Wireless link between CPU and graphics adapter?

There's another way to go: remote terminal. Unix had it for a long time, and from (around) XP Windows has it too. You can put the client on a battery-powered hand-held device with a large display and a wireless ethernet and you have what you just described. Microsoft advertises this as 'smart display' I think. Take a look at ViewSonics' airdisplay at

Regards, Andras Tantos

option?

Assuming you're using an LCD screen, so slow refresh isn't an issue, you're looking at something like: 24 bits per pixel x 1024 pixels wide x 768 pixels high x 60 Hz Which is 1132462080 bits per second, or about 20 times faster than 802.11g.

option?

I think it would be pretty difficult to put enough instrumentation into a device driver to measure that traffic effectively. A PCI bus analyzer could probably easily measure the traffic level, though.

Well, you can estimate by looking at what constitutes an "acceptable" interface for a given task. For general office productivity type stuff, an X-Windows connection over 10 Mbit Ethernet is more than adequate. So that's your low-end target. You should be able to get away with even lower than that if you don't have to support the dancing paperclip :-)

On the other end of the spectrum, many 3D games already suffer from a lack of bandwidth to the display hardware, even with an AGP attachment. So, for a wireless connection, you'd probably need on the order of 10-20 Gbits/second for the best performance. I don't know of any commonly-available RF technology that will provide that kind of bandwidth, but I'm not an EE.

Of course, for the "high-end" number, almost all of that bandwidth usage is texture and geometry transfers, rather than actual drawing instructions. If you were to develop a "wireless enhanced display system", you could load the back end up with a LOT of local memory, which would lessen the number of transfers. Note that 128MB is considered "adequate" for an AGP-connected adapter these days.

-Mark

I have a difficult time envisioning this being implemented in RF, and used in commodity hardware. However if you were to discuss a free-air optical link, I'd be more optimistic. It would be inconvenient to set up and maintain beam alignment, but perhaps a single fiber optic cable could be used in difficult situations.

I honestly do not see the problem with the VGA/DVI cable; you still need at least one cable to run power to the device, so one more doesn't really hurt.

This cord *has* actually been cut for a number of wireless PC projectors (from Epson, among others) using standard IEEE 802.11b wireless LAN technology for transmission of display data (though probably not on a framebuffer basis). While there is no technical reason why this wouldn't work for ordinary TFTs as well, the primary purpose of such projectors is the display of presentations where the bandwidth problem usually isn't much of an issue, unlike with PC games and the like.

Michael

HDTV takes less than 20Mhz bandwidth, LCD panel row-column addressing (non serial raster) and current liquid crystal response of at best 20msec mean bandwidth is not a real problem nowdays. Cost and interference no doubt are...

In article , snipped-for-privacy@tele.ntnu.no mentioned...

No. The massively large amount of data would take too much bandwidth.

No, that would require too much hardware and CPU power, i.e. $$$.

Exactly.

What they should do is perfect voice recognition so that the keyboard usage can be drastically reduced.

One guy at our computer club meetings has been trying to sell his unopened package of Dragon Naturally Speaking. A few months ago he had it for sale for $150, and has been marking it down every month, until at the last meeting I think it was down to $50 and there were still no takers.

Leads one to believe that VR is still in its infancy.

In article , andras snipped-for-privacy@tantos.yahoo.com mentioned...

We had a server for that Citrix thin client stuff, and it got dropped, I can't remember why. But basically, these systems have disadvantages that outweigh the advantages. One thing that comes to mind is that if the processing is forced onto the server, then a multiuser system requires a huge and expensive server. And people just won't spend tens of thousands of dollars on a server. (Even tho it may be cheaper on a per user basis.)