Slashdot Mirror
3dfx Voodoo 5 Review
gnewt writes "The guys at The Tech Report have put together a pretty comprehensive review of 3dfx's V5 5500 AGP. There's a comparison against several NVIDIA products, including the GeForce2, on image quality, FSAA, performance, and mip mapping. Overall, a nice introduction to modern 3D accelerators."
The explanation lies in the fact that the human eye can only see X frames per second. If I remember correctly, it is somewhere in the neighborhood of 40-60. So once you get above that range, your eyes/brain can't tell much (if any) of a difference from increased framerates.
IANANeurologist, but, I seem to recall some kinda fringe-sciencey research on increased framerates in film that suggested that higher framerates make a stronger impression on the brain -- they seem more real to the viewer, are more likely to be remembered or to influence his/her thinking. You may not be able to tell the difference between 60 and 80 fps at first glance, but subconsciously, there is an impression. The military was even rumored to be thinking of using faster-than-normal framerates in their training films.
This makes sense to me -- for one thing, whatever framerate the eye allegedly runs at, I doubt it does so with atomic-clock precision. Even if you're only seeing at 40 frames per second, you're not seeing precisely one frame every 40th/second. Framerates that double your rate of perception will seem clearer because there's less of a chance of your nervous system detecting the flicker between frames or other subtleties that betray that what you're seeing is not real.
Also, The nervous system may very well be able to register hundreds of images per second in moments of high stress -- many people in life-threatening situation later recall a sensation of time slowing down, and suddenly registering an outrageous level of detail with surprising clarity. While a heated game of Quake 3 doesn't necessarily qualify as life-threatening, a faster framerate may be better at fooling a brain in a high state of excitement -- it may even help stimulate the brain in and of itself.
--
perl -e '$_="06fde129ae54c1b4c8152374c00";
s/(.)/printf "%c",(10,32,65,67,69,72,
$_="06fde129ae54c1b4c8152374c00"; s/(.)/printf "%c",(10,32,65,67,69,72, (74..76),(78..80),(82..85))[hex $1]/eg;
Only question I have is how hot this thing is going to get? My Voodoo 3000 AGP get blisteringly hot after some games, and it only has a heatsink. This card has two fans!
- I don't care if they globalize against free speech. All my best free thoughts are done in my head.
Ugh! I can't stand to sit-down at other people's computers when they have the refresh rate set at 60Hz: it flickers so badly. I need the refresh rate set at least 72Hz before I stop noticing flicker, 75 better, 85 or above the best.
The thread was talking about optimal frame-per-second rates.
"Hz" reflects how often a phosphor on a CRT (not LCD or plasma) is reheated. If it's close enough to the refresh rate of fluorescent lighting (60Hz NA; 50Hz EU) then you get flickering. It is limited by the bandwidth of the video card+cable (expressed in MHz, typically), or sometimes by the quality of the monitor.
"FPS" is how often the image on screen is changed. In a 2D environment FPS is noncritical -- it's almost always "high enough". In a 3D environment, FPS becomes limited by the CPU, GPU, and memory bandwidth. It is very often not enough.
Obviously, if FPS is higher than Hz, not every frame will actually get displayed on your monitor. Low FPS will not cause flicker, however -- it causes a jumpy image.
cheers,
mike
I think the memory bandwidth problem is best solved by ATI and Matrox. ATI uses its HyperZ archtecture to greatly reduce the number of reads into the Zbuffer, which really helps memory bandwidth. The fact that it can beat a GForce2 in some tests (only one or two now that the Detonator 3 drivers are out. In the others it still loses significantly Check www.sharkyextreme.com for the latest benchmarks in the ATI 32MB Radeon review) even though its runs as a significantly lower clock and has a good deal less fill-rate is a good indication of this. Matrox seems to have the best idea with its DualBus. Put two 128bit DDR SDRAM interfaces on the board and watch the bandwidth problems magicall dissapear.
A deep unwavering belief is a sure sign you're missing something...
You've hit the nail on the head. This is why graphics card reviewers use tests like the old Crusher demo.
:o)
You push the hardware as hard as it can go and you can get a good idea of how low your framerate will go in a frantic 10-people-on-the-screen-why-am-i-reloading mode
I just put together an AMD 900 w/ a GeForce2 MX for a friend and it almost made me consider my personal boycott on nVidia. On my old computer and on the new one, sitting still on the average game, i get a good framerate. But when it heats up, the new computer doesn't lose framerate. And that's the difference between life and death in a game.
So why is it that we've made all these advances in graphics hardware but we still accept 30-40fps as a 'standard'? Honestly, I'd prefer a game that looks a little less glitzy and delivers 40fps at ALL times. The gameplay is what it's about, isn't it?
But Banshee was just basically a 3d/2d hybrid card that combines the Voodoo chip (not Voodoo1, but Voodoo chip architecture) together with a 2d chip. Voodoo3 just extended Banshee, like you said.
As much as I have liked 3DFX products in the past, that they have made a few (fatal?) errors on this product. One, reading the review, you get the defenite impression that this product was cobbled together quickly with many shortcuts that affect the overall quality, and then rushed to market in order to keep afloat against NVIDIA. The product almost has an air of desperation about it. While they do have a few good ideas, like being able to daisy-chain the processors in future products, I would bet heat and power considerations will limit this in the future. Overall, it is almost as if 3DFX is trying to cheat thier customers with a holding action while they, hopefully, deliver a once more superior product in the future.
And this isn't meant as flamebait, but I am sure they will ensue...
You say you want a revolution....
taking a 450hp V8 engine and putting it in a 15-year-old Ford Escort
Hmmm.. Sounds like fun!! I've got a Windsor-block 351 out back, and I'm sure I can get a slightly rusty 'Scort with a blown motor for $200.. But I'd have to beef the frame, knock the firewall back, put in a new rear-end..
Nah, I think I'll just go play another round of Need For Speed IISE on my pimped out 5x86-166 (Peltier, 256M of EDO and a SLI pair of 12M VooDoo II)
.sig: Now legally binding!
Don't worry, the V5 can't get any hotter than the V3 did. An actual temperature reading on a Voodoo 3 3500 was 180F, legally hot enough to cook raw meat in New York City, and the temperature limit set by Intel at which point performance and stability start to degrade. The V3 AGP series was a thermal mistake. The only way for 3dfx to avoid magnetic noise in the bus from the multiple fans, however, was to add on the power connector.
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
From page 10 in the article:
"A note on Linux performance
We managed to get pretty far down the road of testing the Voodoo 5 using XFree86 4.x with Linux before we discovered that 3dfx's current X drivers don't support dual-chip operation with the V5. That means no SLI mode and no FSAA, either. At that point, we decided to forego benchmarking the V5 in Linux until the drivers have some time to mature. 3dfx has committed to open source principles and to Linux support, and we have every reason to believe they'll continue their tradition of decent Linux support. But it ain't there yet."
Gee, no problem, I'm only missing one of my two cpu's, and half the memory. I wonder if that means I still need the external power?
So the v5's linux support is about as good as Win9x's support for a dual processor system, that being you only get half of what your trying to use.
Funny how the GeForce2 GTS's Linux scores are within 99% of the Windows scores.
But then, what do I know?
-- "So they told me that using the download page to download something was not something they anticipated." - Bill Gates
VSA-100 is probably their biggest creation for a long time, but simplicity of the chip, and it's need for parallelism for multiple VSA-100s which basically amounts to SLI on the same card is suggesting that 3dfx should spend less money on those dumb commercials and spend more on actually making something good.
Actually, the modular design is a brilliant idea - it gives built-in design scalability, and gives you huge memory bandwidth gains. Memory bandwidth limits fill rate for games at the resolutions I play at, so this is a big win.
3dfx's mistake was making their individual chips sub-optimal.
The logical way of salvaging what they can is to switch to 0.15 micron (or better if they can get it) and put multiple VSA cores on one die. This gives them a higher maximum clock frequency and lower power dissipation. Have multiple memory busses running from the die (or interleave requests across one bus), and you have a card with decent processing power and enough memory bandwidth to beat the hell out of anyone else. Part count goes down when multiple cores are bundled onto one die, so the card costs a lot less too.
All of this is effectively "for free", as the R&D has already been done in the development of the Voodoo 5 series.
It remains to be seen whether 3dfx has the money left to pursue this course.
Tweaking the chips themselves wouldn't hurt, but is secondary compared to the huge benefits of the modular architecture itself.
I-Max is shot at 48 frames per second, but movies shot at 24 fps are screened at 48 fps. I would recommend trying to get your frame rates above 48fps for your whole experience.
On a 24fps movie the shutter is opened twice per frame. This helps eliminate strobing of the image. Even with this measure, you can see strobing under normal lighting conditions. There is a *technical* reason why movies are shown in the dark.
For comparison NTSC television is shot at 60 fields per second. These 60 fields are interlaced to give 30 frames per second. (Well, actually 29.997 fps)
In any case, you can tell the difference in frame rate. For example Showscan, used for amusement rides, is shot at 60 frames per second, giving a strong "you are there" sensation. The faster the frame rate, the more your brain tends to believe your eyes. (There is an upper limit, but I don't know it.)
Interestingly the guy who created Showscan, Douglass Trumbull, rejects the use of Showscan for storytelling. "It was film that looked too real-too much perhaps like video."(Videography Magazine May 2000 page 30)
There is also the issue of frame rate variance discussed by another poster. When the scene gets 'busy' you want to make sure your frame rate stays high. This is more critical for computer applications because they have an effectively high shutter speed. By this I mean each frame is still and sharp, it lacks motion blur. If it had accurate motion blur a lower frame rate would be more tolerable. (Inaccurate motion blur will confuse the viewers notion of how things are moving.) Without motion blur images are more easily percieved as "strobing" or flickering, which of course detracts from the realism of the environment. Since adding accurate motion blur is hard, you have to pump up those frame rates- and keep them high.
Don't post innacurate information
If you do, I swear by my pretty floral bonnet I will end you.
Of course. Look at Tomb Raider.
----
ADVENTURERS! - ANTIHERO FOR HIRE - CARDMASTER CONFLICT
You are correct, however, on that there is an upper limit on percievable effect of adding more frames per second but it is closer to 80. I saw a nice scientific explaination of it all, but I'm too lazy to look up the reference.
The explanation lies in the fact that the human eye can only see X frames per second. If I remember correctly, it is somewhere in the neighborhood of 40-60. So once you get above that range, your eyes/brain can't tell much (if any) of a difference from increased framerates.
=================================
I pledge allegiance to the flag...
of the Corporate States of America...
Man, you haven't played many games lately, have you? Not only can you notice the difference, but the difference is MUCH bigger than you think. Try it yourself, and then do a quick 180 degree turn in a game.
Even more importantly, it's important to notice that the rate is an AVERAGE. Even if the game has an average of, say, 100 FPS, in moments of intense action the rate still drops WAY WAY down. I wouldn't dare play an intensive game at 40 FPS -- as soon as the action started, you will be likely to drop into the single digits!
Of course, for single player games, 40 is probably perfectly adequate.
Not to mention, with full screen antialiazing just beginning to take off, all of that extra rendering power can now be used to make the images look better! Right now, only the most powerful video cards can perform any anti-aliasing at all, and it's not even particularly good anti-aliasing. There is still a ways to go yet...
As others have mentioned, the top end is probably closer to 60fps than 40.
More important, though, is the headroom you get with a faster card. A game like Q3 has a standard deviation of about 7fps, which means over 15% of your frames are under 33fps, and about 3% are under 26fps. These are very noticeable slowdowns.
At 80fps mean, your standard deviation may jump to 14 fps (it's not a linear progression in real life, but for argument's sake...), 97% of your frames are at 52fps+, and 99.85% above 38fps. So it's smooth all the time, not just when you're standing around with nothing happening.
And that's why NVidia is still in business.
cheers,
mike
There are also 1 or 2 companies that have announced they will be making PCI versions of their GeForce MX cards. (Don't know the ones off hand)
Subject says it all, the V5 is hardly news: the post smells only of 'the other day we posted GeForce 2 MX and G400 now we gotta post something about the V5 otherwise the 3dfx fanboys are gonna get upset'.
Come on, now you only have to post info about the Radeon and we're all set.
OTOH I wonder why Anand pitted the MX against the G400, maybe he didn't want to piss off Matrox too much ? If the MX totally destroyed the G400 can you imagine what the GeF2 would have done ?
To the moderators: please resist the urge to mod this as flamebait. I have used the V1 (owned), the V3(work), the V5(work), the G400(own a Max) and the GeForce2(work), so I feel I am fairly neutral here.
-- the cake is a lie
Perceptually, there is no difference when looking. The difference comes with strain on the eye. Eye has to work harder to fool the brain at lower refresh rates. This causes all kinds of muscle strain.
Of course, when graphics cards can put up one frame per CRT scan, that is the utmost limit and beyond which there can be no perceieved difference simply because the monitor is the limiting factor.
Just because your eyes cannot discern between events .05 secs apart (in the best case) does not mean that there is no perceived difference. For instance, stand outside at night, stare into a camera flash (this takes less than .1 secs if you are close enough to the camera) and then try to see anything for 1 hr (assuming you didn't blow your rods and cones). Admittedly, this is the most extreme of the situation but it illustrates the point that it isn't the fact that the eye can only "reset" the retina 10-20 times a second, but that the mechanics of the eye that become important for a pleasureable experience.
-----------------------
Nicotine free Amish .sig.
What if you wanted two monitors? I dunno about multiple monitor support for these, but easily one can do two video cards (one in the AGP slot, one in PCI) with no problem...
"Pinky, you've left the lens cap of your mind on again." - P&TB
"I can see my house from here!" - ST:
I can tell the difference between 60 and 50 fps. I find after playing at above 60fps that 50fps or lower is jerky, and much harder to play well at. You get used to it after a while, but it is noticeable.
A lot of people talk about visible framerates, and compare games with film. This simply cannot be done. Film contains temporal anti-aliasing, which smooths out a lot of the visual jerkiness that a game would exhibit rendering the same scene at the same speed.
VSA-100 is probably their biggest creation for a long time, but simplicity of the chip, and it's need for parallelism for multiple VSA-100s which basically amounts to SLI on the same card is suggesting that 3dfx should spend less money on those dumb commercials and spend more on actually making something good.
Those little tricks don't work anymore when you're starting to become the underdog, with companies like nVidia and ATI nipping at your heels...
You can definitely notice improvements above 40fps. True, you get diminishing returns after the 30-40fps point, but there's something silky-smooth about 60+ fps that puts 40fps to shame.
Obviously the eye and the brain can process information MUCH faster than 30 or 40hz.... why do you think monitors look better at 80hz than 60hz? I'm not sure where that conventional wisdom (anything over 30-40fps doesn't matter) got started, but it's totally wrong.
Even if you couldn't tell the difference between 40 and 60fps, a card with excess power is going to be more future-proof. A card that can only run today's games at 30fps is going to have a hard time with next year's games. Today's monster card that runs today's games at 200fps is going to work nicely with next year's games, too. :-)
Whoa, that troll almost bit my hand off when I fed him!
OtakuBooty.com: Smart, funny, sexy nerds.
Above an average of about 40FPS, nobody notices anymore - they can't!
Maybe not conciously, but it DOES make a difference. Ever seen an I-Max movie? They're shot at 48FPS instead of 24, and it definately makes a difference.