Slashdot Mirror
More on the GeForce 3
Tom has extensive poop on the GeForce 3. Some history, and a ton of why this card may or may not be worth the $600 price tag. Lots of stuff on the Vertex Shaders (which look to be a truly amazing idea). Very worth a read if you're interested in state-of-the-art 3D, or just wanna sneak
preview at what next year's FPS will be doing.
I don't care what the context is, I'm always a little apprehensive about clicking on a link labeled as 'extensive poop'.
How much does a basic PC cost these days? Not a whole lot more than the G3, right? How long will it be before video card manufacturers are throwing in a free PC with the purchase of one of their cards?
So it`s ok to be aware of market tactics principles and to approve or disapprove, but to actually NOT buy a graphics card which is superior to others because it makes the other businesses go slower ?
That`s definately NOT the right angle. In fact, if nVidia gets to sell this card as hot as the previous 2 versions, it can set (and raise) the standard in 3d lighting again, and frankly that`s what I want. Ofcourse monopolistic situations like e.g. Soundblaster are absolutely bad to competition (and quality) in the market, but that`s because a Soundblaster (could als have used Windows here) is a good product, just not top of the line. That doesn`t mean there are no other soundcards out there which are actually better, only that you`ll have to pay more for those and go out and look for them.
I support ATI and to a lesser extend, Matrox, because they are the only rivals left in the field. But If I had to buy a card today, it wouldn`t be either of those 2, because I simply want a 'standard' compliant full fledged and top of the line game experience, not the feeling that I did something good for market competition. In the end I might financially regret that choice, but if nVidia creates the best cards at the end of the day, I`m only happy to spend some cash on them. If someone else can do the same or top them AND has less expensive cards, obviously that`s a thing to consider. But today I cheer for nVidia, as I have more pro than con.
With great power comes great electricity bills.
There's no reason for that to happen, now that Nvidia has no real competition. Why not keep prices on the GF3 high? People who want it that badly will pay it, those who won't will have to buy some other Nvidia product if they want to keep up with the pack. Nvidia wins either way.
This sig is xenon coated, and will glow red when in the presence of aliens
I find it a little worrying that so much of the work that has gone into the GeForce3 has been implementing unprecedented new features such as these vertex shaders, rather than improving more general stuff such as fillrate or transformation and lighting. This leads me to believe that Nvidia's goal with this chipset is not to improve the 3D gaming experience of their customers, but rather to lure developers into using these (admittedly excellent) new features.
How is this a bad thing, I hear you ask? Well, it looks to me like an "embrace & extend" tactic. If the developers use vertex shaders to make their games look cooler, then other 3D chipmakers have to either scramble to provide the same features, or all the cool new games will run like ass on anything non-Nvidia. Only Nvidia can get away with a tactic like this because of their present dominance of the market. Witness ATI's Radeon - they added some very innovative features (like all the z-buffer accelerating) tricks but they were all dedicated to improving performance with current software. They couldn't introduce radical new features because nobody would use them, supported as they were only by a minority chipmaker.
If you don't want to see the 3D industry completely monopolised by a single player, avoid the GeForce3, and avoid any games written to depend on its features. Support chipmakers that are seeking to make everything run better, like ATI and PowerVR.
-----------------------
Nicotine free Amish .sig.
Another massive, expensive upgrade, that all the latest games will require you to use (after all, they won't run on old cards 'cause they can't be programmed)...
Screw it. I'm not paying more than two hundred for a video card. Anyone who'd shell out six hundred for one of these is insane. You can get another box for that much, pre-rebate.
You can try going with a company that uses unified drivers.
For instance NVidia offers the same driver set for everything from the TNT, TNT2, up to the GEForce2 Ultra (and maybe GEForce3).
Matrox is also famous for offering unified driver sets going back to the original matrox millenium.
ATI is doing a poor job of supporting Win2000 even with their latest and greatest. There was a recent comparison of Win2000 vs Win98 drivers from ATI and NVidia, and the ATI drivers performed terribly! In some cases 1/2 the speed when tested under Win2000. Plus buggy (some tests could not be completed).
I think the biggest problem is driver obsolesence. I have an "ancient" ATI Rage 128 video card (an AIW to be precise) and ATI has never delivered more than a "beta" set of drivers and applications (TV, etc) for the AIW128 cards under Windows 2000. I'm doubting there will EVER be another set of drivers or tuner software for this card from ATI.
The video card people seem to have like three people that write drivers, and they're always busy writing drivers for the "current" crop of cards, until the cards actually are available, at which point they switch to writing drivers for the "next" slate of cards and the "old" cards simply do not get new or improved drivers written for them. A new OS often means getting stuck with the OEM drivers provided in the OS.
I'm perfectly happy with my ATI-128 performance in the games I've played it with. I've toyed with hunting down a $120 GEForce2 card, but for the reasons you stated I'm missing the why part other than getting drivers more modern and optimized drivers than I'll ever see for my existing card.
How did you get "insightful" mod points from that??
A: You can tell the difference between 30 and 200 fps. Maybe not between 70 and 200, but 30 and 200, yes. And a system that gives you 30 fps in one place will bog down to 10 fps in another. If you can get 70 fps, it will likely only bog down to 30 fps when things get ugly.
B: Getting tech like this out there allows game developers to push the boundaries even further. Now granted, we didn't need the explosion of colored lights that happened when the voodoo2 came out, but still, the point is good. As the tech grows, the developers can use a toolset much richer than they had before. Look at the differences between Quake 1 and Quake 3. The difference between a Voodoo 1 and GeForce2. Imagine that level of difference from what we have today....
C: Your example uses 1024x768. Why should we settle for such a lame approximation of reality. My desktop is 1600x1200. I drop to 1024x768 for my gaming, because anything higher causes noticable performance degredation. I used to settle for 512x384. Now I can't imagine going back to that. And in a few years, I won't imagine being forced to go back to 1024x768.
Nobody's forcing you to buy these new toys. Not everyone needs them. Personally, I can't see spending 10 grand on a home stereo -- after a certain level, they all sound the same to me. But I surely don't say it's "against all common sense" that someone might. I buy my toys, you buy yours, and we'll all live happily ever after.
Not representing or approved by my company or anybody else.
Pretty impressive really. Anyone who might feel a twinge of conscience when following a link to a $600 video card they're thinking of buying is almost immediately comforted with a charity banner where they can assuage said conscience.
--- Hot Shot City is particularly good.
"I often wonder why people spend an absolute fortune buying the lates video cards when the simple fact is that the card will not be used to its utmost capability for several years"
;)
Why do people buy SUVs or luxury cars when a Geo Metro or Mini will do? Why do people buy the latest fashions of clothes when they could get last years in sales?
A lot of people want to play the latest games when they first come out, and have the best machine there. If you don't feel that pressure, then I congratulate you: you're in a sensible minority.
"Noone can tell the difference between 30fps and 200fps anyway"
That's so untrue. I can tell the difference between 50 and 60, no problem. After playing at 60 or above, a drop to 50 or below is very noticable and definitely not as smooth and hampers playing ability until one adjusts.
"Another problem with video cards is that the performance is becoming optimal anyway"
No. There's a very long way to go yet. With more power, there are so many features they can add. Go and read something about 3D graphics, and you will realise how limited this cards are still.
"At 50fps this is approximatelt 37million pixels per second. So it is intuitively obvious to all that a video card with a performance in excess of 37million polygons per second will not provide any better performance under those conditions. Why pay extra for something you can't see?"
You can't base polygon count on pixel count. Some polygons get rendered, but then are obscured by polygons in front of them. So yes, you do have to pay for something you can't see
"It is like insisting on a 500kbit sampling rate, when 70kbit sampling rates are perfect to the human ear"
Not all sounds are sensed via the ear.
In general, I do agree with your sentiment. I bought an original GeForce 256 DDR when it first came out. I'm still trying to justify the expense. If I had waited, I could have got a better GeForce for less. I'll do that again. I'm sure I'll buy a GeForce 3 eventually, but I'll wait until the Ultra model is cheap enough (I just know that there will be several generations of the card).
The whole point of the modern "3D accelerator" was to bring 3D graphics to the consumer at modest prices as compared to 3DLabs, SGI, and their ilk. Now, it looks like nVidia is either knowingly or unwittingly attempting to enter that territory by increasing prices to the "professional" 3D range.
:)
Nobody is going to program exclusively for this card until it saturates the user base. Which, at this price level, ain't gonna happen soon.
Wonder if the "professionals" will strike back
As a hardware designer, I'm very pleasantly surprised by the functionality and likely performance of the new chip. The new memory controller architecture and usage optimisations are particularly impressive; the array of new functionality is mind-blowing. However, they still seem to be missing some basic things. The 2D quality is only going to be "the same as GeForce2" - i.e. well below ATI and especially Matrox at high resolutions. There is no multi-monitor support. All the anti-aliasing modes use uniform sampling (which is very bad). The filtering used in anti-aliasing and mipmap generation seems to be plain averaging (i.e. a hideously ugly box filter) - when very little extra effort would permit gaussian or even programmable filters, providing much better image quality. The video features don't stack up to those of ATI's Radeon. As truly stunning as I find the 3D and memory architecture of the card, it's really going to be difficult to justify buying one unless the machine is used exclusively for games. For anyone with a monitor bigger than 17" it really won't cut the mustard for anything else - same as GeForce and GeForce2. I hope they hurry up with the GeForce3 MX and pray that it's not too badly crippled - no-one seems to want to make a card that's suitable for work _and_ high-end 3D gaming, since Matrox released the G400 back in 1999. Sigh.
What if the "objects" are intersecting? How do you sort them in a meaningful way? And how do you properly render these intersecting objects if you've turned off z-buffering? (perhaps you draw intersecting objects separately, with the z-buffer on)
"It's overkill, of course. But you can never have too much overkill." - Anonymous Slashdot Coward
I plan on buying five of them, and then having fun by taking them out with my shotgun while sailing on my 120 ft. yaught. Later on, myself and five gorgeous women will laugh about the poor geeks, drink a fine wine, and perhaps top the night off (well not really) with a nice cuban cigar.
Some people use all the neato features of a card like this to get faster OpenGL performance out of Lightwave or Maya or some such.
It is great that we can use it for games too, but that isn't the point for many. I am sure there will be an even more expensive version of this in Nvidia's Quadro line, it'll have greater throughput and more processing power...so it'll get bought. It'll make DOOM 3 scream, but that isn't why you buy it.
Unless you are a, "soul of independent means."
Don't post innacurate information
If you do, I swear by my pretty floral bonnet I will end you.
There's a list of previews here Tom's is one of the many out there...
I think one of the coolest things about the GF3 is the ability to get a pixel write count back for a polygon. i.e. you draw a polygon and it tells you how many pixels were "written" to the frame buffer due to alpha and z-buffer test.
And since you can turn off color and z writes, you can test visibility with no changes to the frame buffer. This is perfect for a portal game where you can cull entire rooms if they are not visible because of things you traditionally couldn't compute. If there is a big fireball in front of your face, or a character/pillar is blocking the view. If you have a few monsters that require a significant amount of time to draw, then you can test to see if they are visible first by rendering a coverage polygon first.
You can use this to test the visibility of lens flares so they fade smoothly in and out as they go behind other objects.
You can also use this in game logic in combination with a shadow map to tell how much "in the shadows" characters are. This can make the AI more realistic.
Getting back pixel write counts from the hardware has a very long latency, so it can't be preformed super frequently - but it's a lot faster than trying to read and process the z-buffer yourself.
And for those of you not wanting to spend $600 for a GF3, just wait for the Xbox - it's including almost exactly the same hardware for half the price.
-- Virtual Windows Project
http://www.bluesnews.com/plans/1/
Carmack has quite a bit to say on the subject as this .plan update is rather long (a little too long for a /. comment I think).
mr.nobody
--Don't you wanna go where nobody knows your name?
In case anyone wants a quick link to the other big reviews...i dia_geforce3_preview/
t ech/i
Sharky Extreme: http://www.sharkyextreme.com/hardware/articles/nv
AnandTech: http://www.anandtech.com/showdoc.html?i=1426
HardOCP: http://www.hardocp.com/articles/nvidia_stuff/gf3_
-pepermil
No offense, but having worked with this architecture for a while now, I have to say that the NV2x approach isn't an attempt to hamstring the graphics industry. It's an attempt to raise the bar of hardware design and bring the industry to a new level of verisimilitude in graphics rendering. Criticising nVidia for being a monopoly because they have the technical smarts to develop a revolutionary rather than an evolutionary solution just doesn't make any sense.
With the programmability of the vertex and pixel shaders, graphics applications are now free to create a whole new engine architecture... one that's free from the idea of fixed-format vertex data but instead is purely representation-driven. Because you can pass any binary data that you want to the vertex shader, you no longer have to represent the properties of your surfaces in an implicit format whose characteristics are defined by the fixed capabilities of the hardware. Now that this programmability is available, you can encode surface data in a format that actually stores exactly what you want. The NV2x is the first hardware engine that I feel can be called a "GPU" in more than just name - its capabilities will allow application developers to craft graphics engines that just aren't possible on a card that is "dedicated to improving performance with current software" as you cite you'd like to see. ATI recognize this - witness their Pixel Tapestry technology for pixel shading.
NV2x is the same kind of advance over GeForce2 that the original 3dfx Voodoo cards were over the prevailing PowerVR and VIRGE chips back in (1997?). You didn't see anybody complaining that 3dfx were trying to lock people into their proprietary technology back then, for the simple reason that everyone recognised the potential that was inherent in the change of focus. It took a while for games to become "3D Card Required"... but I'm 100% certain that nobody wants to go back to Quake II-era rendering. The benefit to the application programmer and the consumer is obvious. NV2x may not be the winning solution in the new space that's opening up - but it's a damn good opening salvo.
To paraphrase your post: If you don't want to see the capabilities of 3D graphics engines advance beyond the current status quo, avoid the GeForce3. And miss out.
- Butcher
P.S. You'll be amazed when you see what we can do with this technology. This is a great time to be a game designer or game player.
W does *not* stand for weight, Tom !
W stands for homogenous coordinates. Pg 204, Computer Graphic Principles, Foley & van Dam)
(PreDX7 had something called "rhw" which stands for "reciprocal homogenous W")
- One-pixel polygons are good. Perhaps not in yesteryear's games, which feature huge flat surfaces, but various forms of higher-order surfaces with curves are definitely the trend today. Pixar render their movies using subdivision surfaces tesselated until each polygon is less than one pixel in the final image. We want that.
- Multi-pass rendering is good. Many effects in games are achieved by rendering each pixel more than one time. 3dfx realized this back with the Voodoo2, and added support for multiple textures per pixel. That is good, because it allows you to send a triangle to the hardware once, but get it textured twice. This saves one pass of geometry transform. As soon as an effect requires an additional pass, that reduces your effective polygon throughout quite a lot, of course. So, my point is that even though it might sound excessive based on simple "# of pixels on screen" arguments, huge polygon and pixel fillrates are good, because they allow more passes and thereby more flexibility and coolness in effects.
I don't think quality issues can be compared straight across between audio and video. Or, rather, I'm not good enough with audio to see where seemingly superfluous performance (oxymoron?) can be put to use.main(O){10<putchar(4^--O?77-(15&5128 >>4*O):10)&&main(2+O);}
- Modern 3D hardware supports 32 bits per pixel (RGBA with 8 bits per channel), at least 24 bits per pixel of Z (depth) buffer, and minimum 8 bits of stencil. These are all handy to have around when creating good-looking graphics (although more than 8 bits per channel would be a Good Thing).
Thanks for the heads-up.main(O){10<putchar(4^--O?77-(15&5128 >>4*O):10)&&main(2+O);}
Well, I for one would really like to see that become a reality. Now I know every one who has ever touched a compiler (and often many who don't even know what a compiler does) will give a different opinion on how this is possible, how it has been done, or how it will never be done, etc. And most say that now programmers need to step up and use the features. (I agree with this one) But my question is, are API's keeping up with it. How is DirectX handling more advanced features (not just is it 'supported' but is it clean and efficient) How about OpenGL? How are the projects for middle ware coming like WINE with DirectX and so forth?
I am really impressed with the graphics detail and performance out there right now. I personally want to see more stability. While I am aware of the argument (and it does hold some water I admit) about a higher fps giving better all around performance, it is still common for intense graphics scenes to chug on your machine. I wouldn't mind seeing the ability to average out the fps better, as set by the user. Some method on the hardware to reduce the quality or certain methods only if it detects a forrest of high quality polygons and its own slow speed.
And I would REALLY like to see some better AI, and maybe some api's that make use of hardware driven ai... ok just kidding on that one, but perhaps a set of advanced ai libs with an API isn't too far out there. Tie these in with any of the methods for network distributed processing and you have an amazing LAN party set up. Throw some together on your home server farm and now you have your game server set to go... ahhh. Soon it will be like the old BBS days... but with better graphics and real time interaction.
Well, end of wish list... maybe the internet fairy will bless this if it is seen and make it real.
I seek not only to follow in the footsteps of the men of old, I seek the things they sought.
that's the kind of upgrade that could really increases my productivity, I mean all those xterms could render way faster with that card in my box. I have to talk to my boss about this.
--
Je t'aime Stéphanie