More or less. The actual production cost of CPUs is pretty small. What costs big bucks is design, and keeping top-end foundry research going on.
Basically, you shouldn't think of CPU economics like fuel or something, where most of the cost of the product is direct cost of manufacture, but instead like a newspaper, where the cost to produce one more product is a minuscule fraction of price, but the cost to keep the design (content) current is why they can't be sold for cheaper.
Yes, I agree. But I assume that the development cost is mostly connected to designing the actual CPU and GPU cores. Adding more cores ought to be pretty cheap to do design-wise. So if they already have a 4C GT2 Haswell, making a 6C with 1.5x the amount of GT2 EUs should not require that much extra design costs.
AMD doesnt even sell 1 million 8 cores a year for consumers.
Do you have a source or is it home made numbers.?..:sneaky:
your opinion is appreciated but what about a source:\Not really a stretch when they don't have any real OEM deals.
your opinion is appreciated but what about a source:\
Thanks a lot for the long explanation. So in short it is a simp matter of "they cannot do it" / missing know-how.In general, companies that opt to go gate-first instead of gate-last are companies who simply don't have the resources necessary to make a production-worthy gate-last integration scheme on the timeline necessary for such a node to be relevant to the market when it finally comes out.
(...)
The problem is that the 8C AMD CPUs are slower than the 4C Intel CPUs.
But is there any reason a reasonably priced mainstream 6/8C Haswell would not sell?
The problem is that the 8C AMD CPUs are slower than the 4C Intel CPUs.
But is there any reason a reasonably priced mainstream 6/8C Haswell would not sell?
Now that you said it, you surely know the numbers. How thin exactly are they, Mr. Insider?The console APUs margins are also razor thin. Not much money to be made there at all.
Oh really? How many occurences can you state to make up your version of history?And history have shown that console winners usually flop with their nextgen due to resource starved teams.
Intel introduced its first consumer 8-Threaded CPU (Core i7 920) in November 2008 at $284. Nehalem die was 263mm2 at 45nm and it was released first in Desktops before Servers.
If Intel could release Core i7 920 in the consumer segment back at 2008 they can release a 6-core 12-Threads 256mm2 at 22nm in 2013, 5 years later
How do you know that? Their top end mainstream 3770K/4770K has been selling well, despite being $100 more expensive and "only" about 20-30% faster than the 3570K/4670K (depending on type of workload).Not in a quantity that would bring it to LGA1150 in any meaningfuld way.
How do you know that? Their top end mainstream 3770K/4770K has been selling well, despite being $100 more expensive and "only" about 20-30% faster than the 3570K/4670K (depending on type of workload).
In your pricerange:
AMD is abandoning it.
Intel is not having it.
How do you define selling well? Whats the consumer base thats supposed to buy it, and not willing to pay the extra $ for the LGA2011? You are asking for a special mask, just to satisfy some niche of a niche segment.
What about TDP and clocks? You wanna redesign the entire platform to 130W? Or will the 6C CPUs be significant slower than the 4C? AMD for example publicly stated 65W will be the new future limit.
Now that you said it, you surely know the numbers. How thin exactly are they, Mr. Insider?
Oh really? How many occurences can you state to make up your version of history?
Start with the 4670K increase everything by 50% and price it based on die area. I.e. you'll get 6C, 1.5x GT2 EUs, 84x1.5=126W TDP, 1.5x$242=$363. Possibly price it a bit higher to not cannibalize on the 4770K.
Optionally do a similar one but without any IGPU. That ought to bring down both the price and TDP quite a lot, since the IGPU accounts for quite a large portion of the die area.
Its not hard to look at one of their previous product mixes. They sold 113K 8C desktop CPUs in Q1 2012, 120K in Q2 and 127K in Q3. In the meantime their amount of CPUs shipped is lower, more CPUs moved to mobile. Hell, they might not even ship half a million 8C.
The 6C number is not much higher either. And that explains why AMD is completely abandoning the failed attempt to sell people 6C and 8C CPUs with the extremely low volume they got.
Intel is not selling 6C LGA2011 to the desktop because its cheaper to make a LGA1150 alternative. Its doing so because its the only way to get healthy economics out of it. And AMD is dropping 6 and 8 cores for the same reason. Economics. Or if you want, not enough people want to buy it.
It doesnt work that way. You just cant take an exsiting design and add 50%. For that to be even remotely true, you also have to sell the same amount as the one you use. And I dont think some 100-150million 6C CPus will be sold.
The IGP doesnt even use nearly the same as 1 core. And it acts mostly as dark silcion to the cores. IGP used or not.And even if we use your number. 126W? So the entire platform needs to be changed and revalidated.
A new 6C mask as you want, IGP or not. And you would most likely see a price of 500$, if not more. Hence rendering it useless.
Intel is not selling 6C LGA2011 to the desktop because its cheaper to make a LGA1150 alternative. Its doing so because its the only way to get healthy economics out of it. And AMD is dropping 6 and 8 cores for the same reason. Economics. Or if you want, not enough people want to buy it.
According to AMD AM3+ CPUs shipements will be 30%
with the remaining 70 being FM2 , this for Q4 2013 ,
so we can assume that currently FXs are roughly 30%
of DT CPUs , in line with ATenra s number of 2millions/quarter
or 8 millions for 2013 , FX8XXX are likely to represent half
of thoses 8 millions , far from your thin air extracted numbers.
http://www.tomshardware.com/news/amd-roadmap-kabini-beema-excavator,24029.html
Which connects nicely to the topic of multithreading, like this one:
http://forums.anandtech.com/showthread.php?t=2339967
New games are still using 2 threads. There is very little reason to go more than 4 T/C. Yet, some are still yelling that it is AMDs bad engineering, not a software developers fault.
I'll quote AtenRa:
"Intel introduced its first consumer 8-Threaded CPU (Core i7 920) in November 2008 at $284. Nehalem die was 263mm2 at 45nm and it was released first in Desktops before Servers.
If Intel could release Core i7 920 in the consumer segment back at 2008 they can release a 6-core 12-Threads 256mm2 at 22nm in 2013, 5 years later"
Also, if a 6C Haswell was priced based on die area it would cost around the price of a 4770K. And since people are buying the 4770K at that price despite being slower compared to what a 6C Haswell would be, I don't see how you can say there is no interest in a 6C Haswell CPU. If a 6C would exist at the price of a 4770K people would get that instead. And then there would be enough sales volume to justify it.
Intel may have reasons to price it higher than that (to segment the market or not cannibalize on other CPUs). But then you cannot take that as an argument for consumers not wanting a 6C Haswell CPU. It's like if all 4C Haswell CPUs would be priced at 5x the price of the 2C CPUs due to similar reasons, and then you could claim that consumers are not interested in 4C CPUs...