Lucid Logix HYDRA tech brings together any GPUs for powerful matrimony

There's a lot of tech speak to cut through, but if Lucid Logix's claims are true, we might be looking at a revolution in the high-end gaming segment. Lucid is building a new "real-time distributed processing engine" system on a chip called HYDRA, which can mix and match any GPU from any manufacturer and work with any chipset, and piles it all together for performance scaling that Lucid claims is "near-linear" or "above-linear." It's the above-linear part that particularly makes no sense -- how can you squeeze more power out of cards than is there to begin with? -- but we're gonna give Lucid the benefit of the doubt for the moment and wait for the benchmarks. HYDRA is slated to come to market in the first half of 2009.

[g00n @ Jul 14th 2008 12:59PM]

Ya, Ya, but can it play Crysis?


lol

[Mr. Fax Sender @ Jul 14th 2008 1:00PM]

Most definitely, son. Most definitely.

[Pochi @ Jul 14th 2008 1:31PM]

LOLOL!!!!!!

[Arjy @ Jul 14th 2008 1:01PM]

4870x2+GTX280?

[octoberasian @ Jul 14th 2008 1:16PM]

If this is real, the combined performance of 4870 X2 and GTX 280 combined would be staggering or any other power GPU released to date.

You get the best of both worlds from Nvidia and ATI-- each one complementing their strengths and weaknesses.

Let's hope it's affordable or if any motherboard manufacturer actually implements this.

[From My Cube @ Jul 14th 2008 1:25PM]

I feel like such a nerd...knowing thats a math problem involving three gfx cards and not a math problem involving three variables

[PCIV @ Jul 14th 2008 5:01PM]

Not a nerd... the 4870 X2 is just one card. SHAME!

[Tom @ Jul 15th 2008 12:06PM]

@PCIV:
You've never taken a math class, have you?

[Tom @ Jul 15th 2008 12:10PM]

I hereby retract my premature statement. It was you who read things correctly, PCIV, and obviously not me.

However the 4870x2 card has two GPUs, which could have led to Cube's statement. I'm going to assume he knows what he was talking about, just used a poor choice of words. Maybe he hit "Add Your Comments" as he realized what was really going on, like I just did.

[Juaquin @ Jul 14th 2008 1:03PM]

Yeah...above-linear? Doubtful, unless the chip itself is doing some processing too. Ten bucks says it's going to fail if they're already making claims like that.

[Chris Macdonald @ Jul 14th 2008 9:20PM]

Man, it uses that new flux capacitor technology, it goes ahead about 4 seconds in time to see what the game is going to do next.

[TRAFFICBLOWS @ Jul 15th 2008 9:41AM]

let's see....

above linear = upconverting

[Ekaf Eman @ Aug 20th 2008 7:16PM]

Super linear speedups are theoretically possible when you take into account the increased aggregate memory bandwidth. Such superlinear speedups can be routinely seen in MPP running embarassingly parallel application (though in those cases it is often due to larger aggregate cache size).

I agree with your underlying skepticism, however.

[Kurian @ Jul 14th 2008 1:19PM]

I can't begin to imagine how this works. I mean, most cards are like hardwired to send the output of a lot of GPU functions to the display ports etc.

I reckon the only way is through interfaces like CUDA etc, which will be EXTREMELY inefficient at processing graphics data and returning it back to the CPU instead of outputting it via the display port.

[maveric101 @ Jul 14th 2008 7:49PM]

will this be another chip on the motherboard? or a PCI(e) card?

[kjb434 @ Jul 14th 2008 1:20PM]

If this proves out, video card fanboy BS will be just that: BS.

[rick @ Jul 14th 2008 1:23PM]

For above linear performance:

Maybe an intelligent routing algorithm which plays to each cards strengths while avoiding its weaknesses? Would that slightly increase the performance of each card and be considered above linear?

Or is it comparing the cards all daisy chained together? (See how this solution works vs the blank screen you get when they are hooked up incorrectly)

I'm not saying it can be done (or done well) in realtime.

[Harlan @ Jul 14th 2008 1:35PM]

Perhaps the above-linear involves caching things that'd have been duplicate efforts?

[Ted Capiendo @ Jul 14th 2008 5:36PM]

above linear performance is possible, but only under some special conditions involving scheduling, process priorities, etc. I remember studying this in some comp sci class, but it was so long ago...

[Insider @ Jul 14th 2008 1:24PM]

Better than linear is hard to achieve but in the CPU world keeping one processor purely in decoding and another in encoding is better than both switching logic between encoding and decoding. Processors are smart and that cache close to the CPU (or GPU in this case) certainly like seeing the same math being done over and over again...

[birdoprey @ Jul 14th 2008 1:26PM]

Are they the next Bitboys Oy?

[Razor @ Jul 14th 2008 6:02PM]

Man, I thought I was the only person who still remembered those guys =P

[iptydafu @ Jul 14th 2008 1:32PM]

So is this basically making a cell-processor out of a bunch of old graphics cards?

[IT-Accountant @ Jul 14th 2008 1:54PM]

Too good to be true, my friends. I'm calling vaporware on this one.

[Za @ Jul 14th 2008 3:15PM]

Well, you don't have to *call* it vaporware for it to be so. Vaporware just means they have claims not currently backed by demonstrable hardware. That is the case here. Thank you for *calling* it. Also, the sky in New York is currently blue, with some clouds.

[broli @ Jul 14th 2008 1:55PM]

If you're smart buy as many stocks as you can of this company when it joins the stock trade.

[wrabbit @ Jul 14th 2008 1:59PM]

Wow wow, hold on there a second! You telling me I'm gonna have to deal with both nvidia and ati drivers - meaning double the bugs and double the headaches? :)

[udif @ Jul 14th 2008 2:02PM]

You can achieve above-linear speedup in the rare cases where breaking a job to multiple processors allows you to have simpler code for each processor (i.e. split the code such that each card does different functionality that is only part of the unified code) that just fits in the cache, while the single processor version is larger than the cache, causing thrashing and slowdown due to cache misses.

[Jamo @ Jul 14th 2008 2:07PM]

I see this as being particularly environmentally friendly. Think, when you upgrade your video card (which for hardcore gamers is fairly frequently) you will be able to grab the processing power of your new card as well as that of your old one. This is a decent solution to complete obsolescence, whereby your old card might have ended up in the trash.

[broli @ Jul 14th 2008 3:07PM]

Yes because when I game that's all I think about...the environment.

[Kyle @ Jul 14th 2008 2:38PM]

Will this be in the form of software, hardware or what? Hopefully if its hardware its like PCI or something, I wanna have all my PCIe spots saved for my outdated cards that can led themselves to a bit of extra power!

[ShadowKain @ Jul 14th 2008 3:38PM]

This makes me think of times when Nvidia fans wanted to mix an older 8800 GTS G80 based chip with the newer G90 (or G92 is it?) chip. You can't mix chips. After reading through their site, it will be interesting if this works, but I won't keep my hopes high...

[Amrosorma @ Jul 14th 2008 3:57PM]

"It's the above-linear part that particularly makes no sense -- how can you squeeze more power out of cards than is there to begin with?"

Emergent properties! It's the way of the future!

[no1youknowz @ Jul 14th 2008 4:37PM]

Wasnt this already done with Alienware?

[madgamer @ Jul 14th 2008 3:33PM]

Not sure why you are downrated as it was a legit question.. Alienware did have a board that could do SLI or Crossfire, but it couldn't do both between different cards. It was one or the other.

[no1youknowz @ Jul 14th 2008 5:02PM]

Again, didnt Alienware do this?

http://www.xbitlabs.com/news/other/display/20040519040727.html

Thats right. Thing is, I been around pre the 3DFX days (yes S3 and Virge) and can remember odd bits here and there!

Anyway, Im sure Alienware managed to get Nvidia and ATi cards to work in tandem in an SLI mode.

But whatever, the majority here im sure are joe six pack and only think that Crossfire and SLI is where its at (rolls-eyes)!

[Agent42 @ Jul 15th 2008 1:03AM]

The solution is that you could run ether an SLI setup OR crossfire.

You couldn't run two cards of different brands in parallel

[cjwild @ Jul 14th 2008 5:08PM]

"Maybe an intelligent routing algorithm which plays to each cards strengths while avoiding its weaknesses? Would that slightly increase the performance of each card and be considered above linear?"
-rick


I was hoping the same thing, does anyone know when the hell this is going to be ready to do some real benchmarks?

[Conor Maher @ Jul 14th 2008 5:18PM]

M-M-M-m-m-m-m-m-Microstuttering!

It might work for things like CUDA but it will fail at games because of latencies etc

[no1youknowz @ Jul 14th 2008 5:18PM]

E3 2004: Alienware Video Array and X2

http://uk.gear.ign.com/articles/515/515363p1.html

"The scheme involves two video cards -- they don't have to be identical cards -- and a tertiary PCI card. The outputs from the video cards run to the third card, which sends the signal to your monitor. Basically, each card renders one-half of the scene and then throws it to the array card, which assembles the final image."

There you go.. pure ownage.

Next time low modders, try and validate your info before you mark up/down. ;)

[icingdeath88 @ Jul 16th 2008 1:54AM]

Good memory/research skills.

But, looks like it never came to market. Must not have worked out as well as expected. Not surprised seeing as how ndivia and ati are still having a hard time getting sli/crossfire to scale well, despite having spent tons of time and resources on it. cool that they tried it anyway.

I would like to read more about exactly why it never made it, though.

[jay.viz @ Jul 14th 2008 5:45PM]

I would like to see this work on Vista, considering the new WDDM breaks mixed graphics drivers. They could make money off of something that would just let the different cards operate at the same time, let alone together.

[Amir @ Jul 14th 2008 6:02PM]

will it blend?

[albi @ Jul 14th 2008 6:16PM]

ok, so any chipset, any gpu, ...but what kind of monitor does it take?

[nutrigrainbar04 @ Jul 14th 2008 7:47PM]

Progression of compatibility is always nice to see.

[Darwin @ Jul 14th 2008 7:53PM]

Didn't ATI already use the Hydra name commercially?

[Rami Zamsky @ Jul 16th 2008 1:22AM]

I saw some scepticisem in the comments from people who knows very little about the gaming industry or they are only "just users". Well I saw it in operation and I believe that next year 2009 the market will enjoy the product. I agree with the one who recommanded to buy this company shares when they will be floted.