Source"Most of the problems of getting these systems running on these multicore processors are not solved. They are doctoral theses, not known implementation problems. So it's not even clear that over the lifespan of these next generation systems that they will be solved problems. The amount of time it takes to get a good multicore engine running, the Xbox 360 might not even be on the market any longer. That should scare the crap out of everybody."
I'm just laughing at people who bought dual core CPUs expecting them to work in a few months on games
No, but seriously. This is a big let down as of right now. We were all expecting games to start using 2 CPUs. Though, nvidia has been tinkering with their video drivers so that they will use 2 cpus. Hmm, time will only tell. Hopefully it isn't too long and people might be able to put their X2s to use in a matter of months and not years.
Currently building my i7 920 rig
i7 920 - Gigabyte x58-UD3R - 6GB Corsair DDR 1600
GTX285 - Intel X25-M + 1.2TB - Watercooled
Heh, Gabe is talking out of his ass. Do you remember sega saturn? That had dual cpus and they werent working in parallel either which made them more complex to program for, but the jap devs still did it. Its just that Gabe is lazy and doesnt want to learn. Jap developers had no problems pumping out saturn games. An example of how great multi threaded games will be is Unreal Engine 3.0
Doom 3 was multithreaded...Not very much of it, but the physics WERE set to core 2 if you had a dual core/dual chip setup.
Quake 3 was also multi-threaded, but verrrrry little.
Unreal Tournament 2k7(and all games on UE3) is set to be multi-threaded as well, physics and AI I believe are set to core 2, while the rest is run off the first core.
Havok is going multi-threaded as well.
There's a few reasons to have a dual core system, but it sounds like games will take ALOT longer to be programmed if multi-threaded. Then again, who knows how long they've been working on unreal engine 3.
Well one thing Gabe told was truth - that programming for multyCPUs is absolutly different task than for single CPU. But i doubt that Valve will be out of buisness just cos Gabe sad this is very difficult.
Hmm I see Valve needs new programmers to compet this task
Exactly Gabe is just lazy and incompetentOriginally Posted by DilTech
In the world outside of games, we have been coding for multi threads for many, many years... Games are a decent sized market and will catch up....that Gabe guy, whoever he is, is a bit of an idiot...
"Foldin, Foldin, Foldin...keep those benchers foldin..." (Lyrics by Angra, Music is Rawhide)
BOYCOTT MIR's
As if a picture of him didn't tell you that already?
LOL
Currently building my i7 920 rig
i7 920 - Gigabyte x58-UD3R - 6GB Corsair DDR 1600
GTX285 - Intel X25-M + 1.2TB - Watercooled
All this Gabe h8
Its his fault for making such dumb statements
I don't hate him, just think he's a twit, there's a definite difference...if I hate someone, I like to ruin their life, if I think they are just an idiot, I lick marshmallows and stick them them in their hair...
Gabe falls in the latter category....
"Foldin, Foldin, Foldin...keep those benchers foldin..." (Lyrics by Angra, Music is Rawhide)
BOYCOTT MIR's
He's typical MicroShaft, that's all really.
Forums are the Opiate of the Masses
Atari Jaguar had 5 CPU's Link and it flopped just like the saturn
I thought this was old news...........i was under the impression we wouldnt see multithreaded (truly) games with this generation of dualies.
programming for multi processors is, in general, an open research question, but there is a large body of work that will let game programmers do some basic thing with dual core CPU pretty quickly. The problem is, at least with the technologies in play here, the software must be written to take advantage of the multiple processors.
There is research underway that alleviates this somewhat - i.e. allows non multi-processor code to achieve some instruction level parallelism, but, really the most interesting work in this area (that I am aware of) is in the domain of 16-512 cores per CPU. There are some really interesting techniques for speculative instruction-level parallelism out there if you can afford to "waste" some of your cycles.
anyway, for just 2 or even 4 cores, it really should not take long for game writers and also video driver developers to get some significant benefit. don't expect 2x improvement of course, but things will get better a lot faster than a lot of folks realize.
if you want to see some really interesting parallelism in action, take a look at what people are doing to harness the power of the GPUs to do non graphics computing....
Think of this way, I've been running x86 mutithreaded apps (like oracle) since Compaq released the systempro XL fileserver with 2 486 DX 50's. This is not a new field and there's a lot more technology available from mid range iron even earlier in history. This is just new to game dev's...they'll get it in a hurry or they will get fired when the competition pulls ahead
"Foldin, Foldin, Foldin...keep those benchers foldin..." (Lyrics by Angra, Music is Rawhide)
BOYCOTT MIR's
I think it matters little what we think here. Gabe still has hl2 under his belt and HL3 under his sleeve...so if he says multi-core isnt practical right now, most likely we'll see a delay in an all out industry support for multi-core games. But yes, they are inevitable...seeing as how AMD and Intel are both heading for a completely multi-core processor lineup by 2007.
Perkam
as usual, I totally agree.Think of this way, I've been running x86 mutithreaded apps (like oracle) since Compaq released the systempro XL fileserver with 2 486 DX 50's. This is not a new field and there's a lot more technology available from mid range iron even earlier in history. This is just new to game dev's...they'll get it in a hurry or they will get fired when the competition pulls ahead
One thing I'm curious about is the executable. Will there be different exes for uni proc and SMP/multicore? I wouldn't think that you would need that, but I imagine the code could be better optimized this way. Bear in mind, I've never programmed anything other than QBasic with DOS 5/6.xx and batch files.
And, as far as games and the possibility of multiple exes, would that throw off cheat detections? Or, would it be wiser to detect the way PB does for et and doing separate MD5 checks?
not at all brock. the vast majority of the programs you use on a daily basis are ALREADY multi-threaded. if there is any performance lost to multi-threading, it is obviously minimal compared to the benefits, otherwise we wouldn't be seeing it done even with single procs.
iirc, in single procs all the multi-threading does is add another thread for the cpu to rotate through on each clock cycle.
We have performance problems now in the out-of-order universe because we have programmers who can't figure out
why the changes they made caused the system to behave the way it does.
Me --> Ha Ha try MS$ software ......... or any other SW for that matter !
What a statement !
Your Existing Code? Throw it Away.
Posting Permissions
Reply With Quote
Bookmarks