What are the real differences between GF3 ti500 and ti200
1 - 15 of 15 Posts
Joined
·
8,586 Posts
Check here for benchmark comparisons and technical aspects of the Geforce3 Ti line of video cards..
http://www.anandtech.com/showdoc.html?i=1539
http://www.anandtech.com/showdoc.html?i=1539
Joined
·
5,568 Posts
Actually that's how the chip business works most of the time, its all about marketing. To know what this means lets take a simple over view of the chip manufaturing process. Semiconductor devices such as srams, clock chips, graphics chips and so on are usually processed through CLASS BINNING. Now I know that statement doesn't mean anything but lets go on. For example after one wafer of die is processed in the fab, it goes through SORTING, sorting is usually done to seperate good die from bad ones and these die's are grouped by BINS. After that, the BAD die are either inked or the wafermap (as the name implies it contains the x and y coordinates of dice with their respective bins) is stored.
After this, the wafer goes to die bank where it is stored, then later withdrawed to be packaged at assembly. So the wafer is put on assembly and ALL GOOD BINS are packaged. After assembly and packaging, the UNITS either goes to BURN-IN or Package test, which one goes first how many package class steps it goes thourgh, is prescribed buy the current process flow used by the device and varies by type of device, as well as the GRADE (commerical, industrial, military) of the device. So ALL BINS, GOOD or BAD go through this process. Package class test, screens out all type of fail modes not screenable by WAFER SORT, this includes contact (due to assembly), gross functional, ac, dc, timing failures etc. After all these the one's that survive these processes get to go though the FINISH steps (i.e. lead conditioning , marking (if needed), package checking, lead checking, tape and reel (if needed), box packaging etc...) then finally it goes to finished goods where the marketing guys withdraw them.
I know you are wondering about what the hell this 'story' is all about. But if you track back, you will see that (if the device is speed binned) lower speed parts go to EXACTLY the same steps are the higher speed parts. Which means, that the manufaturing costs for both high speed and low speed parts are EXACTLY THE SAME. Which basically means that if you have a higher speed part the revenue generated from that part is much higher than the lowe speed parts. So in actuality, gunning for higher speed parts by stricter class testing does not make the manufacturing cost higher. It is true that it IS harder to produce higher speed parts, but most of the time it is not costlier to produce them. Evil? well, that's not even half of it. Some semiconductor chips (im not going to mention any), are processed with only ONE GOOD BIN or SINGLE BINNED. When the time to sell them comes they are just marked with whatever speed is needed by the custormer. So if the customer needs a 200MHz part then the units are marked as 200Mhz if they need a 400Mhz part, then the units are marked as 400Mhz and priced much much higher that the 200Mhz part. The customers just don't know that the 200Mhz part is just as good as the 400Mhz part. Evil? well like I said at first.. ITS ALL ABOUT MARKETING...
After this, the wafer goes to die bank where it is stored, then later withdrawed to be packaged at assembly. So the wafer is put on assembly and ALL GOOD BINS are packaged. After assembly and packaging, the UNITS either goes to BURN-IN or Package test, which one goes first how many package class steps it goes thourgh, is prescribed buy the current process flow used by the device and varies by type of device, as well as the GRADE (commerical, industrial, military) of the device. So ALL BINS, GOOD or BAD go through this process. Package class test, screens out all type of fail modes not screenable by WAFER SORT, this includes contact (due to assembly), gross functional, ac, dc, timing failures etc. After all these the one's that survive these processes get to go though the FINISH steps (i.e. lead conditioning , marking (if needed), package checking, lead checking, tape and reel (if needed), box packaging etc...) then finally it goes to finished goods where the marketing guys withdraw them.
I know you are wondering about what the hell this 'story' is all about. But if you track back, you will see that (if the device is speed binned) lower speed parts go to EXACTLY the same steps are the higher speed parts. Which means, that the manufaturing costs for both high speed and low speed parts are EXACTLY THE SAME. Which basically means that if you have a higher speed part the revenue generated from that part is much higher than the lowe speed parts. So in actuality, gunning for higher speed parts by stricter class testing does not make the manufacturing cost higher. It is true that it IS harder to produce higher speed parts, but most of the time it is not costlier to produce them. Evil? well, that's not even half of it. Some semiconductor chips (im not going to mention any), are processed with only ONE GOOD BIN or SINGLE BINNED. When the time to sell them comes they are just marked with whatever speed is needed by the custormer. So if the customer needs a 200MHz part then the units are marked as 200Mhz if they need a 400Mhz part, then the units are marked as 400Mhz and priced much much higher that the 200Mhz part. The customers just don't know that the 200Mhz part is just as good as the 400Mhz part. Evil? well like I said at first.. ITS ALL ABOUT MARKETING...
Joined
·
2,712 Posts
Interesting story. I never knew that semiconductors industry was too lazy to produce different clocked chips. I only heard about Intel sorting their processors about this way, but I never knew the whole industry did that. You work in this area, right Xeven? You should know how much these devices we buy really cost.
Joined
·
5,568 Posts
> You work in this area, right Xeven?
Yeah, and I get to see a lot of very nasty stuff going on behind the scenes..
About the cost, I can't really generalize for all type of devices (or chips if you like), but lets take an example of one device i work on.. 1 unit costs ~$80 to manufacture, we sell it for $600 .. :heh:, actually the cost is well you guessed it, mostly dictated by marketing and as well as the market segment where the device is being sold. Take for actual case of one clock chip that is used to be sold in the DATACOM market, it cost $25 per unit, when the marketing transferred the device to another market sector the cost went down to $8, why? because chip costs (any type of chip) in this market segment are generally much lower than the DATACOM market, so in order for the device to make it in this market it has to conform with the current price range of the market segment... guess it still sucks... but that's just how it goes sometimes.
Yeah, and I get to see a lot of very nasty stuff going on behind the scenes..
About the cost, I can't really generalize for all type of devices (or chips if you like), but lets take an example of one device i work on.. 1 unit costs ~$80 to manufacture, we sell it for $600 .. :heh:, actually the cost is well you guessed it, mostly dictated by marketing and as well as the market segment where the device is being sold. Take for actual case of one clock chip that is used to be sold in the DATACOM market, it cost $25 per unit, when the marketing transferred the device to another market sector the cost went down to $8, why? because chip costs (any type of chip) in this market segment are generally much lower than the DATACOM market, so in order for the device to make it in this market it has to conform with the current price range of the market segment... guess it still sucks... but that's just how it goes sometimes.
Joined
·
2,289 Posts
So basically, we aren't paying for the product, we are paying for...the marketing?!?!
I though when making a chip, you had some good and some bad, the goods were sold at higher costs because of it's higher speed, and the bad were labeled at lower speeds and sold accordingly. But making the chips at the same cost and just labeling them as the market needed is just plain evil. The cost is the same and marketing just throws the price up, because of a label?!?
Hmmm, I think that's why overclocking goes so well
Once I o/c my P3 600 to 750 and it worked fine
I though when making a chip, you had some good and some bad, the goods were sold at higher costs because of it's higher speed, and the bad were labeled at lower speeds and sold accordingly. But making the chips at the same cost and just labeling them as the market needed is just plain evil. The cost is the same and marketing just throws the price up, because of a label?!?
Hmmm, I think that's why overclocking goes so well
Once I o/c my P3 600 to 750 and it worked fine
1 - 15 of 15 Posts
