CPU manufacturers set a target speed they want their chips to run at (ie 3.06 GHz) within an appropriate yield. While probably all 3.06 GHz will probably run fine at 3.1 Ghz, some can run at 3.2 Ghz, some at 3.3 Ghz, etc. However, if they were to sell 3.2 Ghz or 3.3 Ghz chips, they would have a lot less chips that can run at that speed, and thus much lower yield at that speed.
They also want to leave themselves some headroom so they can increase CPU speed later with minor tweaks to the manufacturing process. If they release a 3.3 GHz chip today then it will be tough to release a faster chip for awhile since yields at that speed are so low.
With respect to the marketing - their top chip will always cost ~$500-700 (of course sometimes it's a bit higher or lower, depending on the circumstances). Very few people are going to buy a $1000 CPU (in the mainstream anyways). So if they release a 3.2 Ghz chip at the top, it will invariably cost ~$600, and drive down the price of the 3.06 GHz CPU. When they do get high enough yields to economically make 3.3 Ghz chips, for example, the 3.06 Ghz chip will drop in price under $500, and the 3.3 GHz chip will occupy the top spot in their CPU lineup, right at that $600 hot spot.
They also want to leave themselves some headroom so they can increase CPU speed later with minor tweaks to the manufacturing process. If they release a 3.3 GHz chip today then it will be tough to release a faster chip for awhile since yields at that speed are so low.
With respect to the marketing - their top chip will always cost ~$500-700 (of course sometimes it's a bit higher or lower, depending on the circumstances). Very few people are going to buy a $1000 CPU (in the mainstream anyways). So if they release a 3.2 Ghz chip at the top, it will invariably cost ~$600, and drive down the price of the 3.06 GHz CPU. When they do get high enough yields to economically make 3.3 Ghz chips, for example, the 3.06 Ghz chip will drop in price under $500, and the 3.3 GHz chip will occupy the top spot in their CPU lineup, right at that $600 hot spot.
pm has posted about this on more than a few occasions. You may want to do a search, as Patrick's explanation (novella) is much more comprehensive than my little post will be... 
At any rate, when semiconductor manufacturers test chips, they test them in two ways. First to basically stress test and see if the part will fail. Then to see what speed they can be binned at.
Now, when they test for bin splits, they do it at the extremes... Temperatures (both hot and cold), poor cooling, poor system components (i.e. bare minimum power supply), etc... And they do all of that while using applications that will stress the cpu more than any end user could ever do. And then they must consider some headroom when making the final speed grade decision.
You see, the average AT enthusiast uses quality components and takes care of their PC. When binning cpu's, the manufacturer cannot count on the vast majority of users to do the same.
And if an enthusiast's cpu crashes while running it out of spec, it really isn't much of an issue. But if a customer is expecting their mission critical application to run, any hitches are completely unacceptable.
So, there ya go.
At any rate, when semiconductor manufacturers test chips, they test them in two ways. First to basically stress test and see if the part will fail. Then to see what speed they can be binned at.
Now, when they test for bin splits, they do it at the extremes... Temperatures (both hot and cold), poor cooling, poor system components (i.e. bare minimum power supply), etc... And they do all of that while using applications that will stress the cpu more than any end user could ever do. And then they must consider some headroom when making the final speed grade decision.
You see, the average AT enthusiast uses quality components and takes care of their PC. When binning cpu's, the manufacturer cannot count on the vast majority of users to do the same.
And if an enthusiast's cpu crashes while running it out of spec, it really isn't much of an issue. But if a customer is expecting their mission critical application to run, any hitches are completely unacceptable.
So, there ya go.
ok, sounds funny, one reason why people overclock is they want to be faster then everyone elses. So even if they made it 10Ghz, there would still be people trying to bump that to 11Ghz just for the hell of saying they did it and bragging rights
Heh.... you must be a young'un. I still remember when the top chip cost over $1000 and the sweet spot for CPU's was about $200. This wasn't all that long ago, either.
I personally think the pricing structure has more to do with the sweet spot for maximizing profits. The speed at which maximum yield is attained will be the set to the sweet spot for selling. Similiar to how video cards are differentiated by pricing. I think nowadays, the sweet spot for CPU's dropped so far when the bottom fell outta the market because the average consumer has come to expect more for his dollar.
Anyways, the way I see pricing structures, the speed at which most processors become binned becomes the sweet spot where the company can get maximum profit by selling slightly less chips for much higher profit margins. Every other speed grade gets priced relative to this point. So, if, say, 30% of your yield hits 2GHz, then that speed is sold at, say, $125 (I have no idea where the sweet spot is) so the 1.6 hits $75, the 2.4 hits $175, and so on. If people want speed, they pay a premium for a CPU picked out of smaller supply.
As for why everybody seems to be getting such high overclocks, I think you need to take a closer look at your data source. Those who will post overclocks tend to be because the overclock had good results. Few, if any, will be posting chips who don't overclock too far, or chips that are underclocked (Think fanless). Add to that the fact that anybody who seriously gets into overclocking beyond just a simple bump in speed in the BIOS is already quite proficient with computers (aka, not the average Joe). Which means, you're looking at something like 15% of the overclockers community getting these nice overclocks, and the community as a whole comprising of some 5% of the total market. After factoring all that in, it's not surprising that the results you see are usually pretty damn good. Plus, I haven't even considered those who value quiet computers, those who don't overvolt as much, and those with more exotic cooling methods than just plain air. Oh, yes, and not everybody you see tends to drop upwards of $50 just to cool their CPU.
Originally posted by: Paveslave
How come CPU manufacturers don't just make thier chips run as fast as people overclock them? Why would a major CPU company come out with a 3.06 Ghz processor? Why not make it an even 3.10, or a 3.20, or even 3.50?
It's called marketing, and bleeding our wallets dry
What fun is it when you get your CPU basically already overclocked? half the fun is the gamble.. Getting a new CPU and overclocking it when you get home is great fun
The difference is and has been typically durability and longevity. Overclocking puts more stress on a chip, and will shorten its life (and in many cases other components). The reason so many do it anyway is because shortening its life might mean going from 10 years to 7 years. In this case, most PC enthusiasts will have moved on to the next generation of chip. Intel or AMD however, must think about more than just the PC enthusiasts. The chips are designed to function for long periods of time, longer than most of us keep them around. A specific goal is set. Certain chips manufactured meet this goal and are sold as a certain speed. If the CPU manufactures changed their criteria, there would probably instantly be higher speed chips on the market dispite the fact they would be physically the same as the old "slower" chips.
Well, this is my first attempt at building a computer and I wanted to be sure that I could stay constant for a while without having to upgrade in say a month. I like to play games like Age of Empires II and when I am playing with 6 enemies and everyone has 200 players running around shooting up my huge castle walls, it tends to eat up a lot of my RAM. To tell you the truth, my wife and I were playing the other day (our computers are linked) and I was hosting the game. We had three enemies against us and we spent a long time building a huge castle area and during some major battles the game started lagging a bit. I think I should have went higher in RAM actually, but couldn't afford it. I think my board is maxed out anyway, I don't know. Games like FS2002 play nice too, so I think I'll stick with the 2 Gigs of RAM anyday over any other mod that I have. I could have done better probably, but I have no complaints right now, except I should have just went with the 3.06 instead of the 2.8. Who knows, maybe I could O/C...lol, J/K
Assimilator1
Elite Member
- Nov 4, 1999
- 24,165
- 524
- 126
I can't believe the lagging is down to RAM ,there's no way that 2Gb of RAM would be used up for a game!
What OS are you using?
How much RAM has your 9700 got?
What OS are you using?
How much RAM has your 9700 got?
This question is the same as asking in a race car forum "Why make car go faster, when you can buy a faster car" .. lol
Pulsar
Diamond Member
- Mar 3, 2003
- 5,224
- 306
- 126
Ok, I'll give you the real answer... it's simple statistics.
First, they make the chips. Next, they test the chips. They bin them into separate speeds <-------- this is the key to your question.
See, they don't just say "Ok, xxx passed this test, but just barely". They have to include basic statistical models, i.e., 6 standard deviations.
For instance. Say I make 10 processors. 9 run at a max of 2ghz, and 1 runs slower. That ONE that runs slower statistically affects choosing the entire population's clock speed. If 1 in 100 processors failed in the field, companies like dell would be spending literally millions on warranty costs.
So, they HAVE to clock far underneath their theoretical maximum to:
#1 maximize chip yield.
#2 ensure reliability.
In the end, #1 and #2 can be boiled down to: "It's all about the money".
This garbage about not flooding the market is just that - garbage. Those companies turn out processors that are fast as they can be while still meeting yield (cash) and reliability (cash) requirements.
The End.
First, they make the chips. Next, they test the chips. They bin them into separate speeds <-------- this is the key to your question.
See, they don't just say "Ok, xxx passed this test, but just barely". They have to include basic statistical models, i.e., 6 standard deviations.
For instance. Say I make 10 processors. 9 run at a max of 2ghz, and 1 runs slower. That ONE that runs slower statistically affects choosing the entire population's clock speed. If 1 in 100 processors failed in the field, companies like dell would be spending literally millions on warranty costs.
So, they HAVE to clock far underneath their theoretical maximum to:
#1 maximize chip yield.
#2 ensure reliability.
In the end, #1 and #2 can be boiled down to: "It's all about the money".
This garbage about not flooding the market is just that - garbage. Those companies turn out processors that are fast as they can be while still meeting yield (cash) and reliability (cash) requirements.
The End.
Originally posted by: Assimilator1
I can't believe the lagging is down to RAM ,there's no way that 2Gb of RAM would be used up for a game!
What OS are you using?
How much RAM has your 9700 got?
I'm using XP PRO Corporate Edition which I installed onto a New Hard Drive that had no previous windows version on it. My 9700 Pro is the standard as far as I know, with 128 MB RAM, unless gigabyte added to that I am assuming it's the same as the normal ATI card. I don't know why it would lag either, but it did and sometimes does during games where there is a lot on the board. I am not talking about just a couple of people moving around at the same time, I am talking about somewhere in the neigborhood of 1000 at any given time. I guess that just seems a lot to me but I could be wrong. The last computer I had was the one that my wife plays on and it is an old gateway 400 AMD with 256MB RAM with stock gateway crap. Now that lagged quite a bit on it's own, but it doesn't do too bad when I'm hosting the game, but it still lags. Hey, as long as my wife says she doesn't care, I'm cool with it.
I am going to build her one in the near future so I am taking this as a learning experience so I can do it better the next time. how can u sell to the ridiculously large overclocker's market when your chips are already maxed out?!
I'm not sure if that was sarcastic or not... But the PC enthusiasts make up a very, very small demographic.
I've decided that I will only overclock as far as the stock voltage will take me from now on. Before, I didn't care how long the computer lasted after 6 months, now I am looking for longevity as well as speed due to lack of funding.
It's all about Marketing & Profit
Also, try looking up some information about the bell-curve. Applies to alot of stuff like this.
The largest demand (and profits due to volume) is in the middle range of the CPU's, reasonable price & reasonable performance. There is also good profit (per unit sold) on the high end by limiting availability. What's left over is sold on the low end with last years chips.
All the different speeds of a CPU (in a given model/stepping) come of the same silicon & have the same potential speed. However, the process is not perfect, so the actual speed of the chips will vary.
Sometimes the engineering gets ahead of the marketing & almost all the chips will run at the top speed, yet marketing needs slower chips. That's when you get great overclocker chips like the Celeron 300a, P4 1.6a, etc.
This also happened with the slot A Athlons. I had a CPU module marked 700 Mhz (and ran at 700 Mhz), but the actual CPU (inside) was marked 800 Mhz. I also have a Slot A thunderbird marked 750 Mhz that had a 950 Mhz CPU inside <G>. Needless to say, they where both great overclockers <G>
Also, try looking up some information about the bell-curve. Applies to alot of stuff like this.
The largest demand (and profits due to volume) is in the middle range of the CPU's, reasonable price & reasonable performance. There is also good profit (per unit sold) on the high end by limiting availability. What's left over is sold on the low end with last years chips.
All the different speeds of a CPU (in a given model/stepping) come of the same silicon & have the same potential speed. However, the process is not perfect, so the actual speed of the chips will vary.
Sometimes the engineering gets ahead of the marketing & almost all the chips will run at the top speed, yet marketing needs slower chips. That's when you get great overclocker chips like the Celeron 300a, P4 1.6a, etc.
This also happened with the slot A Athlons. I had a CPU module marked 700 Mhz (and ran at 700 Mhz), but the actual CPU (inside) was marked 800 Mhz. I also have a Slot A thunderbird marked 750 Mhz that had a 950 Mhz CPU inside <G>. Needless to say, they where both great overclockers <G>
Assimilator1
Elite Member
- Nov 4, 1999
- 24,165
- 524
- 126
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter