Kalos' Chronicles >> One, Two, Four: StarCraft II and Multicore >> (1) | (2) | (3)




'Dumb multithreading'

I've decided to coin a phrase to explain why multithreading is not just theoretically better or even potentially better down the line from now, but is actually beneficial here and now. Dumb multithreading. Basically, we already run at least two heavily demanding, complex programs when we are playing our PC games. The game itself is one, and the other is the OS itself and all the components that make it up. The OS has become a vast and heavy resource consumer (as many people are quick to mention Windows Vista is), it is always there and always demanding some processing time. On the old 'single core, single thread' computers most of us use today, as we are gaming the processor is constantly switching between tasks hundreds of times every second, in between the game and any other process that is running. Just hit Ctrl-alt-delete to get into your Task Manager and see how many dozens of system processes you have going on, there can be hundreds of these things going on.

Dumb Multithreading in its simplest form can be explained with a dual core. A decent OS, XP but Vista as well (hopefully I won't get too many complaints about this being Windows-centric) is smart enough to realise as an intensive task is running on one core of the processor to recognise to divert all other minor processes to the idling second core, including its own operation. It does this automatically, to different degrees of effectiveness nonetheless, to yield performance benefits to the user over using a single core processor for gaming uses today. Going back to the road analogy, it would be like pushing all the bicycles and horses on the roads into a lane of their own to allow the car to proceed unhindered in the parallel lane without having to slow down and respond to their presence.

This is one of the reasons why having a dual core processor would yield a large boost in performance over its single core cohorts in most tasks but a quad core does not. The background tasks don't actually overwhelm a whole processing core; they just chip away at its potential for dealing with an important task at the same time and when you only have the one core that can cause quite a diminished performance for certain games that were already limited by the processor's capabilities.


Should I run out and buy an multicore processor then?

The answer to this question is not simple. While it is clear there are advantages, and as time goes on things will only become more and more stacked in favour of the Dual/Multiple core approach with a giant like Intel throwing its weight behind the movement, it is clear from the past that jumping to the latest hot thing before you need it isn't a great idea. Many people rushed to buy early Pentium 4 processors back in 2000 while cheaper and quicker Pentium III CPUs were neglected simply due to the little number that followed the word 'Pentium', had they stopped to look at benchmarks to see the actual performance instead of looking at the Ghz label they'd have realised it was not quicker at all. Going from clueless Joes to the techwise minority, several people purchased Athlon 64 based computers back in 2003-4 due to the belief that having 64 bit extensions would make it compatible with future versions of Windows, allowing them to use it far longer than the competition's Pentium 4 processor. However by the time Windows Vista showed up not only were most versions 32 bit, compatible with the Pentium 4, both the Athlon 64 and Pentium 4 were looking long in the tooth and fairly ragged compared with the Athlon X2 and Core 2 series that had replaced them; the futureproofing didn't work. By the time a 64 bit-only OS will be released, Athlon 64s will most likely be too weak to run it if Vista is anything to go by. Buying for the assumed future doesn't work, and it might not work with multi-cores either.

My personal perspective is to assess the age of your current setup, find out about it, and most importantly decide if it is currently compromising your gaming. If it you're having trouble or just don't know much about your hardware, there are many bored individuals such as myself who exist to part such knowledge and insight if you wish. If you don't need to change things as it is, don't. I happily used a single core right up until this summer despite them being out over three years. Was it the bleeding edge of gaming? No. But it saved me a ton of money on the multicore jump. If I had made that move in 2005 when they launched rather than now, I'd have ended up with an Athlon X2 rather than a Core 2 Quad for the same price, and it wouldn't have really improved anything over the Athlon 64 single core that was in there already. Equally you shouldn't sit on the hedge waiting forever if you have a degrading Pentium II running Windows 98, that's a pretty valid reason for getting a new one in that situation.

Certainly, don't just jump to these multiple core processors because they aren't critical yet, they aren't at a level where they make or break the game experience, and they aren't going to suddenly become more pricey, quite the opposite in fact. This year will see Intel releasing a new series of processors with the eventual aim to replace Core 2 as the high end processor tree, currently known as Bloomfield. It is projected for two, four, and eight core monsters to be made, using a technology known as hyperthreading to create the illusion of there being double the cores available to the PC by treating each core as two separate virtual ones. In essence, next year we'll be seeing the 16 core processor coming out. Well beyond what we'll be able to afford or even think of using at first, but it'll make the multicore designs of today look underpowered and small, and that perception will stimulate lower prices to the consumers for those now-weak looking chips. Hence, why rush there if you barely need it?

Final conclusions - StarCraft II to take Multithread Advantage? On Page 3.

You must be registered and logged in to post comments.