I was about to type a post about how RAM should be integrated directly onto the CPU to eliminate the bottleneck with the front side bus, but I finally found something on the Internet indicating that this is being done.
See http://techfreep.com/intels-tsv-connects-processors-to-memory.htm .
Wow. It’s too bad we can’t buy that stuff now.
My belief that it will be on the market within 5 years = 0%.
Most likely we’ll see Dual-core processor dominate until this year.
This year until 2009, Quad-core processor will dominate.
2010, we’ll probably see introduction of Oct-core processors… and they will only be available for high-end servers until 2012-2013 or so.
As multi-core becomes the new single-core, does that mean we’re all going to have to get really good at writing parallel code if we want to get the most out of them?
There won’t be any sudden increase in our abilities to write multithreaded code. It’s fundamentally hard, even with the nice bits Java gives us to help out. So I expect that most of these cores are going to go largely unused in most people’s PCs; and this leads me to predict that dual core will dominate on the desktop for well over a decade.
Cas 
Isn’t this where parrallel aware languages like Fortress come in? So we don’t have to worry (as much) about multi-core.
Kev
Sounds like a typical functional programming language, they have existed for decades, and I’ve yet to see any application written using them. I seriously doubt that Fortress, when and if it is released, will make any difference. Most likely we’ll see the current imperative/OO languages advance further to better facilitate parallel processing. At least I’m not worrying about this for the next 10 years or so.
Regarding the n-core processors, I’m pretty sure that dual-core will become the “economy” processor while the quad-core will become the more expensive version. I doubt there will be much need for oct-core for consumers, probably only for enterprise purposes.
hmm…ram and cpu integrated. There was also recent rumors that AMD due to their recent purchase of ATI is building a cpu and gfx card in one chip.
Could it be that future computer designs are moving toward just fusing it all into one chip? it was already attempted with the ps3 cell chip where they initially intended to have the gfx done on the cell, but dropped the idea later as it wasn’t working out. Maybe we will see a ram + n-core cpu + gfx in one chip one day.
@ multicore
Well, I guess my games will remain that simple that even a single core wouldn’t have much to do. I mean… such a core could run Quake3’s logic at 2000hz without much problems. So wtf am I supposed to do with all that processing power by myself… or how would I manage to require even more?
But it’s surely great that you can run other stuff at the same time and that things such as ogg streaming (which is relatively expensive) is basically free. Oh yea and newer graphic drivers put the load onto the other core if possible automatically.
So as far as I’m concerned… everything I write will run great enough on those fast machines anyways and multi threading everything to death will only make it run worse on older machines. And any performance gains on those plenty fast machines won’t make any difference (no one cares if it runs a few % faster if it would already reach several hundred fps with a single thread).
Yeah. Other cores are best left to automagically doing things under the hood without any actual explicit programming from us. Eg. garbage collection, network IO, streaming audio, O/S tasks running in the background.
Cas
Yup, threading it like we do with GPUs these days. Somebody writes an API that does basicly everything, and we run shader-like programs (kernels) on the individual cores.
About having too much CPU power: (oNyx)
I somehow can’t get enough cpu-cyles… ever. Think about realtime radiosity, realtime volume rendering (scattering), real vegetation (not a few patches of grass every other meter, nor intersecting swinging grass-quads), all eye-candy requires loads of cycles. Sure, you can write games that do without, but don’t you like pretty graphics? You can’t get true immersion in PacMan (or Q3 for that matter).
Oblivion wouldn’t have been sold so much if it wasn’t for the graphics, so surely a lot of people care about eye-candy. That’d be a big market to ignore, for an indie
Well, the thing is… I a) wont manage to write that kind of complicated stuff and b) content is already a big problem with simple 2d games.
My life simply isn’t long enough. I can’t even afford to pixelate around for a year for some mario esque game. And making gigabytes of crispy photo realistic textures with half a zillion nifty shaders on top? Won’t happen. Sure… it would be extra awesome with strawberries on top, but… it won’t happen. I don’t even have one artist and I would need like… 20.
With nature it’s mainly growth-algorithms, a few textures (1 per plant-type) and you’re done.
Same for clouds: simulate water-vapor and you’re done. Real fluid water anyone?
“realworld” didn’t require more than one ‘artist’ either
just read somewhere that the music in spores will be fully procedural so no need for a music artist, it claims you won’t ever hear the same tune again. Sounds like a really awesome algorithm, hope the music it produces is good though, if its being used for such a big game must be at least nice. An open implementation of this would be really good for Indies.
afaik quadcores are making there way in the high end bit so they should be mainstream in 3-6 months. Clovertown… sure ppl argue that the thing is 2 processors glued togetter. but it doesn’t really make a difference for us programmers, does it? I think the line is a lot steeper. oh well speculation is speculation.
I read to parts of the specification of fortress, I really dislike the whole ‘end’ syntax constructs. Mathemetical notation is fine and yes its proven and wide spread… but the syntax used for the flow is just… eeew.
I sort of agree with this point. Hobby game-programmers are at least a decade away from the commercial games regarding art stuff. One “army-of-one” game developer simply cannot compete with a team of 20-30 specialists working on a game 10 hours a day for 2-3 years. You could dedicate your whole life to it maybe, but by the time you’re done computers as we know them wouldn’t exist.
In fact, game companies are hitting a “peak” today. There is so much competition that game developers are looking for ways to actually make some money. I read somewhere that only 1 out of 9 games actually make profit. We’re gonna see budgets go down.
20 years ago 1-2 game developers could make a commercial game, today you need anywhere from (minimum) 20 to 200 (big commercial ones). In another 20 years, how many will it need? Didn’t Microsoft say that around 100.000 developers put an effort into creating Windows 2000/XP/Vista? I wonder if big games in 20 years will require thousands of developers to work on.