• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Huge CPU.. Why not?

A

AMDoido

Limp Gawd
Joined
Sep 8, 2002
Messages
195
Well...
As i saw in my last thread that some people here do not speak english very we :rolleyes: so i´ll try to explain very carryfully...

If the AMD or Intel do BIGGER cpu (the double size of today´s cpus) would the be more powerful?
And why dont they do it?

Like the video cards... i can´t tell for sure... but today´s cards are bigger then the older ones...
 
It looks like you are the one who does not speak english very well. I really have no idea what you are talking about. Processors have steadily been getting smaller and that is how they have been able to make them faster. They make everything smaller so the signals have a shorter distance to travel, which lets them fire the signals faster. And today's video cards are smaller than the ones from back in the day. The used to use a shitload of logic chips, but now all that is down to just a few chips.
 
I got no idea what you are talking about.

Videocards used to be massive, almost double the size of todays cards.

Are you saying that the physical size of a cpu dictates how "powerful" it is???
 
I always thought they were able to fit more (switches or something?) in a chip thats the same size. but I'm not even sure about that, much less know anything else about the workings of a CPU.
 
If you make a bigger CPU, the electricity has to take longer to travel, so it's slower. If you can fit the same amount of transistors and whatnot on a smaller chip, it's faster.

Essentially:

bigger = slower
smaller = faster

Although, the size of the CPU is not the only factor in it's speed. CPU's have nifty little things on them that make them faster, like a prefetch queue that fetches a bunch of bytes ahead of the current instruction at once to minimize the time of getting them.

But that was implemented in most processors years ago, and the primary speed determiner is the size of the processor.
 
Originally posted by AMDoido
Well...
As i saw in my last thread that some people here do not speak english very we :rolleyes: so i´ll try to explain very carryfully...

If the AMD or Intel do BIGGER cpu (the double size of today´s cpus) would the be more powerful?
And why dont they do it?

Like the video cards... i can´t tell for sure... but today´s cards are bigger then the older ones...
Click to expand...
Spelling: B-
Grammar: D-
Overall grade: C-
 
Originally posted by Arkenzo
Since when did the [H] start giving out report cards?
Click to expand...

Wen the viktom started kritizizing uthers about how they speek english. :)
 
Hey we should give him a break. English could be a second language for him. In which case hes is obviously still learning to speak it. Which brings up another pioint. He would speak two languages (albiet he sucks at one of them). However if english is his first language then he needs to start over cause something, somewhere seriously got FUBAR'd

Now back to the topic at hand...

General Rule of Thumb-Bigger, is not always better. A larger core would do absoulutely nothing to make a processor faster. The trend is as things get smaller they are getting better and better. Look at it this way, 45 years ago a computer would have taken up a good city block and probably two stories high. That little box sitting next to you is about 500 times more powerful than any computer built back then. Which one do you want?
 
I think what he is getting at is more pins.

more pins = more bandwidth to motherboard...

Also if AMD went in and put two 2500+ cores side by side and doubled the pins and then put double the cooling it'd own anything out there. It's basically like putting two cpus into one, less expensive than dual boards/cpus to make and double the speed. I'd buy it.

~Adam
 
Originally posted by CleanSlate
I think what he is getting at is more pins.

more pins = more bandwidth to motherboard...

Also if AMD went in and put two 2500+ cores side by side and doubled the pins and then put double the cooling it'd own anything out there. It's basically like putting two cpus into one, less expensive than dual boards/cpus to make and double the speed. I'd buy it.

~Adam
Click to expand...

No, I'm pretty sure you're giving him too much credit. And while the size of the card might have gotten bigger, the cores of the cards have gotten tiny. The video cards simply need more real estate to be able to cram all the memory and power components onto a card, as well as space to cool everything.

What I do think he's talking about (more or less) is not simply making a larger chip, but increasing the transistor count, so you have a core 2x as big as a p4 with 4x the transistor count. And the cpu cores don't scale like that, but slapping on oodles more l2 and l3 cache on die (ie in the core) can help, as show with the p4ee.
 
The whole point to making CPUs smaller is to be able to fit the same amount of transisitors in a smaller area. The reasoning behind AMD and Intel not making their CPUs "bigger" is because it would require an entire redesign of the core. When AMD created the Athlon (K7) core, it planned on sticking with that design for awhile, the same goes for Intel and their P4 core. Over time the cores have indeed become smaller (.17 micron to .15 to .13 to .09 soon), but the reason for this is because they can make the transistors that make up their core smaller, not to fill the same space with 2x the transistors. For example, the exact same transistors and layout that made up an older .17 micron K7 core can be fit into a much smaller area using a .13 micron process which requires almost no redesigning on the part of AMD (less development cost).

Sometimes Intel and AMD will make use of this extra space gained by using a smaller process by introducing relatively "cheap" redesigns in the core by the way of cache enlargements. Take a look at the Barton core or the P4 EE, they are exactly identical to other AthlonXP or P4 cores except they have extra L2 or L3 cache on-board. This is because memory is pretty much the easiest layout and design of transistors they can engineer. If you see any "new and smaller process" for transistor creation in the news (ie- recently Toshiba and IBM have announced .065 micron parts) you can be almost positive that these prototype parts are memory. These require almost no logic on the part of the cicuitry, only that they be able to store an electrical charge that can be read as 1's and 0's. To add extra logic and processing transistor power to the CPU would require intense engineering much harder and more costly than AMD or Intel would be willing to spend because it would mean you would have to pay more to buy the chips as well.

What I think you are getting at is actually where CPUs are headed, however; what you speak of is similar to the "multi-core" approach that Intel and AMD will both be using in the near future. This method of design involves placing two of the same core on the same CPU die. Theoretically it works like this: in place of a CPU die the size of a .15 micron transistor based core you can fit two .09 micron based cores. This has the possibility of doubling the power of the CPU without taking up much more space than CPUs did a few year ago with a single core. Not only does it make it a mutiprocessor system on one chip, but it also allows AMD or Intel to reuse the same core designs without significantly changing the design of the cores themselves. That is, instead of buying two separate CPUs to put in a system and needing a multiprocessor capable motherboard, these CPUs will have that same functionality using just one CPU package. This can save both the manufacturer and the consumer money.

I think this is what your question was asking originally, if it wasn't sorry for rambling.
 
Altho adding transistors of any kind is quite expensive, and when you're talking about adding 2mb of on die cache, thats a pretty penny, the p4ee does really cost a lot more to produce than a p4.

And the "problem" with multicore designs at this point is heat, a dual core .09u processor, assuming all things are relative, would put out in excess of 250W, with something like 100mm^2 of contact area.
 
Lollerberries....

I took out my voodoo5 and put in a gf4. The voodoo5 stretched from the back of case to under my hard drive. It was massive. The gf4 is a fraction of the size.

As for CPU's... the covering is many times the size of the actual chip anyhow. Unless you tear it apart and measure the dimensions of the chip itself.. simply looking at the covering is worthless.
 
Well, the VIA chips that put out very little heat could be put into a quad setup on-die, making heat less of an issue. I mean AMD and Intel both are looking to make their cores less hot, right? Why not create a core that's very very cool(or use VIA's 1 ghz design) and just stack 'em on the die in large numbers to get insane clock speeds(with very little heat creation). Hell I'd buy that too. This would prbably be more expensive to do, considering the engineering cost of getting a nearly room-temp high ghz processor to stack up. Seems like fun to me.

~Adam
 
Originally posted by CleanSlate
Well, the VIA chips that put out very little heat could be put into a quad setup on-die, making heat less of an issue. I mean AMD and Intel both are looking to make their cores less hot, right? Why not create a core that's very very cool(or use VIA's 1 ghz design) and just stack 'em on the die in large numbers to get insane clock speeds(with very little heat creation). Hell I'd buy that too. This would prbably be more expensive to do, considering the engineering cost of getting a nearly room-temp high ghz processor to stack up. Seems like fun to me.

~Adam
Click to expand...

dude your last post was aight, you should know better than to equate ghz with performance.
 
Ahh yes, well assuming that more mhz would mean more performance (because it would if you went the AMD way, maxing your performance to mhz ratio) then that would be the way to go, no?

~Adam
 
Via's Cyrix chips do indeed run relatively cool, but they're SLOOOOOOW. 1GHz Cyrix != 1Ghz Athlon/P3. And Intel and AMD processors are not running cooler, they're getting hotter as we go, which is one of the biggest problems facing future die shrinks, the heat disapation, when the contact area gets tiny, and yet has a huge thermal output.

Also, the high end V5, if memory serves, had 4 graphics cores spread out along its length, which meant it was basically 4 graphics cards in one, explaining its huge size.
 
has anyone ever used a velocity 4400 video card? they chose to use a larger sized chip, and even with a 1" tall 2" wide fan on the stupid thing, the paint is discolored on the heatsink from how hot it gets.
 
More transistors not only uses more wafer, leading to fewer dies per wafer, but it also increases the odds of having a defect in an individual die, which further decreases useable yield per wafer.

Both Intel and AMD already sell CPUs that have portions of existing cache disabled, presumably due to defects in the disabled portion of the cache. They can salvage (with a lower selling price) dies with defects in the cache, but if something gets borked in the processing portion of the die that there's much hope of salvage... A core could be designed that way, but would the ROI be worth it? Probably not.

I imagine that when dual core per die CPUs become widely available that at some point versions with only a single fuctional core will be sold (at a reduced price) to increase overall yield.
 
Originally posted by Engi
Um no... it travels at the speed of light...constant.
Click to expand...
I think he meant that since the signals have to travel farther then the chip will be slower, not the electricity. And the speed of light is constant in a vacuum, but then when you put it through other mediums such as air and water it goes slower.
 
Originally posted by jpmkm
I think he meant that since the signals have to travel farther then the chip will be slower, not the electricity. And the speed of light is constant in a vacuum, but then when you put it through other mediums such as air and water it goes slower.
Click to expand...

Yes, it has to do with resistance, not speed of anything.

~Adam
 
You must log in or register to reply here.
Back
Top