What, 2 ATI cards isant enough....

[markkleb]

Check this out
http://www.dailytech.com/article.aspx?newsid=2716

Now I know why my Jetway GT4 mobo has 3 PCIX slots

[Kougar]

Yep, ATI is pushing their Crossfire + Physics setup, and they are expecting it to slowly become mainstream over the next year. If physics ever gets off the ground it should become pretty hot stuff... say have an x1900XT in Crossfire, and then a leftover x1600XT running physics in the 3rd slot. There won't be any bandwidth limiting as there is with Aegia's PCI based cards.

[GIBSON]

This does sound good. Finally a use for that old card.

[dukeman]

it does sound good but we are going to be needing 750watt psu in the near future if this keeps up. the wife is going to hate our electric bills.

[Scott]

I personally think that game developers need to start taking advantage of the dual core cpu's for physics. Why make us go buy another card when you can use the second core.

[Bio-Hazard]

I personally think that game developers need to start taking advantage of the dual core cpu's for physics. Why make us go buy another card when you can use the second core.
Why, that answer is pretty simple, MONEY, simple as that. More cards means more sales, even if the cards aren't needed, to many people just jump right on all the marketing hype...................:twisted:

[PP Mguire]

I personaly believe this is a mistake. Nobody is giving these poor PhysX cards a chance. I believe if they would hurry up and finaly release the PCI-E Ageia PhysX card and GDDR3 256mb version it will not only be non bandwidth limited but could be of some better use. You gotta look at things in a non logical kinda stupid way. GPU does not = PPU. p!=G (did i do that right?) An X1600 in theory could do physics but its only still made to do graphics and minor physics whereas something like the PhysX card is made to do heavy physics processing which today the CPU is made to handle. It take all that load off the CPU and it has nothing to do basicaly cept crunch its regular #s.
Also, just because you have a bigger PSU dosent neccesarily mean you have a bigger electric bill. Just because that PSU says 750watts dosent mean your using all 750watts constantly. Normal computer wattage is still normal computer wattage no matter how much wattage your PSU is rated.

[GIBSON]

I personaly believe this is a mistake. Nobody is giving these poor PhysX cards a chance. I believe if they would hurry up and finaly release the PCI-E Ageia PhysX card and GDDR3 256mb version it will not only be non bandwidth limited but could be of some better use. You gotta look at things in a non logical kinda stupid way. GPU does not = PPU. p!=G (did i do that right?) An X1600 in theory could do physics but its only still made to do graphics and minor physics whereas something like the PhysX card is made to do heavy physics processing which today the CPU is made to handle. It take all that load off the CPU and it has nothing to do basicaly cept crunch its regular #s.
Also, just because you have a bigger PSU dosent neccesarily mean you have a bigger electric bill. Just because that PSU says 750watts dosent mean your using all 750watts constantly. Normal computer wattage is still normal computer wattage no matter how much wattage your PSU is rated.
Thing is, with three of those cards, you will be using near the full potential of that PSU. (well, at load that is of course)

[nam-ng]

Why, that answer is pretty simple, MONEY, simple as that. More cards means more sales, even if the cards aren't needed, to many people just jump right on all the marketing hype...................:twisted:
Especially when the PPU is nothing more than a simpler, feature limited GPU. :)

[werty316]

The key word is if it will become mainstream. Maybe Cellfactor will be a turning point since GRAW wasn't.

[Kougar]

I agree, by the end of 2007 quad-cores will be around so much that they will almost be mainstream. Why not have 1 core for Vista, 1 core for the game, 1 core for physics work, and 1 core to handle sound/tcp/ip overhead, AV, and everything that is left over or running as a background process? Hell, or even folding@home while you are gaming... :)

I'd have to disagree that a ATI GPU is "just a GPU" though... ATI is turning their GPU card into a massively parallel, simple, processor that can offer some serious multiple crunching, because it is becoming more and more programmable. This is why the Folding@Home people are finally on the verge of releasing a F@H client for ATI graphics cards.

At a glance Aegia's physical PPU unit doesn't appear to be more than a lightly modified and optimized low end GPU core such as for instance a Chrome GPU, it doesn't look highly specialized for, or that much different, then a low-end GPU core.

[PP Mguire]

Yea, but you would have to be playing a game 24/7 to crank power through all 3 of those card all the time.

[Kougar]

Yea, but you would have to be playing a game 24/7 to crank power through all 3 of those card all the time.

Why not four instances of F@H on your quad-core CPU, and then three more instances on your three ATI cards? ;) According to the F@H people a x1900XT class card would fold circles around any high-end CPU...

[GIBSON]

This is why the Folding@Home people are finally on the verge of releasing a F@H client for ATI graphics cards.
Only for ATI cards? Man, that would suck.

[Kougar]

Only for ATI cards? Man, that would suck.

Afraid so, nVidia didn't see the need to build features into their GPUs that aren't being used, and while I have heard repeatedly they are working on their own kind of "physics" for their own GPUs, that's as far as I know.

ATI has slowly been expanding on making the various parts of their GPUs more programmable and accesible for non GPU rendering uses, such as folding. :) nVidia just hasn't done this, hence why there will likely not be F@H support on nVidia hardware for a pretty long time. It's taken ATI from the x100 series to the x1000 series, with some preliminary work having been started during the 9x00 series even to get to this point with making their GPUs "programmable". That's the best I understand or can explain it anyway. Here's the latest info from Stanford's website on it:

August 2006

We're running on the cluster and starting to ramp up. Vishal's new method to build cores should yield a FAH core soon, but it will take a lot of Q&A (GPU folding is much more sensitive to machine configuration) so it will still take some time before we release the core. It's pretty exciting for us to see the calculation racing away on the GPU cluster! Prof. Pande also formally announced FAH's GPU client at his award lecture at the Protein Society meeting, with a great response for the GPU work. We will be making a more formal announcement of the GPU client and some other surprises relatively soon. Note the update above regarding GPUs -- we currently plan to support only recent generation ATI GPUs (see above).

Source (http://folding.stanford.edu/FAQ-highperformance.html)

[PP Mguire]

Why not four instances of F@H on your quad-core CPU, and then three more instances on your three ATI cards? ;) According to the F@H people a x1900XT class card would fold circles around any high-end CPU... Oh my post was for a previous power supply post. I thought i hit quote but i guess not. Was really tired last night.

[zachig]

I personally think that the AGEIA Physics Cards will eventually become mainstream, we just need to give it a chance...

As currently only few games support it, I think it will become much more popular and demand will increase during year 2007, when much more games will support it. :roll:

Anyhow, it's good to see some boards that support both Crossfire and AGEIA (like the Jetway mobo that was mentioned in the first post in this thread). :grin:

Oh...it was also mentioned that we'll need a very strong, high wattage PSU for future Graphics/Video/Physics cards...I guess this is true, as we already start to see some 900W and 1000W PSU...the only problem at the moment is that these PSUs cost a fortune nowadays, but when the demand for such PSUs will increase, I'm pretty sure prices will drop down drastically...:wink:

And if we're talking about super-demanding Video Cards, here another example:
ACCORDING TO PC WATCH, the next round of GPUs are going to be really smoking, not because they are fast, but because they suck power like it is going out of style.

If you read the article (see link below), you will see the damage that will be done to the world's petroleum reserves.
The short story is this, G80 form Nvidia is looking to suck down 175W on a 90nm process, probably a bit less on 80nm. R600 from ATI is looking to take about 200W in 80nm, and less if they pull off a shrink to 65nm. Either way, this is simply too much. The current 100+W is borderline insane, and adding half again to it is just silly, speed is OK, but I think we have just passed the point of being ludicrous. When you are at more than twice the power of the CPU, it is time to hit yourself on the head with the same high velocity cluestick that AMD and Intel have already experienced
Click HERE (http://translate.google.com/translate?u=http%3A%2F%2Fpc.watch.impress.co.jp%2F docs%2F2006%2F0727%2Fkaigai291.htm&langpair=ja%7Cen&hl=en&ie=UTF8) to read the FULL (translated) article.

[Kougar]

First problem with that article is nVidia has officially stated they will NOT be die-shrinking their GPUs to the 80mm node. They will be nice and hot at 90nm for a awhile longer...

ATI is at least going to build the new DX10 cards with the new 80nm die-shrink, so their cards might actually use less juice than nVidia's for a change... :lol: They also did state 65nm was their goal for sometime in 2007. Info from various ArsTechnica's articles.

[zachig]

Yeah...I think that ATI are going to top over nVidia, at least for the near future and especially with their new X1950 to come soon...:roll:
I think that ATI has a big advatage over nVidia after being purchased by AMD...:wink:

[DragonMaster]

This is why the Folding@Home people are finally on the verge of releasing a F@H client for ATI graphics cards.

THAT's a flexible GPU.

Why can't we use a simple RD482 chipset and use one GPU for physics, and one for graphics?

[nam-ng]

THAT's a flexible GPU.
It did improved F@H once ATI drop the stupid 24 bit floating point hack on the 9800s, and used the same standard floating point nVIDIA was using on their newer 1000 series graphic hardware.
Why can't we use a simple RD482 chipset and use one GPU for physics, and one for graphics?
Ofcourse you can, but not at all efficient. The first thing is the stupid Multi-Processing Dongle hack needed to be rid of, Crossfire's Multi-Processing needed to be more SLI like nVIDIA through the PCI-Express Scalable Link Inteface.

And even better if Crossfire also had the same thing like nVIDIA's Scalable Link Interface called the MIO-bridge which would really aleviate bandwidth problem for inter-processor dependency ops.

Since everything then used Scalable Link Intefaces, programmers wouldn't have to write multiple hack code versions for different connections like the stupid Multi-Processing Dongle.