The Winchester core is capable of higher overclocks (at lower voltages) than the Newcastle, so why do we not see a 3800+, 4000+, FX-53, and FX-55 at 90nm? Why has AMD not released anything clocking over 2.2GHz on 90nm? Is it possible that there are some hidden problems when clocking Winchesters are high rates? Perhaps they are waiting for a new core revision. Perhaps there is some kind of electromigration problem or other reliability and/or other processing issues that remain hidden as of now. Maybe everyone who is overclocking the Winchester well over 2.2GHz will be stuck with a dead processor in a few months. Any thoughts?
Why has AMD only released low-end 90nm A64s?
- Thread starter imported_thomase
- Start date
Companies always test out new processes on low-end parts.
ATI, nvidia, AMD, etc
If it proves to be safe and bug free, they move it up to high-end, flagship parts.
companies like amd and intel are not supportive of overclocking, but amd will release the 90nm for the higher end cpu's soon. because the fx-55 is perfectly fine at 130nm, they have no current need to bring it to 90nm yet. overclocking a 3000+ to 2.6ghz defeats the purposed of them even making and selling the 3200+'s and higher to DIY consumers and builders, and because of the high overclocks on the lower end cpu's, they lose money. note that the 3000+ overclocks about the same as the 3500+ right now, as seen on a review (i think anandtech's).
ATI, Nvidia is not a good example. ATI for example when moving from 150nm to 130nm did it with a smaller and less complex core. They didn't want to deal with the double whammy of a new design AND a new process.
In AMD's case, the Winchester is functionally equivalent (clock for clock) with comparable 130nm parts. In addition, as we can see, people are clocking the Winchester higher than the highest released processors at 130nm. Perhaps it is simply a matter of capacity and economics? Since they sell a lot more low-end parts then high end parts, and since their cost per die is lowest with 90nm, they want to dedicate as much capacity as possible to the parts that sell the most. They technically could produce 4000+ at 90nm but its more cost effective to use the limited 90nm capacity for high volume parts, especially if there is headroom left at 130nm. Since the price of a high-end part is relatively constant over time (4000+ will cost what 3800+ used to cost, etc.), it makes the most economical sense to ONLY release faster processors in order to keep ahead of Intel. If they just decided to come out with a 4500+ they would have to lower prices on 4000+ and that would eat into their margins. Since Intel messed up, they have the opportunty to hold their aces and rake in the cash. However, since the core can do what the core can do, tech savy consumers such as ourselves can take advantage of the untapped potential of AMD's processors by overclock. I hope that this is the case and that their are no hidden caveats (electromigration or something worse).
In AMD's case, the Winchester is functionally equivalent (clock for clock) with comparable 130nm parts. In addition, as we can see, people are clocking the Winchester higher than the highest released processors at 130nm. Perhaps it is simply a matter of capacity and economics? Since they sell a lot more low-end parts then high end parts, and since their cost per die is lowest with 90nm, they want to dedicate as much capacity as possible to the parts that sell the most. They technically could produce 4000+ at 90nm but its more cost effective to use the limited 90nm capacity for high volume parts, especially if there is headroom left at 130nm. Since the price of a high-end part is relatively constant over time (4000+ will cost what 3800+ used to cost, etc.), it makes the most economical sense to ONLY release faster processors in order to keep ahead of Intel. If they just decided to come out with a 4500+ they would have to lower prices on 4000+ and that would eat into their margins. Since Intel messed up, they have the opportunty to hold their aces and rake in the cash. However, since the core can do what the core can do, tech savy consumers such as ourselves can take advantage of the untapped potential of AMD's processors by overclock. I hope that this is the case and that their are no hidden caveats (electromigration or something worse).
dont forget that many people are easily hitting 2.9-3.0ghz on air with the fx-55 because of its strained silicon. i believe that is one of the reasons why amd hasnt brought the fx-55 and maybe the 4000+ to 90nm yet. but ya, 90nm cpu's JUST came out to market for amd, so give it a little time and they'll be out.
So ATI using the 9600, and NV using the 6600 for their new process is not an example of a company using low-end, less expensive parts for testing a new process?
Companies use low-end, and inexpensive (or less exensive) products to test a new process.
Also they sell more low end parts and since it hypothetically costs less to produce more they have higher profits. On the highend, however, margins are already enourmous and switching right away isn't much of a priority, as long as the old process is capable of making the new product. In fact, i would say that is stupid to test a midrange processor at lower scale,/new tech because if it doesn't work, your revenue is fucked.
wrong. they do have a current need to bring high end to 90 nm quickly. it costs them a lot less to make the same part in 90 than 130 and amd is not exactly financially healthy, so they need to do it asap.
i think it is mainly that they don't have the capacity to move everything to 90 nm yet, not that they don't have the technology. they switch over their production capacity in steps so they know it works before they switch everything over.
also they might not be sure of the yields of 90 nm parts yet, and it's easier to get slow parts than fast ones
also they might not be sure of the yields of 90 nm parts yet, and it's easier to get slow parts than fast ones
I think Zebo is the winner here. Companies have a need to clear out old stock, and if they went top to bottom with 90nm right off the bat then demand for the 130nm chips would be zero. As it currently stands, people who do not overclock (which is the majority of customers) have no choice but to help AMD clear out it's older stock.
Why push the 90nm process and the less than mature yields when the FX-55 on strained silicon on a mature 130nm yield process gets you to 2900-3000 on air alone. Within a few months I bet AMD could release the FX-57 on the 130nm process with no problem and continue to maximize their 90nm yield maturity until it is maximized. They only have so much capacity until Dresden comes online so they have to be very careful I would imagine with their production lines and wouldn't want to cause processor shortages in any way.
Of course you could get the 90nm 3000+ and overclock it to 2600 for five times less money than the FX-55 but that is another story LOL.
Of course you could get the 90nm 3000+ and overclock it to 2600 for five times less money than the FX-55 but that is another story LOL.
I don't think AMD is making FX-55 capable chips in that great quantity right now. There's a reason for it's price; it's the pick of the litter.
You make it sound like AMD is pumping out the air capable 3 GHz chips in copious volumes.
Also, "within a few months" is probably the same amount of time it will take for the 90nm process to mature and surpass .13 in every way. I'd expect FX-57 to be on 90nm.
I agree with you completely. I was simply stating if Intel were pushing them with 4.2 and 4.4 parts they could get them out the door. 2.6 on 130nm is pretty dang impressive on an AMD cpu. They aren't going to sell many FX-55 or FX-57 but I meant that the FX-55 is showing more overclockability than the 4000+ which I find fascinating and must be traced to the strained silicon and that if AMD didn't have 90nm that in four months or so they could release pick of the litter FX-57 130nm capable chips. AMD engineers deserve a pat on the back for 130nm goodness
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: 22K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
News NVIDIA and Intel to Develop AI Infrastructure and Personal Computing Products- Started by poke01
- Replies: 397
Facebook
Twitter