@minator

Quote:

minator wrote: I agree that Cell utilisation has hardly begun but I don't agree that the 360 has hit it's limit. Getting something to run on the 360 is easier but in order to get the best out of it will be just as difficult - if not more so - than the PS3, that's going to take quite some time.

Wow, I guess the majority of the people on this forum (and outside) have been trying to convince otherwise...

@GregS

Quote:

The whole point of HDTV 1900x1200 was to replace 35mm film based on the threshold of human perception of detail. A one pixel difference difference at this resolution is not physically visible.

Also depends on the artwork. HD "mega-textures" techniques requires massive amounts of GPU bandwidth and VRAM.

Note that ~25GFLOPS Fold@Home for PS3/CELL vs Radeon X1900’s ~58 GFLOPS (~360 GFLOPS peak) basically refects the bandwidth difference between the two streaming systems. The amount of memory bandwidth dictates the sustained streaming computation performance.

Last edited by Hammer on 08-Apr-2007 at 12:21 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@BrianK
Quote:

I agree that AA should be optional. But, no game on the PS3 has yet to use all of it's godlike processing power. Why not use AA and use some of that untapped potential? If a game is using 50% of the console it makes little sense to me not use 75% and remove jaggies w/ AA. It's not a waste if it wasn't going to be used anyway.

As for untapped potential, factor in gaming middleware that targets SPUs.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@hatschi

Quote:

Sorry, but that's very inaccurate. There is no "biological threshold" for the perception of pixels vs. natural blur that can be made solely based on x*y resolution. On sufficiently large screens, you would have no problems to spot single pixels. So you need to look at dpi instead and factor in viewing distance.

The 16:9 ratio was fixed as such precisely because it fills the visual cortex. Positioning oneself that the entire screen occupies nearly the entire field of vision, creates"visual immersion".

That creates a biological threshold.

Of course using magnifying glasses, telescopes, binoculars and standing inches from the screen, tends to negate the effect (as stated elsewhere in this thread, in terms of screen aspect ratio and its purpose).

DPI is a density scale. There is a separate issue. At some point the DPI becomes so dense the unaided human eye will not be able to detect any dots no matter how close you get. In the industry this is usually referred to as "a total waste of time" because DPI is independent of screen size the number of pixels grows exponentially with the increase in screen size, presents no achievable minimal resolution, would if the screen was big enough require a super-computer or two to fill it - may not therefore be practical for several lifetimes.

Instead they found that 1900x1200 viewed in 16:9 is sufficient to do the trick.

Clearly for the hopelessly anally retentive, some sort of barrier should be supplied to stop them nosing too close to the screen. I would suggest an elderly woman who nags "you'll ruin your eyes. You're too close to the screen" could be given away with sets, or a measured stick, one end attached to the viewers head the other to the set, or some transparent barrier, might be sufficient.

Medical treatment may also help. Ask your doctor about "chronic disagreement", "recurring pedantry", and "reactive myopia"
.

_________________
Greg Schofield, Perth Australia

@Hammer
OK... does this mean that the PS3 could do it, even if current games might not be doing it?

I seriously would not know. I am hoping the PS3 is sufficient to produce good 3D moving images at 1900x1200.

I have presumed that the Cell is in its early days as for making good use of the SPEs, and that practical techniques for effectively distributing code within the CPU may also be in their early days of development.

I have assumed that given this chip architecture some 3D rendering might be better done by abandoning some methods and creating new ones to suit the use of co-processors and RAM access involved. That the role of the GPU might subtly change in the process, but obviously my notions are more based on ignorance than knowledge.

_________________
Greg Schofield, Perth Australia

@GregS

Quote:

Instead they found that 1900x1200 viewed in 16:9 is sufficient to do the trick.

At what size? At 24", maybe, at 50", no way.

It depends on the artwork. Test it yourself, go up to a resolution this high and draw a diagonal line with no anti-aliasing, using a pixel paint program. It is by definition sharp and the jaggies will be more than visible. Maybe you're referring to how a video recording looks at such a high resolution but computer generated graphics is hardly video playback, detail-wise.

Last edited by AMiGR on 08-Apr-2007 at 02:00 AM.

_________________

@Hammer

Quote:

Hammer wrote: Note that ~25GFLOPS Fold@Home for PS3/CELL vs Radeon X1900’s ~58 GFLOPS (~360 GFLOPS peak) basically refects the bandwidth difference between the two streaming systems. The amount of memory bandwidth dictates the sustained streaming computation performance.

Are you sure on that? I read it was because the GPU calculations are only for a subset of the calculations of Folding. Cell can do more types of calculations and I guess that's also one of the reasons why the Cell doesn't get that high.

If you would extrapolate that to normal CPU's, that would mean their bandwidth is really lousy. I couldn't get the exact quotes that I tried to find, but below will have enough information, I hope.

From the PS3 Folding FAQ

What type of calculations the PS3 client is capable of running?
The PS3 right now runs what are called implicit solvation calculations, including some simple ones (sigmodal dependent dielectric) and some more sophisticated ones (AGBNP, a type of Generalized Born method from Prof. Ron Levy's group at Rutgers). In this respect, the PS3 client is much like our GPU client. However, the PS3 client is more flexible, in that it can also run explicit solvent calculations as well, although not at the same speed increase relative to PC's. We are working to increase the speed of explicit solvent on the PS3 and would then run these calculations on the PS3 as well. In a nutshell, the PS3 takes the middle ground between GPU's (extreme speed, but at limited types of WU's) and CPU's (less speed, but more flexibility in types of WU's).

@AMiGR

I am not making this up, the size is relative only.

Think about it, 24" or 50" makes no difference at all - it is the distance between the viewer and the screen at 1900x1200/16:9 the resolution in that proportioned viewed so that the screen fills the core and peripheral vision, but not too much further forward.

The fact that size comes into this only shows that there is basic misunderstanding going on.

A smaller 24" with the same pixel resolution has smaller pixels. A larger screen has bigger pixels, the distance between the screen and the viewer for the former is smaller than with the latter, so as long as physics holds the relative pixel size remains the same.

That is if both screens, or indeed a 100" set, is actually displaying 1900x1200/16:9.

It cannot make any difference, the size of single pixel, at the optimum viewing distance, is invisible to the naked eye. That is the biological limit that full HDTV was designed to exploit.

Seeing jaggies means:
The software is rendering a difference greater than a single pixel.
The hardware is not truly capable of 1900x1200.
The monitor is not what it pretends to be.
The neuro-specialists, optomitrists etc got their science wrong in the first place (a bit unlikely).

If there is some other factor involved I would like to know.

_________________
Greg Schofield, Perth Australia

@GregS

Quote:

I am not making this up, the size is relative only.

I never implied that you did, I just disagree with it.

Basically, if you have a diagonal line and you are further back from the monitor, the jaggies are not very noticeable but they are visible for shallow angles. The reason being, at 45 degrees, for instance, you get very small steps and as you correctly said, people usually don't notice single pixels. At shallow angles however the steps are very large and the illusion does not work anymore. The jaggies are **very** visible there, since the line does not look like a diagonal anymore, it looks like a set of other visible lines.

On a 20.1" 1680x1050 (16:10) monitor, almost no jaggies are visible for 45 degrees at 2m away from the screen. For an angle of 5 degrees, however, I can even count the steps.

Anti-aliasing smoothes these steps to get a more photorealistic image.

Think what all these scientists had in mind first and foremost when researching and developing a system to replace SDTV: Movies. I doubt video games (and as such real-time CGI) was very high up on the priority list, we're talking about a world-wide TV standard.

A movie will never have a sharp line of a single solid colour a pixel all, so the steps won't be visible. As I said, in CGI, most lines are sharp unless if you use anti-aliasing or jump some years ahead in time and get nice new hardware that will be able to render stupid amount of detail so that AA will not be needed.

Note that I agree that if there are significant jaggies visible on fatter lines than of a single pixel, yikes, scaling is going on.

Last edited by AMiGR on 08-Apr-2007 at 02:57 AM.
Last edited by AMiGR on 08-Apr-2007 at 02:56 AM.
Last edited by AMiGR on 08-Apr-2007 at 02:46 AM.
Last edited by AMiGR on 08-Apr-2007 at 02:39 AM.

_________________

@jiyong

From http://folding.stanford.edu/FAQ-ATI.html

What scientific cores are supported by the FAH GPU client? Only Gromacs cores? Others cores like Amber?
We will support a special core for streaming processors only (Core_10); this core has elements of Gromacs (mainly for the "bookkeeping", but has a completely rewritten set of inner loop core -- the part which does all the work). Other core support (Amber or Tinker) is not planned, but is in principle possible, if the science requires it.

Does the FAH GPU client run the same WU’s as the regular FAH client?
No, the GPU will run a set of WU’s specially constructed for Core_10 ’s new functionality. While the Core_10 WU's use the same file format as Gromacs WU's, the scientific code which performs the calculation is different and the WU's for Core_10 will yield incorrect results if run with Gromacs (and vice versa).
------
Obtaining correct results is an issue for the current GPUs. R580 can't calculate integers (added with DX10's Shader Model 4.0) and DP floats (may have to use hybrid CPU correction/GPU model).

I don't think FAH uses PeakStream APIs and VM....

Last edited by Hammer on 08-Apr-2007 at 03:05 AM.
Last edited by Hammer on 08-Apr-2007 at 02:59 AM.
Last edited by Hammer on 08-Apr-2007 at 02:52 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@jiyong

Quote:

Wow, I guess the majority of the people on this forum (and outside) have been trying to convince otherwise...

In general the 360 will be easier to program - to get the best from it is going to be a different matter.

_________________
Whyzzat?

@AMiGR

Quote:

At shallow angles however the steps are very large and the illusion does not work anymore.

Well its not an illusion but a physical limit. The tendency of pixels to line up in shallow angles and step down may not exist in nature, however the step down should, if rendered correctly (which is a big if) only be in unperceivable single pixel amounts.

In nature this is the same as a perfectly perceived circle that is in fact produced not by a perfect circle but of numerous angled flat planes. Get close to nature and the smooth arc of a leaf may be the result of counter scallops across its defining edge. The eye does not see this and it is seen as a smooth arc.

The threshold is all important, the eye will smooth and blend what it cannot precisely see.

I don't know if the threshold is met with a 1680x1050 screen. I do know that outside this threshold the normal human eye is extremely good at seeing imperfections. The threshold, and 1900x1200 is designed to be just beyond, rather than on this threshold. Being a threshold, whatever its exact resolution, like water turning into ice, the final process is not incremental but relatively sudden.

I have not a doubt that what you are saying is correct. But the unknown bit is just how far away this is from the threshold of perception. The only other condition I can think that would effect this is extreme long-vision (a lens defect). 2 metres on a 20inch screen should be far enough, however, there is also a point where the effect should vanish, that might be only a short distance more (or should be 2 m is already big).

If you want to find the threshold, get a real clear jaggy and slid back slowly until it disappears and smooths out. If you are long-visioned, your eye's lens bulges and can be sharply focused at a considerable distance, you may also need to ensure that the whole centre of the screen is also in focus as concentrating on a detail effectively makes your eyes into binoculars.

Its not jaggies but the same thing can happen watching a film where some play of light or arrangement of objects forms an unusual pattern, concentrating on this and defocusing on the rest of the picture can produce a large distortion, making something look like a fault that is in fact an illusion.

_________________
Greg Schofield, Perth Australia

@GregS

Quote:

Instead they found that 1900x1200 viewed in 16:9 is sufficient to do the trick. Clearly for the hopelessly anally retentive, some sort of barrier should be supplied to stop them nosing too close to the screen.

Odd, because i even noticed the jaggies of motorstorm even from quite a distance. No offense, but maybe my eyesight is better than yours?

You just cannot compare CGI with real life film, as it is just not upto the task yet.
AA will be needed for quite some time still.

If you dont see jaggies, then there is either something bad with your monitor/tv or your eyesight.
I am not talking about standing 100 meters from the screen though, i am talking about normal viewing distances.

Last edited by Tomas on 08-Apr-2007 at 03:47 AM.
Last edited by Tomas on 08-Apr-2007 at 03:45 AM.

@Hammer

I don't know what exactly is your point with that last reply, but I think it was a bit silly by trying to prove your point with those Folding benchmarks.

As I said before, that would mean Intel CPU's really have lousy bandwidth.

The fact that other Folding calculations are possible on a GPU, doesn't necessarily mean they will also achieve the same speed, as can be seen from the PS3 FAQ.

In my opinion it's more like RC5. Altivec really knows how to handle those, but not because of the bandwidth.

@Tomas
I am not talking about my personal experience - my set won't touch 1080p and I have no software even pretending to reach this resolution.

I have no idea whether any PS3 software, or the HW is itself up to the task of true HDTV realtime rendering. What I do know, as a matter of scientific fact, that 1900x1200 means pixel sizes are below human perception (extreme long sightedness and over-focussing perhaps exceptions). That is ALL I have been saying ever.

If CGI is able to render every object 1:1 with the resolution, it should not have jaggies. It is a big if, as I know damn well most games enlarge image parts instead of rerendering them. Even if perfectly rendered for the resolution, if an object is enlarged the pixelation will increase, without AA jaggies will appear The one pixel difference will jump to 2, then 4 etc.,.

If you have a true HDTV monitor, load Linux at 1080p, turn off AA fonts and print at 1:1 the following "A line of any text in Times" at 12, 10, 9, 8, 7, 6, 5 and 4 point sizes.

7 point times should be readable (with strain) the letters should be perfectly formed, but the "serifs" should be invisible, just as they are in print version. At 6 the line should be recognisably print but unreadable, at 5 not even really recognisable, at 4 just blobs.

These are the print thresholds sizes 5 or less being deemed micro-printing, only readable when they are examined very close up or using a magnifier. If the papers written on HDTV during the development of 1900x1200 are true this should also hold for a computer screen, except no matter how close you get to the micro-printing it will not become any clearer, while 6 point will be very malformed, and 7 only resembling the letters crudely.

At 8 points the letters should be near perfectly rendered and 9 point, deemed the smallest readable print, should be crystal clear, serifs included.

AA will not, or should not, theoretically improve the font rendering.

Unless the whole theory behind HDTV is fundamentally flawed, whatever it is that causes problems lies elsewhere.

The reason I have waited so long for this technology to develop (true HDTV into a domestic device) is as a point of convergence, where computer generated content (however much of it needs to improve) is not disadvantaged to either film or print at the viewing end of things.

My real interest is in the printed word, that reading a screen is different from paper (obviously) but not a lesser form as it is now. Perception thresholds are critical in such long term convergence.

_________________
Greg Schofield, Perth Australia

@Tomas

Quote:

Odd, because i even noticed the jaggies of motorstorm even from quite a distance. No offense, but maybe my eyesight is better than yours?

Note that Motorstorm is still a 720p launch title, but still visually the most stunning complex racing game I played so far.

@GregS

It should meet the threshold, it's got the same density as its bigger brother, the 1080p 24" panel, so viewing it at the same or a larger distance should bear the same effects.

_________________

Interesting clip on mini keyboard for XBox 360 controllers:
http://image.com.com/gamespot/images/2007/news/04/05/text_screen002.jpg

I wonder if we will see a webbrowser.

@jiyong
Quote:

As I said before, that would mean Intel CPU's really have lousy bandwidth.

Wintel (CPU) scores could range from Pentium II @66Mhz FSB to Intel Core 2 @1333FSB.

With PS3’s CELL and Radeons, they are limited to certain configurations.

Quote:

The fact that other Folding calculations are possible on a GPU, doesn't necessarily mean they will also achieve the same speed, as can be seen from the PS3 FAQ.

Well, potential GFLOPS is higher for R580....

Last edited by Hammer on 08-Apr-2007 at 11:56 PM.
Last edited by Hammer on 08-Apr-2007 at 11:50 PM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

I have to comment to everyone here that I just love the fanboy comments from both camps (you know who you are), it's been a fun read (sorta like the old amazing computing Rumor Column)..

From the out of context commentaries, to the people who think they know the machine inside out and comparisons about how the PS3's cel technology is superior and the the Xbox 360 is running out of steam.

Never in a million years did I ever think I'd be asking people who are suggesting system functionality and speed, and how much better one platform is over the other if they are developers or had ever developed for either platform.

I wonder if they ever developed a game in their entire life or if they had at least worked with the XBox SDK (XNA, Managed DirectX), or even looked that the developer specifications for graphics cards from ATI and Nvidia. aka the SDK and hardware manuals published by the manufacturers.

I am not trying to be mean here but a lot of what is being said here about hardware and software on both PS3 and XBox 360 and what people think they know about HD broadcast hardware I would actually call into queston.

Mostly because this discussion is finding some comments here (well quite a few) that are not based on fact and are based on emotional investment in the platform.. I would just caution everyone here to stick to what you absolutely know to be true.