By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Soleron said:
TheJimbo1234 said:
Soleron said:
TheJimbo1234 said:
Ok, that is nice, but very wrong.
Technology that reduces the transistor size, which currently is being driven by the smartphone industry, will always result in old technology being cheaper. This causes higher efficiency and less waste per silicon wafer, thus the price of production goes down. Look at sticks of DDR3. Reduction in growth allows for larger sticks, but the high efficiency also creates dirt cheap sticks.

Did you read my words? "Transistor size reduction" technology is slowing down considerably at all foundries producing smartphone parts (TSMC, Samsung, GF) because of high implementation costs.


Have you been in a clean room before? Have you fabricated your own products? No? The listen up - it's all sensationalist crap.

It isn't slowing down. If anything, the 40-32nm bridge was harder than what we have. With Intel making their $6 billion clean room factory (!), and ARM designing yet more sub 30nm transistors, there is no problem....for the moment. Sub 15nm and yes, we have an issue which will require some insane engineering, but that is a long way off, and simple restructuring of proc designs will compensate for this, or we could move onto carbon trans, apply graphene in different ways, or use magnetic transistors. What ever the solution, there are solutions.

And no, the cost part is crap. A poor PC back in the mid 1990's cost £800+. Now look how cheap and powerful a £300 laptop is. Hell, most phones are now twice as powerful as my first PC and cost less WITHOUT adjusting for inflation! This is because fabrication techniques are backwards compatible. Being able to produce 22nm trans. means making a 32nm one is almost 100% efficient, hence why electronics deflate in value so quickly.

I am a physicist, I have taken basic EE courses, and of course I follow those news sources who DO have specific experience.

I said Intel was fine and isn't affected yet. Intel is irrelevant to PS4/720. ARM have nothing to do with the fabrication side. Pick better examples if you wish to demonstrate what you are saying.

Yes, different transistor technologies will allow us to progress further than we can see now. But these have much higher R&D costs than the shrinks of the past, because they require more engineers for more time in order to bring into production.

Yes, costs have come down in the last 10 years. I even said, the shrinks during PS2 and PS3 were basically free. It's now that that stops. An HD4850 cost $99 in 2009. Now in 2013, the graphics cards you can buy for $99 are barely any faster.

New shrinks will come slower, and they will NOT be "100% efficient". Look at the 28nm shrink of recent graphics. It's more like 70% area and 50% power reduction, and those ratios will get worse the longer EUV, Finfet and 450mm wafers are pushed back by the foundries.

Do you deny that TSMC won't have 22nm ready in 2013 like they promised? Do you deny that they cancelled Finfet? Do you deny that TI, Renesas, AMD, Fujitsu, Infineon and so on have been forced to go fabless at 45nm or 32nm because of rising costs?


*facepalm*

ARM come up with the method and design which is the whole issue. In labs we can fabricate ~5nm trans (hell, a 7 atom/12 angstroms trans was made), but that is in labs. ARM can design ways of mass producing them which is what the challenge is. So I would do some homework before trying to put someone down.

 

And the market is bigger and more lucrative than before, due to people wanting computers in their phones.

As for gpu's etc; actually with inflation added, gpus are cheaper eg a 4850s equal, the 6670 is £66.50 and will use less power. Now the 4850 was pretty weak and came in at low end gamer gpu. Just look at this. http://www.videocardbenchmark.net/video_lookup.php?gpu=Radeon+HD+4850  One generation later you get a better card (5670) for a lower price. If you added a few $'s on for inflation, you can get a gtx 650 for $106 which would demolish a HD 4850. Also as I have said, look at RAM. That has pretty much just crashed in price for the last few years. What use to cost £80 now costs £30 at a push. 

No initial shrink is 100% and I did not say this. Iirc, they are viable for marketint at ~30%+ as they work on the concpet of getting it to ~96% within the year. Also you have to reduce power otherwise you will simply burn up the silicon/have thermal runaway. The first sandybridge had this along with the gt 4xx series, but of  course this was solved with simple redesigns of the layout rather than fabrication.

TSMC may not, but Intel are already producing and selling 22nm procs - it's the Ivy Bridge range  :)

Well the GT 600 series is using 28nm, and they will hold off as long as possible simply because they know that large gpu performance boosts can be gained from complex redesigns which are cheaper than shrinking their trans, but this is out of a quest for profit rather than it is too expensive to do. Also there has been no demand for more powerful gpus, and if you can make 2 series of cards rather than one, you'll drag it out for as long as you can.