Slashdot Mirror
Rumors of a GeForceFX 5800 Ultra Cancelation?
chris_oat writes "It seems that nVidia's GeForceFX 5800 Ultra may never see the light of day after months of super-hype and annoying delays. This article on megarad.com suggests that poor manufacturing yields are causing nVidia to rethink plans for its (new?) flagship part. Lack of an "Ultra" type solution from nVidia would leave ATI's Radeon9700 uncontested as the defacto performance part."
Lack of an "Ultra" type solution from nVidia would leave ATI's Radeon9700 uncontested as the defacto performance part
The Radeon 9900 is expected out next month, with the new R350 core.
I am glad I don't have Nvidia stock right about now.
I work at Best Buy (unfortunately) and we were instructed to stop selling all Preorder GeForce FX's and destory the boxes and give all the free stuff to the employees or whoever wanted them. Apparently at least the pre-orders will be fulfilled but I don't think the card is going to make it to the stores for quite some time do to "extremely limited supplies" (according to the store memo). At least I got a free Nvida t-shirt and Hat out of it. :o)
-macado
Just a couple of things to add. For me at least, 30 to 40 fps is a minimum requirement. However, that is the lowest it should possibly go. If it gets below 30 fps it starts turning into a slide show.
ALso, aside from just the visual effects, more powerful hardware gives you better performance in game litterally. Example, the quake3 engine. In the quake 3 engine, you can jump much further with 150 fps then you can with 30 fps. The way it was coded if you were to jump, the game checks on a frame by frame basis to see where the jump is going. I think it was designed with a baseline of around 90 fps if I remember right. Which means that if you are going under that, your jumps will be shorter and over it, longer. Also, on Return to Castle Wolfenstein, if you fps ever drops below 30, then you rate of fire actually slows down. So, just FYI, fps can mean more then simply "how pretty" it looks.
I guess that most /.ers would care more about the video card. The URL you posted didn't work. Try this:
l e= article&sid=1270
/ ar ticle_4.asp
http://www.megarad.com/modules.php?name=News&fi
Or this:
http://www.globalcomment.com/science&technology
It's a spoof article unfortunately, but a pretty good one.
HH
But, it's not just a rumor anymore. When it first came to [H], everyone regarded it as BS. It was a rumor posted on a board that spread incredibly rapidly. But, apparently it's been confirmed by either OEMs or nVidia itself to those with good contacts. BFG has stopped taking preorders, AFAIK, because...
"According to an e-mail John Malley sent out a couple of days ago, BFG is concerned that pre-sales may exceed their allocation of units."
So, yes, the 5800Ultra is gone. Oh well. NV35 in June, according to some.
---
nV News
And 16 million colours is more than the eye can see, and 44,100 samples per second is more than the ear can hear . Throughout the march of technology we've heard these ridiculously arbitrary "limits" of our senses, and invariably they are discounted at a future time. In essence you can consider them a sort of justification.
These limits aren't arbitrary. You can test them the same way you proposed that frame rate limits be tested.
For colour gradations, make a picture that has a very gradual colour ramp from 0-255 in each colour (or one that sweeps across colour tones, but that changes at most one component by at most one between adjacent bands).
When I tried this with an old VGA card that used 18-bit colour, I could see banding. I had to stare for a while to let my eyes adjust, but I could see it.
When I try it on a modern card with 24-bit colour, I see no bands if the monitor's gamma correction is properly adjusted.
A monitor without gamma correction will end up expanding some brightness ranges and compressing others, with the result that gradations will not be visible at all in some areas and will be (barely) visible in others. Check your configuration before complaining.
The 24-bit argument applies to distinguishing colours. Similar experiments (not performed by me) have shown that you get about 10 bits of depth in greyscale, as humans have more sensitive black and white vision than colour (which is why everything appears in shades of grey at night with poor lighting; go for an evening walk and look for badly-lit stop signs some time).
You can do the same kind of tests with sound. It's actually more difficult with modern sound cards, as they have low-pass filters that cut off everything above about 22 kHz (nyquist rate of 44 kHz), but a PC speaker works. Or use a piezo buzzer and a signal generator if you're worried about the speaker efficiency dropping at high frequencies. My hearing, last time I tested it (and last time it was tested by a doctor), dropped out about about 18 kHz.
The reason why higher frequencies are relevant at all is because of nonlinear behavior both in the speakers and in the human ear. Beat frequencies between high-frequency tones can turn into audible frequencies when interacting with nonlinear systems (this is how that two-tone ultrasonic speaker linked to a while back worked). However, the key is that the final tone you hear is in the audible frequency range. This means you can duplicate the sound perfectly by using a microphone that acts more like the human ear when recording (i.e. that has similar nonlinear effects), or by recording at high frequencies and applying appropriate transformations before downsampling.
The fact remains that if I played a 20 kHz pure tone at you right now, you wouldn't hear it. And this is easy to verify by experiment.
In summary, while you're most definitely right about frame rates, your other objections about limits are unfounded.
Actually the reason 60hz is hard to view on a monitor while 24fps on a movie screen is fine, is that the entire movie frame is drawn at once, while each of those 60hz consist of a tiny dot drawing the whole screen. Flicker is therefore much more pronounced on a TV or monitor. LCDs by contrast do not do line-drawing and therefore look smashing at 60hz.
Jeremy
I'm sure it doesn't look as bad as 24fps on a monitor. Again, perhaps the only reason movies are watchable at all is that the bluriness at the frame transitions might make it easier for the brain to "add in" the extra information to interpolate
Actually, one of the reason why movies are so horribly jerky is that the actual refresh rate is 48fps, even though the frame rate is 24fps. Each frame is projected twice. The reason for this is to reduce flickering and to protect the film.(projector lights are HOT). Unfortunately, this double exposure messes up the brains visual prediction system, much in the same way a 30fps game on a 60Hz screen, only more so. Since there is a tangible delay between capturing an image in the optic nerve and feeding it to the brain, a lot of prediction is carried out to predict what things are going to look like when you receive the visual stimuli.
I agree that even a monitor at 48Hz would look worse than a movie theatre, but I expect this has something to do with the relatively low contrast movie screens have. A darker image takes longer to "see" than a bright one, not unlike how a photographer needs a longer exposure to take a picture in a dark environment.
Ever seen "3d-glasses" that have one dark glass and one perfectly transparent, instead of the normal red and blue/green? Those work on that principle, and the effect is best when the camera rotates clockwise around an object or pans across a landscape from right to left. If you reverse the direction, the 3D-effect is also reversed.
But i digress:
My point is that the human vision is incredibly advanced with a lot of special adaptations. There is no framerate of the eye. Fighter pilots have been shown to be able to not only see but also correctly identify a picture of a plane even when the image is displayed just in a 200Hz flash.
The ideal frame rate is the same rate as the monitor refresh, and to have a constant framerate. I'd much rather have 75fps at 75Hz than 80fps at 85Hz.
A witty
Am I the only one that sees how freakin' poetic this is? This card was touted as the first real tangible result of the marriage between NVidia and 3Dfx (one of the reasons for the "FX" moniker, supposedly), and the company's having the exact same problems as 3Dfx did with their Voodoo 4 and 5's. Namely, that they're not as fast as people expected, they use too much power and generate too much heat. And their competition is passing them by.
Still, I don't see NVidia in the same precarious position as 3Dfx was at the time. NVidia likes to point out that after the latest Radeons were released by ATI, NVidia's market share actually went up, not down. The super-performance market is actually a very small market, and NVidia still offers the best value out there for mainstream users in the GeForce 4 Ti4200. For most people, the extra $250 they'd spend on a Radeon 9700 Pro vs. a Ti4200 is just not worth it - the extra few frames per second you'd get in most games are generally not even that noticeable, and there are a lot of better ways to spend that money. I don't really think NVidia's got a lot to worry about, then - unless the performance gulf and manufacturing problems become so pronounced that public perception (or misperception) filters down to even the mainstream products (as has been ATI's bugaboo over the years).
Still, it looks like the GeForce FX has been NVidia's first real dud in some time. No doubt the "stock" FX 5800's will be a good value once the NV35 is released (just as the Ti4200's are a good value now), but at the moment the card doesn't seem to really fit in any niche. Performance gamers will choose the Radeon 9700 Pro, mainstream gamers will choose the Ti4200, and low-end or business users will continue choosing ultra low-cost but perfectly capable cards like the GeForce 2 Ti.
Just to clarify, Quake3Arena wasn't specifically coded to do this, it was actually a bug, and it only affected the jumping physics, nothing else in the game was affected (it was not intentional behaviour, in fact the game was specifically designed to try to NOT have the physics dependent on the frame rate). (You could jump a little bit higher and in some maps this gave a big advantage, e.g. DM13, since you could take a shortcut to the megahealth). The bug was fixed in one of the last patches (I think they made it optional though).
The jumping performance also wasn't proportional to the frame rate, the bug occurred around specific frame rates, such as 120 fps.
Having a physics engine be dependent on the current framerate shows a flaw in the game's design, and it is just one more reason to stop using the sorely outdated Q3 engine to benchmark new hardware.
Just to clarify, again .. this WAS a bug in Quake3Arena. However, it WAS NOT a bug in the "Quake 3 engine". It was a bug in the Quake3 game code. The "Quake3 game" is separate from (and built on top of) the "Quake 3 engine". The engine is the basic graphics and network system, source code NOT available, while the Quake 3 game itself was built essentially as the "default mod" for this game, and the source code is available for it.
The slightly-frame-rate-dependent jumping in Quake3 was a bug in the game code, and ONLY affected the jumping. The bug was fixed in one of the Quake3 patches. The game was intended to be designed so the physics were NOT frame-rate dependent. As you said, this would be a major flaw in a game design.
If the physics in a game were frame-rate dependent, you would see a HUGE difference in physics performance between 30, 60 and 90 fps. These sorts of rates affect (badly designed) game physics in a big way - you would notice it quickly. No major commercial game intentionally has such flaws.
You don't understand how PCI IRQ sharing works; typically you will NOT be sharing an interrupt between the AGP slot and the PCI slot next to it. Furthermore, interrupt sharing doesn't necessarily "cripple your system performance".
PCI does have 4 IRQ lines, INTA#, INTB#, INTC#, and INTD#. However, each card has access to all of them, and is allowed to use any of them if it likes. Furthermore, the spec leaves it up to the motherboard implementor how to connect the lines up. You can tie all four together and service them via a single interrupt if you like. PCI is designed to allow almost any interrupt architecture.
In PowerMacs for example, each PCI slot has a single dedicated interrupt line. It is connected to all four PCI interrupt pins. There is no IRQ sharing between slots in a PowerMac no matter what.
Most PCs have 4 IRQ lines from the interrupt controller going to PCI slots. As the lines pass through each slot, the order is swizzled:
int 1 --> slot1 INTA# --> slot2 INTB# --> slot3 INTC#
int 2 --> slot1 INTB# --> slot2 INTC# --> slot3 INTD#
int 3 --> slot1 INTC# --> slot2 INTD# --> slot3 INTA#
int 4 --> slot1 INTD# --> slot2 INTA# --> slot3 INTB#
Most cards use only one of the four interrupt lines, INTA#. Therefore, interrupt sharing (if it happens on your motherboard -- as per above, on some it does not, especially not on architectures not hobbled by backwards compatability with the IBM PC AT interrupt controller) takes place with a periodicity of 4 slots. Meaning that the AGP slot and the PCI slot next to it probably do NOT share an interrupt in most cases.