computers of future discussion

[jasong]

Maybe it's my imagination, but it seems like sometime around the turn of the century, the Hz-per-second increase went through a major slowdown. I was just wondering if people agree with this observation.

Also, where do you think computer technology is headed? For instance, I just bought a Tivo recently, and I have all my savings for the next 2-3 months going towards networking and buying another Tivo with more capacity. The thing is, the Tivo only has an 850MHz something-or-other under the hood coupled with a specialized graphics board, so it's super-specialized for what it needs to do.

So, where do YOU think computers are headed? Do you think this slowdown is permanent, or do you think new technology will put things in high gear again(like nanotubes might in the next few years).

What about GHz? Do you think we'll see 10-15GHz anytime soon, or will the advancements be in a separate area?

Post your thoughts.

Edit: typo

[rshepard]

I think two things are happening
1. The current technology has just about peaked- I recall an article I read a while back that said something to the effect that heat was becoming a limiting factor; they can build faster processors, but there's no way to cool them adequately/economically.
2. The demand for faster machines is starting to level off--- present company excepted, most people don't need a 10GHz box. Until the next "killer app" comes out that needs that much power and that everyone simply must have, people aren't going to be that interested in more power.

[Fozzie]

that they have gone as far as they can go with Silicon and are looking at other materials and methods of miniturisation.

I think the scientists in question were ooking at creating circuits at a molecular level in carbon.

Before you say you're taliking :bs: I am going on something I seem to remember from last year.

It said that due to the slow down in processor development there was a significant knock on effect to chip suppliers and manufacturers and all the associated industries that would cause an economic implosion similar to the last stock market issues a few years ago. Except it won't be the dot.coms that get hit but the hardware and ancilliaries that will.

[jasong]

Originally posted by rshepard
The current technology has just about peaked- I recall an article I read a while back that said something to the effect that heat was becoming a limiting factor; they can build faster processors, but there's no way to cool them adequately/economically.

I can certainly agree with that. I read an article awhile ago where the main draw for DCers(it was a DCing forum, I think) for a certain chip was that it used .09 nanometer(Edit: later post says it should be 90)technology which gave you the same performance with less concern necessary for heating. Basically the circuits were thinner, but just as many, which apparently helped with cooling. Wish I knew more.

Edit: for clarity

[CaptainMooseInc]

check news.com.

Intel is stopping in making huge jumps in GHz. Instead they plan on increasing the cache on the processors.

Theoretically they claim they can run a 1GHz processor with a large cache, use less electricity and have less heat, and it'd run the same as a 3GHz processor.

So maybe instead of increasing GHz Intel can just keep resetting the clock at 5GHz and add more cache. :)

Eventually there'll be too much cache on the processors to make any difference in the chips and then they'll move on with GHz again. ;-p

Just give me a quantum computer dammit!

-Jeff

[PY 222]

jasong, the decimal is indeed wrong. It should be 90 nanometer.

For what its worth, the Gigahertz myth that has been pushing Intel and AMD for so many years have started to wane with the public as people get more educated about how processors work.

AMD got out of the GHz race when they moved to the rating system, ie. 2800+ 3200+. Now that Intel knows that its getting more and more difficult to pump out more GHz without a giant heatsink, they too have started a rating system.

I believe IPC (instructions per cycle) is what that matters in future processors. The higher the IPC, the more efficent the processor is at processing information. So, in light of that, I doubt we will see anything in the 10GHz range for a long time but rather multicore processors with better architecture (Pentium M comes to mind).

[jasong]

What about the FX processors? I generally decide what Athlon to buy by the XXXX+ rating, so I have no idea what the value of the FXs are.

Is the architecture so radically different that they wanted a new(FX "XX") rating system. Or maybe they're so radically overpriced that they wanted to confuse the issue?

Educate me, please?

[PCZ]

The rating Sytem is abused by AMD.

A current case in point is the Sempron which everyone knows is Athlon XP renamed.

Sempron 2200+ 1500mhz
Athlon XP 2200+ 1800mhz

So the bus speed is 166mhz for the Sempron against 133mhz for the Athlon XP, that doesn't make up for the huge difference in clock speed.

[jasong]

Does anybody think that we'll ever have economical RAM that can remember things even after being powered down?

Lately, I've been wondering why they don't make computers that boot from flash memory. If you're not a DCer, the ability to boot up and down quickly would be wonderful. You could keep your "main" stuff in Flash, plus backed up on the hard drive and keep the large, not technically necessary, files on your hard drive. That way, if all you want to do is passively surf the Internet or watch a DVD, the hard drive wouldn't need to power up at all.

Doesn't that sound nice?

[Shish]

There are some new memory technologies getting close, primarily from Rambus and Hitachi with even more radical stuff coming out of IBM and several of the other big research labs which is a bit further out.
The processor race as far as Ghz has more or less topped out as they`re looking at what they can get the consumption, therefor the heat produced, down to somewhere reasonable.
Efficiency of processors depends more on their design than their speed and has a lot of things to take into consideration including the length of the processor "pipe", the fpu and caches etc., and memory needs to catch up significantly to the processor speeds to be more efficient. Hence the expansion of the caches and that makes cache prediction an important part of the cpu process. That also means bigger caches can speed things up cos the steps required in order of speed are cpu to cache at cpu speed, then second level cache then memory then hard drive. As you get the memory access times down, so you increase the speed of the whole process.
As far as a "fast" system goes, you need to look at the whole of it to optimise all the parts. So a faster system can be achieved by speeding up from the slowest parts, which in order are - Hard drive to memory to cache to cpu.
One of the reasons some pro systems are better than others, all things being more or less equal, is that they have large amounts of memory so obviating the need for the very slow, in comparison, disk access. Plenty of memory means getting rid of the swap file on the drives then bigger cache takes it one step further by allowing you to have more chance of having the data the cpu needs when it needs it and so on.
That was one of the reasons that Xeons (PII and PIII) with the larger caches were so good at Seti and some other DC projects. They could cram their cache with almost a full chunk of the data they were working on and didn`t need to keep going back and forwards to memory and drives which were significantly slower than they are now.
But memory speed (and FSB) still have a way to go to catch up to cpu speed so I guess we`ll be playing catch up for now and I reckon that`s where the advances will be soon.