The online racing simulator
Are Intel slipping a few stages?
(10 posts, started )
#1 - dadge
Are Intel slipping a few stages?
Sadly that each of those cores are about as powerful as an Atom core. Not very high power.
But even 48 2GHz Atom cores are still going to be able to crank out mental amounts of calculations And obviously they're going to keep advancing the tech as much as they can. I wouldn't say no if they offered one to me.

But nVidia with 512 cores? These numbers are getting insane.
Although some things just can't be parallelized easily, as well as parallel programming is more difficult, so it's not as perfect as you'd think. It's just a different GHz war. Like come on, with the i9's + Mac Pro (Which I can assure you will happen VERY soon), you will have Windows (and OSX) seeing 24 SEPERATE CORES. FFS, is that really needed!?
I guess it depends. You're definitely not going to NEED that sort of core count unless you're a render farm or running complex calculations for NASA or something.

On the other hand, if you design games for those cores from the start you could get some stunning performance gains. Imagine splitting physics calcs, AI routines, particle effects etc across multiple cores. You'd be able to increase the amount of stuff happening on-screen without compromising graphic quality or frame rate. It has incredible potential.
Especially as the CPU is the bottleneck now, I just doubt that many calculations can be parallelized as effectively as you'd think. I know I'll be proven wrong eventually, but at this point.. it's too complex.
Yup, I noticed a few years ago that speeds weren't increasing as quickly as they did before. I guess hardware capacitors have reached their limit. But you never hear about anyone researching light-triggered ones any more.
Contain's 1.3 billion transistors.

That's just mad.
#9 - wien
Quote from DTrott :Contain's 1.3 billion transistors.

That's just mad.

Not really. The latest Radeons have almost twice that, at a smaller process to boot.
There will reach a point where increasing parallelisation has diminshing returns, even once software has been engineered to be as parallel as possible. It's going to be an interesting decade ahead though.

Imagine two pass video encoding where all the keyframe points are decided first, then each short video segment is encoded concurrently. On a 48 core machine, that will encode video far faster than machines today. Assuming other hardware doesn't bottleneck it (like memory and storage), in fact I see other hardware like that having to go parallel to keep up. 100+ cores will be permanentaly waiting for data without a RAID array, or ridiculous quantities of memory.

Are Intel slipping a few stages?
(10 posts, started )