What do you really want from a CPU?

Not really, the name is desktop for a reason, it's built for non-portability so that's not a fair disadvantage. Like a PS2 and PSP, the PS2 is made to be stationed a television and not carrying around, you get much more power for doing so.

 

1. Don't inhibit my GPU's performance.
2. Don't overheat.

All I ask.

 

I want 8 cores, 5Ghz per core, 1GB L2 Cashe, 25nm, max at 15C, be smaller, and to run off the power of its own heat.

And cheaper prices...

 

Yeah, sure.

 

I don't understand what you mean by "not a fair disadvantage" in this context, but never mind -- the point is that the most important reason for people to use laptops is that they're portable. There are others; for example, laptops use far less power than desktops, but the most important one is that one can easily ferry most laptops between work and home, take it on trips and so on whereas this is not true of a desktop.

This was true in the beginning, when only people with a corporate budget or a military one (the US Air Force bought a whole lot of them) could afford them, but I don't think it is quite as true anymore today. I work in a scientific lab and everyone (no exceptions!) I know of who works in our field has at least one laptop. It's gotten to the point where if you go to a conference, they don't even bother collecting the presentations anymore -- you just plug your laptop into their screen and you're good to go. Similarly, I don't remember a single college student who lived on campus and went without a laptop.

Only if you're willing not to install any new applications on it and use an outdated OS. If all you want to do is browse the internet and basic word processing, then yes, you can get away with having the same computer for a decade (not more than one because they break and good luck finding parts or expertise for fixing 10+ year old hardware). But if you're going to use it for photographs or video (even at the casual level) or programming or anything of the sort, you'll still want to buy a new PC once every 5-7 years.

It's true that this is not the ridiculous pace at which gaming PCs must be upgraded, but "decades" is an pretty large overstatement.

It sounds like you underestimate the number of scientific, military and business applications for computing. It's not just mathematicians who need faster CPUs, it's everyone who relies on a computer model of anything non-trivial. Whenever you want to simulate something from real life, you're going to need a CPU far more powerful than it is currently available regardless of whether it is a biological process, the weather, a bomb explosion or a particular corner of the stock market. Because a partially accurate simulation is better than none, we make various approximations and drop higher order effects, but the resulting models rather far from what people actually want and faster processors go a long way towards alleviating this.

It is possible that at some point in the near future, people who don't play games will no longer need faster processors for personal usage, but I can't see that happening in professional settings.

 

Poooowwwwerrrrrr!!!!!!!!!!

 

Right, that's what laptops are good for. But desktops are not made for portability, so saying that's a disadvantage is redundant. Laptop power doesnt compare to them. Same as saying an 82" HDTV is not portable so that's a disadvantage since my mini dvd player can be carried around anywhere. But picture quality also doesn't look nearly as good. Well, duh? No point restating the obvious really.

 

When CPUs can do wPrime 1024m in less than 1 second, we've hit the limit of silicon technology.

 

I can't argue with you Althernai - very well put.
+ Rep for you.

 

The point where disagree is that laptop power does compare to desktop power. It's true that desktops are more powerful (both on average and at the high end), but the difference between the two is not like the difference between an 82" HDTV and a mini DVD player. I would estimate it to be closer to a factor of 2-3 overall (including both performance and price) as far as most components (CPU, RAM, hard drive) are concerned.

Thanks.

 

The point is, you can get a desktop for $800 that outperforms by factors of 1.5-2x what a $2500 laptop can do, and for $400 you can get one that outperforms a $1000 laptop. Lack of portability is presupposing that a 10 pound, 2 and a half inch thick laptop is 'portable'. The only place you can compare is at the bottom end.

 

Yes i think so.

But by that time, we will have Hyper-Silicon.

 

I'd want adaptability, as in, the CPU dynamically adjusts (I don't mean discrete jumps, I mean continual adjustments) its power to handle tasks.

Then again, this is just to save power and heat.

 

The technology is already in CPUs. It's called SpeedStep on Intel CPUs and "Cool n' Quiet" on AMD CPUs.

 

I think Intel speed step is discrete and not continous.

But then:
Wouldn't adding too many steps or even making it truly continous use more energy than 1 or 2 steps?

 

Really? Can you explain in exactly for what purposes the desktops are that much better than the laptops? I can think of very few, the main one being gaming. Also, I was not talking about the 10 pound laptops, but rather the standard 15.4 inch ones or even the ones that are 14 inches and they shouldn't cost more than $1500.

CPU-wise, the desktops have a substantial advantage in that it is far easier to get quad core processors, but since very few things fully use them at the moment, this is not that big a deal. By the time support for more than 2 cores becomes common, I'd wager we'll have quad cores in mainstream laptops (they're a year or two away at most while the software will take at least 3 to fully catch up). As things currently stand, the dual core laptop CPUs are weaker and more expensive than their desktop counterparts, but not by that much -- I'd say a factor of 1.5 to 2 overall.

In terms of RAM, there is no practical difference. The desktop RAM is faster and you can stuff an insane amount of it into the newer motherboards, but you'd have to be benchmarking just to notice a difference.

Desktops win by about a factor of 2 in hard drives -- you get more capacity at the same speed or the same capacity at higher speeds. This is evolving pretty quickly though and once the SSDs take over the difference won't be that large.

That pretty much covers what the majority of computer users (among the general population, not on these boards) care about. I've left out the one component for which the performance is a full order of magnitude better in desktops at the cost of consuming a full order of magnitude more power and thereby generating that much more heat, but that situation is rather perverse and it isn't likely to last long.

 

What it really comes down to is personal intentions, so both are flawed and have their own advantages. There are hundreds of companies who use desktops for professional work/research, whatever the industry may be, even the gaming industry. Core usage is mostly up to whatever applications the user is using, so it's not that they aren't a big deal, they are just not a big deal to you. Video producers and 3D modelers in Hollywood don't complain when they are using a system with more cores, and it is definitely a big deal for them, and once again not you (hence you feeling it's not a big deal, when in fact, it just isn't a big deal to your world).

When laptop quad's become the mainstream, desktops will have octocores, or six minimum (which is releasing next year) mainstream for cheap. The current QX9300 is over $1000 retail, at 2.53GHz and once overclocked will likely cook an egg. The current Q6600, which costs $190, can be easily overclocked to 4GHz, and even though it generates more heat, it's a solid performer and unlikely to fry your system with a heatsink. I've still yet to understand the justification of laptop quads and dont think it will be getting better anytime soon. Intel i7 mobile doesn't sound cheap, especially with the current QX9300 being so much.

Dunno about this one, my new Antec 1200 case can hold 12 hard drives, at the capacity and speed I want. A typical laptop only holds 1x500GB, if you're lucky, and at an expensive cost. Desktop HDDs also go past 10000RPM, if necessary, while laptops don't, and at terabytes of storage for whatever the heck I need. And a WD Raptor will outperform many of the SSDs out today.

True, electricity bill is higher, but heat is something you don't really care about or complain about with a decent case and cpu heatsink. Even so, it's still up to the user to determine whats better for him.

 

Well, the idea for me is that once laptop quads come down in price sometime next year/decade, I'll pick one up and help keep my laptop going for that much longer (running Windows 7 by then).

Guess where all that heat is still going? Not a problem in cold climates but in Singapore or other hot places... Further air-con bills here.

 

This is actually a problem. CPUs stopped at 3 issue(4 for Core 2) because there are codes that are inherently not parallelizable instruction-level so they went for more cores. But now experts are talking about programs that might not get much beyond couple of threads, having a limitation on TLP.

Increasing single thread performance is very hard. Very parallelizable codes are akin to having a computer add numbers like 1+1, and very hard to parallelizable codes are having a computer tell how many fruits are there when its apple+orange. They have different colors, slightly varying sizes, different weights, nothing like adding numbers.

Clockspeed is one way to increase single thread performance relatively easily and consistently. Future architectures like Sandy Bridge will continue to focus on enhancing the Turbo Mode(dynamically increases clock speed) on Nehalem to increase performance while abiding the power consumption limits.

 

I have been wondering for a while now... why doesnt Intel create a single socket for both desktop and notebook? Is there really any difference between a 2.0Ghz desktop and notebook CPU?

That way we benefit from much lower TDP, heat and noise.

 

I think desktop CPUs may be cheaper to manufacture.

It should be easier to design a CPU on which you know that you can waste say 20W compared to a CPU on which you know you've got to conserve energy.

 

I wouldn't say overclocking a Q6600 to 4ghz is easy. Aside from Engineering Samples and Week 39 2007 super chips there aren't a ton out there that will do 4ghz without pretty extreme measures, and the newer ones seem to be getting worse again. My G0 stepping chip does 3.85 ghz with water cooling and .3 volts more than stock, that's not easy and mine relatively speaking is a pretty good chip. But regardless 3ghz is cake on almost any Q6600, most at stock voltage, and that's still enough to cream most laptop quads. And you certainly aren't cramming a 15k rpm scsi drive or a 1tb hard drive in a laptop, let alone both. My 350 dollar G50 with 100 dollars in upgrades scores a whopping 1600 points on 3dmark03, my desktop 2 years ago scored 22,000 (x1900xtx/Opteron 165) and my current one does 45,000 (Q6600/8800 GTS). I spent about a grand on each of those rigs for 15-30x the performance at twice the price.

As for the chips, desktops have more space to waste so making a bigger chip with more pins and more features/performance that takes up a square inch or more isn't a big deal, in a notebook you've only got so much space on the board to cram a chip on and still fit all the other stuff you need inside the case.

 

This won't work because desktop CPUs are much larger than notebook ones. Then you will have to sacrifice performance or pay a lot more.