Pierre Terdiman, a developer on nVidia's PhysX, have put up a blog post called "The evolution of PhysX" with benchmarks over several PhysX versions and Bullet, an open source physics engine.

"I felt compelled to also include the last version of Bullet (2.81), to provide an objective external reference. The thing is, there are many people out there who still think that “PhysX is not optimized for CPU”, “PhysX does not use SIMD”, “PhysX is crippled on purpose”, and so on."

Let me first make it clear that I'm far from an expert in physics engines, and I'm not even that up to date with the latest in the business. I'm pretty sure though that PhysX is still CUDA only as far as GPUs go, meaning it's nVidia hardware only. The rest of my argument rests on that basis.

Though the benchmarks clearly proves that each version of PhysX is faster on the CPU, it doesn't adequately disprove the idea that nVidia have engaged in 'crippling of PhysX' in games.

The problem here is that you probably don't want to run your physics on the CPU to begin with, and as far as I know, there is no publicly available OpenCL backend for PhysX, though I'm sure they have one internally. When I've heard PhysX critiqued, it is this that is the point. It's all very well that PhysX is faster than the competition on the CPU, but what matters to gamers is scenarios where you have CPU and GPU available.

In gaming, nVidia and AMD have about 85% of the market between them. Both losing ground to Intel recently. Simply put, if you belong to the 33% that run an AMD GPU and have your physics run on the CPU where the nVidia owners get to run on the GPU, certainly it's valid to say that PhysX is crippled in this respect, especially if the competition will run on the GPU on both nV and AMD.

The benchmarks as presented only show that PhysX is fast on the CPU in the best case, but it doesn't show that it can't be made to be slow, for instance by artificially resource-starving it. For instance, in a "TWIMTBP" game, nVidia may have instructed the physics to use a maximum of one or two extra threads on the CPU, even if more were readily available. In fact, that they're fast on the CPU could encourage them to cripple an integration just to make sure the CPU path can't catch up with the GPU path even on the fastest of machines.

"Some early investigations into PhysX performance showed that the library uses only a single thread when it runs on a CPU. This is a shocker for two reasons. First, the workload is highly parallelizable, so there's no technical reason for it not to use as many threads as possible; and second, it uses hundreds of threads when it runs on an NVIDIA GPU. So the fact that it runs single-threaded on the CPU is evidence of neglect on NVIDIA's part at the very least, and possibly malign neglect at that." -- Ars Technica, 2010

It would take someone posting game and/or driver disassembly AND benchmarks from a recent title to prove nVidia in the wrong. Again.

PhysX on Next-Gen

It's worth thinking about this in the context of the "next-generation of consoles" rolling out later this year. We know with a fair amount of certainty that they will both use AMD APU parts, so PhysX will almost certainly NOT run on the GPU on them as long as nVidia insists that PhysX should be CUDA only.

I can see three scenarios.

1. nVidia gives up on the physics engine market.
2. nVidia provides an OpenCL backend for all platforms.
3. nVidia special-license GPU-version for the consoles.

(1) is unlikely, (2) would be great for everyone involved, but nVidia being nVidia I wouldn't be surprised to see them take the third options; crippling PhysX on the PC where they (think they) can use it for marketing, but giving in where required to be able to compete at all.

Actually a (4) where everyone come together around a common API would be the best outcome, but it's my view that only MS could push something like that through, and then it'll be some DirectX specific hackery that'll be useless for cross-platform and then what was the point?

On these things and more, the future will tell.

EDIT: @Ken_Addison linked to this press release: "NVIDIA Announces PhysX and APEX Support for Sony PlayStation 4". From the wording it appears as though PhysX will run on the CPU only on the PS4. I should have listed this as an option since it's the most obvious one. Let's call it option 0.

The review is almost unreadable. Graphics; The choice to use a small image instead of industry standard text is very odd. It's also very short, at less than 150 words (estimated) you can read it in well under one minute. On the upside, loading times are great. Sound design is non-existant, and the silent music track gets repetitive after a while.

I give it a ten out of ten.

There's a growing trend with the type of anonymous online gaming found in games like Journey and Demon/Dark Souls. I just recently found out about this cool little game called "Way" through the Let's Play by Deceased Crab (part 1, part 2), though it seems to have been around since at least 2011.

"Way" is a (forced) cooperative platforming'n'puzzle game where you can only make it through the game by getting help from a stranger on the internet. The trick is that you don't know who your partner is, and the game only allow you to communicate through avatar gestures, similar to those in LittleBigPlanet, where you can change your facial expression and point your arms independently of each other.

I recommend watching those videos, UNLESS you're the kind of person who likes going in blind of course.

Way homepage where you can download the alpha, or read more on the project page at CMU.

Is there a good word for this kind of multiplayer? Suggestions!

Did one man alone resist when Zynga bought out Draw Something developer Omgpop? His reasoning is one I share, especially with regard to being able to fully own and control projects I make on my own time.

So this is a thing(follow through to the shop). Anyone tried them?

Something just came across my screen which made me feel both sad and a bit angry.

Morgan Webb, host on G4TV and reviewer for X-Play, tweeted:

I've been meaning to write about this since I first completed Mass Effect 2 early this year, but with the new Assassin's Creed: Brotherhood video demo (which basically showcases the idea, though probably a bit less ambitious) out I'll cut my own ambition about a larger article short and settle for this blog post.

CENTRAL ARGUMENT: Mass Effect 3 need to have a team away-mission management component.
 

No more hang arounds

Core mechanics

Risk vs Reward

The Final Piece

Graphic Adventures - Being a Mostly Correct History of the Adventure Game Classics by Lucasfilm, Sierra and Others, From the Pages of Wikipedia

Just discovered this amazing DMA Design history project site. DMA Design deserve their own page IMO, I'd create it on the spot if I knew how.

  Like Gish (and Penumbra and game called Lugaru) before it, the source for Aquaria was recently released. Unlike Gish however, with its quite horrible C, Aquaria is written in C++ and the code base looks to be in great shape! It is also quite a bit larger and more difficult to understand in some ways, but that's them deals.

 It's interesting comparing the two games, codewise. Aquaria can be studied for how a modern (but smallish) game looks internally; class hierarchies ("Quad", "BitmapText" is-a "RenderObject"), state-machines, rendering-layers, shaders and speciall effects like bloom. Gish should probably not be studied in the same way, but remembered for the lesson it teaches us about how you can reach a goal with lesser means. Is Aquaria more fun than Gish? I don't know, I haven't played either, but there's probably no strong correlation between good code and fun game. That said, if you can, I say you go for good code :-)
 
 The one thing I'm missing with these releases are documentation. Especially for Aquarius you'd think there were some architecture- or design-notes that could make the release? Speaking of which, I think these developers could probably more than double their income if they'd released their games together with a book about their development and architecture. Game development is hot now, and I think that kind of book -- where the core is an honest-to-deities commercial game -- could sail the charts quite nicely.

 Don't think I'm underestimating the effort though. I know it's a huge investment (primarily time-wise) to write a book. I'm just saying I think there's a market here which no-one is satisfying. The first one to make that investment could find himself handsomely rewarded.
 
 I'm saying that because it's what I've been thinking about doing, writing a simple game and documenting it while I'm doing it, then cleaning that up into a book (probably a free one though). I've always found that kind of "okay, so this is how I did this..." material super interesting.
 
 I'm never going to get anything worthwhile done though, so no, don't wait around for that. Tell you what, just pretend like you never read that last sentence.

Anyhoo... Here's a trailer forAquaria (with game-play):