Review Index:
Feedback

Lucid HYDRA 200 Multi-GPU Technology Performance Preview

Author: Ryan Shrout
Manufacturer: General
Tagged:

Analysis, Stability, Future plans and Closing Thoughts

Stability and Software Concerns

Now that we have looked at performance we have to note our
experience during the whole testing process.  In general, the HYDRA
Engine software and tool was a breeze to work with and
enabling/disabling the software worked seamlessly and without a hitch. 
I could switch modes without a need to reboot and if you wanted to
disable HYDRA and enable CrossFire instead in the ATI control panel,
you could do that as well.  Overall I think the software is fairly
mature and should be ready to find its way into the hands of
enthusiasts.

There were a couple of hiccups to note though; in our multi-vendor
testing some of the DX10 titles had some significant coloring issues
that didn’t show up for other media that had used the same
demonstration machine.  Some simple reinstalling of ATI/NVIDIA drivers
seemed to fix the issue but it did revert one time.  Obviously this is
something Lucid needs to work out before they unveil the official
multi-vendor driver this winter, but for an early beta it was workable.

I was also a bit let down to see the software lists that existed in
terms of what software Lucid was comfortable with us benchmarking. 
Batman, for example, wasn’t on the list.  That being said, the team
told me over and over that this was only intended to be an “early look”
at the HYDRA technology and performance and that they wanted to save
some goodies for the final release of reviews this fall/winter.  I
found that, as of my meeting with them, over 55 games had been
qualified by their team to work very well with HYDRA technology.  That
may not seem like a lot of games, compared to the hundreds and hundreds
for SLI and CrossFire, but for a small team in a short period that
number is acceptable.

Performance

Based on the performance results you saw above I have to say that
our initial impressions of the HYDRA technology are very positive.  And
yes, before I get the countless emails, I realize we only tested a
handful of games on a handful of graphics cards and Lucid and its
partners still have a lot to prove to make this technology truly
consumer ready.  But for just an afternoon, my general impressions of
the type of scaling Lucid’s HYDRA was providing impressed and should be
able to give both SLI and CrossFire a run for their money.



MSI's upcoming "Big Bang" motherboard

Identical GPU scaling performance of 80% doesn’t quite hit the term
“linear” that was thrown around during Lucid’s development but in my
experience that is better than SLI and CrossFire native configurations
in every case we have tested.  And since we don’t have any other
options to compare it to, the different-GPU scaling of around 50% of
the “slower” card also seems pretty reasonable based on the additional
overhead.  The days of throwing away your “old” graphics card, even of
a different graphics vendor, might soon be over.

Lucid did point out that in terms of same-GPU HYDRA versus
SLI/CrossFire comparisons, there will be some cases where HYDRA is
faster and other cases where the NVIDIA or ATI native solution is
faster by about 10-20% in either direction.  But obviously neither
NVIDIA nor ATI are doing different-GPU scaling at all, so Lucid will
definitely see the advantage there.  Also, Lucid pointed out some newer
games that did not yet have any SLI or CrossFire scaling with the
latest WHQL drivers from NVIDIA and ATI – we have seen this over the
years where it takes time for the GPU vendors to release updated
drivers and profiles to get GPU scaling working.  In those cases,
because Lucid’s software doesn’t require profiles in that sense, the
HYDRA solution provided scaling when native options could not.  If
Lucid’s software team lives up to its claims it could mean
day-of-release scaling for any PC title on HYDRA-enabled systems. 

Other Notes – Random information I couldn’t figure out where to place

As of this writing, dual GPU graphics cards like the GeForce GTX
295 are not going to work with Lucid’s HYDRA technology in the way you
want it to.  Essentially, because the two GPUs behind the PCI Express
bridge chips on the graphics boards (both NVIDIA and ATI) are “hidden”
to the system, the HYDRA driver will only be able access one of them. 
This is a bit of a letdown for us as the idea of having a single GTX
295 and then throwing in an HD 5870 sounded very appealing, it doesn’t
appear Lucid will be able to get that working. 

What my theory is, and I hope to test soon, is that you might be
able to enable SLI on the GTX 295 and then enable HYDRA on the “two”
graphics cards in the computer then.  (The SLI’d GTX 295 being ONE card
and the HD 5870 being the second.)  Sounds like a recipe for awesome or
disaster. 



On the "Big Bang" motherboard the Lucid HYDRA 200 and P55 chipset sit side by side.

Some users will also want to know why NVIDIA or ATI did not take
this route on their own years ago with scaling technology.  Both
companies have been fixed in their reasoning that you could not load
balance GPUs of different performance (with only a few very minor
exceptions) in any way that was acceptable to them.  Obviously the
amount of benefit you get from the “slower” GPU would be less than when
paired with identical GPUs (we saw the 80% of GTX 260s fall to 50% or
so with different-GPU cards) but surely even those big boys can see the
benefits of “something is better than nothing” right?  It wouldn’t
surprise me if either of the major GPU vendors suddenly found
themselves opening up the options for more GPU combinations in the near
future.  One thing you still will never find from that source though is
the ability to use cross-vendor combinations – Lucid will have the
advantage there forever.

Finally, Lucid is pursuing other avenues for their HYDRA technology
than just gaming.  Think corporate and high performance computing usage
models and very high profit margins.  If you are in to that kind of
thing keep your eyes peeled on the upcoming supercomputing conventions…

Future Plans

The immediate future for Lucid will be to get an actual product
from one of its partners out the door and to get better drivers and
game support for those early adopting consumers.  To do otherwise would
be corporate suicide and would alienate any good will they have built
with their unique and impressive technology.  We have to take them at
their word that they will accomplish this for now but it is something
we can directly measure as 2010 progresses. 

The CEO did talk to us about some other very interesting uses of
the HYDRA technology for the consumer spaces besides standard game
scaling.  One demo we worked with showed a dual-monitor configuration
on the HD 4890 / GTX 260+ setup (one monitor for each GPU) with HD
video playback going on the GTX 260 driven display.  While that was
running, a benchmark iteration of Call of Juarez could be run that
would still show performance gains (though slightly lower than with out
the HD video playback) over the results from ONLY the HD 4890. 
Essentially the HYDRA engine is able to tap into the GPU’s power even
if it is being utilized by another application; this could mean great
things for Folding@Home users or gamers that want to use some
GPU-encoding applications while gaming at the same time. 



Are you a no-good-dirty cheater?

Another feature the team has discussed but that is in the very
distant future are unique implementations of secondary rendering. 
Imagine two monitors hooked up to your computer with the primary one
showing the normal screen for your Modern Warfare 2 game while a second
display shows you the reverse angle looking behind you.  Sure, it’s
probably going to be called cheating by most but the technology sounds
impressive.  Simply put, because the HYDRA software engine has to
interpret DirectX data to work its scaling magic already, it has access
to the entire spectrum of 3D data in the game at any given time and
could, in theory, render other angles of the game that the developer
did not intend.  This could be used for rear view cameras in racing
games, showing HUD-only information on a secondary panel or even
zooming in on your gun sights.  The potential is impressive though the
moral implications are as well.  And my guess is that applications like
PunkBuster would interpret these types of actions as illegal anyway. 
Maybe a question for us in late 2010…

Closing Thoughts

Our time with Lucid has grown longer and longer with each meeting
and we were glad to FINALLY be able to go hands on with the hardware
and be able to report on it.  It was tough getting a glimpse of the
scaling power of the solution at IDF this past fall and not tell all of
our readers about it!  From the beginning we knew that what Lucid was
doing had the potential to be revolutionary to both the consumer and
the graphics industry as a whole and as we get closer and closer
(release it already!) to retail availability it seems that I am more
confident that this will indeed be the case. 

There are hurdles though – the most dire of which is likely the
technology itself.  Can the Lucid software and QA team, obviously
orders of magnitude smaller than those at NVIDIA or AMD, keep up with
the new PC gaming titles and graphics hardware as it is released? 
Lucid likes to say their driver is completely software and hardware
agnostic, but we know that is simply not the case otherwise we could
have tested ANY game and ANY graphics cards in our time with them this
week.  There are going to be bugs and glitches and missing textures and
performance issues to deal with for the life of the technology and
Lucid has to prove to the community, and to partners like MSI, that
they are willing and able to keep their end of the bargain. 

It would also appear that another hurdle is other entities in the
world of technology now as well.  MSI has delayed the Big Bang
motherboard, the first product offering to include the Lucid HYDRA chip
on it, and rumors persist that NVIDIA was at hand in the reasoning for
it.  NVIDIA would likely be pressured then to quit the SLI licensing
program completely if HYDRA is a success as any motherboard vendor
could just include the Lucid technology and allow gamers to run
dual-GPUs that way instead, but with MUCH improved flexibility. 
Obviously the HYDRA option is going to be more expensive and
complicated to integrate but how much so has yet to be seen.

After my day with the team all I can say is that I am going to
start preparing our labs for some in-depth analysis of the Lucid HYDRA
multi-GPU scaling technology very soon.  There are going to be
limitless combinations of potential testing scenarios, and though we
can’t possibly hit them all, we are going to hit as many as possible so
that we can get a final verdict on whether or not the solution is ready
for consumers.  Thus far I think Lucid is on the right path and the
early look we have shown you today proved that the technology is real,
working and potentially industry rattling.

Related Reading:

No comments posted yet.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.