Is Moore's Law irrelevant these days?

[barkmonster]

I know it's about transistor count, not performance but if doubling the transistor count doesn't double CPU power, it's no indication of anything but fabrication bragging rights. If you look at CPUs over the past few years, they're hitting a brick wall of trading clock speed for core-count or vice versa. Every CPU from laptop i7s to Xeon systems float around the same performance in cross-platform benchmarks at the low end, the i7 often comes out on top because it can ramp up it's clockspeed. On raw CPU benchmarks like Geekbench and there just isn't the big leaps in CPU power every 3 or 4 years like there was. From the tests on recent Haswell-based PCs/Macs, it's just token increases of maybe 10% more CPU power from one generation to the other coupled with lower electrical requirements and better onboard GPUs, the Xeon workstation CPUs are only powerful because they can have 10+ cores and multiple CPUs.

For cost reasons, I've had long waits between upgrades, I've only had 4 computers since having an Atari ST till the 90s and that means a big leap every time because I went from an 8Mhz 68000 with 1Mb > 300Mhz G3 with 192Mb > Dual 800Mhz G4 with 1.5Gb> 2.53Ghz Core 2 Duo with 8Gb. Even if my next system is at least a 2.6Ghz Core i7 or the 2013 Mac Pro equivalent to a 3.2Ghz Xeon, the one after that won't offer 3 or 4x the CPU power and it won't matter about GPU power because it doesn't affect any of the software I use and if I was that bothered about games, I'd own a console.

 

[maflynn]

I'd say Moore's law which was remarkable at how long it stood the test of time but how much more transistors they can cram into a small sliver of silicon.

From the wikipedia page it appears to be in force but I'd say we're probably at the tail end. I'd also say we're probably beyond the point of where more transistors translates into raw power as well. There are other competing forces that impact performance at this stage of the game

From a performance perspective, my 2009 Mac Mini with sufficient ram and SSD performs many of the tasks that my 2012 rMBP does. Sure the more demanding apps like photoshop, aperture, etc benefit from the faster cpu/gpu but my point is we're not seeing huge improvements in performance. Moore's law did translate into more raw power in the early days but now as the technology has matured the speed gains are less dramatic.

 

[barkmonster]

I'd say Moore's law which was remarkable at how long it stood the test of time but how much more transistors they can cram into a small sliver of silicon.

From the wikipedia page it appears to be in force but I'd say we're probably at the tail end. I'd also say we're probably beyond the point of where more transistors translates into raw power as well. There are other competing forces that impact performance at this stage of the game

From a performance perspective, my 2009 Mac Mini with sufficient ram and SSD performs many of the tasks that my 2012 rMBP does. Sure the more demanding apps like photoshop, aperture, etc benefit from the faster cpu/gpu but my point is we're not seeing huge improvements in performance. Moore's law did translate into more raw power in the early days but now as the technology has matured the speed gains are less dramatic.

Have a maxed out 2009 Mac Mini booting off an SSD. I thought it was quick in Snow Leopard but it's even faster still with Mountain Lion. The thing is, with modern systems all getting either PCIe SSDs or 6Gb/s SSDs, once I do upgrade, the raw CPU power I'd need to upgrade again would be something like the current top end Mac Pro but for Mac Mini prices.