- ‹ Prev
- Next ›
A Multi-GPU Wish Come True?** Updated as of 29th January 2010 **
We've updated the article with findings from testing the newer Hydra driver version 1.4 and we've added these thoughts on our concluding page. Most important are more GPU and games support in multi-vendor, multi-GPU configuration.
A Multi-GPU Wish Come True?
Two different graphics cards from two rivaling firms working in tandem? That's certainly fantasy stuff deserving of a spot on the New Year's wish list for graphics junkies. Except that it's now come true with the MSI Big Bang Fuzion P55 motherboard. In fact, we were promised this late last year but the Fuzion couldn't make it for Christmas.
Instead, MSI placated us somewhat with the impressively equipped Trinergy, which had an NVIDIA nForce 200 chip onboard to enable 3-way SLI and CrossFireX, making it a pseudo X58 equivalent. That board was very nice, very expensive and the start of MSI's ultra high-end Big Bang motherboard series but what we really wanted was cross-vendor, multi-GPU support. Fortunately, the Fuzion is finally ready to be unveiled to the world.
MSI collaborated with fabless semiconductor firm Lucid to implement the latter's vendor-agnostic Hydra 200 chip. According to Lucid, the Hydra is "a real-time distributed processing engine (system-on-a-chip) that allows efficient load-balancing of multi-GPU environments using any GPU vendor." Plainly put, it allows consumers to combine graphics cards from either of the two major graphics chipmakers on a single motherboard, though you'll find later that there are some restrictions on what cards are compatible.
The standard P55 motherboard already supports 2-way SLI or CrossFireX after the proper licensing and certifications. However, such multi-GPU technologies typically limits users to pairing only GPUs that are practically identical, with some allowance. ATI's CrossFireX for example has a limited ability to mix two different GPUs of a similar series/architecture while NVIDIA's SLI limits you to the same GPU model but works with varying clock speeds and memory sizes within that restriction.
Hence, the main attraction of Lucid's Hydra is its mix and match capability. Lucid claims that its distributed processing or load-balancing technology allows the addition of a newer generation graphics card to an older one while maintaining a decent amount of performance boast. This is done by having its ASIC chip sit in between the CPU and the GPUs and 'meditate' or distribute the workload via PCIe according to the GPUs' capabilities in real time. It's something that ATI or NVIDIA cannot do currently with their Alternate Frame rendering method of splitting the workload between two GPUs, which requires a similar level of performance from both. The generic nature of Lucid's technology also means that it doesn't matter what architecture the GPU belongs to, which hence allows for ATI and NVIDIA GPUs to work together.
Unfortunately, our excitement over its potential immediately got doused with cold, hard reality upon opening up the Big Bang Fuzion package. Mixing two graphics cards from the two vendors is only possible for a handful of models, and none of the latest Radeon HD 5000 cards. The bad news to come is that the application support for such a configuration is at the moment, quite paltry.
Here's what you need for the multi-vendor, multi-GPU to work:
- Windows 7 32-bit or 64-bit as the OS
- Single GPU Radeon HD 4500/4600/4700/4800 with single GPU GeForce GTX200/G90
- Supported Catalyst and ForceWare driver versions (Catalyst 9.9 and ForceWare 191.07 are the newest drivers quoted in the manual)
And the supported applications for mixing an ATI graphics card with an NVIDIA one currently are:
This list expands greatly, reaching around 70 games if you're using GPUs from the same chipmaker and it includes some recent and notable ones. For GPUs from the same chipmaker, you can use Windows Vista (32 or 64-bit). The Windows 7 requirement is for multi-vendor configurations. Of course, the important thing is how does Hydra perform in these games and that's what we'll be exploring in this article.
- ‹ Prev
- Next ›