Forums | MacLife

So where does Apple's move to Intel leave the UT gamer?

Ryan reports at IMG.

Ryan G wrote:

Ryan Gordon, Epic Games: From a game development (rather, a game porting) viewpoint, this will be a huge win once we get the majority of users over to these systems, both in terms of developer expertise and end-user performance. Most games we deal with are already running on Windows/x86, and were optimized with the x86 in mind, so "porting" these Mac games is turning off the byte swapping and turning back on the SSE codepaths. Not having to write anymore Altivec code is a GOOD THING for everyone involved. All my bitching about having 30 windows developers and one me are a non-issue in terms of optimization.

I could probably get, say, ut2004 up and running on an x86 Mac within...well, the time it takes to change a few lines in a Makefile and recompile the game, and I'd have optimizations suddenly enabled that were never previously feasible to put into the Mac version.

It'll be interesting to see what games in the future ship as Fat Binaries and have some small issues in the PPC build that vanish in the Intel version, where some nagging byteswap issues cease to be a problem.

Not to mention that all sorts of things become feasible:

1) Pixomatic could run on a Mac, for what that would be worth...Havok and such loses all technical excuses for not showing up, as do plenty of other middleware packages.

2) FreeBSD has had a "Linux Binary Compatibility layer" for years...it is more than sufficient to run the x86 Linux version of UT2004 on x86 FreeBSD flawlessly and without significant overhead...I suspect something similiar will show up in Darwin (if not a straight port of the FreeBSD code). This gets you a few good apps that weren't previously practical to port to the Mac, and basically all the Linux dedicated servers. Could this be something that moves the game hosting industry to Macs? Well, probably not, but it at least makes this a feasible endeavor for those that are interested.

3) Games that were inexplicably network incompatible with the PC version, or couldn't move savefiles between platforms, etc, suddenly work.

I mean, I don't assume we're going to see a bigendian x86 chip in the Mac or an Altivec unit at all (SSE2 will be the thing, I assume), so really, in terms of porting, this actually makes the job of moving games to the Mac easier...you're now dealing with moving code to Mac frameworks and not to an otherwise alien architecture. And really, the PowerPC has been holding the Mac back, both in technical terms and mindshare. I mean, there are many things I like about the architecture, but I'm looking forward to thinking about it the way we currently think of the 68k.

It's not like Windows apps are going to run on Macs without porting, regardless of the CPU. Also, the thing that is well and truly killing Mac gaming in the triple-AAA field is that even the G5s aren't competitive with the high-end PC gaming rigs, so in this sense it can only be better to level the playing field (and really, level it, not in that "the G5 will level the playing field" rhetoric way. Still, gaming is just like theater, open source and cockroaches...you can't kill any of these off, and if you think you have, they always show up again. Atari's collapse in the 1980's didn't kill gaming any more than the Catholic church killed theater in the Dark Ages, despite their best efforts to murder playwrights and burn scripts. Eventually, Nintendo manifested. Gaming is an inevitability. When all the whores have run to the consoles because "it makes financial sense," the void will be immediately filled by independents that now have a worldwide market of powerful computers, a distribution channel on the internet, and no competition...it's not like people are going to stop playing games on their computers, regardless of what makes "financial sense" to the game houses. Will these indies be shipping Half-Life 3? Probably not, at least not at first, but then again, the Catholic church eventually made dramatic performance--that is, theater--part of their ceremony, so expect these things to rise up right on the spot where they were supposedly killed.

Oh, a few more random thoughts about the Intel switch:

- I'm calling BS on Rosetta. While I'm sure it'll run Office well (specifically, the Must Have With No Viable Alternative Apps), I refuse to believe that it'll be the panacea people at WWDC are already calling it, and shame on everyone for not knowing better. Even if the Intel chips are EMT64s running in longword mode, you get 16 general purpose registers, and must emulate 64 on a PowerPC (96 if you count the Altivec registers). That being said, PowerPC is much saner to emulate than x86 is, so a comparison between Rosetta and VirtualPC wouldn't be fair. I can't imagine they'll emulate 64-bit apps ("oh well, there weren't many of them, and they can be recompiled!" or something), and I imagine Classic is just flat out gone. Whether that's okay or not is to be judged against Win64 dropping DOS and Win16 support. I'm sure we all instinctively know that Rosetta will be worthless for gaming.

- Anything that works on Windows (or hell, Linux) and Mac will have NO problem moving to x86 MacOS. Anything that is Mac only (for the few that are) could have some serious issues that Apple makes look easy...("hey, all our stuff works great on the x86!" ... well, yeah, you had five years and planned it from the start. I bet finding the byteswap issues in iPhoto would take a lot longer if they started from scratch porting it today.)

- Can't wait for someone to hack up one of these boxes to run PC-intended video cards. There goes that 500 dollar GeForce market, and good riddance to smurfy exploitation.

- In many cases, the bottleneck isn't the CPU, it's the graphics card.

- Someone needs to find out what the 64-bit plans are for Apple now...Intel has a 64-bit chip with 32-bit compatibility, much like the G5 is to the G4, but will Apple ship this? Will everything be in longword mode by default (unlike the PowerPC, you have to pick a mode and jump through some hoops to support 32-bit apps thereafter, making a later switch a little more tramatic than the G5 shipping was).

- Apple's been a good contributor to GCC, and it'll be nice to have their engineers optimizing the compiler's x86 output...good benefit to Linux, if nothing else.

- SSE3 is way more flexible and feature-rich than Altivec. I don't care what the haters say, that's the truth. In terms of vectorization, SSE will be a better instruction set. There are lots of places were you get the feeling that Altivec should be a good optimization, but you can't find a way to coerce it to do what you want without a lot of effort (and frequently, without making the program _slower_). SSE has wider application. It just does, even without names like "Velocity Engine".

- As usual, Don't Say You Weren't Warned. Apple just spent five years saying "stop using CodeWarrior" and even without a clear support path or G5 support, people still wouldn't stop using it. Now, they'll pay for it. Generally, I find when Apple suggests you do or don't handle development a specific way, you ignore them at your own peril, and this was another great example.

- I want one of these machines before I conclude any crap that I spew about this is remotely accurate. But really, if we all survive the transition, it's removing the piece of the Mac that many people mock, accurately or not, and then permently killing some myths that "Apples are slow." Can't be all bad!