easily driven into a flying rage by blithering idiots"
-cr0bar [The god of this domain]

OpenGL specs include the ability to specify the position of "lights" that the polygons get lit by, but most gaming cards don't support this.

Calculating the lighing at each pixel in real time would have been computationally infeasible, so games had to resort to multitexturing to do lighting, calculating bitmaps representing the light intensity at each pixel on the polygons beforehand, and then blending these lightmaps with the texture of that polygon to get the resulting pixel colour values. This works fine for light that stays the same, but for dynamic lighting it is bad - the fixed lights in quake 1 and 2 appeared OK, but the light from the rocket was just added to any texel within a range of that rocket. (remember quake 1 and 2 where you could see the light from explosions against the other side of the wall?)

Real-time calculation of the light maps would look more realistic, and that is what the lighting section of nVidias T&L is all about.

I don't know how sophisticated nVidias GPU lighting model is, but from a programmers view, this is a step in the right direction.

Lizzy, Dustman says in laymans terms "OpenGL doesnt get to take full advantage of GeForce" and thats really all you need to know...

does anyone find it a bit ironic that the l33t hax0r has to ask the lam0r to explain stuff?

Still, though, nVidia would probably ace the Quadro with a low-cost (comparatively speaking) gaming card in six montns or a year. The main reason I buy truly serious hardware is to make upgrading fairly pointless for the next few years (although playing with the big irons is nice too).

Oh, and btw, I bet it wouldn't take much pressure to get nVidia to post a database of games and the extent of their GeForce2 support. I remember 3Dfx had that a while back (and may still have it), and nVidia's already got an article touting how over 100 games take advantage of the GPU...

what's funny, though...I know some people who, about half a year ago, were still waiting for BitBoy's Glaze3D! How far behind is its release by now...a decade? heh

I saw a satirical preview on the Glaze3D on somethingawful.com somewhere..."More powerful than Jesus Christ!" Yet still no more than a diagram of the chip's internals. Funny as hell.

Did you hear that the voodoo5 5500 was delayed another week? jesus, this card must have bugs out the ass it's being rushed so much. I'm less and less inclined to go with it now. the gf2 is much more likely to have fewer bugs being based on a tried and true design

------------------
aventari
"witty one-liner"

------------------
Just your generic meaningless signature. Mix with 2 quarts water and stir till evenly coated.

The part of a solid, such as a cone or pyramid, between two parallel planes cutting the solid,
especially the section between the base and a plane parallel to the base.

frustum \Frus"tum\, n.; pl. L. Frusta, E. Frustums. [L. fruslum piece, bit.] 1. (Geom.) The part of a
solid next the base, formed by cutting off the, top; or the part of any solid, as of a cone, pyramid, etc.,
between two planes, which may be either parallel or inclined to each other.

You'll probably get a card that can do that by the end of this year or the beginning of next year. By then, you'll want a card that can do 2048x1536x32bpp with a solid 100 frames per second for your 22 inch monitor.

------------------
Just your generic meaningless signature. Mix with 2 quarts water and stir till evenly coated.

quote:

---

nVidia Makes Bold Claim, Gets Ass Handed Back

In a recent press release on the GeForce2, nVidia claimed real-time Pixar-level animation was capable. A Pixar employee responded by calling the GeForce2 a "toy".

In a press release posted on Silicon Investor, Jen-Hsun Huang, President of nVidia, claimed, "Achieving Pixar-level animation in real-time has been an industry dream for years. With twice the performance of the GeForce 256 and per-pixel shading technology, the GeForce2 GTS is a major step toward achieving that goal."

In response, Tom Duff from Pixar responded, "Do you really believe that their toy is a million times faster than one of the CPUs on our UltraSparc server?"

Duff then goes on about the throughput of the AGP bus in a typical PC, by asking, "Don't forget that the scene descriptions of (Toy Story2) frames average between 500MB and 1GB. The data rate required to read the data in real time is at least 96Gb/sec. Think your AGP port can do that?"

The Pixar employee then points out the time constraints of the ludicrous statements the marketing team at nVidia by saying in 10 to 15 years, nVidia may have Pixar-level animation in real-time, but not in this day and age.

While this doesn't mean nVidia engineers are ignorant of what goes on in the rendering world, it does point fingers at the marketing teams of the graphics world and how ignorant they must think users really are. This smells like Apple marketing schemes.

---

------------------
jim
< Brews and Cues >

quote:

---

nVidia Makes Bold Claim, Gets Ass Handed Back

In a recent press release on the GeForce2, nVidia claimed real-time Pixar-level animation was capable. A Pixar employee responded by calling the GeForce2 a "toy".

In a press release posted on Silicon Investor, Jen-Hsun Huang, President of nVidia, claimed, "Achieving Pixar-level animation in real-time has been an industry dream for years. With twice the performance of the GeForce 256 and per-pixel shading technology, the GeForce2 GTS is a major step toward achieving that goal."

In response, Tom Duff from Pixar responded, "Do you really believe that their toy is a million times faster than one of the CPUs on our UltraSparc server?"

Duff then goes on about the throughput of the AGP bus in a typical PC, by asking, "Don't forget that the scene descriptions of (Toy Story2) frames average between 500MB and 1GB. The data rate required to read the data in real time is at least 96Gb/sec. Think your AGP port can do that?"

The Pixar employee then points out the time constraints of the ludicrous statements the marketing team at nVidia by saying in 10 to 15 years, nVidia may have Pixar-level animation in real-time, but not in this day and age.

While this doesn't mean nVidia engineers are ignorant of what goes on in the rendering world, it does point fingers at the marketing teams of the graphics world and how ignorant they must think users really are. This smells like Apple marketing schemes.

---

------------------
jim
< Brews and Cues >

------------------
jim
< Brews and Cues >

------------------
Just your generic meaningless signature. Mix with 2 quarts water and stir till evenly coated.

------------------
aventari
"witty one-liner"

------------------
s | s | 5 | 9

[This message has been edited by ss59 (edited May 30, 2000).]

------------------
Just your generic meaningless signature. Mix with 2 quarts water and stir till evenly coated.

quote:

---

Originally posted by Hellraiser:
Just got some new drivers for my Geforce 2 GTS, and like I thought, there's room for improvement. I just ran the Q3 timedemo 1 and got 63.3FPS @ 1600x1200, up from 61 with the old drivers. It's a great card, each day I play with it I like it more.

---

------------------
Just your generic meaningless signature. Mix with 2 quarts water and stir till evenly coated.

------------------
Just your generic meaningless signature. Mix with 2 quarts water and stir till evenly coated.

Anyways, on my earlier post...yes, I had a brain fart of some kind. I noticed the posts about "nv15?" and remembered a little late. The GeForce is the nv10, sorry

Anyways...I have to comment on 3Dfx's FSAA approach. IMHO, it's just brute-force applied to age-old techniques, which is why the performance hit is so severe. The GF2 will (or does now, if its driver has already been refined enough) do FSAA in a different manner...one that actually takes little or no performance hit! I hear part of it is due to a dedicated FSAA chip (on same die as GPU?), a solution 3Dfx's FSAA technique doesn't have room for. Score another one for the GF2...

Hah...and the GF2 marketing blunder was funny as hell Speaking of marketing blunders...

On 3Dfx's Q&A about the V5 6000, one question is like, "Is the 128MB memory used to hold replicate data for all four processors?" 3Dfx's answer (and I almost quote): "No, only the texture memory is replicated between the processors."

ONLY the texture memory? I laughed my ass off. I like the way they said, "Only." "Only" 90\% of all the memory needed anyways has to be wasted duplicating data for all four processors! What a crock.

------------------
"Welcome to my Neon Dream."

Speaking of the 90\% of wasted memory.. couldn't they gain a large amount of performance on all theier multi-processor cards by simple rewriting the drivers so that instead of replicating duplicate graphics data it does it in a more effecient manner? could that be a marketing stragety.. advertise this card that's subpar with the a lower price.. and know that it's only going out to 3dfx loyalists.. then preform a MAJOR driver upgrade giving a HUGE boost in performance? They did something similiar with the Voodoo3 and Graphics Quality.. i wouldnt put it past them to try something similiar with performance.

[This message has been edited by xaustinx (edited June 01, 2000).]

------------------
Just your generic meaningless signature. Mix with 2 quarts water and stir till evenly coated.

found that presentation somewhat interesting.. mebbe i'll wait for that.. doesnt the DreamCast use PowerVR for it's graphics?

[This message has been edited by xaustinx (edited June 02, 2000).]

---SNiP--

Main Specs
CPU SH4 - RISC CPU with 128-bit graphics engine (200MHz, 360 MIPS/1.4 GFLOPS)

Graphics Engine - Power VR Second-Generation (CG performance of over 3 million polygons per second)

------------------
"Welcome to my Neon Dream."

If the Playstation 2 ever gets to the Western Hemisphere, it's going to outright humiliate the Dreamcast. Hell, half of us here probably already have PC gaming rigs capable of that.

------------------
"Welcome to my Neon Dream."

the thing about gaming platforms like dreamcast and ps2 is not always about superior graphics and stuff (which truthfully shouldnt have much to do with why you like a game anyway)its about simplicity. With ps2 or dreamcast, when a developer makes a game, they know it will work on all nondefective units. If someone calls up and has a problem, they dont have to figure out if someone needs more ram or HD space or whatever, because its all the same. This is the same reason i have a psx and stuff...i mean yeah, its fun to boast about how my computer has more gonads than my friends, but my computer is a high maintenence machine, i plug stuff in and turn it on, no thinking required. You'll note that those of lesser intelligence often flock to the video game console and declare it king, that or theyre smart and poor, and so theyre forced into the position. Either way, its a set pattern. Its about 12:30, im really tired, if this makes no sense, im sorry, bye

quote:

---

Originally posted by Anztac:
I would suggest the Geforce GTS2 all the way. The V5 looks to suck comparitavly. The V5 does not have a GPU (thus does not assist with the CPU processes). The Geforce also has more/faster memory, with better everything. I myself though would wait for ATI's Raedeon. It's fast, supports all the bump mappings, has a great texturing pipeline. That is the shortest way I could say it.

---

What are you smoking? The GPU, or hardware texture and lighting only works in _some_ games and only provides a benefit in _some_ circumstances and _none_ at all on very high end processors. The Geforce2 only has one processor and as such _less_ memory bandwidth than a V5 5500. You spend more b/c nVidia went w/ DDR RAM and not 2 processors. The ATI Radeon will suck b/c it has only one pipeline and ATI always has terrible drivers, at least as far as gaming.

quote:

---

Originally posted by Anztac:
what visualal effects? t-buffer? HA! Who wants a blurry far ground, or blurry, confusing trail. The only features that the T-cuffer brings that matter are soft reflections and FSAA. FSAA works on the G FTS2. Soft-reflections are easily done. The speed/price of the V5 sucks ass.

---

Again, what are you smoking? While depth of field and motion blur might be undesirable in first person shooters, I'm willing to wait and see them myself. FSAA on the V5 uses rotated grid supersampling and as such is faster, looks better, and supports a lot more games. If you don't believe me, check out any major hardware review site. Soft shadows and reflections may be doable in software, but they will certainly be slower. If anything, the V5 5500 is cheaper than a GF2.

quote:

---

Originally posted by aventari:
thats what i think too

Did you hear that the voodoo5 5500 was delayed another week? jesus, this card must have bugs out the ass it's being rushed so much. I'm less and less inclined to go with it now. the gf2 is much more likely to have fewer bugs being based on a tried and true design

---

The VSA-100 is still based on the Voodoo architecture, which has been around for an entire generation more than nVidia's first entrant into the 3d card realm.

quote:

---

Originally posted by simulacrum:
Also the drivers for the GF2 are much more developed than the 5500's. From what I understand, the drivers for the GF1 and GF2 are basically the same, and some of the later drivers for the GF1 are the same as the early drivers for the GF2.

---

Even though 3dfx's drivers are only in the 3rd or so release the V5 5500 and the GF2 are neck and neck in the benchmarks. With such new drivers, 3dfx has more room to improve than nVidia.

quote:

---

Originally posted by TiresiasX:

Anyways...I have to comment on 3Dfx's FSAA approach. IMHO, it's just brute-force applied to age-old techniques, which is why the performance hit is so severe. The GF2 will (or does now, if its driver has already been refined enough) do FSAA in a different manner...one that actually takes little or no performance hit! I hear part of it is due to a dedicated FSAA chip (on same die as GPU?), a solution 3Dfx's FSAA technique doesn't have room for. Score another one for the GF2...

---

You're totally misinformed. 3dfx's RGSS method is _new_! That's why it's actually faster than a GF2. The performance hit is _less_! The GF2 has nothing on hardware that supports FSAA, while the V5 does. nVidia never intended to have FSAA until 3dfx announced it. They just added it to their drivers as an option. They use an older method (OGSS), that doesn't eliminate pixel popping and requires a higher fill rate than RGSS. That's why it only works with certain APIs and certain games.

1: about having more or less processors, and types of ram...
just because theres only one processor, doesnt mean its inferior, if i have a p3 1Ghz, it will always win over an 8086..same thing, just a more exadgerated difference. about DDR, DDR is not that much more expensive than SDR, and nVidia didnt go with DDR, the companies did because its faster. i have an SDR geForce, and it runs fucking fast anyway.

2:about tbuffer and v5 being cheaper...yes, the tbuffer is nice, but at what expense?, the v5 6000 is the closest performing card to the gf2, and its 600 dollars, so how is that less expensive, when most geforces are around 350 at the most?...i mean sure im in high math classes so i might think im better at most in math, but most people would probably agree that 600 is more than 350...

3:about still being based on the voodoo architechture, just because its been around longer, does not make it better. 64 bit processors will be faster than 32 bit, and they wont be based on the x86 architechture, and thats good, so your argument is null

4:about being neck and neck in benchmarks...i dont know what benchmarks you see, but every comparison ive seen to date of geforce2 and v5 cards have the geforce2 beating the v5 marginally, it only gets close in the super high resolutions.

5: about FSAA, if you would care to note, the person said "IMHO" meaning an opinion, and besides that, it is supported via hardware in geforce 2, just not too well. and just because nVidia tried to do it based on the fact 3Dfx did, doesnt make that bad, its called competition, and thats what the whole MS thing is all about.

basically, you needa do two things, support yourself with research and examples, and you needa combine things, note i only have 1 post to your 5

------------------
Just your generic meaningless signature. Mix with 2 quarts water and stir till evenly coated.

quote:

---

Originally posted by Sithiee:
holy shit man. post it all into one fucking post. and what are you? 3Dfx's poster boy?...just because a company was good, doesnt mean it cant get bad. ie, lucasarts used to be the unchallenged leader in most computer game types, of late though, the games are starting to suck. allow me to prove you wrong in each of your posts....

---

From years of experience in newsgroups, I've learned that ad hominem attacks are usually only employed by people who have lost an argument.

quote:

---

1: about having more or less processors, and types of ram...
just because theres only one processor, doesnt mean its inferior, if i have a p3 1Ghz, it will always win over an 8086..same thing, just a more exadgerated difference. about DDR, DDR is not that much more expensive than SDR, and nVidia didnt go with DDR, the companies did because its faster. i have an SDR geForce, and it runs fucking fast anyway.

---

I'm not arguing that the GF2 is inferior b/c it only has one processor. I'm arguing that b/c the V5 5500 has 2 processors, it has double the memory bandwidth than it would have had with a single processor and doesn't need DDR. That gives it greater memory bandwidth at a lower price. The companies didn't have any other choice but to go with DDR. If they had used SDR, the GF2 wouldn't have been any faster than a GF1. Try running your SDR GF1 above 1024x768x32. The bandwidth crunch will hit and framerates will collapse.

quote:

---

2:about tbuffer and v5 being cheaper...yes, the tbuffer is nice, but at what expense?, the v5 6000 is the closest performing card to the gf2, and its 600 dollars, so how is that less expensive, when most geforces are around 350 at the most?...i mean sure im in high math classes so i might think im better at most in math, but most people would probably agree that 600 is more than 350...

---

The V5 5500 is the closest performing card to the GF2 and it costs slightly less on average. nVidia's closest card to the V5 6000, < , will cost over 񘈨.

quote:

---

3:about still being based on the voodoo architechture, just because its been around longer, does not make it better. 64 bit processors will be faster than 32 bit, and they wont be based on the x86 architechture, and thats good, so your argument is null

---

I wasn't arguing that the V5 is better b/c the architecture has been around longer. On the contrary, someone I replied to argued that the GF2 is better b/c it's architecture has been around longer. I was correcting him.
Yes, most 64 bit processors are faster than 32 bit processors, except for Intel's Itanium flop.

quote:

---

4:about being neck and neck in benchmarks...i dont know what benchmarks you see, but every comparison ive seen to date of geforce2 and v5 cards have the geforce2 beating the v5 marginally, it only gets close in the super high resolutions.

---

The vast majority of GF2 and V5 reviews use only Q3, which is one of the very few games w/ hardware T&L support (and the V5 is still usually higher in high resolutions), and synthetic benchmarks that have little or no correlation with real world performance.

quote:

---

5: about FSAA, if you would care to note, the person said "IMHO" meaning an opinion, and besides that, it is supported via hardware in geforce 2, just not too well. and just because nVidia tried to do it based on the fact 3Dfx did, doesnt make that bad, its called competition, and thats what the whole MS thing is all about.

---

However, the fact that nVidia lies that the GF2 supports FSAA in every app is bad.

here >

"Full scene anti-aliasing is completely transparent to applications; end users can simply enable it with the NVIDIA control panel and experience FSAA in any title."

nVidia's FSAA is still a poor implementation and is done mostly in software.

<

"Suffice to say that, on average, a 4X Ordered Grid anti-aliasing should be very identical to a 2X Rotated Grid anti-aliasing in terms of quality. A 4X Rotated Grid anti-aliasing would be even better."

According to Riva3D >, an nVidia fan site,

"OGSS also is accomplished largely through software, which is why just about any 3D card can utilize this method if it is powerful enough to do so."

quote:

---

basically, you needa do two things, support yourself with research and examples, and you needa combine things, note i only have 1 post to your 5

---

Why don't you support your arguments w/ research and examples? What's wrong with replying to each person individually? It's a lot nicer than a post beginning with "You're all wrong" or something along those lines.

Posted by TiresiasX on Jun. 10 2000,20:18

Concerning memory bandwidth:

Two processors accessing shared memory does not equate to twice the sustained bandwidth. Memory access between requesting processors is often mutually exclusive, especially when memory writes are involved. Having multi-port memory chips makes it a little easier, but certain situations (especially two chips trying to write to the same memory block) still have to be done on separate memory clock cycles so that one chip doesn't fux0r the memory circuits or the other chip's work.

Even having simultaneous memory reads is difficult, as it requires an extra management layer around the memory ICs and may even require higher memory latency assumptions to allow for this management to take place.

And as for 3Dfx's anti-aliasing approach, their approach plus their aging architecture requires implementing more than one processor unit, if you read 3Dfx's own white paper on it. 3Dfx considers it a reasonable assumption that it can be done on one chip in the future (still with a severe performance hit), but they either haven't developed their drivers that far or do not have the capability in hardware. This is why the single-chip Voodoo4 has FSAA disabled.

------------------
"Welcome to my Neon Dream."

---

Posted by Sithiee on Jun. 11 2000,00:06

first id like to point out i did research this, and i did use examples, so shut your fucking mouth. second id like to point out that nowhere in my reply did i say "youre all wrong", so again, shut your fucking mouth...on the topic of ad hominem arguments, wasnt that what yours was?..and besides, what do you want? a pre-emptive response to what you havent already said?...gimme a break, im only human.

i do consistently run my geforce SDR above 1024x768x32, and i rarely have framerate drops, mind you this is on a p2 450, so you dont know what your talking about. maybe you shoudl start running tests before you start stating things about my computer.

how do you figure the closest will be the 128 board?...considering the memory on the v5 6000 has to be duplicated per processor, that means that the 128 must be split down to 32...wait...isnt thats whats on the v5 6000?...hmmm....

q3 has some T&L support, but it also is one of the most known games, and most graphics intensive. if people used something like SS2 for benchmarks, it would be dumb, because its not a limits pusher, and it wouldnt give anyone a very good idea of how the card performs. and besides, if 3Dfx cant see that T&L is teh future, then their benchmarks should be hampered.

about nVidia lying, sure that may not be the greatest thing, but can you seriously tell me that you would expect them to say "we have FSAA, but it doesnt work at all, and will not work with anything"?..no, so they say it does work, and hope that someone will make it work

---

Posted by Avenger on Jun. 12 2000,03:20

quote:

---

Originally posted by TiresiasX:
Concerning memory bandwidth:

Two processors accessing shared memory does not equate to twice the sustained bandwidth. Memory access between requesting

---

That may be true, but the GF2 and V5 5500 have about the same theoretical memory bandwidth, 5.2 GB/s. However, in high resolutions, when memory bandwidth is crucial, the V5 usually wins out.

quote:

---

And as for 3Dfx's anti-aliasing approach, their approach plus their aging architecture requires implementing more than one processor unit, if you read 3Dfx's own white paper on it. 3Dfx considers it a reasonable assumption that it can be done on one chip in the future (still with a severe performance hit), but they either haven't developed their drivers that far or do not have the capability in hardware. This is why the single-chip Voodoo4 has FSAA disabled.

---

The fact that the V5 requires 2 processors for FSAA is irrelevant to the end user. The V4 is a low end card, designed largely for OEMs and meant to compete with cards that don't offer FSAA. 3dfx's architecture may be old, but look at what Intel has done with the 5 year old ppro core. AFAIK, they still hold over \%70 of the x86 market.

Thanks for the reasonable and rational post.

---

Posted by Avenger on Jun. 12 2000,03:48

quote:

---

Originally posted by Sithiee:
first id like to point out i did research this, and i did use examples, so shut your fucking mouth. second id like to point out that nowhere in my reply did i say "youre all wrong", so again, shut your fucking mouth...on the topic of ad hominem arguments, wasnt that what yours was?..and besides, what do you want? a pre-emptive response to what you havent already said?...gimme a break, im only human.

---

You didn't have a single reference in your post. It was simply a mixture of common knowledge and opinionated heresay. I never accused you of saying "you're all wrong". I was pointing out that individual responses from _me_ were better than _me_ posting a message attacking the whole board. I attacked arguments, not the people who originated them. Why are you being so hypocritical? What makes you think that I asked for a response to something I have yet to say?

quote:

---

i do consistently run my geforce SDR above 1024x768x32, and i rarely have framerate drops, mind you this is on a p2 450, so you dont know what your talking about. maybe you shoudl start running tests before you start stating things about my computer.

---

At those resolutions it's still a lot slower than a DDR Geforce.

quote:

---

how do you figure the closest will be the 128 board?...considering the memory on the v5 6000 has to be duplicated per processor, that means that the 128 must be split down to 32...wait...isnt thats whats on the v5 6000?...hmmm....

---

Even the 128mb GF2 won't approach the V5 6000 and it still costs more. Only the textures need to be duplicated for each processor. As a result, 128mb is plenty of memory, especially with the texture compression employed by 3dfx, except for insanely high resolutions such as 2048x1536x32 and nothing until the NV20 or Rampage has the fillrate to handle that.

quote:

---

q3 has some T&L support, but it also is one of the most known games, and most graphics intensive. if people used something like SS2 for benchmarks, it would be dumb, because its not a limits pusher, and it wouldnt give anyone a very good idea of how the card performs. and besides, if 3Dfx cant see that T&L is teh future, then their benchmarks should be hampered.

---

There are dozens of other system intensive games that can be tested. IMHO, tile based rendering is the future.

quote:

---

about nVidia lying, sure that may not be the greatest thing, but can you seriously tell me that you would expect them to say "we have FSAA, but it doesnt work at all, and will not work with anything"?..no, so they say it does work, and hope that someone will make it work

---

I think they should be honest and admit that they added FSAA as an afterthought and it only works on certain games. They designed the GF2 and are responsable for making it work as designed; no one else is.

---

Posted by TiresiasX on Jun. 12 2000,06:44

well Justice, you could get yourself an SGI IRIX workstation...if you happen to feel like selling your house for it, LoL

and an interesting thing...if you check at Tom's Hardware, when he's not (rightfully) knocking RDRAM, he's got a pretty decent review of the GF2 that actually puts it side-by-side with the Quadro, nVidia's professional openGL chipset...

The Quadro consistently performs slightly worse than the standard GF2!

I suppose it's not so surprising...the Quadro's probably optimized for T&L and had some of its fill rate capabilities scrapped to cut cost. Makes sense, since most 3D designers (me included) prefer to test-render in wireframe and do the final render completely in software. Real shame though...looked like the Quadro might have been the perfect chipset for professional OpenGL and gaming combined.

------------------
"Welcome to my Neon Dream."

---

Posted by TiresiasX on Jun. 12 2000,06:58

Oh, btw, SGI's been marketing Wintel/Linux boxes for a while (with a seriously marked up proce, heh). They used to sell very nice quad Xeon workstations with their own proprietary peripheral bus, which is superior to PCI. Back then, they were using a Number Nine vid accelerator...now they've got single-CPU boxes with VPro graphics chipset.

Anyone ever heard of VPro? I haven't...proprietary SGI company, perhaps? And what happened to Number Nine? They had some seriously kickass products a year ago, and now they've "ceased operation." It's like they were just completely forgotten.

------------------
"Welcome to my Neon Dream."

---

Posted by JusticeDenied on Jun. 12 2000,17:32

I gotta agree with avenger on these.... I think for the most part he has it nailed.

Drivers are already being patched see < www.voodooextreme.com > LOD patch is sweet

I wont be happy till I have a card that can do 1600x1200x32 true fsaa and all the visual eyecandy bells and whistles at 90 frames a sec... now who is gonna build it for me ?

---