Sneak-peek at Nvidia's Fermi Review - PAGE 2

The most impressive demo of the day shown was the new tech demo called "Supersonic Sled." This demo really brought everything together for the gaming potential of GF100. The premise of this demo is based on yonder-century daredevil maniacs who it was a fantastic idea to attach themselves to jet rocket, on a sled attached to a railroad, and launch themselves at speeds humans don't normally go. That's the setup for this tech-demo: man in a sled. The nice catch is though that you have some control here: just how much boost is the jet capable of producing? No human should be traveling that fast in only a helmet; poor guy.

This is one of the best Nvidia tech demos yet. The most impressive thing here was that it wasn't an animation: PhysX was used extensively in just about every obejct. The only thing that was animated was the sled-driver's poor, expressive mug, as his doomed mission begins. Each object in the demo has simulated physical properties; so, depending on the forces involved, such as wind, that come out of the speed of the sled, there is potential for each sled launch to run just a bit differently. Often, to the determinent of the sledder's health.

At end of the demo we were shown how many objects the GF100 can handle at once. While we don't have an exact number to offer here, let's just say it was a lot. You can expect explosions to look quite...explosive with a GF100 running in the years upcoming DX11 titles. While there are many physics AP's around, Nvidia's PhysX will continue to offer solid physics performance in supported games that are due to arrive on the scene this year, such as the post-nuke war shooter-RPG-survival-horror game, Metro 2033.

We are getting to the point where hair can be rendered almost perfectly. And that's much easier said than done.

Although it wasn't the sexiest part of the day's proceedings, and probably not the part that excite gamers the most, NEXUS, and Nvidia's CUDA development strategy, was also one topic of discussion and example given in the conference. CUDA (read more about it here) is a core strategy of the future of Nvidia's GPU design team. It is an effort to open up the parallel processing nature of the GPU to software developers. Nvidia is expending many resources on creating tools, documentation, and support that will encourage and invite software makers to really tap into the power of the GPU. These days, GPU's have more transistors than CPUs, and represent a untapped computing resource that is commonly found in desktops everywhere. The hope is with the C++ language support in the Fermi generation, that a the burgeoning field of CUDA development will continue to expand, as it has massively already, this year.

NEXUS is another tool that is used in conjunction with the ubiquitous Microsoft Development Studio package to lower the barrier, so to speak, of developing apps specifically to take advantage of the GPU. In a demonstration I found incredible, the debugging potential of NEXUS was show. The example code was run in a virtual machine environment -- this is almost a necessity with CUDA development, as if there is a serious crash, you are libel to hang your video card, which means that your not going to be able to recover, and you'll have to reboot. NEXUS was able to freeze the virtual machine running the example code, and the user was then able to mouse-over parts of the rendered scene, in attempt to find out what bugs were in the code. Each pixel or triangle could be traced back from draw routines. I know, I know -- this might not sound exciting for you. But for many of programmers out there, the gaming side of GF100 will be ho-hum, while this new tool is going to knock off their socks. With NEXUS, CUDA looks to crack things wide open; perhaps indeed the GPU will become a necessary co-processor in our well-performing desktops of the future, even for non-visual tasks.

And to wrap this report up, there was one final surprise shown to us. Get ready for 3D Vision Surround. That's right: 3D Vision is pretty cool, but what about three monitors running 3D Vision at once?

Being 3D images, this shot didn't turn out really well -- but trust me, it was looking good.

We were treated to the demo of a triple-display 3D racing game, and it was fairly impressive. This new technology can obviously be seen as a counter to ATI's Eyefinity technology, which came out with the HD 5000 series. In fact, you'll be able to run three displays with Fermi without the 3D. This will not be something that many will be able to get into, because of the price. Three 120Hz displays will not be cheap. Furthermore, the video card will have about roughly 6x the work load of just one 2D display (three displays, times two for 3D Vision, as twice the frames have to be rendered), so we imagine that realistically speaking, only the craziest gaming rigs will be able to handle games with sufficient framerates. Think SLI GTX 380 cards -- of course though, that is just a guess for now. We'll have to see.

For the system that they were running the 3D Vision Surround demo on, the game was impressive. While not every game would benefit much from three displays, some games, such as the racing game demoed, with the huge panoramic view, certainly benefited.

Well that about wraps up my report of the look at what Fermi is finally shaping up to be. I eagerly look forward to the video card manifesting, in order to test it out, and see what work Nvidia has been up to with this novel new design. It should be interesting!

Thanks for reading.