cheapchips
Veteran
Meant Tflops obviously!
Predict: Next gen console tech (9th iteration and 10th iteration edition) [2014 - 2017]
- Thread starter Shifty Geezer
- Start date
- Status
USGS Germanium (pdf): http://minerals.usgs.gov/minerals/pubs/commodity/germanium/mcs-2015-germa.pdf
USGS Copper(pdf): http://minerals.usgs.gov/minerals/pubs/commodity/copper/mcs-2015-coppe.pdf
Germanium is ridiculously expensive and rare mineral is any process that relies on it 'commercial' in the way copper is? From the USGS you're looking at roughly $7.40 per kilo for copper and $1,900.00 per kilo for germanium (yields at ~3% per tonne of zinc ore) or ~256 times more expensive per kilo as well as being incredibly rare with only ~2,300 tonnes total available in ground let alone as ready to use ore.
These processes, like SiGe and Gallium Nitride, are for use in expensive bespoke applications like the transmitter elements in Active Electronically Scanned radars et al, they will never be used for mass market semiconductors like games consoles.
I was kind of disappointed the Ars Technica article did not make this clear, it's like running an article announcing the amazing clarity of a screen made from pure diamond, great and all but irrelevant to the rest of us.
USGS Copper(pdf): http://minerals.usgs.gov/minerals/pubs/commodity/copper/mcs-2015-coppe.pdf
Germanium is ridiculously expensive and rare mineral is any process that relies on it 'commercial' in the way copper is? From the USGS you're looking at roughly $7.40 per kilo for copper and $1,900.00 per kilo for germanium (yields at ~3% per tonne of zinc ore) or ~256 times more expensive per kilo as well as being incredibly rare with only ~2,300 tonnes total available in ground let alone as ready to use ore.
These processes, like SiGe and Gallium Nitride, are for use in expensive bespoke applications like the transmitter elements in Active Electronically Scanned radars et al, they will never be used for mass market semiconductors like games consoles.
I was kind of disappointed the Ars Technica article did not make this clear, it's like running an article announcing the amazing clarity of a screen made from pure diamond, great and all but irrelevant to the rest of us.
It's using it for the transistor channel material, a structure whose dimensions are measured in nanometers. The cost of the element itself is unlikely to dominate. The complex chemistry of modern processes is such that there are other expensively rare elements in physically small quantities per chip.
upnorthsox
Veteran
From the link:
So it sounds like its not actively being collected to any extensive degree, most likely because there's little market. That could/would obviously change. Also, stacking should also be taken into account where layers can be at different processes so that registers which wouldn't benefit much from the 7nm could be on its own 10nm or higher layer.
At 0.01g per chip, that's <2 cents. And it probably uses a lot less than that per chip!
ArcticCircle
Newcomer
Gold is currently priced over 30000 euros per kilo. I bet most of us adults living in westen nations have some gold item like necklace, ring or ear ring. Germanium rarity is about 1 atom per million silicon atoms, about the same as tungsten. Silver and mercury are more rare. Gold is thousand times more rare. Obviously opening new mines are expensive and there might be price hike for germanium, but it won't be too bad. And hey, the way world economy is going, we could use new "gold rush" 
If you're going to be wrong be very wrong
Cheers for the pointers, not sure where I got the idea that it would be a cost prohibitive but as you point out there's less gallium per chip than would make a pin head. It'll be interesting to see how quickly the process gets used in mass production, even Intel has been struggling to get 14nm up to speed at the same rate they deployed 22nm, to say nothing of TSMC having to skip a node and try and go straight to 16nm.
Cheers for the pointers, not sure where I got the idea that it would be a cost prohibitive but as you point out there's less gallium per chip than would make a pin head. It'll be interesting to see how quickly the process gets used in mass production, even Intel has been struggling to get 14nm up to speed at the same rate they deployed 22nm, to say nothing of TSMC having to skip a node and try and go straight to 16nm.
Megadrive1988
Veteran
I love speculation regarding new hardware, especially Nintendo's
I still think NX won't be a single device, but a whole family, starting with a handheld with a 4,5-5,0" screen (720p max.) and around WiiU performance (4 CPU cores, 150-200GFLOPS GPU, 2GB RAM). This should be doable in 2016 for ~150€. Games will be cross-releases with WiiU games (Zelda will be a launch title), they will also share account system.
The home console will come a year later (2017) with 8 CPU cores (or threads, depends on the architecture), 1024 GCN shader units and 8GB of HBM2.
Both will use flash cards like the 3DS which will allow for a smaller, cheaper packaging of the devices.
Agreed.
I expect something like this:
2019 announcement, late 2019 - early 2020 release, specs for Microsoft and Sony platforms
2019 announcement, late 2019 - early 2020 release, specs for Microsoft and Sony platforms
- 1TB SSD
- 32-64GB HMB3 unified ram for CPU/GPU 1.5-2.0 TB bandwidth
- AMD APU with 16 cores at 2.0-2.8 ghz and GPU performance comparable to a Titan X + 20%
- Possibly Digital Only without an optical drive, retail copies will include a code to be activated on PSN or XLIVE (like steam)
- Three months after release digital renting will be possible through PSN/XLIVE for every game 3 $ per day, 7.5 $ for 3 days and 15$ for a week, 30$ for a month, after a year these prices are halved
- Same thing will be possible through your friends at a lower cost, when you rent a game from your friend you pay a reduced amount for the rental (2$/day, 5$/3 days, 12$/week, 20$/month) and the game is deactivated for that time on your friend
- PS+ will allow for one 3 day rental per month
- Price around 400-500 $
D
Deleted member 13524
Guest
I agree that optical is starting to be a mess regarding size and transfer speeds, but I'm hoping for stacked nand to transition soon-ish towards external memory cards, making the cost/GByte low enough that console manufacturers consider using flash cards for retail games during the next generation.
At least for consoles, I still enjoy that weird sense of "possession", and I believe that used games will still be a huge market for many, many people.
At least for consoles, I still enjoy that weird sense of "possession", and I believe that used games will still be a huge market for many, many people.
D
Deleted member 11852
Guest
Actually you're not hugely wrong. The amount of material in a finished product produced from a complex chemical process is far from commensurate to the amount of material required to make the product. There have been, and continue to be, huge strides in chemical processes to increase efficiencies particular when dealing with rare earths (because of the inherent costs), but the truth is processes are still hugely inefficient - and expensive in their own right because they require bespoke chemical processes to be developed - there is a whole chemical processing chain that isn't immediately obvious when you read about a new material used in chips.
A lot of electronics used in non-pressured avionics are made from more exotic materials or worse, composites of exotic materials, because you need the physical properties of those materials given they operate better in different physical environments. The costs of the avionics systems were often approach the cost of the structures and payloads.
In addition to the actual costs of commercial manufacture you can also bet your arse that a ton (not literally!) of material has been used during the development and all of these costs need to be recovered as well.
It's laying down a very thin layer of GeSi, which may use more Ge than what Intel uses to add strain to its channel. Hafnium goes into the gate dielectric, and that is in the same league of expense.
The biggest cost adder from adding something like Ge as part of the channel material would come from the complexity required to use it in the process, rather than its material cost. Things like quad-patterning and EUV are the bigger question marks in terms of economics, but those wouldn't be blamed on the Germanium.
The biggest cost adder from adding something like Ge as part of the channel material would come from the complexity required to use it in the process, rather than its material cost. Things like quad-patterning and EUV are the bigger question marks in terms of economics, but those wouldn't be blamed on the Germanium.
D
Deleted member 11852
Guest
Hafnium is quite a bit more expensive than germanium per gram but hafnium is orders of magnitudes easier to work in terms of chemical process. Like I said above, the costs will come primarily from the processes.
Infinisearch
Veteran
Anybody know whatever happened to gallium arsenide...? I remember back in the 80's it being hailed as silicon's superior successor. That is if I'm remembering things correctly.
cheapchips
Veteran
It is used in some specialist chips but it's about x1000 more expensive than silicon for an 8 inch wafer. An article on some Stanford research suggest that they could get the cost down to x10 for use in more efficient solar panels. Don't know whatever would mean for chips.
Here's my go at it.
2019
7 or 10nm process
4-8 core ZMD Zen x86-64 CPU
8-10 TFLOPS GPU
32GB HBM2/3
No disc drive (uses up too much space)
usb 3.1
bluetooth 4
802.11ac wifi
2tb hdd
128gb ssd acting as a cache
backwards compatibility out of the box
~$449 (pre-inflation 2015 dollars)
2019
7 or 10nm process
4-8 core ZMD Zen x86-64 CPU
8-10 TFLOPS GPU
32GB HBM2/3
No disc drive (uses up too much space)
usb 3.1
bluetooth 4
802.11ac wifi
2tb hdd
128gb ssd acting as a cache
backwards compatibility out of the box
~$449 (pre-inflation 2015 dollars)
Not that I mean to start the argument in the wrong place but I discarded a post on the matter as I could not remained focus but I think that there is an issue, consoles are the only CE devices that are affected by inflation. CE devices prices go down, PC are getting cheaper, tv too, etc. You get a lot lot more for the same price and more often than not you can find better than what you are replacing for cheaper.
The price dynamic on the market is really at odd with trends in CE devices, I believe it has huge implication on how sustainable the model is.
The price of the hardware significantly lessens the reach of consoles, no matter the good sales. It also promotes the creation of games that are extremely costly which in turn in an attempt to manage risk is affecting productivity and creativity.
The reach the consoles in the countries with lesser income is really low. The industry is really at odd with the entertainment industry too, Hollywood blockbusters are reaching the "third" world a lot more widely than consoles games.
I've no idea about how that can or will play out but I think that the market is clearly unbalanced and so a lot less solid than it looks. It performs well now. It seems that the console environment is trying to escape the trend toward commodity that affects CE devices. It really acts as a gravity field and I'm not sure that console can reach what would be the "escape velocity" matching that gravity field.
Last edited:
- Status