BlackHole wrote:Not buying a dual-core processor is completely devoid of reason. Not only are the vast majority of processors dual core anyway, dual-core processors will have effects when you are multitasking. For example, i used to be able to do almost nothing while installing a program. Now i can do whatever i want while installing a program. It does make a difference, often in places you'd never think it would.
Also, i don't think you truly comprehend just how much stuff goes on in the background. You have norton, scanning every goddamn process and file transfer that occurs, your anti-spyware program scanning every single internet connection, MSN (or whatever), and approximetely 120 windows processes doing god knows what.
Yeah, it makes a difference. Not buying a dual core is like shooting yourself in one arm and saying it was your bad arm.
No-one said it was a bad idea.......
(NB: installing a program the primary lag would be the data bus and disk transfer speeds, anyways, so I doubt dual core has much effect; to a lesser degree this could be true of anti-virus programs)
When considering multi-core performance relative to single core and whether it is equal to the increased price (which is what we're doing), there are several things worth considering.
One is that parts outside the datapath may be shared - such as the memory controller on Opteron chips, or L2 cache on, IIRC, Intel ones. Single threaded programs are unlikely to have considered this (well, maybe hyperthreaded ones - it's worth noting HT suffered a lot of performance hits for non optimized code because it shared a single memory resource between its hardware threads/datapaths)
Another thing worth noting is that individual cores tend to be slower than a single, same-priced core; which means that the performance of
any one application is handicapped by the core speed unless it supports multi-threading across >1 core.
The OS is also handicapped because it can't load-balance effectively with single threaded apps; if you have 100% usage on one core and 50% usage on the other, then you've lost 25% of potential performance. That's presuming deadlocks can't occur, of course. And i'm not sure how well windows itself is optimised to multi-thread (in older versions, presumably Vista should handle it and maybe XP) as it's been developed from a legacy single-thread base.
Plus, it's natural that Intels newest dual cores have significant performance improvements; for one thing they can probably drop the ridiculous 20+ stage pipelines (pipelining reaches peak efficiency at 15 stages, after that point it loses efficiency due to misses, prediction cost, and other terms whose precise name escapes me), which I can only assume existed to make the GHz rating higher (and dual cores aren't advertised in such a way now). Plus, y'know, march of technology and all that.
All of which is beside the point. The point is (or was), multi-core processors don't have the software support and thus performance to be worth the money, relative to the cost of an equivalent single core. There's oodles of benchmarks out there for that, that show individual applications aren't helped (or insignificantly at best) by multicore. Whereas if you scaled 'up' the performance a single core (let's assume it's technologically feasible for point of comparison) to the same price point, you would see a more signficant advantage.
But i'm still buying one because eventually in a few years time there will be the (necessarily ubiquitous) software support to make them properly good value, and at that point I should be able plop out my then-outdated chip and hopefully slot in a nice new quad or perhaps even oct core one of the same socket type.
(FL7 is multi-threaded; IIRC multiple cores are especially suited to DSP tasks)