Where AFR Is Mediocre, and How Hydra Can Be Better
Perhaps it’s best that we first start with a discussion on how modern multi-GPU linking is handled by NVIDIA and AMD. After some earlier experimentation, both have settled on a method called Alternate Frame Rendering (AFR), which as the name implies has each card render a different frame.
The advantage of AFR is that it’s relatively easy to implement – each card doesn’t need to know what the other card is doing beyond simple frame synchronization. The driver in turn needs to do some work managing things in order to keep each GPU fed and timed correctly (not to mention coaxing another frame out of the CPU for rendering).
However even as simple as AFR is, it isn’t foolproof and it isn’t flawless. Making it work at its peak level of performance requires some understanding of the game being run, which is why for even such a “dumb” method we still have game profiles. Furthermore it comes with a few inherent drawbacks
- Each GPU needs to get a frame done in the same amount of time as the other GPUs.
- Because of the timing requirement, the GPUs can’t differ in processing capabilities. AFR works best when they are perfectly alike.
- Dealing with games where the next frame is dependent on the previous one is hard.
- Even with matching GPUs, if your driver gets the timing wrong, it can render frames at an uneven pace. Frames need to be spaced apart equally – when this fails to happen you get microstuttering.
- AFR has multiple GPUs working on different frames, not the same frame. This means that frame throughput increases, but not the latency for any individual frame. So if a single card gets 30fps and takes 16.6ms to render a frame, a pair of cards in AFR get 60fps but still take 16.6ms to render a frame.
Despite those drawbacks, for the most part AFR works. Particularly if you’re not highly sensitive to lag or microstuttering, it can get very close to doubling the framerate in a 2-card configuration (and less efficient with more cards).
Lucid believes they can do better, particularly when it comes to matching cards. AFR needs matching cards for timing reasons, because it can’t actually split up a single frame. With Hydra, Lucid is splitting up frames and gives them two big advantages over AFR: Rendering can be done by dissimilar GPUs, and rendering latency is reduced.
Right now, the ability to use dissimilar GPUs is the primary marketing focus behind the Hydra technology. Lucid and MSI will both be focusing almost exclusively on that ability when it comes to pushing the Hydra and the Fuzion. What you won’t see them focusing on is the performance versus AFR, the difference in latency, or game compatibility for that matter. The ability to use dissimilar GPUs is the big selling point for the Hydra & Fuzion right now.
So how does the Hydra work? We covered this last year when Lucid first announced the Hydra, so we’re not going to cover this completely in depth again. However here’s a quick refresher for you.
As the Hydra technology is based upon splitting up the job of rendering the objects in a frame, the first task is to intercept all Direct3D or OpenGL calls, and to make some determinations about what is going to be rendered. This is the job of Lucid’s driver, and this is where most of the “magic” is in the Hydra technology. The driver needs to determine roughly how much work will be needed for each object, also look at inter-frame dependences, and finally look at the relative power of each GPU.
Once the driver has determined how to best split up the frame, it then interfaces with the video card’s driver and hands it a partial frame composed of only the bits it needs to render. This is followed by the Hydra then reading back the partial frames, and compositing them into one whole frame. Finally the complete frame is sent out to the primary GPU (the GPU the monitor is plugged into) to be displayed.
All of this analysis and compositing is quite difficult to do (which is in part why AMD and NVIDIA moved away from frame-splitting schemes) which is what makes Hydra’s method the “hard” method. Compared to AFR, it takes a great deal more work to split up a frame by objects and to render them on different GPUs.
As with AFR, this method has some drawbacks:
- You can still microstutter if you get the object allocation wrong. Some frames may put too much work on the weaker GPU
- Since you can use mismatched cards, you can’t always use “special” features like Coverage Sampling Anti-Aliasing unless both cards have the feature.
- Synchronization still matters.
- Individual GPUs need to be addressable. This technology doesn’t work with multi-GPU cards like the Radeon 5970 or the GeForce GTX 295.
This is also a good time to quickly mention the hardware component of the Hydra. The Hydra 200 is a combination PCIe bridge chip, RISC processor, and compositing engine. Lucid won’t tell us too much about it, but we know the RISC processor contained in it runs at 300MHz, and is based on Tensilica’s Diamond architecture. The version of the Hydra being used in the Fuzion is their highest-end part, the LT24102, which features 48 PCIe 2.0 lanes (16 up, 32 down). This chip is 23mm2 and consumes 5.5W. We do not have any pictures of the die or know the transistor count, but you can count on it using relatively few transistors (perhaps 100M?)
Ultimately in a perfect world, the Hydra method is superior – it can be just as good as AFR with matching cards, and you can use dissimilar cards. In a practical world, the devil’s in the details.
Facebook
Twitter
RSS
47 Comments
View All Comments
shin0bi272 - Thursday, January 7, 2010 - link
I know thats what I was saying. The technology was supposed to be more worthwhile than this. Plus you cant mix gpus with a regular motherboard so Id have to get another 8800gtx to match mine on my sli supported motherboard. Or if I wanted to go with ati's new card Id have to get 2x5870's ($400ea)and a new crossfire mobo($150) to go crossfire instead. That's more expensive than getting this $350 mobo and adding a 5870 to my 8800gtx. Even if I went with 2 5850's at $300 each its still more expensive than buying this $350 mobo and one 5850. So you see why I really was hoping this tech would work better than it does.This would really do well in the budget mobo market IMO, so that people who didnt want to pay 300+ dollars for a motherboard then buy two video cards could use an old card and get better performance than they would have by just getting the low end mobo and using their old gpu.
If they can get it to be truly scalar (or close to it) like they originally claimed it would be then maybe some other motherboard makers will pick it up but if they dont get it fixed it will end up falling by the wayside as a tech that missed its time (sort of like the hardware physx cards).
Then again the crossfire 5850 in the call of juarez test got nearly scalar performance increases itself which is sort of new isnt it? isnt it the norm for crossfire and sli setups to do 40-50% better than single cards not 94%? Could just be my erroneous recollection but I dont recall near perfect doubling of fps with sli or crossfire before.
GourdFreeMan - Thursday, January 7, 2010 - link
It is an amazing technological feat that they got this working at all, but in the few games in which it does work properly the performance is frankly terrible. Look at what happens when you pair a 5850 and a GTX280 -- equal or worse performance to a 5850 by itself, when theoretically you should get a ~75% gain over a single card.FATCamaro - Thursday, January 7, 2010 - link
This technology had fail written all over it. They unleashed a big sack of fail...danger22 - Thursday, January 7, 2010 - link
maybe the amd 5000 cards are to new to have support for hyrda? what about trying some older lower end cards? just for interest... i know you wouldn't put them in a $350 mobochizow - Thursday, January 7, 2010 - link
Sadly, I think Lucid missed their window of opportunity as the need for Hydra largely evaporated with X58 and certainly P55's mainstream launch, offering support for both CF and SLI on the same platform. The only real hope for Hydra was the prospect of vendor-agnostic multi-GPU with better-than-AFR scaling.Those lofty goals seem to be unrealistic now that we've seen the tech, and with its current slew of problems and its incredibly high price tag, I just don't see the technology gaining any significant traction over the established, supported multi-GPU AFR methods.
The article touched on many of the key problems early on, but never really drilled down on them, hopefully we'll see more on this in the next installment, especially IQ and latency:
Guru3D did an extensive review as well and found CF scaled significantly better than Hydra without fail. Add in the various vendor-specific feature compatibility questions and an additional layer of driver profiles that need to be supported and synchronized between potentially 3 parties (Nvidia, ATI sync'd to each Lucid release) and you've got yourself a real nightmare from an end-user perspective.
I'm impressed they got their core technology to work, I was highly skeptical in that regard, but I don't think we'll be hearing too much about this technology going forward. Its too expensive, too overly complicated and suffers from poor performance and compatibility. I don't see the situation improving any time soon and its clearly going to be an uphill struggle to get their drivers/profiles in order with both older titles and new releases.
sheh - Thursday, January 7, 2010 - link
I agree. It's interesting from a technical standpoint but not many would want to go through all the fuss of SLI/CF (noise, heat, power) plus having to worry about the compatiblity of two or three sets of drivers at the same time. And that's assuming costs weren't high, and performance was better.I suspect in 1-2 years NV or ATI will be buying this company.
(I'm somewhat surprised even SLI/CF exists, but maybe the development costs aren't too high or it's the bragging rights. :))
chizow - Thursday, January 7, 2010 - link
Not so sure about that, Intel has actually been pumping venture capital into Lucid for years, so I'm sure they're significantly vested in their future at this point. I actually felt Lucid's Hydra was going to serve as Intel's CF/SLI answer not so much as a straight performance alternative, but rather a vessel to make Larrabee look not so....underwhelming.
Think about it, sell a Larrabee for $200-$300 that on its own, is a terrible 3D rasterizer and pair it up with an established, competent GPU from Nvidia or ATI and you'd actually get respectable gaming results. Now that Larrabee has been scrapped for the foreseeable future, I'd say Intel's financial backing and plans for Hydra are also in flux. As it is now, Hydra is in direct competition with the PCIe controllers they provide for little added cost that support both SLI and CF natively (licensing fee needed for SLI). In comparison, the Hydra 200 chip reportedly costs an additional $80!
TemjinGold - Thursday, January 7, 2010 - link
I think the issue I see is that X-mode will most commonly be used by people looking to save a few bucks when upgrading by combining the card they already have with a new one they buy. Unfortunately, I seriously doubt that this is the same crowd that would shell out $350 for a mobo. That just leaves A and N modes, which the Hydra currently loses horribly to CF/SLI.If the Hydra was put on a cheap mobo, I might see where it could be appealing. But someone who spends $350 on a mobo will most likely just shell out for 2-3 new gfx cards at the same time rather than going "gee, I could put this $50 to use if I reuse my old video card."
AznBoi36 - Thursday, January 7, 2010 - link
Right on. If I were to spend that much on a mobo, I wouldn't be thinking about saving some money by using an old video card, and in no way would I be mis-matching cards anyway. Seeing all this performance issues, I wonder how 3-way would be like...ExarKun333 - Thursday, January 7, 2010 - link
3-way would be ideal. ;)