Niode wrote:Yep and in 10 years time when we look back at our quad core 3.6GHz systems with 4 GB ram and 2 TB hard drive space. We will laugh that we paid £500 for all of that when £250 will get all that on a device no bigger than todays mobile phones.
I hope you're right, but I'm becoming pretty pessimistic about Moore's law holding too much longer.
On one hand I completely agree with you. We do seem to be stagnating. However moore's law also covers the price of technology. So the technology could stagnate in processing speed/power (as in the number of transistors on an integrated circuit) but it still fits in regards to the shrinking/affordability of ICs. So it's not particular a failing in moore's law, it still works. We're just going through a phase of shrinking/refining before the next big leap in technology.
I do feel we have hit the theoretical maximum for transistor technology. It's nigh-on impossible for chips to hit speeds faster than 4GHz stable at reasonable temperatures. It's one of the reasons why clock speed has been dialled back and we are shooting for more parallel processing power - IE quad core, six core, eight core, twelve core etc. However, only certain operations benefit from parallelization so that is why we haven't seen the huge jump in speed that we have seen in previous years.
Anayo wrote:I have to admit what made me post a topic about this was reading the article about "scariest unsolved mysteries episodes" on X-entertainment. XD
Yes unfortunately not everything is parallelizable. It is a very interesting topic, back in school we used a library for C++ to write code to run across a cluster. Some of the issues at hand:
1) Finding a problem that can take advantage of multiple cores (or physical machines)
2) Appropriate way to break up the problem space
3) Ability to intercommunicate between cores or machines
I wrote a clustered bruteforce password cracker as my final project for the class. I took a simplified view of the world and assumed the password was 8 alphanumberic characters. You can then define the keyspace from the possible permutations of the characters. Then have the lead program give out work to worker threads would attempt to crack the password. Finally, communicate back to the master if they found it, or ask for more work.
I know this is getting completely off topic but thought I would share my ramblings.
E: Few things are stupidly parallel, ie, can take advantage of a ton of cores thrown at the problem. Certain math, science, weather calculations etc are stupidly parallel. But I'm sure the crowd here mostly cares about gaming, which certainly has parallelizable features.
Ziggy587 wrote:That's the way computers (and most electronics) work. They get smaller, faster and cheaper.
I predict that in 10 years computers will become twice as powerful, ten thousand times larger, and so expensive that only the five richest kings of Europe will be able to afford them.
Yes unfortunately not everything is parallelizable.
Amdahl's Law is a bitch.
Blizzard Entertainment Software Developer - All comments and views are my own and not representative of the company.
That's another thing, use for the consumer is quite finite as far as parallel computing goes. Aside from video games and video rendering and encoding what else is there that benefit from it?
Quite a lot of creative tools don't particularly benefit from access to multiple core environments. Take a look at benchmarks for Cinebench. That benchmark much prefers higher core clock speed than multiple cores. A 5 year old 4GHz dual core CPU will beat a 1 year old 4 core 2GHz CPU hands down. It favours higher clock speed. A lot of software out there just hasn't been coded for parallel computing yet. With the birth of CUDA openCL ATI Stream etc we will see much more software built to take advantage of all the Parallel Processing Units in a system (PPU). One thing I like about the new OS X is the grand central dispatch, it designates parts of the system that will handle certain tasks better than others. Optimising the use of the computer to complete tasks as fast and as efficiently as possible.
I think desktops and laptops used by the service industry have been serving in the same capacity for many years now, with the real limit being humans' limitations when it comes to multitasking and the outdated input mechanisms. How about this to increase efficiency: A thread for Facebook, another for Sharepoint, and a third for Solitaire
Pulsar_t wrote:I think desktops and laptops used by the service industry have been serving in the same capacity for many years now, with the real limit being humans' limitations when it comes to multitasking and the outdated input mechanisms. How about this to increase efficiency: A thread for Facebook, another for Sharepoint, and a third for Solitaire
The pricing is still so crazy on technology stuff. $250 for a ipod, which isn't the most impressive bit of technology, though it is small. Compared to $300 for a PS3, which has one of the most impressive processors around inside of it.
