Review Index:

HYDRA Engine by Lucid - Multi-GPU Technology with No Strings Attached

Author: Ryan Shrout
Manufacturer: General

HYDRA Engine Hardware Implementation

From a purely hardware perspective, the HYDRA chip takes in a
single PCIe x16 connection and outputs two full PCIe 2.0 x16
connections.  Depending on the partner's implementation method, that
could connect to two GPUs or split into four x8 PCIe 2.0 connections
for four GPUs.  What might you find the HYDRA chip on in the future? 
There are two likely scenarios for potential designs: on a motherboard
or on a graphics board. 

On a motherboard, including a HYDRA Engine chip would allow ANY
chipset to support BOTH SLI and CrossFire technology since it is
completely chipset independent and doesn't require SLI or CrossFire
licensing.  That would enable said motherboard to offer 2-4 GPU scaling
with NVIDIA or AMD graphics cards - a VERY compelling solution but also
likely an expensive one. 

The HYDRA technology would also likely find its way onto custom
design graphics boards in place of the standard PCIe bridge - ala the
Radeon HD 4870 X2.  Lucid is claiming nearly linear scaling on up to 4
GPUs compared to 50-70% with SLI or CrossFire and thus a board vendor
could really make a top performing part and stand out from the crowd or
potential build one with slower chips for a new price-performance

I was told implementations would likely be available early in 2009 and we are eager to see what becomes of them.

As for the chip itself, obviously Lucid is being very close lipped
about it.  The chip runs very cool and draws just about 5 watts of
power.  Inside the chip you will find small RISC processor and the
custom (secret sauce) logic behind the algorithm powering the HYDRA
Engine.  The production chip was JUST finished yesterday and will be
sampling to partners soon - though they wouldn't indicate WHO those
partners were.

Questions We Asked

Here are some more details we inquired about that you are likely interested in:

How can Lucid be sure their task based distribution methods accurately represent what the game designers intended? 
An interesting dilemma - with the company essentially taking control of
the graphics pipeline there are all kinds of ways for the company to
accidently screw some things up.  Lucid answered this by telling us
their quality assurance program was already well under way.  In fact,
they use a pixel-by-pixel comparison engine comparing the HYDRA images
to a single GPU render to check for errors or problems.

What about the possibility of combining an AMD and NVIDIA GPU to work together to render one image? 
This was my biggest hope after hearing the initial introduction. 
Apparently it won't be happening though as the operating system
prevents multiple graphics drivers from running 3D applications at the
same time.  Since a 6800 and 9800 use the same driver from NVIDIA, both
may operate in 3D mode without a hassle but combining AMD and NVIDIA
just won't work.  Let's hope for the future...

How will tasks be distributed in greatly exaggerated card matchups? 
For example, in that 6800 and 9800 combination mentioned above?  With
the HYDRA Engine it is apparently very easy - the software and chip
recognize the potential for each and distribute the appropriate number
of tasks to each card.  Maybe 5 tasks to 1 or something like that; the
results are then combined by the HYDRA chip and sent to a single GPU
for output.

Will integrated graphics cores be supported? 
Nope - Lucid says that isn't in the card though I can't help but think
that it would be trivial to add and beneficial for low-cost systems. 

Can HYDRA support multiple monitors while in rendering mode?  YES! 
This was one of the cooler parts of the demo; while playing Crysis at
over 60 FPS on a pair of 9800 GTs we had a web browser window open on
the second monitor.  Even more impressive was that Lucid claimed we
could be playing 1080p video back on that other monitor and the HYDRA
system would adapt easily.  With one GPU doing HD decoding it would
simply end up with less "tasks" assigned to it as its load would be
higher than the other GPU(s) inside the system.  Scaling would be lower
of course, but the transition would be completely transparent to the
end user. 

What DirectX versions are supported or will be supported and what about OpenGL? 
now, only DX9 is working though DX10.1 will be ready by the end of the
year.  With DX10 and DX11's implementations of multi-GPU data improving
and adding to the HYDRA Engine technology will only get easier for team
compared to the work they had to do on DX9.  OpenGL is supported by the HYDRA Engine as well. 

Could this technology be applied to GPGPU work as well? 
though that is still far into the future.  One area the team did say
would be easily taken advantage of by their technology is ray tracing
with its incredibly task-oriented workflow. 

Can HYDRA really be used with ANY graphics technology? 
Yep, Lucid said you could even setup some VIA S3 Chrome cards if you really wanted too...

Questions We Have

Of course, not all is golden for Lucid quite yet - we have some
questions and concerns about the technology that we hope will be
addressed as the technology matures.  Top on my list is the support
that Lucid will be required to maintain if the technology succeeds. 
While much of the HYDRA Engine is automated there will be times when
new games, new game engines and new rendering methods will be
implemented by game developers that will require continual updating and
tweaking on the driver side of the technology.  With as large as
NVIDIA's and AMD's driver teams are, even they cannot always keep up
with the many games that are released throughout the year. 

My other major concern is that this technology could end up like
AGEIA's PhysX - great potential but gobbled up by one of the
mega-players rather than turning into a product on its own.  Honestly
after hearing the entire presentation I was curious why NVIDIA or AMD
hadn't already thought of this - the potential for being bought up is
extremely high here. 

Final Thoughts

Overall I have to say I was very impressed by Lucid and their HYDRA
Engine technology and the potential it holds for the future of scalable
graphics.  Though the current implementation supports as many as four
GPUs they say the technology can scale into the thousands and maintain
the claimed linear scalability.  Obviously we are months away from
seeing a product that we can actually test and beat on to gauge
performance and reliability of their claims so any kind of final
judgment is at least that far off.  For now I remain cautiously
enthusiastic about what HYDRA could mean to graphics in 2009.

I want your feedback on this - stop in this thread of our forums to discuss!!