The next DirectX has some significant changes to it - what kind of approach does a company like Futuremark have for dealing with this? A rolling training program for all of the team or do you rely on the programmers themselves at picking this up?

Every time a new DirectX is on the horizon, we make sure that everyone at Futuremark knows what is happening. Of course it is difficult to keep everyone on track with all the smaller changes to any specifications, but we have dedicated key persons who know exactly what is going on. Our programmers are always on top of things, and we try to keep the rest of the teams in the loop as well.

The art work in 3DMark appears to have increased at a substantial rate over the past few releases. How many man-hours are put into the design and creation of such material for each game test? How accurately do you think this level of detail matches what near-future and future games may be like - for example, over the next few years we're more likely to see greater use of high res normal maps with offset parallax mapping rather than very high polygon counts (the latter being something that 3DMark has stuck to rigidly).

It’s actually next to impossible to say the exact man-hours we put into the design and creation of every test since our folks sometimes tend to spend extra personal time to make all look beautiful, but it is A LOT! I believe that the man hours done for the artwork has grown 10 fold only in the last couple of years. The uses of different materials, normal maps, high-detailed models, complex scenes, animations etc. are things that simply require a lot of work. I’m not saying that the programming side would have got any easier, but the user-expectations in cool graphics have gone through the roof which makes the lives of the artists very tough. It also seems to be a common problem to find talented 3D artists, and especially technical artists. If you know any, let me know!

Techniques such as parallax mapping are interesting and useful in certain situations (brick walls etc.), but I don’t think they will "fix" the problem. I have read quite lot of papers on various implementations of parallax mapping, and though they all look neat, there are still a lot of situations where it simply isn't feasible (e.g. edges, round objects etc.). The idea behind parallax mapping is still great and I believe that it will be used to a certain extent in games, but it doesn’t work in all places which means you need to pump out those polygons to get the visuals look decent. We have made some research in using parallax mapping but I am not sure if we will end up using them, at least in the next 3DMark.

The use of normal maps has been in 3DMark since 3DMark03, so that's nothing new really. In the next 3DMark we use high resolution normal maps on almost all surfaces. By combining normal maps and a handful on polygons you get much better looking artwork than simply doing high resolution normal maps with low polygon stuff. Personally I am not that fond of blocky heads for the characters, as it seems to be the case in many games these days. With the next 3DMark we have tried not to go too crazy with the number of triangles, but instead tried to increase the workload on the pixel shaders respectively.

With so much artwork being produced, have you ever thought about producing in mini-games, perhaps downloadable for free (or a very small fee) for registered 3DMark owners?  The car game was very popular in 3DMark2001 and for many people, helped to identify the relationship that 3DMark had between synthetic benchmarking and real-time graphics scenarios found in modern games.

Mini-games are a blast to produce & play and we'd love to do more of them! Let me just say that "the next 3DMark will be different". ;)