Why doesn't AMD just make a 2000mhz?

[hkgonra]

I really wonder why Amd , instead of going with the new "performace " based naming, didn't just come out with a true 1800mhz chip, think how much AMD would smoke Intel then?!

Or even go all the way to 2000 !!! I just don't understand?

[train22]

They are moving as fast as thier little financial boundries let them, believe me they are trying their hardest to screw Intel but it doesn't just happen overnight.

Expect the 2000 around Chrismas.

[hkgonra]

2000xp , or a true 2000mhz ?

[Thelemac]

Well, they haven't yet because they can't. AMD is sacrificing being able to have a high MHz to have a high performance chip. If they would have been able to get that out, they would. Even still they probably would have introduced the PR ratings..because then they'd have an Athlon 2200+ or something like that.

[hkgonra]

I have always heard that chip makers are way ahead of what they release. They use (from what I've heard) a plan that they release a new chip, then wait until sales slack off, then drop a slightly higher one, and so on down the line. Is this true?

[Pitspawn]

Yes mostly.

Chip makers ARE always further ahead than what you think. They can make very very fast processors, but cannot make them commercially cheap.

This means you could have your true 2Ghz processor, but it would cost $1000 easily and would have serious heat problems. Thats another thing that the chip makers spend a lot of time working on, reducing power requirements. This is normally done through reducing the size of the processor (usually measured in microns or nano-meters). Current sizes are in 0.18-0.15 micron. Both AMD and Intel are working on 0.13 micron sizes for even higher clock speeds and less power requirements.

[ZeroWing29]

There's rumors of a TRUE Athlon MP 1.8 Ghz (NOT just PR) so I don't see why they can't pull off 2.0 on the XPs...

What about even lower microns, like 0.10 or something does that exist yet???

[Pitspawn]

I heard of some places experimenting with 0.09 and even 0.08 micron technologies but this is CNN stuff. The only realistic fabrication size you will see in the next few months is anything above 0.13 micron.

Like I said b4, its easy to get the processor size down to really low micron sizes but making the process cheap is the hard part. The military have some really scary stuff when it comes to processor technology.

[CrackedSKull]

also the .13 micron or smaller stuff is expensive, thats why .13 just starting to be introduced and why the athlon xp is still .18 because of cost. The new P4's(soon to be 2.2ghz) will be fast, but just like when every P4 was introduced, they will carry a heavy price tag.

[DaveB]

IMO opinion, AMD simply dropped the ball on the die shrink to 0.13u. For a relatively small outfit (compared to Intel anyway), they had too many different projects going and then kept ramping up the price war, which only reduced profits, further limiting the cash they could invest. The die shrink should have been their highest priority but the got sidetracked into multiprocessing (the MP) and other efforts with limited potential for immediate profits.

So now their strategy is the stupid PR system that Cyrix tried and failed miserably with several years ago. This won't work so they must intensify their getting down to 0.13u fabrication. New processor architectures are nice, but speed is what sells and you need to shrink the die size to get there.

[Tomas]

I think this will be the dead for amd...... I think the XP may be one of the last amd cpus if they dont do something soon. A bit scary that a PIIII 1800 is cheaper than a XP 1.53g"1800+"

[mbigna]

Tomas said:

A bit scary that a PIIII 1800 is cheaper than a XP 1.53g"1800+"

I just checked http://www.pricewatch.com/

Pentium IV 1.8 GHz $244
Athlon XP 1800 (1.53 GHz) $219

[trulyred]

I've worked for companies that installs the equipment that makes the chips for the big chip makers. It's a huge chain of command ,and nothing moves without a lot of people knowing about it and giving it thier blessing.So what i'm saying is they started planning for the xp two-three yrs ago .lots of work involved. not only do they have to come up with a working cpu design but they also have build the building for the fab and order the machines that will build these chips. They do it this way in order to be profitable.Making two hundred platters a day with 1000 chips per platter takes up alot of room and costs alot of money up front to get the ball rolling.
So why didn't they just do a 2k chip? Architecture.Each chip design is different on the inside,each has it's own constraints and design limits..013u being one of them.The next design in cpu's will be @ the moleculer level.(I know an outfit thats building a fab for that right now) and from what I hear it wont be like anything you have ever seen...............TR

[mbigna]

I'm surprised that this fact isn't being mentioned in this thread, as I am sure that several people who frequent this board know this:

Whenever production of a new chip design begins, the process is rather unrefined. AMD probably makes all its Palomino's (Athlon XP's) on the same line. It is when the individual chips are cut from the wafer that the chips are tested for speed, heat, and performance--indeed if they even run at all, that they are rated and then packaged. More simply, chips coming off the same wafer may be rated 1500, 1600, 1700, or 1800 depending upon how well they test. Chips that don't rate for 1500 may be saved for later or just trashed. Some chips may rate even faster, but economics may say that AMD can make more money selling a slower rated chip at a lower price point. So, even though a chip may rate faster (meet 1800 specs), AMD may want to package and sell the chip at a lower rating (say 1600) to sell more of them. This isn't to say that AMD tested the chip at the 1800 level, just that they made sure that it passed at 1600. [I know the numbers don't mean clock speed here]. Furthermore, there may be chips that passed the 1800 level that could pass 2000 (or +) level tests. But if there are so few that could pass the more stringent tests, the manufacturer may decide to throw all better performing chips into the 1800 bin.

As time goes by, the die process is refined to a point where more 2000 (or +) rated chips can be culled at a rate high enough to meet a particular demand and warrant the announcement of a higher frequency chip. [Remember that Intel tends to announce higher speed chips before they reach a point where they can satisfy demand.]

This is why we try to keep track of steppings and such, and why we attempt to overclock at all.

[BigPappaP]

10ghz processors and higher are completely possible, its government restrictions that hold it back. If companies would produce 20ghz processors that performed 30gflops then any country could go pick one up for about a million dollars, design new weapons with it, bomb us. I think this is a good idea, for one there really is no reason for us to have computers that fast - current software is expensive enough and even now doesnt completely use. Some of us would like to have it for bragging rights but its pretty much useless.

[trulyred]

your right ,the fed gov does put restrictions on cpu speed ,but they do relent,as evidenced by the speeds we have today.I remember when they worried about the 486 and the capacity it had to crunch the numbers need for "bomb" making . and that they really didn't want it out on the market . But I guess now we dont have to worrie about that anymore do we? Our adviseries are more street smart than they are scientific smart which is turning out to be quite the curse ,which makes them unpredictable and harder to catch.....................TR.

[Tomas]

Originally posted by BigPappaP
10ghz processors and higher are completely possible, its government restrictions that hold it back. If companies would produce 20ghz processors that performed 30gflops then any country could go pick one up for about a million dollars, design new weapons with it, bomb us. I think this is a good idea, for one there really is no reason for us to have computers that fast - current software is expensive enough and even now doesnt completely use. Some of us would like to have it for bragging rights but its pretty much useless.

Would have helped alot for video encoding:) which is what i mainly use my pc for.

[mbigna]

BigPappaP said:

If companies would produce 20ghz processors that performed 30gflops then any country could go pick one up for about a million dollars, design new weapons with it, bomb us. I think this is a good idea, for one there really is no reason for us to have computers that fast - current software is expensive enough and even now doesnt completely use.

Tried to run Aquanox yet? What about Unreal][ or Unreal]|[ when it comes? You give me the GHz, and I'll find a way to use them.

[Alex99]

Originally posted by hkgonra
I really wonder why Amd , instead of going with the new "performace " based naming, didn't just come out with a true 1800mhz chip, think how much AMD would smoke Intel then?!

Or even go all the way to 2000 !!! I just don't understand?

In terms of "pure" clock speed AMD is ahead of Intel. Remember, the P4 is use a 20 stage pipline architecture, in order to boost their clock speed. While Athlons and P3 uses a 10 stage pipeline. If you recall, Intel tried to release a 1.13 Ghz P3 and failed miserably when they had to recall them back because of instability issues. Intels fastest 10 stage CPU is the new 1.2 Ghz Tualatin. That's all they can mustar for now, 1.2 Ghz. And this is based on a .13 micron process. AMD on the other is able to pump out a 1.53 Ghz CPU based on the 10 stage architecture! And it's a .18 micron process!! Wait til they jump to .13 micron baby.

If you look at true performance caliber RISC processors such as Alpha and SGI, they don't run very high MHZ, but these processors are not even 10 stage pipelines, they're 6 stage. A 700 Mhz RISC processor will blow a 2 GHz P4 right out of the water. Here's a quote from Anantech concerning the performance issue with long pipelines.

"Modern day CPUs attempt to increase the efficiency of their pipelines by predicting what they will be asked to do next, this is a simplified explanation of the term Branch Tree Prediction. When a processor predicts correctly, everything goes according to plan but when an incorrect prediction is made the processing cycle must start all over at the beginning of the pipeline. Because of this, a processor with a 10 stage pipeline has a lower penalty for a mis-predicted branch than that of a processor with a 20 stage pipeline. The longer the pipeline, the further back in the process you have to start over in order to make up for a mis-predicted branch."

[Slaught]

Well one factor i've so far seen un mentioned is that as you get smaller and smaller production sizes, i.e. .25. 13 micron etc, you begin to get problems, as the electrons can start jumping between the tracks. This means that it cant function properly, so new processes need to be made, and this will be expensive and time consuming. Im unsure of how accurate this is, but its what ive seen talked about in various places.

[Sl1m1thy]

I hear that the 1gig athlon is a lot more powerfull than the 1.8gig p4 ???? and faster (apparently it puts the p4 to shame)
Pentium is more popular due to the amount of money they put into marketing-how often do you see an Amd advert on the telly ?