Lucid HYDRA 200 Multi-GPU Technology Performance Preview

[inferKNOX]

After all the talk of nVidia blocking Lucid's Hydra and much speculation about whether there may be reasons to fear it for nVidia because of quality control reasons, we finally see some performance stats.
Well, all this of nVidia's block has been thankfully recanted, so here's a preview of what we can expect at this point from Hydra: http://www.pcper.com/article.php?aid=815

Comments?

 

[afw]

Hmmm .... it works thats the most important thing ....

but when will this technology will be made available for the users ??
eventhough nVidia refused saying it had nothing to with the MSI big bang MOBO release being delayed ... i seriously doubt that ... (its just ma opinion ... dont get me wrong guys.. ) ... nVidia is damaging its image lately ... :shadedshu

 

[pantherx12]

Nice, might grab a 8800gt or 9800 gt and slap that with a 5850 when I eventually upgrade.

 

[Deleted member 3]

I'd love to see the chip being implemented on addon cards, perhaps a videocard with the chip integrated to allow it to pair with another card. They said this would be possible, it would allow you to use basically any motherboard.

 

[InTeL-iNsIdE]

Nice, might grab a 8800gt or 9800 gt and slap that with a 5850 when I eventually upgrade.

Then you wouldnt be able to run dx11 when games start appearing without disabling hydra, which defeats the object

 

[DRDNA]

Hydra will be a benefit to everyone.....this has so much potential...looks what is does already in its early beta stage....very nice technology!

 

[pantherx12]

Then you wouldnt be able to run dx11 when games start appearing without disabling hydra, which defeats the object

I thought the 5850 would handle DX11 effects D:

 

[DRDNA]

I thought the 5850 would handle DX11 effects D:

Yes Sir the 5000 series is DX 11 ready.

 

[MilkyWay]

so with this fish n chips you can run a 8800gt for example and a 4870 and scale them?

i dont understand! can i run 2 completely different chips together and process the same thing? i thought they had to be the same graphics core?

 

[pantherx12]

That's the magic about this new tech : ]

Any gpu two GPUs will scale.

Two identical scales best how ever.

 

[Binge]

what was up with those gpu configs that did not use the same card? Totally strange numbers there. Nothing worked as well as GTX260 with GTX260.

 

[qamulek]

we need micro-stutter-begone

I just hope they have good split frame rendering algorithms down to help prevent micro stutter. I can imagine a gts250 combined with a gtx260 in alternate frame rendering would produce some wonky micro stutter. I remember trying out SFR(when it was still an option in the nvidia control panel), and I found some games would produce a line across the screen :shadedshu

 

[DRDNA]

Iand I found some games would produce a line across the screen :shadedshu

I found if you use vert sync then the line in the middle will be gone..(tearing i think)

I find it odd I have not seen any microstuddering in my builds since my X1800XTX crossfire build.

 

[grimeleven]

I thought the 5850 would handle DX11 effects D:

What he meant is that the game won't run DX11 enabled while a 8800GT (older directX) is enabled, it will scale down and the Hydra software would run the game in DX10 for example.

so with this fish n chips you can run a 8800gt for example and a 4870 and scale them?

i dont understand! can i run 2 completely different chips together and process the same thing? i thought they had to be the same graphics core?

Yes it will work, it works in a way to redirect the rendering of the scene on both GPU and the processor in the chip calculates which one of the GPU should handle the majority of the graphic load. It does so in various ways, among one it calculate the time it takes for a GPU to finish a task and "answer" back, the delay then decides which card is more powerful.

I just hope they have good split frame rendering algorithms down to help prevent micro stutter. I can imagine a gts250 combined with a gtx260 in alternate frame rendering would produce some wonky micro stutter. I remember trying out SFR(when it was still an option in the nvidia control panel), and I found some games would produce a line across the screen :shadedshu

They do not use AFR they use "Real Time Distributed Processing" http://www.lucidlogix.com/download/WP-Multi%20GPU.pdf

Check this out http://jonpeddie.com/publications/whitepapers/multi-gpu/

 

["Born"_*Leader**]

This stuff aie it da BOMB LOL

 

[wahdangun]

i hope nvdia didn't screw this time,because i can't wait to buy this hydra thing

 

[wahdangun]

do anyone know how the external PCIe bus work(like hydra demo), it's really interesting if that can be adopted to laptop, imagine you did'nt have to buy a new laptop every year just for the video card

 

[OnBoard]

I'd love to see the chip being implemented on addon cards, perhaps a videocard with the chip integrated to allow it to pair with another card. They said this would be possible, it would allow you to use basically any motherboard.

Yeah, that would be nice. If it could work with PCI-E 1x, then it would fit most motherboards and still room for two cards.

Reason being, the big bang motherboard won't be cheap and you could just use your current motherboard and get 2 identical card with the money you save.

Now you need to get that motherboard first (and new CPU and new memory) and after that you don't have any money left to get a new GPU to do NVIDIATI fusion.

Anyhow, it's nice that it works, I was already happy for X58 boards doing both stuff and then cheaper P55 boards. Soon we have a third option and even though I've never used multi-gpu, would be fun to try (if I had that motherboard already).

 

[human_error]

do anyone know how the external PCIe bus work(like hydra demo), it's really interesting if that can be adopted to laptop, imagine you did'nt have to buy a new laptop every year just for the video card

ATI and nvidia have messed with that sort of technology for years - the cost of the new GPUs with their own power supply and cooling wasn't going to sell enough units to warrent commercialising the work they did with external PCI-E graphics cards for laptops. The demo by hydra just used a standard PCI-E extension cable - you can get those in the shops for desktop mobos now, there's just not that much use for them so you don't see many people using them in desktops.

As for the hydra demo - good to see it working and still due for release - i wonder if ATI will drop the PLX PCI-E bridges on their X2 cards and replace them with customised hydra chips?