During Computex, we took the opportunity of talking with Lucid on the subject of the report that we devoted to Virtu MVP, a software solution designed for the combination of two GPUs of different brands within the same system, something that is particularly the case with the Intel socket 1155 platform. To recap, its HyperFormance component hadn’t really convinced us as it looked too much like a cheat to increase 3D Mark scores.
HyperFormance consists in cancelling the full rendering of images which won’t be displayed (because the fps is too high). However our observations in 3D Mark highlighted something different: all the frames were displayed twice with in between a modification of the interface’s fps counter superimposed on the frames. It’s difficult not to describe such a practice as cheating and yet Futuremark decided not to describe it as such, settling to say that the results thus obtained would not be taken into account in its official classification.
Lucid of course wasn't all that pleased with our report and we had several rather tense discussions with the company's bosses who refuted our result’s without however responding with anything very convincing. Last month, in Taipei, we were able to speak frankly with Moshe Steiner, CEO of Lucid.
A special benchmark mode
In contrast to our previous discussions, he wanted to cut to the chase and told us that there had been a misunderstanding: two versions of HyperFormance had been set up in Virtu MVP. However, in focussing our attention on the case of 3D Mark, we used the special HyperFormance version designed to simulate within the benchmarks a representation of the gains that the technology can give in games. This was a shortcut that the company would have preferred to avoid but without which it would have been difficult to convince motherboard manufacturers… This special HyperFormance version is activated in all 3D Marks as well as in Unigine.
The standard version, used for games, doesn’t however ever kick in under 60 fps, or more exactly under the refresh rate, which is what Virtu MVP uses as a parameter. It functions in this way as planned in trying to increase the responsiveness of the display with variable success depending on different cases.
What is difficult to understand is that Futuremark didn’t pick up on this because indeed using a special benchmark mode is in practice forbidden in its charter… even more so when this special mode manipulates the frames which are supposed to be calculated. It looks clear to us that the weight of Intel, which supports Lucid, cooled the usually irrepressible Futuremark zeal when it comes to combating anything that looks like a cheat.
At the end of the day, Lucid is insisting on the fact that whatever we might think with respect to how it works in certain benchmarks, HyperForce does really work correctly in games, at least if they are supported (a profile is required).
Lucid also told us that the demo of the latest developments concerning XLR8, a version of its group of technologies optimised for systems equipped with a single GPU and more particularly simply an integrated graphics core. The goal of XLR8 is to improve as much as possible the playability of gaming on systems that are slow in terms of graphics, on, say, an Ultrabook which just has an Intel IGP.
In addition to HyperFormance (not very useful given that these machines generally don’t run at over 60 fps) and Virtual V-Sync which makes the rendering slightly more fluid and prevents effects such as tearing, Dynamix is now being finalised. This third technique consists in dynamically reducing (frame by frame) the resolution of certain rendering surfaces (render targets or RT), so as to try to maintain a certain fps rate.
In comparison to a fixed reduction of the rendering resolution, Dynamix allows you to preserve certain elements of resolution reduction. The standard example consists in conserving the interface in full resolution and reducing the quality of the rest of the image, which is what certain games such as ArmA 2 and its derivatives do already.
Lucid had two demos to show us. The first concerned Diablo III on an Ultrabook. Very jumpy and disagreeable without Dynamix, with this technololgy it was at +/- 30 fps with sufficient gaming comfort. The quality did drop somewhat however, but was acceptable given that the interface and the HUD remained at full resolution. We were however less convinced by Battlefield 3 which did become more fluid but at the price of lower quality.
Lucid hopes to finalise Dynamix in the next few months and support a maximum of games, a specific profile once again being required. Currently a demo of Dynamix is available for Skyrim. While Lucid may manage to support many of the big hits rapidly and make them playable on Ultrabooks, at the price of lower quality, Dynamix will probably be the most interesting Virtu MVP component for gamers. And we’ve had an undertaking, Dynamix won’t be activated in 3DMark!