22" 9.2-Million Pixel Display

chrisd writes: "Just noticed this article over on Yahoo news. It described a research project that Intel and Stanford university developed that concentrated on next-gen displays. The result? A 22 inch display that displays 9.2 million pixels (they use the odious 'megapixel' descriptor in the article), needs 16 processors and 2 GB of ram to run it and costs $200,000US. So it's a little spendy. This is a big step up from my first 12" amber screen though, that's for sure." Ah, the march of progress ... I'm happy with anything that will help drive down the cost of 17" and 18.1" LCD displays, no matter how indirectly.

Comments

Re:High-DPI monitors need resolution-independent G

[FFFish]

"But it seems like no one (besides MS) is working on resolution-independent GUI frameworks."

Hey, buddy, you misspelled "Apple," there. OS X is thoroughly vector-based.

--

you don't need $200k to drive 9.6mp

[jonbrewer]

I have a Matrox G200MMS Quad-head that will do 5.2 megapixel over four monitors, and IIRC, it was only $800. I'm using it now to drive a pair of Samsung 770TFT panels. At $950 each for a 1280*1024 flat panel, they're a bit more reasonable than most.

I'm a bit disappointed in their manufacturing though. I have one born in Oct 2000 and another born in March 2001. Unfortunately the older one has a different white point and neither have hardware color temperature adjustment. But it was the low cost that allowed me to get away with having a 2560*1024 (soon to be 2560*2048!) flat panel desktop, so I can't complain too much. :-)

Now if you want to display a bunch of DVDs on those monitors, you're going to need a heck of a lot more bandwidth than the PCI bus can provide you. (G200MMS is PCI card) And you'll need something to decode with too. But if you didn't read the article you should know that you don't need that kind of cash to get 9.6 million pixels up on a monitor, as long as the pixels aren't doing much.

Yum.

[Mike Schiraldi]

I had the Mexapixel at Taco Bell this week.

--

So...

[nakaduct]

Where do I get 9.2 megapixel porn?

What's that you say? It doesn't exist? Well, what's this contraption good for then?

cheers,
mike

Re:Two questions

[randombit]

1) How long will it take to see I/O bandwidth improve to where it can handle real time streaming of such huge pictures from storage.

Quite a while - a gig per second is way more than even 66 Mhz 64-bit PCI can handle (I belive that will only go up to ~600 (?) Mb/s). Also, there really isn't much need for it. I mean this monitor is nice and all, but with a 20K pricetag and no real applications for it (for the average user, anyway - maybe this would be useful in some 3d modeling scenarios, perhaps), I doubt we'll be seeing technolgy to support this kind of thing in even high end workstations for at least a few years. Both Intel and AMD are working on new buses to replace PCI so maybe that problem (the bus) will go away fairly soon. Actually getting a gig per second off a disk - that could take a while.

2) Is there any hardware out there right now that can handle such high I/O bandwidths?

Surely. But once you're looking for that, you're talking about big machines with a lot of hardware RAID. I'd bet a E10K or S/390 could handle that kind of I/O. If you meant something you could actually afford - probably not. ^_^

Stanford project page

[Ryu2]

is here, known as the FLASH graphics system. I worked closely with those folks, not on FLASH itself, but on an ancillary project to visualize various parts and parameters of the computer system (bus utilization, cache latency, etc).

Very cool stuff that's just starting to get commercialized -- this is what you'll be seeing in your GeForce 4s or whatever.

It's big, but not cheaper per pixel

[Animats]

The display conference was this week, and everybody is hyping their favorite technology. Actually, I was expecting a "plastic transistor" article, that being one of the hyped technologies.

Incidentally, the reason displays don't obey Moore's Law, even though they're made using photolithography, is that making transistors smaller doesn't help displays.

As the article points out, displays are getting cheaper at about 8% per year. That's a lot better than almost any other non-IC product. Be patient.

If you're near SF, incidentally, visit the Sony Metreon, which is being used to show off Sony flat panel displays of various flavors. They're everywhere. Little ones. Big ones. LCDs. Plasma panels. No Jumbotron, though.

Re:High-DPI monitors need resolution-independent G

[Dahan]

You mean no one besides Apple? OS X's Quartz is vector-based.

Re:Megapixel

[istartedi]

It's annoying to those of us who realize that the "megapixel" number rises faster than the dimensions of the display. The dimensions are linear, and increase linearly. The megapixel number is a product, and thus increases at a rate proportional to the square. It's pure marketspeak and psychological manipulation. They figured that idiots would go "ooooohhh look at all those megapixels". Meanwhile, those of us who know better have to guess the aspect ratio, and back it out to find out what we really want to know.

The display in question, were it square, would be roughly 3033 pixels on a side because that's the square root of 9.2 million.

Now... what's the aspect ratio... umm... the article doesn't say. So... Let's say the horizontal resolution is 4096, then the vertical could be 2246. 4096*2249=9199616. Close enough for government work.... But... We just don't know. That, my fellow Slashdotter, is why "megapixel" is annoying.

"Hoarders... cannot help their neighbors" --RMS

Possible application: 35 MM

[tylerh]

At 9 megapixel, you are at 35 MM photography resolution (assuming enough color depth - betwee 16 and 24 bit, if memory serves). With this display, the need for chemical photography evaporates for all but niche applications. True, for motion, this is complete overkill. The human eye/mind would be hard pressed to absorb this many pixels at 24 frames/seconds.

Think big: put two dozen in your house, network them, and every day you could live in a different world class museum. Or have you photo album be available via voice command instead of having to get it of a dusty shelf. or be surrounded by stunning high production value or porn. or whatever.

The resesearchers reached for 9 megapixel for a definite reason: 150 years of chemical photography says this is a resolution people really like for still pictures.

Re:High-DPI monitors need resolution-independent G

[Wesley Felter]

But it's not resolution-independent. For example, all the widgets are bitmaps, so on a high-dpi monitor, they would become tiny.

Re:Why?

[homer_ca]

The answer is eyestrain. It's much easier to read 300dpi text than 72dpi text. Just like it's easier on the eyes to read a printed book than a monitor screen.

Comparison with apple 22" cinema display

[Coulson]

For comparison, Apple's 22" Cinema display, the best consumer flat-panel on the market (it's gorgeous!), has only 1600x1024 pixels, or 1.5 megapixels.... then again, the cinema display only requires 1 CPU, and costs about $2500.

One of the largest problems facing manufacturers of large-dimension LCD screens is the high rate of failure during the manufacturing process. This means that each batch of fabricated monitors yields a low number of functional units, driving up the cost per unit. I wonder how the researchers were able to combat this, while at the same time increasing the pixel density by 7x?

Can someone who's more familiar with the industry give approximate numbers on the failure rate of LCD manufacturing? Are we talking 1 bad screen in 20, 1 bad in 2000, or...?

Re:Comparison with apple 22" cinema display

[MOSSey0T0]

LCD manufacturing yields must be over 40% in order to make a profit. Volume is generally not a problem because most of the existing fabs are overbooked. The recent introduction of Taiwanese manufacturers into the Japanese dominanted LCD manufacturing industry was seen as being sucessful (read: profitable) when they reported yields of 50%. Japanese manufacturers typically produce devices at 70% yield efficiency. Since the glass substrates used to produce LCD devices are a little under a metre squared (600*700 mm), a single substrate should be able to produce 4 17" screens or 6 14" screens. It is generally accepted that LCD yields will never approach IC yields, which are typically around 90%. The upper theoretical limit for conventional LCD production is probably something like 80%.

A 1 in 20 failure rate would be a 95% yield. So, in short, we're looking at a 3 in 10 failure rate in real life with a 70% yield.


If you want to read the rest of my long-winded post, go on. If you've had enough, I suggest you go play your favorite video game.


People don't realize how complex an LCD is. The traditional example is two glass substrates surrounding a layer containing the liquid crystals, which is itself sandwiched between two polarized layers. Then there is a backlight and an active matrix of transistors used to address each pixel. Because contaminants will kill pixels in the display, this layer must be filled under ultra-clean, high vacuum conditions. The glass substrates (AND every single layer of the LCD)must be manufactured to precise planar dimensions to prevent dimensional variations across the screen. Since larger substrates approach a square metre in size, this is not an easy task. Advanced LCD technology has taken advantage of polymer coatings on each layer to separate them, prevent contamination, act as internal reflectors, etc. This adds more layers, and adds more complexity to the process.

The active matrix of transistors itself consists of a grid which is vapour deposited under ultra-high vacuum. This layer consists of at least three layers itself: anode, cathode, and at least one active layer, since active matrix LCD screens rely on field effect to twist the LCDs.

Perhaps it is not evident to the reader that high vacuum and fast manufacturing processes don't exactly mix. Even if you achieve vacuum, you are basically racing against time to complete your manufacturing/analysis before residual impurities hopelessly contaminate what you are working on.

This is why plasma and OLED displays (hopefully soon to be my research topic) are being pumped up with research dollars. The accepted theoretical limit (economically) for current LCD active matrix displays is 30", and the market is clearly looking towards very massive wall-mounted units in the future. LCD will probably dominate the POS, business and computing display markets as people become more enamoured with flat panel units, but due to sheer complexity, I think it's only a matter of time before they are eclipsed by either OLEDs or plasma.

This is long. whew. sorry, guys. pixel density was increased as a function of more efficient transistor drives for each LCD cell. It's relatively easy to pattern pixel densities that small (look at CRT monitors), but with better field effect transistors, the ghosting and cross talk present in early displays was eliminated. This was accomplished by adoption of amorphous silicon in the transistor layer as opposed to the original CdSe thin films. That in turn was enabled through advances in physical and chemical vapour deposition which allows manufacturers to pattern the transistor matrix more precisely.

That's the best I can do right now. I mainly regard LCD wrt to comparisons with OLEDs, which I am more familiar with, so I apologize for any screw-ups in here.

Superman said...

[been42]

If you can somehow trick them into saying "mexapixel" backward, they'll be sent back to their own dimension. Maybe we can grab their $200,000 display as they're fading away!

Them: "Lexipaxem...Oh, crap!"
Us (grabbing display): "See you at the pawn shop...SUCKER!"

Re:A good step, but not there yet...

[Xylantiel]

Just so you know, printers need those high resolutions so they can DITHER to get colors other than Cyan, Magenta, Yellow, Black (CMYK) or white. Monitors don't have to dither and can actually display the color value needed with coincident illuminated elements whose brightness can be adjusted. For example, on a printer if you assume that you need an area of about 6x6 =36 pixels to actually get to any color, your effective resolution is: 1200/6=200 dpi. This 6x6 value is actually dependent on the picture you're trying to display, where you are in that picture, and probably your dithering algorithm.

It's useless going past about 300 dpi on a black and white (two-color, no-grayscale) printed page as that is the limit of the human eye's resolution at about 10 inches. The printer needs that extra resolution to dither for grayscale and color. With a monitor, 10 inches is pretty close, back off to 15 inches and you only need 200 dpi. No dithering necessary so that's it, you're done, optimal display. (Actually this is generally overkill because the eye's resolution gets worse for things that aren't just black-on-white, but if the display will be used a lot to just read text like the black on white here on slashdot it's better to be safe.)

Just clearing up some misconceptions.

Re:Read that article?

[RedWizzard]

I think they're talking about a setup that would run those 16 DVDs at once costing that much. Normal use as say, a computer monitor would only require the one PC - and the cost would be a LOT cheaper... Although they don't mention the cost of the monitor itself.

They mention in the article that the display itself is made by IBM. It is probably one of the ones mentioned in this /. article last year. What Intel have done is the 16 simultaneous DVDs, which sounds like a waste of time to me. They're not using all the display either: 4x720 by 4x480 is 2880 by 1920, only 5.5 of the 9.2 million pixels available. The monitor's resolution is 3820x2400 (info here) so they should have been able to display 25 DVDs at 720x480.

From the wording, I'd say it'll be in the neighborhood of at least $2000-4000:

More than that. IBM's T210 (20.8", 2048x1536) is around $6000. I'd say this monitor would be at least $20,000-$30,000. If you can buy it at all.

3840 x 2400 (was Re:Megapixel)

[foobar104]

f4 ?0. (Or four 1920x1200s, if you prefer.)

I would say RTFA, but the FA completely failed to link to any relevant info. I had to search myself; IBM's page is here.

not really -- compare vs. print

[Preposterous Coward]

Displays are really used for two quite different things (even though they happen on the same screen): Displaying images and displaying text. For images it's arguable that current displays have sufficient pixel density for normal use, but this is definitely not the case for text.

Think about print. Early laser printers were 300-dpi resolution, which is pretty good -- good enough for entry-level, mass-market desktop publishing -- but is still quite low by traditional printing standards. The professional (print) graphics shop I used to work for used to run its phototypesetters ($100,000 behemoths) at 1,200 and 2,400 dpi. (And that's in both dimensions, so 1,200 dpi output has 16 times the information content of 300 dpi output.)

600 dpi print is pretty high-quality and acceptable to most applications, particularly with the availability of resolution enhancement, which is roughly the printer equivalent of anti-aliasing. And personally I couldn't tell any difference going past 1,200 dpi.

Anyway, my point is that even 200-300 dpi isn't as good as we might really hope for. Still, it's a vast improvement over the resolution of today's displays. I hope this will have an impact on the well-documented fact that people read (current) computer screens more slowly than printed material and find them more uncomfortable. There's actually a lot of subtle but very helpful detail contained in type that gets lost at lower resolutions such as those used by most displays today.

In a way, graphics are more forgiving than text, because the human eye-brain combination is pretty good at interpolating what's intended if there is sufficient resolution or sufficient color depth. Of course, artifacts can be pretty nasty too -- we've all seen bad cases of the jaggies. But the bottom line is, I think if you ever get a chance to actually see a display this good you won't doubt that the extra pixel density is wasted.

Possible applications

[Migelikor1]

Like with every new technology reviewed on slashdot, the real question about it is, what does this mean for Quake? Since it's resolution functions are nonexistant, wouldn't the Quake screen just shrink to about two square inches (maybe less)? In all reality, I don't really understand why such ultra-high resolution displays would ever take root in home consumer applications. Nobody has eyes good enough to utilize this invention. Maybe it comes with extremely high magnification glasses or eye gene manipulation tools! It could be useful for wearable technologies, though. The 1/4 inch display in your glasses would look a lot better with if it could display at a normal resolution. In current technology, the number of pixels aren't nearly as important as the refresh rate anyway. Unless you're a graphic pro, it's the headache-giving flicker that needs 200,000 research projects, not the number of tiny dots, and good LCDs take care of that anyway.