I find it funny that a 5.0GHz CPU is still unreachable. After 10 years they only could make CPUs speed 50% better (from 3.0 GHz to 4.5GHz).
From 1995 to 2005 it was from 60MHz(?) to 3000MHz which makes the advancement 5000%.
AMD have a 5 GHz CPU, actually. The FX 9590 gets to 5 GHz out of the box. However, it's a total waste of money. If you want AMD for a budget gaming/video encoding rig, you shouldn't spend for more than an FX 8320.
Anyways, the main issue at play is heat and power. There is nothing stopping AMD or Intel releasing a 7 or 8 GHz CPU, except for the fact you'd need Liquid Nitrogen to cool it and a massive PSU to power it. This is primarily because of transistor count. As transistor count in CPUs increase in accordance with Moore's law, the heat the CPU produces rises just as much if they are operating higher than certain speeds. We only need look at how badly CPU heat and power consumption scales when overclocking, especially on the AMD side. 10-20% increase in clock speed, a 50-100% increase in power consumption.
Besides, because of the higher transistor counts, the need to increase clock speed is greatly reduced because CPUs are capable of powering through more workloads in the same amount of time.
Linked to the previous, another way to increase performance without increasing clock speed is parallel processing through multithreaded software. Parallel processing is a way to reduce the amount of total time used to compute tasks since it's spread out over different cores. GPUs are actually a fantastic example of this in action. A GPU has much more horsepower than a CPU since it utilises thousands of small processing units at once. This is why people mine Crypto Currency like Bitcoin with GPUs instead of CPUs. Actually, most serious miners use ASIC cards now but still.
Think of it like a road and your goal is to get as many cars through as possible in your average day. Clock speed represents a speed limit. While increasing the speed limit might be nice if you want to increase productivity and traffic flow, it won't be very effective because it's all running in one lane. Introducing multiple lanes (parallel processing) and smaller cars (transistor count) to handle the traffic produces better results without needing to raise the speed limit.
Basic explanation, I'm sure someone with actual knowledge on the subject could explain better than I could.
But the point I'm making is that I wouldn't use pure clock speed as an indicator of performance. Something running at 4 GHz might be slower than another type of CPU running at 2.5 GHz. Plenty of factors at play.
Anyways, how about that downgrade? heh heh heh
Retour en haut
So I might flick through that, mainly to look at concept art pieces for Ciri and Triss. And Connifer of course, can't forget good old Connifer of Vengerberg.
