Msi P55 The Big Bang - Lucid Hydra Inside - Ati Work With Nvidia

[Jamesrt2004]

man this wont work mixing cards as.....

nvidifail have blocked nvidia cards working if it "sees" a ati card, so only if the lucid chip can get around that will id do that....

but imo the interesting thing will be the scalling of x-fire and sli now... as you will not be able to mix cards (still pissed at nvidia about that..)

we shall see as we still don't know what version of the chip it's using

[trinibwoy]

I wish I could go back to all the topics where people proclaimed "rabble rabble rabble, this will never happen! You're a fool if you think it will!"

I got called quite a few nasty names for my faith. Guess it pays off.

Heh, I wouldn't be so quick to celebrate. Wait until you see if it actually works first

Jon Peddie Research did a little paper on multi-GPU/Lucid and they seem to think it's the second coming but even they are just going on Lucid's promises and not anything concrete.

http://www.jonpeddie.com/special/Whi...ortunities.pdf

[gatecrasherlok]

Yeah I need to actually see it fullyy working with a few different graphics card configurations along with some test results and so forth.

[OBR]

No, it can't.
Hydra 100 has 1x16 upstream port and 2x16 downstream ports. So possible configurations are:
x16+x16 or x16+x8+x8 or x8+x8+x8+x8.
http://www.lucidlogix.com/files/hydr...duct_brief.pdf

This is not Hydra 100, but new improved version (maybe Hydra 200?)

[Tjalve]

First of guys. Hydra wont run Crossfire or SLI. It will (should) run 2 or more graphics cards at the same time for higher performance.
SLI and Crossfire works mostly by letting each GPU work on a singel frame. It can also be done by drawing half the image one one GPU and the other half on the other GPU. The problem with this is that ypu need 2 GPUs with exactly the same performance. And the first GPU needs to wait for the other GPU to finish befor they start on the next frame.

Hydra uses some kind of advanced algorithm to analyze each frame and devide the work on all the GPUs in the system, based on how fast the GPU can put out images. Witch means that if ou pair a GTX 280 with an 9500GT, the 280 will do most of the work in the frame (liek light, water and other shader intesive parts of the image). The slower GPU will render the less intesive parts of the frame. Both images will then be joined togheter to form a complete frame. With two identical GPUs this would mean that the work will be devided equally and scalling will (in theory) be 100%. It also means that there are no obsticals that prevent you from using an ATI and a Nvida card in the same system. However there will be driver issues since both GPUs needs to use there seperate drivers and settings. Another good thing is that if you have 2 idetical cards, and you overclock one of them, you will se an overall performance increase without oveclocking the other card. The overclocked card simply takes more of the work and the frame will be completed faster.

I just cant wait to see how will it performs in reality

[Jamesrt2004]

First of guys. Hydra wont run Crossfire or SLI. It will (should) run 2 or more graphics cards at the same time for higher performance.
SLI and Crossfire works mostly by letting each GPU work on a singel frame. It can also be done by drawing half the image one one GPU and the other half on the other GPU. The problem with this is that ypu need 2 GPUs with exactly the same performance. And the first GPU needs to wait for the other GPU to finish befor they start on the next frame.

Hydra uses some kind of advanced algorithm to analyze each frame and devide the work on all the GPUs in the system, based on how fast the GPU can put out images. Witch means that if ou pair a GTX 280 with an 9500GT, the 280 will do most of the work in the frame (liek light, water and other shader intesive parts of the image). The slower GPU will render the less intesive parts of the frame. Both images will then be joined togheter to form a complete frame. With two identical GPUs this would mean that the work will be devided equally and scalling will (in theory) be 100%. It also means that there are no obsticals that prevent you from using an ATI and a Nvida card in the same system. However there will be driver issues since both GPUs needs to use there seperate drivers and settings. Another good thing is that if you have 2 idetical cards, and you overclock one of them, you will se an overall performance increase without oveclocking the other card. The overclocked card simply takes more of the work and the frame will be completed faster.

I just cant wait to see how will it performs in reality

+1

but the word on the street is lucidd is makign custom drivers

[xman01]

pack of rubbish
4870+4650 rendering the same game lol

[Smartidiot89]

pack of rubbish
4870+4650 rendering the same game lol

why not?

[proxis]

imagine the 3DMark01 fun ... dragos with ATI and nature with nVidia :p

[Dragy2k]

so how does this work with nvidias newer win7 drivers to disable mixed cards ??

[zalbard]

If one has some issues with drivers, they will have twice more using this chip... I'd avoid mixing ATi and Nvidia cards unless you absolutely have to.

[trinibwoy]

I still find it funny that people believe that Lucid can just stick a chip on a motherboard and make Nvidia's and ATI's chips work better than the designers can Can't wait for the first reviews of this thing. I'll be just as impressed as anybody if it works but right now it sounds like a pipedream.

[xman01]

why not?

why ati didnt implement it b4?
why are there limitations to xfire and sli

this magic chip supposed to fix that?

[XSAlliN]

If the Hydra Engine works as advertised, both Crossfire and SLI are useless in these boards.

Damn.. This came a lot faster than I thought.
I just hope a 790GX version is released, so I don't have to change CPU also. I even wonder if an IGP could lift the overall performance, even if just a little bit.

It's not just that, you can mix a Mainstream Card with a High End one based on your needs and it's suppose to be a better alternative to Sli/CF.

[>HyperlogiK<]

I thought that ELSA was already selling boxes for render farms that used Hydra 100 or did these get delayed/cancelled?

[OBR]

asked MSI and this is only prototype ... not for sale in first and second wave of P55 boards ... but everything can be changed.

[JAG87]

why ati didnt implement it b4?
why are there limitations to xfire and sli

this magic chip supposed to fix that?

Because from a marketing point of view that's complete failure. And you fail to realize that ATI and nvidia exist for the sole purpose of making money. What good could come from ATI and nvidia investing money and research time to get xfire/SLI working with mixed and matched configurations?

Option A: you can mix ATI with nvidia, so now you go and purchase the competitor's product. FAIL for both companies.

Option B: you can increase the performance of your current 4890/GTX 285, by throwing in your old 3870/8800 GT, rather than being forced to buy ANOTHER 4890/GTX 285. Once again, FAIL.

That's why you need a third party to come in and develop such a thing. And it seems they have found the best way (distributing directx/opengl code rather than distributing frames) to allow cards of different performance levels to scale almost perfectly. Obviously the same technique would work with cards of the same performance level and should show linear scaling.

Whether it works as advertised or not is the question. Will we see terrible micro or macro stutter, will every game engine be supported, and so on.

[Smartidiot89]

why ati didnt implement it b4?
why are there limitations to xfire and sli

this magic chip supposed to fix that?

Because from a marketing point of view that's complete failure. And you fail to realize that ATI and nvidia exist for the sole purpose of making money. What good could come from ATI and nvidia investing money and research time to get xfire/SLI working with mixed and matched configurations?

Option A: you can mix ATI with nvidia, so now you go and purchase the competitor's product. FAIL for both companies.

Option B: you can increase the performance of your current 4890/GTX 285, by throwing in your old 3870/8800 GT, rather than being forced to buy ANOTHER 4890/GTX 285. Once again, FAIL.

That's why you need a third party to come in and develop such a thing. And it seems they have found the best way (distributing directx/opengl code rather than distributing frames) to allow cards of different performance levels to scale almost perfectly. Obviously the same technique would work with cards of the same performance level and should show linear scaling.

Whether it works as advertised or not is the question. Will we see terrible micro or macro stutter, will every game engine be supported, and so on.

JAG87 summed it all up.

I thought that ELSA was already selling boxes for render farms that used Hydra 100 or did these get delayed/cancelled?

I believe they have been for a few months now?

[>HyperlogiK<]

I believe they have been for a few months now?

I can't find reviews, but I presume that must be working somewhat well for a major manufacturer to put out an enterprise product based on it.

Though perhaps that is the most naive thing I'll say all day.

[Chumbucket843]

i would like to see >4 gpu's with this or would that be too much to handle.

[zanzabar]

I still find it funny that people believe that Lucid can just stick a chip on a motherboard and make Nvidia's and ATI's chips work better than the designers can Can't wait for the first reviews of this thing. I'll be just as impressed as anybody if it works but right now it sounds like a pipedream.

at the bare minimum it cant be a worse than the nf200 for slot control

[Xoulz]

Ans the wait is finally over.... (my sig for last 8 months)

87% scaling?

[STaRGaZeR]

Ans the wait is finally over.... (my sig for last 8 months)

87% scaling?

This is nothing more than a PCIe bridge for now, and the mobo may not see the light of day

[Hornet331]

First of guys. Hydra wont run Crossfire or SLI. It will (should) run 2 or more graphics cards at the same time for higher performance.
SLI and Crossfire works mostly by letting each GPU work on a singel frame. It can also be done by drawing half the image one one GPU and the other half on the other GPU. The problem with this is that ypu need 2 GPUs with exactly the same performance. And the first GPU needs to wait for the other GPU to finish befor they start on the next frame.

Hydra uses some kind of advanced algorithm to analyze each frame and devide the work on all the GPUs in the system, based on how fast the GPU can put out images. Witch means that if ou pair a GTX 280 with an 9500GT, the 280 will do most of the work in the frame (liek light, water and other shader intesive parts of the image). The slower GPU will render the less intesive parts of the frame. Both images will then be joined togheter to form a complete frame. With two identical GPUs this would mean that the work will be devided equally and scalling will (in theory) be 100%. It also means that there are no obsticals that prevent you from using an ATI and a Nvida card in the same system. However there will be driver issues since both GPUs needs to use there seperate drivers and settings. Another good thing is that if you have 2 idetical cards, and you overclock one of them, you will se an overall performance increase without oveclocking the other card. The overclocked card simply takes more of the work and the frame will be completed faster.

I just cant wait to see how will it performs in reality

And how you want to tell the gpus what to render.. that would need massive support by ati/nv, or another possibility would be, that there would be a layer of software between dx/ogl and the gfx driver that calculates what high workload is and what not, then split it up, distribute it and rejoin the frames and then send it to the dx/ogl rendere... would need quite a bit calculation power...

Right now thats nothing more then a pipe dream... as already mentioned by some.

[Xoulz]

This is nothing more than a PCIe bridge for now, and the mobo may not see the light of day

That^^ has not been substantiated or confirmed... matter-of-fact, that chip looks alot different than the two previous they showed a year back.

Lucid didn't market their H100 chip, because it would only work on DirectX9 and Lucid already knew DX11 was in the works...

Given the talent, board of members and inverstors... I'm sure they are working on a mature version of this infant technology. They have something like 50+ patents.. and have no need to hurry. Plus, Larrabee isn't out yet and that's when this chip will make the most sense.

RTFM...



Coincidentally, could this be the reason nVidia chose to announce their PhysX restriction the other day...?