Intel Says Goodbye to 4GHz P4 - The Core of the Problem
(Page 4 of 4 )
The problem is when you decrease the feature size of a transistor, some things go down, others increase, and this isn't always a linear process. Going smaller requires less voltage; you don't need as much of a kick to go from a one to a zero, or zero to a one. This means your switching loss also decreases, so you use less power per transistor. But when you go smaller, your leakage current increases. That's the electricity that burrows its way through the channel of the transistor, and it increases on an exponential curve with transistor size decrease. So while up until now leakage current increases have been relatively insignificant (you can hardly measure it at 250nanometer), at 90nanometer, leakage is the dominant power loss. And that generates heat, lots of it.
Another interesting finding is that the normal voltage drop doesn't work. 90nanometer hammers run at 1.4V. 130nm ones are at 1.5V. Those 90nm ones, if things kept up like they have for previous shrinks, should be at 1.2V. Same goes for Prescott, which will only operate at that 1.2V when you decrease the speed significantly.
So instead, everyone is going "dual core", and eventually multicore. Sun saw this a while back, which is why the UltraSPARC IV uses two USIII processors jammed together on the same die. But before Intel can release one of these, they need to finish designing it, and make sure it works. So they're using a stopgap procedure of increasing cache size.
Essentially, they're bringing the EE to the masses, without the price tag. One of the nice things about 90nm (and 300mm wafers) is you can do stuff like that without making your CPU stupidly expensive. Intel will also enable x86-64, likely once a real Windows version that supports it comes out.
Getting back to dual core, is this really an answer? If the software out there that needs more processing power is able to make itself parallel using multiple threads, then yes. As can be seen in database software, more CPUs really are better, to the point that dual processors can show almost a complete 100% advantage over their single CPU counterparts. The same goes for 3D rendering software, and programs like Photoshop and After Effects, when they aren't limited by memory or disk access. Therein lies the problem. CPU power has come so far, so fast, it's left the rest of the computer in the dust.
Dual core isn't going to make the Internet faster (from your end). It's not going to make you type in Word faster. But with certain software, after it's been optimized, it will show advantages. Intel, AMD, IBM, and Sun aren't doing this because they want to; they have to. And everyone else is going to have to play along for it to work. Just something to keep in mind.
| DISCLAIMER: The content provided in this article is not warranted or guaranteed by Developer Shed, Inc. The content provided is intended for entertainment and/or educational purposes in order to introduce to the reader key ideas, concepts, and/or product reviews. As such it is incumbent upon the reader to employ real-world tactics for security and implementation of best practices. We are not liable for any negative consequences that may result from implementing any information covered in our articles or tutorials. If this is a hardware review, it is not recommended to open and/or modify your hardware. |
