Running Video Cards in Parallel

G.A. Wells writes "Ars Technica has the scoop on a new, Alienware-developed graphics subsystem called Video Array that will let users run two PCI-Express graphics cards in parallel on special motherboards. The motherboard component was apparently developed in cooperation with Intel. Now if I could only win the lottery."

Comments

In other news...

[Unloaded]

...Microsoft announced that Clippy had broken the before unheard of 2,000 fps barrier.

Re:In other news...

[david.given]

Microsoft announced that Clippy had broken the before unheard of 2,000 fps barrier.

However, they went on to say that Clippy was still intact. They're going to try again using a bigger catapult and with a concrete-reinforced barrier.

Man am I out of the loop.

[Randolpho]

PCI-Express? What happened to AGP?

Seriously, I've been out of the PC market for too long. Alas, poor wallet. I had cash flow, Horatio.

Re:Man am I out of the loop.

[Laebshade]

PCI-Express is meant to replace AGP. From what little I've read into it, it will require lower voltages than AGP and has a wider bus.

Re:Man am I out of the loop.

[Randolpho]

After I posted that, I did a quick google:

http://www.pcworld.com/news/article/0,aid,94724,00 .asp

Re:Man am I out of the loop.

[Plutor]

You've been out of the PC market for about a decade then, if you've never heard of PCI-Express. It's been proposed and talked about and raved about for years, but it's just now finally coming to market. The best thing is that it's not limited to a single slot per board! That's why this parallel thing is even possible.

Re:Man am I out of the loop.

[Anonymous Coward]

You can have several AGP slots on board. It's just that none of the manufacturers were interested in making such motherboards.

Re:Man am I out of the loop.

[dabug911]

PCI Express is like gigbits faster then AGP is, so it a good thing to have video cards moving in that direction.

Re:Man am I out of the loop.

[Auntie Virus]

There's a White Paper on PCI Express from Dell: Here

Re:Man am I out of the loop.

[The_K4]

Since the PCI-Express spec defines switches (these are like P2P-bridges only they have 2 sub-buses) a mother board manufacturer could add 2 of 3 of these and get 4 PCI-Express Graphics ports (or 7 and get 8 ports) the problem is that every time you do this you have to share the total bandwidth at the highest level. Since PCI-Express does have more bandwidth the AGP 8x and 1/2 of that bandwidth is dedicated up the other 1/2 dedicated down. So the down-stream (where video cards use most of their bandwidth) is greater the AGP8x's TOTAL bandwidth. So this data path bottle next shouldn't be bad if you have 2 cards (might work well for 4 if they use the bus right).

Re:Man am I out of the loop.

[zelphior]

Are there any motherboards out on the market that have dual agp slots? I did a little bit of googling, and couldn't find any, and some sites state that no one makes them.

Re:Man am I out of the loop.

[Xaroth]

I love that the article you mention indicates that 2.5Gib/s is "blazing", but regular PCI's 128MiB/s is "paltry". Especially when the new stuff is only ~2.5x the speed, if the numbers are to be believed.

Re:Man am I out of the loop.

[UID1000000]

way out of the loop. are you still running a 80386 arch?

yeah PCI-express has been the buzz word with Dell and HP for their new PCs. it looks promising.

another note, does anyone know the bandwidth of PC2100 DDR ram? or even PC2800 DDR ram? i'm wondering what the max bw is on DDR RAM vs the bw of the new PCI-express.

Re:Man am I out of the loop.

[canajin56]

Not only that, but PCI is bidirectional, while PCI Express is not. A PCI express pair of wires transmits at 1.25Gib/s, so PCI express is barely faster than PCI at all for applications using mostly unidirectional stuff. Graphics cards, for example, are rarely used to send data back to the CPU. The difference is if, instead of 4 pins for a 1X PCI express connection, with 2.5Gib/s (1.25 each way), you have say, 64 pins for a 16X PCI Express connection, with 40Gib/s (20 each way).

Also note that the 1Gib/s figure for PCI is actually for PCI-X v1.0. Normal PCI like you would find in a desktop is much slower.

The plan is that a desktop motherboard would have a 16X PCI Express slot for the graphics card, some other slots, possibly from 1X-4X, as well as legacy PCI slots. A server, on the other hand, would likely have a bunch of 8X+ slots. (Because 8X is the first time when a PCI express slot is faster than a PCI-X V2.0 QDR slot)

But yes, it is funny to see them talking about "Blazing" avs. "Paltry" for figures that are actually pretty close ;)

Re:Man am I out of the loop.

[Geoff-with-a-G]

You've been out of the PC market for about a decade then, if you've never heard of PCI-Express. It's been proposed and talked about and raved about for years, but it's just now finally coming to market.

He must have been out of the market for a decade, to have never heard of something which is only just now in the market? What?

I've been half way out of the market for about five years, and I only recently heard of PCI-Express, and I didn't have many details about it. Researching new, not yet marketed technologies isn't the same as "being in the market".

Re:Man am I out of the loop.

[Phalse Impressions]

Actually I was just at one of Intel's channel conferences and they surprised me a little.

It seems that the new PCI-E X16 port will support up to 75 watts of power. I'm pretty sure this exceeds current use but feel free to correct me on this.

Re:Man am I out of the loop.

[WuphonsReach]

You've been out of the PC market for about a decade then, if you've never heard of PCI-Express.

That's rather over-stating the case.

Roughly 10 years ago, PCI was finally just supplanting EISA/VESA and ISA boards were still common.

I build a few machines per year, and PCI-Express only just hit my radar screen in the past 12-18 months. Even today, I have yet to see mainstream motherboards or cards for it, so it's still rather ephemeral at this point.

It is an interesting design. Whether or not it will live up to it's promise remains to be seen.

Re:Man am I out of the loop.

[f0rt0r]

Yes, but AGP's bus is longer. All depends on which one the girl prefer's. What were we talking about again?

Re:Man am I out of the loop.

[Frenchy_2001]

Yes, the PCI-E 16x will support more power: ~75W.

No, it is NOT enough for the latest generation graphic cards, as they can use up to ~120W.

Nvidia, for example, has planned to put "only" one molex connector on their GF6 PCI-E cards, instead of 2 on the AGP one...

Re:Man am I out of the loop.

[Frenchy_2001]

another note, does anyone know the bandwidth of PC2100 DDR ram? or even PC2800 DDR ram? i'm wondering what the max bw is on DDR RAM vs the bw of the new PCI-express.

Is that a joke or a trick question? You're answering your own question here...

DDR2100 (or DDR266) has... 2100MBi/s of maximum theoritical bandwidth (64bits * 133MHz * 2 exchange/clock). If you use a dual channel chipset, you double it (2* 64bits bus).

Same for the rest of the memory. The name *IS* the maximum single channel theoritical bandwidth.

PCI-E is 1.25Gib/s per channel. So, a 16x PCI-E port (used for graphics) is 20Gib/s or 2500MiB/s.

You'll need at least DDR2700 memory (DDR333) to keep it fed. As the chipset including PCI-E will *start* with DDR3200 dual channel DDR1 and go up, that shouldn't be a problem in itself...

Re:Man am I out of the loop.

[NoMercy]

It's not quite as clear cut as 'wider bus' since PCIe is actually a lot narrower as it's a serial bus as oposed to the classic parrell PCI and AGP busses, the version for Graphics cards is 16x PCIe, which is essentially 16 serial busses wired into the graphics card, which gives a preformance somewhere between mind boggling and insane.

Realistically, at twice the speed of AGP 8x, which apart from loading textures onto the graphics card memory, that bandwitdh isn't even touched in modern systems most of the time the cards are running, so it should last us a long time to come.

Re:Man am I out of the loop.

[UID1000000]

I was actually thinking back to the day where we had PCI memory bays. I was wondering if you could get quicker access to your RAM through a PCIe bus than reg arch.

I was obviously hit by the stupid tree before I wrote that. The sys arch would most likely always have better access time b/c it's a direct access to the processor. If it were through the PCI bus it would go through PCI then to the proc thus there would be latency.

There I go answering my own question again.

Quad-screen?

[Vrallis]

Hell, I couldn't care less about parallel processing for the video cards.

I want tri-head or quad-head video, but with at least AGP speeds. You can do it now, but only with PCI cards getting involved.

Re:Quad-screen?

[houghi]

I want tri-head or quad-head video, but with at least AGP speeds

So order one now. They are available here at Matrox.

Re:Quad-screen?

[gr8_phk]

I'll second that. Flight simulation begs for 3 screens, as do some driving and other games.

On another note, I suspect the only way it will really accelerate single images is in cases where render-to-texture is used. i.e. per-frame generation of shadow or environment maps. The completed maps could then be passed to the card that actually has the active frame buffer to be used in regular rendering. Two cards could at BEST double performance and nothing ever scales optimally.

Re:Quad-screen?

[RhettLivingston]

As long as you can live with at least one of the screens not running quite as fast (maybe an informational type of screen as opposed to 3D scenery?), 3 screens ia really easy today. Almost all decent AGP cards these days support 2 screens at 1600x1200. Throw in a good PCI card and you've got 3. I've been running this way for years and it works well. Actually, the PCI card isn't shabby.

The only problem I encounter in Windows is an occasional tooltip coming up on the primary monitor instead of a secondary monitor. This is not the fault of the OS, rather the application is constraining the tooltip to be on the primary monitor by forcing it to be within the primary monitor's coordinates.

Note that Matrox's single board AGP solution does not compete with this. Using a high end NVidia for the main two screens provides too much of a performance advantage to give up for Matrox's slow cards. Matrox's cards, even though on AGP, run about like the PCI cards.

Regardless, when these systems become more available, I will be one of the first to put 2 video cards in and run 3 or 4 screens from my PCI Express system. But, though I like playing 3D games this way, I do it for the extra informational surface for programming. It greatly eases things to run your application on one screen and your development environment on all of the others so that you can see everything at once. And with 19" 1920x1440 monitors (which usually manage 1600x1200 with better focus than a 1600x1200) running around $250 a pop, its a very worthwhile investment.

Re:Quad-screen?

[Lumpy]

hows it feel to want? espicalyl about something that has been available for a decade now.

here

Matrox, because all the other cards are merely toys for the kids at home.

Re:Quad-screen?

[RTMFD]

Have you thought about how many trees you're killing or the C02 you're pumping into the atmosphere just to make your programming job easier? You capitalists never think about the feelings of the dolphins when you claim such an extravagance is needed to help you get your job done more efficiently. Next time, be Earth-Friendly and write all of your code on mud tablets first.

</tounge-in-cheek>

Re:Quad-screen?

[wisdom_brewing]

pci EXPRESS... not pci... there is a difference you know... :)

Re:Quad-screen?

[Vrallis]

General uses:

NOC situations (monitoring)
Graphics work (GIMP, Maya, etc)
Coding (reference materials on one screen, IDE/terminals on another, debugging/output on another)

While I do game some, nothing I play has any worthy multi-head capabilities, so this is meaningless for gaming, at least for me.

Re:Quad-screen?

[Vrallis]

I do know about what Matrox has. I used Matrox dual-head for years (G450, I believe), and it sucked--bad--under Linux, using their proprietary drivers or not. I eventually convinced work to let me get a low-end nVidia dual-head, and things greatly improved.

While I do game at home, nothing I play can use dual-head, other than stretching--which is unplayable. Ever tried playing with your crosshair split in half by 4 inches?

I probably will end up doing what others have suggested--just mix a PCI card in with the AGP. I don't have to have 3D acceleration on all the screens--just one (at home at least, none at work).

Re:Quad-screen?

[Rosonowski]

As other posters have pointed out, these cards are horrible at 3d rendering.

Re:Quad-screen?

[GeckoX]

Sucked for years, ok, but compared to what?
Have a better option did you?

If it weren't for Matrox, irregardless of your own issues with drivers on your chosen platform, do you think there would be _any_ multi-monitor cards today? And even if there were, do you think they'd be up to the quality that they are now?

Come on, at least give credit where it is due.

Because of Matrox, you can now get a multi-monitor nVidoa or ATI card that is _better_ than a matrox card for your needs.

Next comes dual AGP graphics.

[Trigun]

dual Nvidia 5950's under Linux!
Just in time for DNF!

Re:Next comes dual AGP graphics.

[DaHat]

Nay, the AGP standard is built around a single slot and a single graphics card. To permit two AGP cards running natively (via the AGP bus) in a single system would be quite difficult if not impossible, far easier to look to the future and a new technology to make it work better then any sort of hack job that could be done today.

Re:Next comes dual AGP graphics.

[Trigun]

Yes, I know. I was going for the Trifecta of things not going to happen.

Although, I've never tried the dual 5950s. I guess we can replace that line with subtle, abstract humor probably grossly misplaced, is understood on Slashdot.

Re:Next comes dual AGP graphics.

[Jeremy Erwin]

It's in the AGP 3.0 spec.

AGP3.0 allows a core-logic implementation to provide multiple AGP3.0 Ports. Each AGP3.0 Port is a bridge device with multiple AGP3.0 devices hanging off the secondary bus. Each Port has a separate Graphics AGP aperture and GART that is independent and not shared with another AGP3.0 Port; however, these are shared across the devices within a single AGP3.0 Port.

Re:Next comes dual AGP graphics.

[Martin Blank]

You and me, both. But I salvaged it with a bit of recycled Slashdot humor, so the original responder seemed satisfied.

Anyway, it will be nice to be able to get back to the old methods of multiple video cards, even if only the highest-end systems will be able to handle them. I really don't see this happening in too many systems based simply on the power draw. nVidia seems to think they can knock the power requirements on the new cards back to needing only a single rail, which is good, but even then the draw is going to be enormous. A 3.06GHz P4 pulls 100W (I don't have numbers for higher, but they're not likely to be any lower), and assuming that nVidia's card can get down to the level of the X800 line (~95W), for a dual-card solution a system will need 300W just to power those, not factoring in any other peripherals. Using an efficiency of roughly 70%, that's a 425W power supply just for those. Throw in a couple of high-performance hard drives, the motherboard, sound card, NIC, and dual opticals, and you're going to be squeezing every last bit of power out of a 550W power supply, if not requiring more.

Re:Next comes dual AGP graphics.

[kabocox]

To permit two AGP cards running natively (via the AGP bus) in a single system would be quite difficult if not impossible, far easier to look to the future and a new technology to make it work better then any sort of hack job that could be done today.

But this is /. That is what would really make /. news. Near impossible hacks on old hardware that are far easier on newer hardward that's designed for the task is the bread and butter of /.

Press Release

[Anonymous Coward]

over here: clicky

Voodoo

[Eu4ria]

Didnt the early voodoo cards allow something similar to this ? I know they had a pass through from your 'normal' video card but i seem to remember the ability of running more and they would each do alternating scan lines.

Re:Voodoo

[scum-e-bag]

The company was 3dFx, and it was thier Voodoo II cards that allowed the use of two cards a few years back, sometime around 1998 IIRC.

Re:Voodoo

[Trigun]

Yes, they did. Unfortunately, at the time they were too expensive and took up all of your extra slots on your mobo. Now, with integrated everything, it's not so bad.

Good idea implemented too early. Such is life.

Re:Voodoo

[naoiseo]

without a special motherboard, yes.

I think you could string something like 4 voodoo rush cards together or something (who knows if you got 4x performance, but I'm sure it went up not down)

Problem was, by the time they put this out there, the tech it was running was months behind cutting edge. 4x something old is easily forgotten.

Re:Voodoo

[UnderScan]

SLI - scan line interleve, was available for 3dfx Voodoo IIs (maybe even Voodoo 1) where the first card would process all the odd lines & the second card would process all the even lines.

Re:Voodoo

[kamelkev]

Voodoo was basically the beginning of the performance PC market, with tons of wierd options and card types.

Benchmarks for the old 3dfx V2 SLI can be seen here:

http://www4.tomshardware.com/graphic/19980204/

I was (and still am, although its in the junk pile) a 3dfx V2 owner, the performance of that card was just amazing at the time. The Voodoo and the Voodoo2 definitely changed the world of 3d gaming.

Also of interest is an API that came out much later for the 3dfx chipsets that actually let you use your 3dfx chipset (they didn't call it a GPU back in the day) as another system processor. If you were an efficient coder you could actually offload geometric and linear calculations to the card for things other than rendering. I can't seem to find the link for that though, it may be gone forever.

Re:Voodoo

[Halthar]

If I recall correctly, there were Voodoo1 SLI type systems, but they were primarily used in arcades, and were produced by Quantum3D. They may not have been actual SLI systems, but they did use multiple 3Dfx chips, at any rate.

Quantum3D also produced the Obsidian3D cards, which were single slot Voodoo2 SLI cards, unfortunately because of the weight of the cards, they had a problem with bending unless you had a way to brace the end farthest from the PCI slot inside your case. Very long cards, very heavy cards

They seem to have been built like tanks though. I still have one at home that I keep around for playing some of the GLIDE only games that came out when 3Dfx was doing well.

While not SLI related, there was a card, I think the Voodoo6 6000, that never got released (at least to the best of my knowledge), it was basically a quad voodoo5, I think. It looked somewhat similar to the Nvidia 6800, including needing two slots worth of space on the MoBo. It also had it's own external power supply that would connect to the back of the card next to the VGA connector.

It makes me wonder how many of those 3Dfx people are still plodding away at Nvidia after the buyout. The GeForceFX cards and the GeForce 6800 cards really do look alot like that beast that 3Dfx was trying to release before they got purchased by Nvidia.

Re:Voodoo

[alex_tibbles]

Well I'm still running my twin Voodoo2s SLIed. Doesn't run any games more recent than Deux Ex (1), but Rollcage runs smoothly. When Doom3 comes out then it'll be retired...

Re:Voodoo

[Rosonowski]

From what I understand, too, that card performed on the same level as the Geforce 3 ti500, which didn't come out for a whole generation of nVidia tech. (presuming good drivers)

Re:Voodoo

[minister of funk]

The VooDoo I chipset was also able to support SLI. I believe there was one company... Orange? that built a solution for around $800.

Light on Info

[the morgawr]

The PR mess is light on information and I don't have flash to view their site. Can someone give some technical information? e.g. How does this work? What does it really do? What can a typical gamer actually expect (surely it doesn't just double your power by sending every other frame to each card)?

Re:Light on Info

[Caesar]

http://arstechnica.com/news/posts/1084398037.html

It splits the screen in half. Alienware claims a ~50% boost.

Re:what will this do?

[m4vrick]

Now you can play solitary and minesweaper at the same time on a maximized screen. :)

interesting technology

[cheese_wallet]

I think it is great that a company has the will to do something like this, even if it doesn't catch on. It's cool to try something new, instead of just hanging back and doing the tried and true.

I'll admit I haven't yet read the whole article, but even though it says that it isn't tied to any one video card, that doesn't say to me that it can have multiple disparate cards. If it is doing something along the lines of SLI, I would guess that the speeds would need to be matched between the two cards. And that would imply having two of the same card, whatever card the user chooses.

But maybe not... maybe it's the advent of asymetric multi video processing.

Re:interesting technology

[jonsmirl]

You can do this today with Chromium.

Chromium replaces your OpenGL library with one that farms the OpenGL drawing out to multiple machines. It's how display walls are built.

You can use the same technique for multiple card in the same box.

Re:interesting technology

[cheese_wallet]

"You can use the same technique for multiple card in the same box."

Yes, but wouldn't you still need extra hardware to merge the display of these multiple cards to a single monitor?

Re:interesting technology

[Soul-Burn666]

How about this:
Each card will recieve a number of scanlines to process according to it's strength, therefore making the rendering speed similar, and after that syncing the speed to the lower one.

There are still problems with different features that might or might not be available on one of the cards such as pixel shaders. Also anti-aliasing can be weird.

Re:interesting technology

[cheese_wallet]

"How about this:
Each card will recieve a number of scanlines to process according to it's strength"

It might be polygonal based too. If they added a third frame buffer that was fed output from the other two cards, and then rastered on the screen, they would have a lot of flexibility in how they comingled the cards. Latency would be a big issue though.

Re:interesting technology

[SeanAhern]

Not really. Chromium contains a binary-swap compositor that allows you to depth-merge the output of several cards. But that's not what a display wall (tiled display) is about.

This is the distinction between sort-first and sort-last rendering. Chromium can do either and both.

Re:interesting technology

[SeanAhern]

I wasn't clear in my post. I'll attempt to be more clear here.

Running a display wall (which is what this was all about) doesn't require any "merging". It requires partitioning the scene before rasterization into a set of bins based on projected screen space. Each bin of geometry is sent to a separate graphics card for rendering. And that's about it. That's the essence of sort-first rendering.

If you do want to merge the output of multiple graphics cards to a single video stream, this is a different class of problem. This requires readback from the framebuffer (which not only stalls the pipe but is also very slow with current AGP cards), some type of composite (depth comparison is very popular), and a final transfer to a frame buffer. This is the essence of sort-last rendering.

The problems both involve parallel rendering. But they are very different problems that each have their own set of challenges and constraints.

I think that's the sound of a pin dropping

[Phidoux]

So they run sound cards in parallel too?

Re:I think that's the sound of a pin dropping

[Phidoux]

The moderator obviously didn't RTFA!

Re:More human-like

[kernelfoobar]

On a more serious note: wouldn't this be great to compare GFX cards and settle the 'nVidia Vs. ATI' wars?

this isn't new

[f13nd]

Alienware didn't invent this
the PCI and PCI Express have had this written into spec
AGP does too, but when was the last time you saw dual AGP slots on a mobo? (they do exist)

Re:this isn't new

[241comp]

Do any non-Mac dual AGP motherboards exist? If so, could you list some or all of them so that I can do some research? Thanks!

Re:this isn't new

[SlamMan]

No Apple motherboards do dual-AGP either.

I can dream...

Re:this isn't new

[BenBenBen]

The AGP port spec lays it out; AGP is a preferred slot on the PCI bus, with four main enhancements (pipeline depth etc) designed to... Accelerate Graphics. Therefore, if you had more than one PCI bus, you could technically have more than one AGP port. However, I cannot find a single motherboard that offers 2 AGP slots, including looking in numerous AV/editing specialists, where I'd expect this osrt of thing to tip up.

Re:this isn't new

[LoudMusic]

AGP does too, but when was the last time you saw dual AGP slots on a mobo? (they do exist)

Would you mind to enlighten us as to where we might find such a board?

Re:this isn't new

[bairy]

Not exactly something you can get from PC World, but this one can have 2 AGP, or there's this one which offers 4 AGP

Re:this isn't new

[Caesar]

PCI-Express does support this, as did AGP. However, those technologies do not include the ability to split the same screen in half and load balance the performance. That is what Alienware is contributing. That is "new," in the sense that this methodology has not been done before (SLI was alternating scan lines).

Re:this isn't new

[marmite]

I think that the Intergraph ZX10 had 2 AGP slots on. Sadly SGI bought the product line and destroyed it.

--ralpht

Big Deal - PCI Express. Any one can add two video

[liquidzero4]

So what technology did Alienware create here? None..

So they have one of the first MB's with two PCI Express slots. Big deal, soon MB's will contain many PCI-Express slots. Hopefully a lot more than 2.

Oh, come on!

[Short Circuit]

All you really need is some way to copy the data in memory from one card to another.

Easy solution? Several high-speed serial connections in parallel between the two cards. With a little bit of circuitry on the card dedicated to keeping the data identical.

Or, with a little bit of a performance hit, you could keep each section of RAM separate, and route misses over the cables.

Re:Oh, come on!

[sindarin2001]

You only seem to be addressing one of the issues of parallelism. There's also the problem of access lockout. If one card is manipulating said memory, how do you address the issue if the other card needs access to it, and then do you allow it to it before or after the other card starts manipulating it? These are questions that have been plagueing parallel structures of all kinds for years, be it parallel processing, multiple threads, or multiple video cards.

one of two things

[Darthmalt]

The way I'm reading it either it makes cards from 2 diff manufactuers work together to display video from one card twice as fast ie beowulf clustering.
Or it make to diff cards support dual screen? But I've been doing that on my computer for years so that couldn't be it.

Nice, A complete Vapor-article.

[Gr8Apes]

From the article: "The answers may have to wait until Q3/Q4". There are no performance numbers, no real statements of how it works, nothing much at all. Just wow, gee whiz, dual graphics cards in parallel. What exactly does "in parallel" mean? That's not even addressed.

Some things I thought of immediately reading this, great - two displays each driven by a separate card, or, better yet, quad displays driven by two cards. Nope, not a word about either possibility. The implication of the PR/article is that 3D graphics will be processed faster. How? Do they have some nifty way of combining two standard off the shelf graphics card signals into a single monitor? (Hint, it's hard enough getting the monitor to properly synch up with a single high performance graphics card!)

Since when does ArsTechnica merely regurgitate PRs? This was 99.999% vacuum.

Re:Nice, A complete Vapor-article.

[ProfBooty]

a couple years back you could link together two voodoo2 cars together to run games at a higher resolution. It did require a special cable as I recal.

Perhaps something like that?

Re:Nice, A complete Vapor-article.

[archen]

Some things I thought of immediately reading this, great - two displays each driven by a separate card

Am I missing something here, because I just set up two video cards up on my machine the other day. Now it's relativly easy to set up a dual display on a dual headed card, but using two cards takes a bit more work, especially if you have two completely different monitors that run different resolutions and refresh rates. Now I use an AGP and PCI card, but I imagine it's the same difference with PCI Express.

So I'm assuming that they're talking about using the power of another GPU to help processing, because there isn't anything really new about using two graphics cards.

Re:Nice, A complete Vapor-article.

[rokzy]

>What exactly does "in parallel" mean?

well this is Alienware so I think they mean that one card is on top of the other and both are perpendicular to the MB, hence they are running "in parallel" with each other.

Re:Nice, A complete Vapor-article.

[Gr8Apes]

Using two cards for dual or quad displays (i.e., showing an expanded scene across 2 or 4 monitors for games and such)

Or, using two random graphics cards to process signals for a single monitor. While voodoo did this, it required specialized hardware/software with the voodoo cards to synch everything up for the single monitor.

Re:Nice, A complete Vapor-article.

[Gr8Apes]

Yep, required specialized hardware (the voodoo cards themselves) and a cable between them to allow them to work as one. Certainly not a generic plug any 2 cards into two slots and have them magically work together. That's what this PR implies.

Re:Nice, A complete Vapor-article.

[Caesar]

E3 is on this week. E3 is dominated by product announcements, including products that won't see the light of day for years. So, the "vapor" aspect of it is what pretty much what happens at most of the trade shows. By that measure, half of the news out this week is "vapor." Sure, some people might blow smoke and tell you about the performance stats, but we all know that when the products actually ship, their "performance stats" will probably have changed as well.

So, you have the press release to go on. And as you noted, I threw up a flag of caution. I didn't call "vapor" on it, however, because as I noted, they are saying Q3 and Q4 for VA and X2, which seems awfully soon to be pimping vapor. I also happen to know that this is in fact happening, so I didn't feel obliged to cast aspersions on it.

From the PR, it's rather clear to me that this is a single-monitor solution. It's also likely the case that this is not just a PCI-Express matter. Call me old fashioned, but Alienware says that they have patent-pending technology, and while I would usually take the language for marketing drivel, I have no reason not to trust Alienware on this matter. In fact, they have tried this before, but the underlying technology just wasn't doable.

Re:Nice, A complete Vapor-article.

[SleazyC]

Well according to Gamespot's article on the sytem, it looks like performance will increase.

Reminiscent of 3dfx's Voodoo2 scan-line interleave (SLI) technology, Alienware's "Video Array" technology allows two video cards to work together while running a game. Unlike 3dfx's SLI, where the two video cards shared the workload by drawing alternating lines on the screen, the video cards in Alienware's Video Array split the screen in two with one card working on the top half and the second card working on the bottom. Load balancing software ensures that the workload is evenly split between the two cards in cases where graphics complexity in one half of the screen is greater than the other. Alienware isn't ready to release benchmark numbers since the hardware is still in alpha, but the company expects a "50+% performance gain over traditional graphics solutions" when the system is ready for release.

Looks like this system could be pretty killer IF the price isn't insanely high that is.

Re:Nice, A complete Vapor-article.

[Gr8Apes]

The price is going to be insanely high... as compared to just getting a better graphics card in the first place (if you're trying to do this with "cheaper" cards, as you'll have to buy the system and the dual graphics cards). Of course, if you're buying top end cards, then it becomes a different story, but still very high. Do you want to spend $1200+ just for your graphics subsystem? (Taking 2 $500 cards and a $200 MB - a pure guestimate, as I imagine it'll be more)

Finally, the article is pure vapor, as it mentions absolutely nothing factual other than hey, you can plug in any two graphics cards and, what? It makes marketing statements without substance. Which was my true point about the article and the post.

In response to your timing, MS (the favorite whipping boy of vapor stuff) announced Cairo when OS/2 2.1 came out as coming out in 6 months, then 12 months, then 2 years... and gee, we're still waiting, but it's now been renamed Longhorn, and it'll be out in... 2 years... yippee. And you'll only need hardware that won't be out for 4 years to run it.(ok, ok, maybe the hardware might be out in 2006, but it won't be any sub $1K system)

Re:Nice, A complete Vapor-article.

[Gr8Apes]

Please go ahead and read this for my view of why this is complete vapor.

Unless the new PCI Express cards have additional default functionality built into the bus that allows some kind of buffering and access to card hardware that PCI/AGP have never allowed, this is pure BS. I have not heard that ATI and NVidia are working on products that have open APIs to hardware which allow their products to interoperate. Heck, I haven't heard that either one of them even have anything like true dual paralleled cards in the works for their own lines! (Note: this would be like the dualism of Voodoo, 2 cards simultaneously rendering on 1 monitor, not two cards in 1 system.)

Re:Nice, A complete Vapor-article.

[Caesar]

Well, people have seen it running at E3. How they split the screen is not yet clear, but the key seems to be in the "software load balancer" that they are using. To me, this implies that the drivers don't see data to pass on to the cards to process before the "software" component. So, in that sense, the cards could be ignorant.

Are we going to need this...

[CodeMonkey4Hire]

for Longhorn?

Intel's Chipset only supports One x16 PCIe

[hattig]

In fact all the first generation PCI-Express chipsets only support one x16 PCIe for graphics controller.

I doubt that Intel is going to make a 2 port one especially for Alienware.

So I expect it means that the second graphics card is plugged into a x4 or x1 PCIe connector.

Anyway, this is nothing special, it is all part of the specification. Hell, you could have two AGP v3 slots in a machine working at the same time - how do you think ATI's integrated graphics can work at the same time as an inserted AGP card's?

Re:Intel's Chipset only supports One x16 PCIe

[DaHat]

Are you quite sure that ATI's integrated graphics are AGP based and not on the PCI bus? Last I checked, there was not a publicly available chipset that could handle more then one AGP video device in a system at a time. If it were possible, I'd throw out my nVidia card today and worship at the alter of ATI... I fear though that you must be mistaken.

Re:Intel's Chipset only supports One x16 PCIe

[strictnein]

In fact all the first generation PCI-Express chipsets only support one x16 PCIe for graphics controller.

Have you seen anything talking about second-generation chipsets that support two 16x PCI-express connectors?

This is what I want and I'm not getting a new computer until it happens.

Re:Intel's Chipset only supports One x16 PCIe

[hattig]

I'm talking about ATI's integrated graphics chipsets, the ones where you can run the integrated graphics (with AGP functionality) at the same time as a graphics card in an AGP slot. Now whilst ATI could have done some clever stuff with a single AGP controller, it would make more sense to simply make use of the AGP v3 functionality that allows more than one AGP device in a system.

But ... sensible ideas are often eschewed in favour of some hair brained system that kinda worked all too often...

Re:Intel's Chipset only supports One x16 PCIe

[hattig]

I haven't seen anything myself from my all-too-regular trawls around hardware sites. I've seen x16 + x4 + x1 however (x4 is 2GB/s total bandwidth). However I'm sure that it is only a matter of time until it does happen.

Re:Intel's Chipset only supports One x16 PCIe

[TeknoHog]

I thought PCI Express is a bus, like the current PCI; even with one slot on the motherboard, you could connect multiple devices in parallel into it.

If it's not a bus but a port, I don't see how it's radically better than AGP.

Re:Intel's Chipset only supports One x16 PCIe

[Short Circuit]

Anyone with that hardware could tell us.

Under Linux, run "lspci" as root, and see if the two cards are on different PCI buses.

You can do something similar under Windows XP:

Go to the device manager, and look at the Location field of your two video devices. The box I'm on only has one, but here's what an AGP card's location field looks like: "PCI bus 1, device 0, function 0"

Re:Intel's Chipset only supports One x16 PCIe

[joib]

My understanding is that it's switch based. I guess the original poster of this particular subthread meant that Intel has saved a couple of cents by making only one port on the PCI Express switch 16x, the others are apparently slower.

Re:Intel's Chipset only supports One x16 PCIe

[IceStorm]

The chipset used as per Homelan's interview is Tumwater.

Tumwater, far as I knew, was only to have a x16 and a x8 PCIe slot off the MCH. Apparently, this is incorrect and it has twin x16 PCIe slot support. While Tumwater is targetted for Xeon workstations, there's nothting that I'm aware of which would block its usage in a P4 system. The P4 and Xeon CPU bus protocols are the same.

The extra that Alienware is adding is their software layer which allows processing of video data by two separate video cards. Metabyte (I think they were formerly Wicked3D), which Erwos mentioned elsehwere in this discussion, had this sort of technology up and running, but far as I know it never shipped.

This isn't just multiple display card support, this is also a software layer allowing load-balancing such that the two work together to render a whole screen. It's entirely possible this could end up being multiple-display-capable if the underlying drivers for each card allow combining of displays, similar to the way Matrox allows their multi-monitor cards (G4xx, G550, P-series, Parhelia) to be presented to the OS as "one" display.

Hopefully other companies will provide something similar. I'd rather have an ATI-only setup that is compatible with any dual x16 PCIe-supporting Tumwater board than be forced to buy Alienware's ugly case. :-)

Everything old is new again?

[Fortunato_NC]

When Windows 98 came out, there was a new feature (that before had pretty much been limited to Matrox cards with a special driver) that would let you use multiple PCI and AGP video cards in the same motherboard with multiple monitors. At first glance, this seems like pretty much the same idea.

The article seems to claim that the cards will be able to split processing duties, even if they're not from the same manufacturer. That particular claim seems very dubious to me for some reason. Other than integrating two PCI-Express slots on a motherboard, I'm not sure Alienware has achieved anything here. Of course, should Alienware want to send me one of these to try out, I'll be happy to post my review on Slashdot.

Re:Everything old is new again?

[skiflyer]

I'm sure this will be said a million times during this thread, but this has nothing to do with multiple heads... it has to do with multiple cards serving 1 head.

Re:Everything old is new again?

[frenetic3]

There's gotta be a sick joke in there somewhere. :P

-fren

Re:Everything old is new again?

[LqqkOut]

Yep, a new way to do an old thing... Patent pending!

Alienware did this before.

[thecombatwombat]

I'm pretty sure it was alienware, when the voodoo 3 was really, really new I saw a little blurb in what I think was a computer gaming world about a system that would run up to 4 voodoo 3 cards. It didn't use SLI like the old voodoo 2s, but would split the screen up among the cards, a 2x2 grid with 4 cards, the blurb went on about crazy quake 2 framerates I think.

Re:Big Deal - PCI Express. Any one can add two vid

[jcostantino]

Like this?

Re:Is that new?

[lucas teh geek]

I belive you need to RTFA, this isnt about dual head setups. one head, two cards

Metabyte PGC

[Erwos]

It looks like the same thing as Metabyte PGC - and Alienware was supposed to be the roll-out partner for that.

Nothing wrong with it, though - PGC actually did work, and was previewed independently by several people (I think Sharky?).

-Erwos

Re:Metabyte PGC

[MrMonty]

I knew I remembered it. Found an article on it though, circa 1999.

http://www.cdmag.com/articles/018/039/metabyte_p cg .html

Re:I don't think the author really got it there...

[Enrique1218]

Maybe they don't have to cooperate. Graphics card gennerally support the same standard (vga/directx/opengl). Perhaps, the video array will have its own driver/software component to receive the game data then parcel the data to each card.

Re:Big Deal - PCI Express. Any one can add two vid

[strictnein]

Can anyone show me an upcoming motherboard with two 16x PCI Express slots? From what I've read most (all? except for this one) will only have one 16x PCI express slot for the video card and then a couple 1x, 2x, and 4x slots PCI-Express slots for everything else. The reason I wanted to upgrade to a new computer this winter is in the hope that I could have two high-end video cards in the same system (instead of having one high-end agp and one mid-range PCI).

But do you need multiple monitors?

[MtViewGuy]

I think the big question we need to ask is do we really need multiple monitor setups?

Besides the obvious issue of hardware cost of multiple graphics cards and multiple monitors, you also have to consider desktop space issues. Even with today's flat-panel LCD's, two monitors will hog a lot of desktop space, something that might not be desirable in many cases.

I think there is a far better case for a single widescreen display instead of multiple displays. Besides having a lot less impact on hogging desktop space widescreen displays allow you to see videos in the original aspect ratio more clearly and also allow for things like seeing more of a spreadsheet, clearer preview of work you do with a desktop publishing program and (in the case of a pivotable display) make the reading of web pages easier and/or single page work with a DTP program easier. Is it small wonder why people so much liked the Apple Cinema Display that uses a 1.85 to 1 (approximately) aspect ratio?

Re:But do you need multiple monitors?

[Jeff DeMaagd]

Two 4:3 displays can be bought at a lower cost than one widescreen display.

Re:But do you need multiple monitors?

[aquabat]

I thought the point was that multiple cards would split up the 3D number crunching calculations between them, and send the results to a master card to be displayed on a single monitor.

Re:But do you need multiple monitors?

[addaon]

So, uh, why not multiple widescreen displays?

Re:But do you need multiple monitors?

[MtViewGuy]

Two 4:3 displays can be bought at a lower cost than one widescreen display.

I agree with that, but the desktop space hogged by two 17" LCD monitors is surprisingly large, far more than what you get with the Apple Cinema Display.

Besides, with large-scale manufacturing of widescreen LCD's the cost would come down very quickly. Remember, most of today's latest graphics cards can easily add display drivers that can support something akin in aspect ratio to the Apple Cinema Display (they're already part way there with the 1280x768 display driver used on some smaller widescreen LCD's).

Re:But do you need multiple monitors?

[MtViewGuy]

I thought the point was that multiple cards would split up the 3D number crunching calculations between them, and send the results to a master card to be displayed on a single monitor.

It's like back to the days of the Voodoo 3 card running in SLI fashion? Such an idea seems superfluous nowadays, especially starting with the ATI R300 and nVidia GeForce FX generation of cards, which already have a tremendous amount of graphics processing power to start with. The very latest ATI X800 and nVidia GeForce FX 6800 series chipsets have graphics processing power far beyond what was available even only a few years ago.

Re:But do you need multiple monitors?

[stephenisu]

mmmm racing sim junkies around the world pop a woody over this stuff

Re:But do you need multiple monitors?

[tjw]

Even with today's flat-panel LCD's, two monitors will hog a lot of desktop space, something that might not be desirable in many cases.

If you would have left that third paragraph off of your post, I would have rated Funny. (I thought you were talking about 'desktop' in terms of root window).

Hell yes, some of us NEED multiple monitor setups. I've been using a dual monitor setup for about 5 years and although I can get by with one, it would make my day to day work much more annoying. It would be like replacing a mechanics impact wrench with a tire iron.

I do hear your point about a single widescreen display. I really wish someone would make a LCD display that is 8;3 (two 4;3 lcd screens stapped together with no gap between them).

Re:But do you need multiple monitors?

[MtViewGuy]

The only places I know where multiple monitor setups are a good idea are in CAD/CAM work (where seeing multiple views of an object being drawn is a good idea) and in the financial industry (where seeing multiple real-time charts and other financial data are a must for equities traders).

Besides, most games are designed with the assumption that you're using only one monitor. Wouldn't it be better for a game to take advantage of a wider aspect ratio display so the view becomes a bit more realistic?

Re:But do you need multiple monitors?

[mike300zx]

You need dual+ monitors for when you want to open two seperate things. It's not that often other than movies, which are better seen in my home theater setup anyhow than a smallish computer monitor (in comparison). Very rarely do I want something to strech across 2 screens or be that wide, however all the time practically it's great to have dual monitors that you can maximize to easily. If you had one big screen when you wanted to maximize your browser it'll take up the entire widescreen so you would end having to manually resize it to only take up a smaller amount. With 2 screen it's simple and in the rare case I want something double size I can strech it across if I want. Seems like a better solution. If I want my calendering program or a document open in one screen and a browser in the other doing research with multiple windows to toggle between not having to manually size them is significantly better. Or say you want some small programs on the sie playing music, signed in to aim, etc. you can leave those on the secondary and do the rest of your work on the primary while still having it be easy to keep an eye on those other apps without having to worry what programs are open trying to cover up everything else and come into the foreground.

Re:But do you need multiple monitors?

[ophix]

you forget other content creation such as video editing, image editing, web development, etc...

you are also implying that it isnt a good idea outside of the fields you mentioned. i would be willing to bet you have never spent any serious time using a dual display setup for work.

i absolutely love dual monitor support for system administration. bring up a vnc or remote desktop connection full screen in the 2nd display while i do other things in the primary display. also i had to do some minor web editing and having a browser open in one window and the editor open in another was quite nice.

having a dual monitor setup is quite a bit more productive for just about any job past spending the day in a word processor. downside is the lost desk space... but unless you are already rather cramped for space i think adding a second display would be very beneficial

as for gaming, well thats something different entirely.

Re:But do you need multiple monitors?

[slashdoter]

I've been using 2 monitors for years now and I have to say that once you've used them for a while you won't go back.

Right now I'm posting to slashdot on my secondary monitor and playing eve-online on my primry. While I've used the two monitor set up for web design and programing you will find uses beyond the standard CAD/programming argument really fast once you've got it in front of you.

Re:But do you need multiple monitors?

[gnu-generation-one]

"I think the big question we need to ask is do we really need multiple monitor setups? Besides the obvious issue of hardware cost of multiple graphics cards and multiple monitors, you also have to consider desktop space issues."

Ironic you mention that on the day that my development computer at work is moved onto a new desk to cope with the 4 monitors (for 4 PCs, admittedly), while our game-playing PCs are filled with multiple-output graphics cards, and with outputs serving multiple monitors. Oh, and the input-selection switches so that we can show 2 channels on one monitor. And the KVM switches to choose between 8 channels on one monitor.

If only we could get around the "1 PC = 1 graphics card = 1 output = 1 monitor" philosophy, such things would be a lot more convenient, and with a lot less messy cabling...

LCDs are good though. You can get stands that hold 8 LCD monitors or so on the desktop without really using up much space.

Re:But do you need multiple monitors?

[chez69]

you can buy yourself a new desk with the money you saved.

Re:But do you need multiple monitors?

[proxima]

I'd be interested in software (for X/KDE, as I'm a Linux user) that allows you to custom-specify the dimensions and positions of a few "pseudo-maximize" zones. Say you have one wide screen. Maybe split it up 1/2 and 1/2 or 2/3 1/3 zones. If the majority of a window is within a particular zone, when you hit the maximize button (or perhaps an additional button instead), it maximizes to that zone.

Maybe someone's done this already, but I've never come across it. In the age of more widescreen laptops and desktops I could see a definite advantage. Maximize a browser in the 2/3 zone and a terminal in the 1/3 zone on a particular virtual desktop, etc.

Re:But do you need multiple monitors?

[Bandwidth_]

Do you know how expensive widescreen displays and the like are?

17" CRTs are cheap. I have 4 running right now. It cost me ~$200 for this setup. That wouldn't even get me a 17" LCD normally.

Re:But do you need multiple monitors?

[rvw14]

I use my multiple monitor setup for video editing.

Re:But do you need multiple monitors?

[alecks]

Two 19" monitors do hog a lot of deskspace, but they sure look cool!

Re:But do you need multiple monitors?

[minister of funk]

Isn't it easier to read things in narrow, long columns than wide short ones? I turned one of my monitors on it's side so I could have another 30 lines in my text editor. It rocked... but then it died.

I don't think it died because of the orientation, rather it was a few years old and laready refurbished.

It didn't really die, the color convergence is off.

Re:But do you need multiple monitors?

[MtViewGuy]

Two 19" monitors do hog a lot of deskspace, but they sure look cool!

If you're talking two 19" (diagonal) CRT's, they also use a lot of power, too. :-(

I still think that unless your needs are very specialized (CAD/CAM, equities tracking, programming work, video editing and a few others), having more than one monitor is kind of overkill. After all, except for a very small group of users most computer users still work with one monitor. The number of games that take advantage of multiple monitors are still very rare, hence the reason why most computer games are design on the basis that the customer will be using one monitor only. I of course assume that the upcoming Doom III will have multiple monitor support, but then Doom III will be pushing the limits of computer hardware to start with.

Re:But do you need multiple monitors?

[MtViewGuy]

Do you know how expensive widescreen displays and the like are?

Of course they are still expensive because very few such widescreen displays are made in LCD form for the computer market (most of them are being made for the consumer TV market such as the Sharp AQUOS line). But once these displays at native resolutions higher than 1280x768 and larger diagonal sizes are widely made, the cost will drop quite a bit (my guess is that with large-scale production they could produce something akin to the Apple Cinema Display but at a slightly small diagonal size to sell retail around US$1,200).

The real question

[241comp]

Is this compatible with Brook and other general-purpose GPU programming techniques? The use I see for it is this:

Imagine an openmosix cluster of dual-processor machines that run bioinformatic calculations and simulations. Lots of matrix math and such - pretty fast (and definitely a lot faster than a single researcher's machine).

Now imagine the same cluster but each machine has 2 or 4 dual-head graphics cards and each algorithm that can be created in Brook or similar is. That gives each machine up to 2 CPU's and maybe 8 GPU's that may be used for processing. The machines are clustered so a group of ~12 commodity machines (1 rack) could have 24 CPU's and 96 GPU's. Now that would be some serious computing power - and relatively cheap too (since 1-generation old dual-head cards are ~$100-$150).

By the way, does anyone know if there is any work going on to create toolkits for Octave and/or MatLab which would utilize the processing power of a GPU for matrix math or other common calculations?

Re:The real question

[theCat]

Wired carried a story on this very topic last year (or maybe the year before). I don't have the issue handy. It was nicely done, and was the first inkling I'd had that these GPUs where some serious hardware looking for a problem to solve.

It would be interesting to see some of the distributed computing efforts (seti@home, etc) take advantage of GPUs if there is anything there of use.

Re:The real question

[joib]

Now imagine the same cluster but each machine has 2 or 4 dual-head graphics cards and each algorithm that can be created in Brook or similar is. That gives each machine up to 2 CPU's and maybe 8 GPU's that may be used for processing. The machines are clustered so a group of ~12 commodity machines (1 rack) could have 24 CPU's and 96 GPU's. Now that would be some serious computing power - and relatively cheap too (since 1-generation old dual-head cards are ~$100-$150).


Agreed, that would be cool. Vector processing on a budget, in essence.


By the way, does anyone know if there is any work going on to create toolkits for Octave and/or MatLab which would utilize the processing power of a GPU for matrix math or other common calculations?


Both matlab and octave use the BLAS library, as do most scientific applications needing dense linear algebra. I guess one could do a BLAS implementation for these GPU processors.

What I don't see happening is that scientists would be eager to write applications with Brook or whatever proprietary language these GPU:s support instead of standard Fortran or C. While the need for performance of course is the reason to use parallel programming in the first place, portability is incredibly important for scientific applications. Supercomputer centers don't tend to have very much vendor loyalty, rather they buy whatever gets them the most bang for the buck at the moment. Nobody has the time (nor wants) to rewrite applications every few years.

Now, what I think could happen is that some smart company would make a Fortran 95 compiler, which would offload array operations to the GPU(s) while doing the more complicated logic on the main CPU(s). Of course, such an approach would require quite a lot of bandwidth between system memory and the graphics card memory. In the short term, a BLAS implementation is about the best one can realistically hope for. That alone could produce a huge increase in bang-for-the-buck for a lot of applications.

Re:Big Deal - PCI Express. Any one can add two vid

[strictnein]

That's PCI-X. PCI-X is different than PCI-Express.

The Return of Voodoo 2 SLI

[MoZ-RedShirt]

History repeating: Who can (or can't) remember

Power to them if they can pull it off!

[l33t-gu3lph1t3]

discrete parallel graphics processing has been around for a while. The most notable example of it is probably 3DFX and their Voodoo-2 cards. However, there's a problem with this tactic, namely, in the "diminishing gains" department.

So here's the question:

-How is pixel processing going to work? For a given frame, there is vertex, texture information, as well as the interesting little shader routines that work their magic on these pixels. How are you going to split up this workload between the 2 GPUs? you can't split a frame up between the GPUs, that would break all texture operations and there would be considerable overhead with the GPUs swapping data over the PCI bus. *MAYBE* having each gpu handle a frame in sequence would do the trick, but, again, it's a dicey issue.

It would appear to me that this dual-card graphics rendering is quite similiar to dual-gpu graphics cards. Except, where in a graphics card you can handle cache/memory coherency and logic arbiting easily due to the proximity of the GPUs, with this discrete solution you run the problem of having to use the PCI Express bus, which, as nice as it is, is certainly not that much faster than AGP.

So I say, power to you Alienware. If you can pull it off with Nvidia, ATi et all, great. It's too bad the cynical side of me thinks this idea reeks of those blue crystals marketing departments love :)

Re:Power to them if they can pull it off!

[Fujamabob]

Well, if it works much like any other numa system, you have a few options - send data back and forth like you suggested, duplicate computations, etc. Then again, the software that does load balancing may look at textures when parcelling out tasks.

In any case, they're only expecting a performance boost of about 50%, according to the update. Whatever solution they've come up with for such operations, we and they both know it's going to involve a less-than-optimal speedup.

Re:How is this different than what I've been doing

[IAmAMacOSXAddict]

If you read the short article. they are talking about pushing all the data for a SINGLE monitor, through two PARALELL cards. Essentially allowing for twice the GPU power to crunch the graphics for the monitor.

You are running a bunch of video cards INDEPENDANT of each other. Clearly NOT THE SAME THING...

More interested in real dual PCI-Express GFX slots

[Performer Guy]

They might be able to make this work for games but I'm personally more interested by the simple fact that Intel chipsets will support dual PCI Express graphics buses. Hopefully this will be possible on a reasonably proces mobo.

Having 3 slots would be ideal but I won't say no to 2 GFX cards so I can drive two monitors from two independent graphics cards at last.

It doesn't say how this technology will combine the two cards and whether it will need software support from the games. Hopefully it won't but the devil is in the details. I'm pretty skeptical about this at the moment. We need more details on the implementation.

Re:Finally! (NOT!!)

[IAmAMacOSXAddict]

Re-read the article your thinking of. M$ Long(wait)horn will require 3-4 times the GPU speed...

Thank would mean 3-4 GPU cards, not to mention the overhead for running them.

No Thanks, Make mine a mac, I have everything now what I've seen them planning for Longhorn. Feels like I'm back in 1984 waiting to see what Micro$oft copies from the mac... AGAIN...

too many standards

[jcostantino]

I'm going back to ISA...

how is this different from my voodoo2 cards?

[kerp11]

i had dual voodoo 2 cards that were linked together with a special cable to make them run in SLI (scan line interlacing) mode. quake certainly ran a lot faster once i got the second one installed.

But will they run?

[strredwolf]

Okay, we have a GeForce and a Radeon in parallel. What's the communication protocol that's running over PCI Express that allows them to do that?

Something tells me you need special drivers AND/OR a standardized graphic card accellerator protocol just to pull it off, otherwize you're stuck with two of the same cards.

Re:But will they run?

[ocelotbob]

I don't see why they wouldn't. You've got two cards with two different PCI IDs and bus numbers. The OS simply needs to say, "Okay, this video packet is going to 1,0,0 and this one is going to 1,0,1". The comm protocols have been there for a very long time; I can remember playing around with using a couple different PCI video cards by different makers, or even the oldschool dual head back in the days when both video cards needed to be in 8 bit slots to work. Yeah, you'll probably get better performance by having two of the same card -- lover driver overhead being the key feature -- but nothing is stopping you from doing it right now.

Re:But will they run?

[Zathrus]

You've got two cards with two different PCI IDs and bus numbers.

That's not the problem.

Presuming they're doing interleaved frame processing (you simply cannot do scan line interleaving -- things as simple as anti-aliasing make that impossible), then you'll have every other frame rendered by each card.

You realize that the cards don't produce the exact same output, right? You'd probably get a headache from trying to visually process the subtle differences that occurred between each successive frame (but weren't too different from the alternating earlier frame). Ow.

And when it comes to randomized effects, such as used in some pixel shader methods, having two wildly different pseudorandom number generator algorithms would cause interesting effects.

I could see using two different cards from the same processor family, and maybe (but doubtfully) two different cards from the same manufacturer with different chipsets (e.g. -- GeForce3 and GeForce4), but beyond that I can't see a way to do it and not cause problems.

Of course, this presumes they're doing frame interleaving. But I can't think of any other way to do anything like this, and any other possible way would cause more problems, not less.

Re:Power hog...

[^_^x]

No kidding... the way 3d cards are heading is asinine, and if it goes mainstream, I guess I have a reason to switch to console gaming. (Not that they're much better with hardware costs doubling every console generation...)

Re:Great News

[goatan]

Um how is depresingly boring games nvidias fault? i don't think they go to games manufacturers and demand that they make inferior games

Not really hard....

[Kjella]

How? Do they have some nifty way of combining two standard off the shelf graphics card signals into a single monitor? (Hint, it's hard enough getting the monitor to properly synch up with a single high performance graphics card!)

Duplicate data stream (should be doable in hardware), have them render half each (every 2nd scanline?) and merge them with a trivial buffer (keep two bools, one "firsthalf=done/not done, secondhalf=done/not done"). You'd limit yourself to the minimum of the two, but since they each paint nearly the same (one scanline off) the performance should be near a 100% doubling.

Basicly, it's going to be an expensive design involving some already expensive cards. But it's definately doable.

Kjella

Re:Not really hard....

[Gr8Apes]

I'll call you on this. Yes, the graphics cards could do that, but how does third party hardware access this buffer on the graphics cards? The claim here is that 2 dissimilar graphics cards can be run together in a system and work togther. That is a pretty major claim, not at all like Voodoo, which had specialized hardware and connections to enable that functionality.

So, exactly how is an ATI 9800 going to render 50% of a screen and send that info to my NVidia GFX that rendered the other 50% to pass onto the monitor connected to the NVidia? Or vice-versa? Seriously, that's why my bullshit meter pegged on this.

Quantum3D did that several years ago

[ggravier]

I used to do this with my Quantum3D Obsidian graphics boards years ago (at the time when 3Dfx meant something to gamers). :)

Re:3dfx has done it again ...

[John Courtland]

I know this is a semi-troll, but SLI was shit and so were Voodoo2's. Any time a voodoo2 would hardlock, they had complete control of the screen, even though the system was still functional. So all you could see was black, or the last frame of your game. I never tried removing the video jumper cable and seeing if the main card was still displaying, but even after you reboot, the voodoo2 would still chug along, preventing the screen from displaying. I had to shut the machine off for a good 20-30 minutes to allow the voodoo2 to discharge to the point that the screen would be visible again.

And SLI doesn't work the way you described. SLI = Scan Line Interleave. Each unit would process every other line, then it would add them together on the screen. The only problem with that is that sometimes they would get out of sync. Pieces of junk.

Who else will obviously support this on mobos...

[Geiger581]

Let us not forget that nVidia and ATI both produce chipsets. Multiple graphics card purchases per system would be a dream for them, and they can help in a direct manner. Although there are not many (read as 'maybe a dozen worldwide') boards with dual AGP, the PCI-Express standard will lead to much easier multi-GPU setups. Also, the newest ATI chipsets with embedded GPUs support multi-monitor if an ATI card is used in the empty AGP slot, so you know that these guys already have to have agendas for PEG in mind.

Hello? Matrox, anyone?

[pla]

Not PCI-X, but Matrox has offered quad-headed machines for years.

For those who don't want it for gaming (ie, don't want to blow a few hundred for the latest and greatest multi-headed AGP card), throw in a few extra el-cheapo PCI cards. Win2K supported up to ten heads - Only one AGP, obviously, but although it can't run as the boot display, both Windows and Linux (X, anyway) can make it the primary after startup so your games will run on it. XP supports even more than that (up to 255, I believe?).

On my Windows development machine, I use a 5+ year old Trident PCI card for the second head, and an older Geforce (3? low-end 4? Don't really care, it works) as the primary. I keep WinAmp, Task Manager, Calculator (a nice graphing one, not the 'doze default), and SI's TCPMon (a network connection monitor that looks much like TaskMan) on the second head, with my actual dev IDE open on the primary. On the rare occasions when I want to play a graphics-intensive game (I far prefer RTS to FPS), it works perfectly. Good frame rates, no glitches (that wouldn't have happened anyway). I do run a single-monitor screen saver I wrote (I've made it publically available on my homepage), that uses a moving-windows style saver so I can see its contents without worying about showing a static screen for too long. I also use a mouse corral (don't recall the name, do a Google search), to keep the mouse from leaving the primary display while playing a fullscreen game (that will cause glitches, just don't go there). Overall, it works great, and that PC doesn't even have a PCI-X bus.


So, not really news, I'd have to say. Someone managed to do with a new tech what we've had the ability to do with the previous gen of hardware. Whoop-de-do. "In other news, the sun has continued to burn hydrogen for the 4.6 billionth year, making a new record for our solar system!".

CAVE is the killer app here

[karmajudgment]

While it is interesting to muse the possibilities of optimizing rendering with multiple graphics cards, I think that the possibilities for 3D and multi-plane projection are much more interesting. Imagine a PC with 8 independent graphics cards and 8 small footprint DLP projectors. You could have a private, 4 plane 3D CAVE virtual reality environment.

An overview of CAVE virtual reality systems>

Multiple video outputs is definitely a step in the right direction for a lot of exciting developments is visualization. Gamers rejoice.

Like most Widescreen monitors...

[Kjella]

Apple's is 16:10. Not that I quite figured out why, but TVs seem to be 16:9, Monitors 16:10. Now, if only there was HDTV content near me (HDTV broadcasts? None of the national broadcasters at least. HD-DVD? Still up in the blue. HDTV newsgroups *cough*? Too large for a measly broadband line).

I just hope that we in Europe can *please* have HDTVs and none of those fucking stupid region codes, yes? Or do I have to wait for DeCSS2 before I can buy any of the special offers??? (ever notice how those with no codes are always more?)

And yeah, I wouldn't mind having a huge TV to double as my monitor either... hehehe. Except if that goatse.cx guy shows up in HDTV while reading slashdot...

Kjella

It's about damn time!!!!

[stoicio]

It's about time we were able to install more than one video card. Being just another PCI device I should be allowed to install as many as I have slots if I wanted. Same goes for sound cards too. This would actually make a PC useful for video and audio production beyond the home hobbyist/consumer market.

Re:Power hog...

[SFBwian]

Funny, I seem to remember some expensive consoles a while back. Sega CD? 32X? TurboGrafx? Jaguar? NeoGeo?

I don't think a Nintendo system has ever debuted above the $200 price point. Systems used to come with pack-in games at launch too. These days that doesn't happen until right before the first round of price cuts.

Admittedly though, you didn't used to have to buy as many extra controllers or memory cards. The games used to cost more though!

ATI will be doing this it's Catalyst drivers soon

[89cents]

Ok, so they havn't explicitly said so, but have been hinting at it

Ati's Terry Makedon says: "Something big is coming for CATALYST in the next 2-3 months. It will take graphic drivers to a brand new level, and of course will be another ATI first. It will be interesting to see how long before other companies will copy the concept after we launch it."

Hmmm... just in time for PCI Express and it's not something specifc to Ati's hardware.

Re:Hello? Matrox, anyone?

[Niddix]

You've missed the point. Multi head is fine and has been around for ages.. 8 years ago I used to use a EGA monitor attached to a card for debugging. This is using the GPU of multiple cards to crunch the numbers for a single display. Very different.

In Depth Interview

[rsmith-mac]

There's also an in depth interview up at HomeLAN, which talks more about the specs of the X2 systems, along with how Alienware's going to handle powering and cooling the beast.

Doesn't look like multihead...

[fikx]

From reading the press releases, it doesn't look like this is related to multiheaded systems. I didn't see any mention of two monitors anywhere in the releases. It looks like something different, although technical details are pretty light. My first thought was maybe a way to switch between multiple video cards without opening the case. But, from reading it, it more seems like it;s a way to use more than one video card to drive a single display. It's kinda like the Vodoo SLI stuff, only this is a generic solution that works for any brand O card. Could be cool. Literally getting the best of ATI and Nvidia at the same time. I like the sound of it!

This is multi-display, not faster, graphics

[Animats]

This is just a motherboard with two slots for graphics boards. Period. This is not about somehow using two GPUs on separate cards to run one display faster.

It's possible to design and build GPUs that will play together to provide higher performance graphics. The Apple 3D Quickdraw Accelerator Card, from the early PowerPC days, does exactly that. If you get two, drawing speed nearly doubles. That device was more of a coprocessor, closer to the CPU than a modern GPU. It didn't drive the display; it just pushed bits into display memory elsewhere.

Dynamic Pictures, before they tanked, made the Oxygen line of boards, with one to four GPUs. Those units tiled the screen, with each GPU working on different tiles. The screen would break up into a checkerboard when the device was having problems, so the division of labor was quite clear.

Softimage, in their heyday, once showed a system at SIGGRAPH which was running the mental ray renderer on two 6' racks of machines with custom accelerator chips, so you could do final-quality rendering in real time. That was too pricey even for Hollywood, but very nice to watch.

Re:This is multi-display, not faster, graphics

[bairy]

This is just a motherboard with two slots for graphics boards. Period. This is not about somehow using two GPUs on separate cards to run one display faster.

But that seems pointless, although you'd get accelerated graphics per screen, it wouldn't be new as most graphics cards these days have a tv- or dvi-out letting you use a second display.

Re:This is multi-display, not faster, graphics

[Caesar]

No, please check the story again.

This is two cards working to draw one screen by splitting the workload in half.

Re:This is multi-display, not faster, graphics

[Animats]

Ah. They've changed the story:

UPDATE: The current implementation splits the screen in half, assigning rendering for each half to one of two cards using a software load balancer to try and ensure proper synchronization. ... Alienware is currently saying that they expect users to see a ~50% performance boost over single card implementations.

So that's a dual head graphics system mapped onto one display. Like picture-in-picture TVs.

With a really wide display, it might be worth it.

Re:This is multi-display, not faster, graphics

[IceStorm]

One may not need wider displays.

Matrox permits the combining of their dual display cards into what appears to the OS to be a virtual wide display. Under XP/W2K you get what appears to be one very wide display instead of two distinct monitors.

If other card makers support this option, and to be honest I haven't checked to see if my ATI cards can do it, then it's possible the vid card driver could present the abstracted single display to the software layer that Alienware is building, such that you do get a wide diplay for gaming on multiple monitors.

Hey, maybe it'll make Parhelia usable as a gaming card. :-)

Re:Hello? Matrox, anyone?

[dead sun]

Different from what the original poster was saying yes, but not different from what's been done. I do agree that most people here have missed the point of the article. However, remember the Voodoo 2s? They worked in parallel rather nicely.

Frankly, though, I could care less about using a pair of cards to crunch graphics faster. I'd rather have more displays. I want PCI express so I can have a quad head setup with all the screens running decently fast. The only way to do this now is with PCI cards which aren't that fast. Actually, I think nVidia made a Quadro4 card that did four displays at decent speed, but it wanted a 64 bit PCI slot. Ah well.

From a non-gaming point I'd rather have more room for stuff than have my desktop refresh faster than it already does. From a gaming point I'd rather have a game that engages me and is fun than have it concentrate solely on graphics to the point it needs two massive cards.

On the other hand, it might allow decent single screen performance on low cost hardware for gaming and switch modes to multi-screen on the desktop. A boost like that would always be welcome, but the article seems a little scarce on details.

Re:3dfx has done it again ...

[LilGuy]

What does this have to do with Alienware's press release?

Sheesh... :P

Version control

[Short Circuit]

The routing circuitry could make two copies of data requested by the GPU. Changes only occur on one copy, until a given number of other read requests occur. At that point, the old copy is freed.

Remote requests are always read from the old copy.

At high enough framerates, you can exceed the refresh rate of your display. In such cases, old data won't be as noticable. That doesn't solve the problem on high-refresh-rate displays, though.

Hardly a new concept

[default luser]

And no, I'm not referring to SLI, which was specifically designed to pair two Voodoo 2s together. I'm talking about technology that can bridge any two cards together. This is nothing more than the complex bridging involved in say Metabyte's TNT 'SLI' solution that consisted of a PCI bridge and software to split the framebuffers. It was never released for two reasons

1. GeForce 256 released shortly after this was announced.
2. PCI bridge required both the AGP and the PCI card to operate in PCI DMA mode. Unfortunately, there never was such a thing as an "AGP bridge".

In any case, other companies have now successfully implemented a simple framebuffer splitting concept on-card, where the bandwidth is more plentiful. The ATI Rage Fury MAXX and the 3dfx VSA-100 come to mind, these chips simply split the framebuffer rendering according to complexity. Beyond that, NOTHING was shared - triangle and texture data were replicated for each chip.

The key to this: on the software side in 3D mode their software automatically splits two framebuffers between the two cards. As for the "special" chipset, whatever scene data is sent to one video card, the same data is sent to the other video card. I can't imagine it being any more complex than this.

I want a Quad X800 setup.

[legomad]

Imagine 300fps on Far Cry with uber max settings.

Doom III

[Angry Black Man]

I hope this technology doesn't delay Doom III anymore. John Carmack will probably want to program in full support for his new engine before it ships off cause he has to make money of licenses until he makes the next one.

Re:More interested in real dual PCI-Express GFX sl

[Znork]

"Having 3 slots would be ideal but I won't say no to 2 GFX cards so I can drive two monitors from two independent graphics cards at last."

What's stopping you from plugging an extra PCI based card in today? I've been running two monitors from two independent graphics for years.

Technically, you could. Practically, not possible.

[Ayanami Rei]

Ideally AGP is specced to run at a rate which is as close to the same speed as the northbridge's bus to RAM (close because AGP is a multiple of the PCI bus rate while the FSB on many motherboards is independant of this). Once you've done that, adding another AGP doesn't help because you'd have to share that fast bus into the AGP-mapped memory space. It would have been nice to run 3D games on dual monitors, but by the time that became pratical, 3D cards had dual monitor outputs anyway. So there was no demand to have a dual AGP port board (while not being able to advertise AGP 8x on both, or whatever the fastest at the time was). Increased manufacturing costs without giving any additional benefit to the user.

And workstation/visualization boxes in the hi-end could use PCI-X on the chipsets with like 4 seperate PCI-X buses.

Ultimately no one even bothered to work out how dual AGP might be presented to the operating system.

It's about time

[hal2814]

Graphics algorithms are some of the easiest on earth to distribute and run in parallel. I'm surprised that this hasn't been more popular already.

VGA

[b100dian]

I supposed they stripped the Graphics part from the VGA when they came up with this?..

Re:3dfx has done it again ...

[GeckoX]

Yes, SLI stands for Scan Line Interleave, but that wasn't the only way that the Voodoo2 could use the connection between 2 cards, if you wanted, you could write code to get the voodoo2's to behave the way the grandparent described.

Secondly, and more importantly, nahh, you're obviously trolling if you can dis the Voodoo2 for what it was at the time...THE card, PERIOD.
(No, not for long, but without any doubt at the time it came out it was by FAR the best consumer card available, also cost me twice as much as any card I've purchased since ;)

Re:MD tag LLY 347

[MindStalker]

THE BELTWAY is an interstate highway
encircling Washington, DC & passing through Maryland and Virginia suburbs

Not quite the same...

[Ayanami Rei]

that's still one AGP slot per "motherboard", those systems can have a variable number of motherboards that are tied to a back plane, which is then partitioned.

Although it is impressive that an SSI can run across them. Very neat.

But I wonder if you can actually run a single X11 session using both screens.

Errrr.... mmm no.

[Ayanami Rei]

There definitely isn't going to be any of that interleaved scanline bullshit of the days of the Voodoo.

Think about it, each card is going to have it's own dual display outputs, as every card that you'll plug into PCI Xpress should have.

Besides, a single graphics card is more than enough to play most games on the market at an acceptable frame rate. ANY PCI-Xpress card will be overkill for almost every game, and fine for the most demanding.
What benefit would you get for having two "working together", when you consider the massive overhead?

So you plug in two, and so you have four monitors. Big deal, Windows already does this.

Maybe this "revolutionary technology" is getting Direct3D to use the whole virtual viewport of all four monitors instead of limiting it to a single card at at time. That'd take some trickery to fool both a generic game engine and the seperate graphic cards w.r.t. viewing frustrums and what's not visible, syncronization of vertex/texel programs, etc.

Re:3dfx has done it again ...

[John Courtland]

It was good when it worked, I'll give you that, but it was buggy. At least the two cards I owned. The RIVA TNT I had was consistently better, in both speed and graphical quality, not to mention I could actually play games ouside of 640x480 and 800x600 (I don't remember what the resolution limitations were on the Voodoo's but I assert that those limits were a nail in their coffin.)

Not really.

[Ayanami Rei]

That's the location of the PCI-bus-space accessible I/O space of the card (and a place to fall back to if the AGP GART is non-functional).

The AGP card itself is attached in some (chipset specific) way that is very close to the northbridge... it is essentially acting as non-cached physical RAM mapped into the memory space of the system routed through the chipset. The bus protocol (like whatever protocol the chipset has to your RAM) has nothing to do with PCI at all.

BTW, RAM tends to appear as I2C devices on the ISA bus. Does that mean RAM sticks are on the ISA bus?

Re:Not really.

[Short Circuit]

Ah. I'll have to look at my sysfs tree at home to see if I can find another mention of AGP in there.

BTW, RAM tends to appear as I2C devices on the ISA bus. Does that mean RAM sticks are on the ISA bus?

Probably not. IIRC, the ISA bus runs at 4.77MHz. RAM has required greater speeds than that for a long time.

Re:Big Deal - PCI Express. Any one can add two vid

[strictnein]

No, they are not.

PCI-X = high-speed backwards compatible PCI replacement

PCI-Express = high-speed PCI and AGP replacement and is not backwards compatible with PCI

Reading the comments...

[ReyTFox]

Makes /. look even more illiterate than usual.

I think 1/2 the posts were about multi-head setups, even though the article DIRECTLY STATES that the two cards in this setup are each rendering one half of the screen :P

Intro to Communications

[Short Circuit]

serial connections in parallel

When you have parallel serial connections isn't it just one parallel connection?

Not if each connection handles different data.

Let's say you've got 8 wires going in one direction. In a parallel connection, you'd be communicating the 8 different bits of a single byte at the same time. Do it seven more times, and you've sent 8 different bytes.

As multiple connections, you're transmitting the first bit of eight different bytes, then the second bit, then the third. After 8 cycles, you've again sent 8 different bytes, one on each wire.

The downside of the multiple serial conections lies in latency. It doesn't matter if you only want one byte, it'll still take eight cycles to transmit. With a parallel connection, it only takes one cycle.

The downside of a of a parallel connection lies in that you have to keep all eight wires in perfect sync. If you don't, you risk mis-matching bits from different bytes, or even not registering signal on all the wires. With multiple serial connections, Each wire can be independent of the other wires.

That's simplified, of course. With RS232 serial communications, the old standard in PC peripheral technology (though still pretty common in manufacturing technology), you had as few as nine wires. With USB, they've cut it down to four wires.

Re:More interested in real dual PCI-Express GFX sl

[Performer Guy]

You're crippling the card. PCI doesn't cut it for me and you can't always get PCI versions of cards either. Even with PCI Express it's not as interesting as PCI Express 16x, so bear that in ming on future PCs that have PCI Express everywhere, you still want the higher bandwidth PCI Express 16X for graphics.

Honey...

[Ayanami Rei]

that was the point I was trying to make.

The logical hardware layout that the operating system presents to the user is not congruent with the physical hardware layout.

That's just how it APPEARS to the operating system so the OS knows how to talk to and control all the parts. If you want to check the size and health of a piece of RAM, you send commands to an emulated ISA bus that the memory controller presents to you.

If you want to adjust the power settings (you know, like sleep mode or suspend) on an AGP card, you would use the virtual PCI bus location to specify the card you mean and send it standard power commands, even though it's not on a PCI bus at all.

You won't see a hypertransport bus in your sysfs tree even if you have an Opteron, but it doesn't mean it's not what the RAM, AGP, and other CPUs are connected to. It's all handled in the hardware, and you don't know the difference.

Now, I mean there may be this "hypertransport controller" that appears on the PCI bus, but that's just your operating systems window into doing stuff to it. But that doesn't mean the hypertransport links are on the PCI bus. It's actually the other way around.

Here's a good example: attach a USB-based flash drive to your system windows/linux system. SCSI system says it's a SCSI disk attached to a SCSI controller. A virtual SCSI controller. Check the USB bus. USB bus says it has a mass storage device. Huh! Well which is it? (Actually the SCSI subsystem would probably call the controller something like USB mass storage host controller or something)

Re:Honey...

[Short Circuit]

Ah. I see. You make (several) excellent points.

Here's a good example: attach a USB-based flash drive to your system windows/linux system. SCSI system says it's a SCSI disk attached to a SCSI controller. A virtual SCSI controller. Check the USB bus. USB bus says it has a mass storage device. Huh! Well which is it? (Actually the SCSI subsystem would probably call the controller something like USB mass storage host controller or something)

That's a software issue. The driver access the device by communicating through the PCI bus, through the OHCI/UHCI/EHCI controller, through the device tree, to the flash disk.

Under Linux, the usb disk driver then provides access to the scsi subsystem and an entry under sysfs.

Somewhere in there, a device node is also available for raw access to the USB device.

Re:3dfx Still Alive! Drivers Still Being Written

[Cutting_Crew]

i am a part of 2 teams that are still writing drivers for 3dfx so please dont give up on them just yet. the newest drivers support up to 8.0 and beta for 9.0. they also allows emulated T&L for playing some of the latest games at a decent FPS and some nice DVD movies with all the special effects. for instance, on my 3dfx Voodoo 3 i can play warcraft 3, UT 2003 and UT 2004, Need FOr speed HP2 and Underground(and of course the older NFS titles) and many more games! you can go to http://www.3dfx.com and click on the voodoo files link to get the updates and/or drivers for the current 3dfx card that you have or would like to buy. Unfortunately i think all the new drivers are compatible with WIN9x, 2k and XP..sorry linux 3dfx fans. anyway the cards are still alive so if you would like to help with GLide drivers or openGL drivers for the cards then please snoop around and ask how you can help. :) oh yes and to respond to one person, a lot of 3dfx fans did stay with NVidia b/c most of the 3dfx engineers went to NV. so youve got a lot of die-hard NV/ex-3dfx fans out there... including me :)

Obligatory beowulf cluster comment

[spannah]

Personally I am looking forward to running my own beowulf cluster made up of 8 Radeon X800 XT XT Super Duper XXX TT Platinum Card, with 64 pixels pipelines each, effectivelly making a 512 node supercomputer. All of this inside my desktop tower.

Re:3dfx has done it again ...

[damiam]

Nothing. You probably have "reparent highly rated comments" turned on. The above comment was replying to this.

Re:Technically, you could. Practically, not possib

[The_K4]

Wow, so many things wrong with that statment.

1) You COULD do this same thing with AGP what they are talking about here is using the 2 cards to process data for 1 display (not for dual head). You COULD do this an AGP because in most cases you would send the SAME data to both cards and since AGP is a multi-drop bus all the cards could listen to the same data, the bus won't take that much of a hit. The only bottle neck would be in returning data from card 2 to card 1 for display, but it wouldn't be that bad because all the processing has already be done, so that data's going to be small

2)AGP didn't do this not becasue of the cost to MOBO makers but because it would require a lot of software and hardware support on AGP (a multi-drop bus).

3) This article isn't about a chipset with 2 PCI-Express (NOT to be confused with PCI-X) ports, it's about a motherboard with them. The board pobabaly uses a PCI-Express switch to get the second port (so they are going to share the bandwidth).

4) How dual AGP would present to the OS is easy. The hard part is how the two cards talk to each other, and I suspect that this will still be an issue for alienware.

Exactly...

[Ayanami Rei]

ultimately EVERY device interaction in the system appears to be a memory access to the driver writer. But in order to support various modes of being to attach arbitrary devices and support unknown features, we adhere to some protocols standardized some time ago. Whether or not these protocols are conveyed over the same type of physical hardware that they originally did when it was introduced is up in the air.

The driver doesn't communicate through the PCI bus at all. Rather:
1) USB-mass storage takes the SCSI commands from the disk subsystem
2) mass storage converts them (trivially) into packets for injection into the USB bus.
3) mass storage informs the host controller subsystem to transfer data.
4) host controller talks using directly to the USB chipset through memory-mapped I/O (how it actually "gets" there, it doesn't care) and lets it know it's about to do DMA
5) host controller schedules DMA transfer with system (ultimately also using memory mapped I/O or maybe OOB I/O)

How the data actually gets moved from memory into the USB chipset is completely up to the motherboard manufacturer. Most likely the USB chipset is hooked up via a real PCI bus. And when you configure the devices, the system would be talking to a real PCI bridge when getting all the address ranges to do that I/O.

But that's not necessarily true. The USB host controller could just as easily be a directly attached member of the northbridge with no physical PCI bus presence at all. It'd still work, provided that the host controller was properly initialized and had the same register map.

It's a software issue, but so's all of it. It looks like that for simplicities sake, and for making the drivers more portable.

The USB hardware likes to talk in a dialect of SCSI, so you make it look like a SCSI device, naturally.

There'd be no reason to attempt to bit-bang the memory-mapped registers of the USB host controller, even though you could.

They're expecting you to layer your drivers and interfaces in that fashion, and the hardware has provisions to help make the end-result (the state of your flash disk) match up with what the software expects.

How the illusion is completed can be all software, all hardware, or a combination of both!

Why build Frankenstein? Because you *can*!

[mactari]

From the article:

The current implementation splits the screen in half, assigning rendering for each half to one of two cards using a software load balancer to try and ensure proper synchronization. (You may recall that 3dfx's SLI technology split the workload by alternating lines. This is a slightly different approach.) The motherboard is being produced for Alienware by iWill, and Alienware is currently saying that they expect users to see a ~50% performance boost over single card implementations.

50% better from an extra card is nothing to sneeze at. And that sounds reasonable from the point of view of all the extra overhead you'd be going through to coordinate getting the top and bottom halves up at the same time. It's still boggling to me that anyone would shell out so much for a new system in order to game (my PS2 does relatively well, thank you, and even my old StarMax with Voodoo 3 plays Team Fortress quite well in 2004), but as far as geeky-cool, this is awfully neat zeroes and ones.

That said, your original point is right on the money -- sorry can't mod right now. You definitely should be able to see a number of artifacts comparing the top of your screen to the bottom, which could be comical in effect, depending. (Hey, let's see what my Rave looks like up next to this X800 XT...)

That said, however, I'm sure the idea is to have two of the same, but to have the demo create Frankenstein's monster if only to show it can be done. Certainly if two different cards work, two of the same should look proverbially beautious.

Come on...

[Ayanami Rei]

1) See my other comment in this article about why I think routing data from one card back to the other is silly, and it is a huge bottleneck.

Also, they talk about using different cards. Last time I checked, an ATI card wouldn't work with an NVidia driver. Not likely. No, they'd have to be seperate address spaces for it to work in any sane fashion.

I can imagine if they were the SAME card the driver would create some overlapping maps for a special "joined" render mode. It have to be more at hoping there's a large correlation in the data both cards need pre-render, but you're not going to have perfect scaling.

2) Well, okay. But I was not assuming it'd be as simple as multidrop, that'd be kind of retarded. Ultimately it's whether or not the consumer has a way to take adventage of it...

3) I was explaining what was available in the past which filled the role that dual AGP might have... multiple expensive PCI-X cards in server mobo/chipsets. They did exist, they still exist. You can put 4 XVR-1200s in a Sun v880z if you want... gives you 8 1920x1200 3d-accelerated displays. There was an existing solution, so no market pressure existed.

4) No, it's not. Direct3d would have to do some heavy lifting to reinterpret what the games' graphic engine wants to display to what each card needs to know. A lot of it would be shared, but there'd be a need to modifying vertex programs, adjusting vertex/transformation data, syncronizing retrace, etc.

The easy solution would to be to add a feature to a card that lets it pretend the viewport is XY, then only render the subset XZ. Each card gets all the data required for XY, but knows which subset it's supposed to render independantly. But I don't think any chipsets out there support that yet. So it'd have to be added. Hence, dual AGP would not be useful for existing hardware. And again, it'd only work for matched cards.

But that's AGP, anyway. We're talking about PCI-express. And they are not multidrop. So clearly an extension to Direct3D is needed. I think that's what Alienware is going to do. Define the interface. Pressure Microsoft into updating DirectX to support it and create a software emulation layer to present the right data to each display. Then hope the driver authors can implement the subset rendering feature.

They'll solve combining multiple cards' output into one monitor by selling an Alienware monitor with two inputs that displays seamlessly side by side... taking a cue from Liebermann

The Real Question...

[SnowCrashed]

Is how massive Alienware's markup on this technology will be? I'm sure that constructive modders, or mobo manufacturers will have this out to the general public (read people who don't want to pay 5k for a pretty case) as soon as it ships.

RTFA

[Atario]

This is not a multi-monitor technology, either. The current implementation splits the screen in half, assigning rendering for each half to one of two cards using a software load balancer to try and ensure proper synchronization. (You may recall that 3dfx's SLI technology split the workload by alternating lines. This is a slightly different approach.)

Not multiple monitors. One monitor, two video cards.

Now, I can see how this would speed things up mightily, but...they keep mentioning that the two cards don't even have to be from the same manufacturer. Won't this cause an obvious quality difference between the two halves of the screen? Or is something going to force the better card to render like the worse one so they match?

Re: What happened to AGP?

[SphericalCrusher]

Nothing. AGP video cards are still running ahead in the video card preformance race. Which goes to make you think... two PCI video cards would be simply awesome, but what about two AGP video cards? How much better would two less powerful PCIs be than one powerful AGP? Just something to think about while reading this...

Re:MD tag LLY 347

[MindStalker]

Actually it was a direct copy paste from a google search (cause honestly I didn't know what the beltway was either :)
http://www.c-span.org/guide/congress/glossary/ belt way.htm

Translation

[Alex Belits]

Alienware just found out that PCI-Express can be used by gamers, so it is going to carry re-branded PCI-Express boards, and an analog monitor splitter.

Re:3dfx has done it again ...

[John Courtland]

I know how the things work man, I owned two of them. Buggy at best.

Re:Why build Frankenstein? Because you *can*!

[Zathrus]

The current implementation splits the screen in half, assigning rendering for each half to one of two cards using a software load balancer to try and ensure proper synchronization.

Ewww.

You definitely should be able to see a number of artifacts comparing the top of your screen to the bottom

You'll be able to see artifacts even with two identical cards. There will be a disconnect between the two, since things like antialiasing, ansiotropic filtering, pixel and vertex shaders, and virtually every other modern feature bases "what does this pixel look like" on what adjacent pixels look like.

It'll get really ugly for some effects too -- waves could abruptly terminate and start, fog and smoke could have different outlines, etc.

It's still boggling to me that anyone would shell out so much for a new system in order to game (my PS2 does relatively well, thank you, and even my old StarMax with Voodoo 3 plays Team Fortress quite well in 2004)

I'm not touting Alienware, but please don't compare apples and oranges. The PS2 runs virtually all games at a wonderful resolution of ~640x240 @60fps. There are a handful of games that it can do 640x480 @60fps with (requires component video cables and an HDTV). And even then it's rendering stuff that can be easily handled by a GeForce. TeamFortress is ancient... it doesn't have anything close to the graphical goodies in the upcoming games.

If you have the time, download the 700MB HL2 trailer from this year's E3 (it's 25 minutes long). We're rapidly closing in on photorealism.

Re:Why build Frankenstein? Because you *can*!

[mactari]

I'm not touting Alienware, but please don't compare apples and oranges.

I'm just saying that I can easily get my fill playing my PS2 -- and that about the only thing missing there is online fps gaming at a level I enjoy, which I easily fill by playing Team Fortress and Action Quake online. You might say that makes me not much of a gamer, but I do play a decent amount, all things considered. Why I'd need to shell out for hardware newer than my 1 GHz iMac or 2 GHz P4 Gateway laptop, I'm not real sure. I can afford to skip Doom 3 for the time being if I have to (and I bet I don't, if I put the rendering quality low enough).

So as a relatively hard-core gamer myself, I'm just not sure how Alienware creates such an audience. If I had $3k to burn, it sure as heck wouldn't be on a souped up gaming box, that's all. I'm happy we're "closing in on photorealism", but don't need to spend Alienware dollars to get there. It'll come to me soon enough at a price I can afford. Until then, the Metal Gear Solids and Madden 2005's will keep me entertained.

That said, as a technical advancement (and as a programmer myself), the dual card set-up is a pretty neat trick with the 0s and 1s! Just don't expect my credit card to vote for the advance.