Germany Carl Zeiss, heart of Dutch ASML Lithography Equipment.


Status
Not open for further replies.

AndrewS

Captain
Registered Member
support it.
If an IC will sell in 100 millions there is same (marginal) benefit to go for small node ( like slightly smaller mobile phone : D ) , but I seriously doubt that there will be any base-station controller on sub 22nm node.
The benefit of smaller nodes is lower electricity usage per calculation.

Smartphones have limited battery capacity.
So a smaller process node means the smartphone lasts longer before it runs out of battery.

With datacentres, it's the lower operating costs of electricity and the associated cooling costs. It costs less money to do the same work with a smaller node
 
Last edited:

Chish

Junior Member
Registered Member
Desktops dying out because there is no speed improvement.

There was a decade long learning curve in phones, and now they dying out as well, the sales flat/decreasing now.

There is no performance improvement in the last decade , so no one wants to buy desktop.

With phones there was a lot of money spent for phone capabilities, new screens/custom ICs and so on.

But it ended as well - there was no need to replace the desktop in the past 10 years, and now there is no need to replace the phone.

Incredible, you try to falsify me with data that actually support my point ?


In the case of data centres 70% of the new gen server CPUs contain ASIC circuits, tailored for specific customer.

They don't get the edge from the underlying technology - they managed to sell the products due to higher and higher customisation.
Desktops have never catch up in China just like cerdit cards. These two are basic necessities in the Western world's everyday life but almost useless in China. China skip a generation of IT communication instead going directly into digital mobil communication. Even beggers use QR code thesedays.
 

Anlsvrthng

Senior Member
Registered Member
The benefit of smaller nodes is lower electricity usage per calculation.

Smartphones have limited battery capacity.
So a smaller process node means the smartphone lasts longer before it runs out of battery.

With datacentres, it's the lower operating costs of electricity and the associated cooling costs. It costs less money to do the same work with a smaller node


Where could I see dramatic improvement in watthour/calculations ?
I see a pitifully maybe 2-3 times improvement over 15 years, most likely contributed to architectural improvements.

The desktop power consumption "only " doubled twofold over 15 years. , so any performance/watt improvement should show itself in performance improvement.

From the above table the only benefit of smaller nodes is to have few mm thinner phones, with 1cm2 die instead of 7-8 cm2.


That is all .


And the difference in the freefall of desktop and laptop sales is simple: the laptop wear quicker than the desktop, so it require more frequent replacement.
 

nlalyst

New Member
Registered Member
Where could I see dramatic improvement in watthour/calculations ?
I see a pitifully maybe 2-3 times improvement over 15 years, most likely contributed to architectural improvements.
From 2008 to 2018, each node has brought about 1.6x improvement in transistor energy efficiency.
rsta20190061f05.jpg
Source: The future of computing beyond Moore's law, John Shalf

If we take architecture into account, the improvements are actually quite substantial (~19x) as can be seen by comparing the Top500 lists, sorted by power efficiency from Nov 2009 and Nov 2019.

Top 500, November 2009. X-axis rank in the Top500 list
eff2009.png

Top 500, November 2019. X-axis rank in the Top500 list
Eff2019.png
 

tidalwave

Senior Member
Registered Member
  • Thread Starter Thread Starter
  • #475
Missing the point here when talking about nodes and architecture.

Going after EUV lithography underlying technology means China will comprehensively bridge the gap that it lacks fundamental science research capability of industrial power houses like US, Japan and Germany.

Through this effort it will bring China entire industry to a different level.

45nm and 28nm optical lithography is nowhere near the extreme physics limitation and challenges of EUV lithography.it's different world.

I see quite number of Russian physicists from Moscow institutes of physics are hired by Western companies in the EUV industry.

Maybe China and Russia should partner up on this EUV Project. Russians traditionally very strong on physics and most importantly the ability of industrialization of it.
 
Last edited:

tidalwave

Senior Member
Registered Member
  • Thread Starter Thread Starter
  • #476
I really think the EUV lithography consortium among China, Russia , Iran will strengthen the alliance. If successfull it will free up the 2nd world from dependence technologically on the west. It has great geopolitical implication. Strangely the current Wassenar agreement is headed by US with its sick mindset to block China , Iran , North Korea from acquiring leading edge commercial tools. Yes, China is being lumped into the group of Iran and North Korea despite being the second biggest economy in the world in the eyes of US. Ironically, russsia and india are members of Wassenar agreement, being accepted by US. I think that's so wrong on so many different levels. How US mindset is so sick and worrying , almost like anyone but China.. Lumped an industrial giant like China into the rank of Iran and North Korea. Full of innate of racism , hatred, and hegemony all mixed together
 

Anlsvrthng

Senior Member
Registered Member
From 2008 to 2018, each node has brought about 1.6x improvement in transistor energy efficiency.
View attachment 57889
Source: The future of computing beyond Moore's law, John Shalf
Thanks, this kind of discussion is the reason I visit these forums : )

Figure 5. The energy consumption of compute and data movement operations at different levels of the compute hierarchy—from the arithmetic logic unit on the left to system-scale data movement across the interconnect on the right. As lithography has improved, the energy efficiency of wires has not improved as fast as the efficiency of transistors. Consequently, moving two operands just 2 mm across a silicon chip consumes more energy than the floating-point operation performed upon them. (Online version in colour.)
Means there is massive penalty for memory operations, small batch multiple operation is the preferred way of computing.

The transistor efficiency increased, but the movement of data consumes as much energy as before .

Explain why the GPUs have 40* performance improvement but only 7 time bandwidth improvement.

Actually the VLIW is quite effective way to overcome this kind of issues.
 
Missing the point here when talking about nodes and architecture.

Going after EUV lithography underlying technology means China will comprehensively bridge the gap that it lacks fundamental science research capability of industrial power houses like US, Japan and Germany.

Through this effort it will bring China entire industry to a different level.

45nm and 28nm optical lithography is nowhere near the extreme physics limitation and challenges of EUV lithography.it's different world.

I see quite number of Russian physicists from Moscow institutes of physics are hired by Western companies in the EUV industry.

Maybe China and Russia should partner up on this EUV Project. Russians traditionally very strong on physics and most importantly the ability of industrialization of it.
Wondering what lithography technology China use for her top notch AESA radar? and who manufacture them?
 

tidalwave

Senior Member
Registered Member
  • Thread Starter Thread Starter
  • #479
Wondering what lithography technology China use for her top notch AESA radar? and who manufacture them?
GaN, GaAs is six inches wafer, whereas digital chip is 12 inches wafer.
Therefore GaN, GaAs using much older technology it stil optical lithography like digital chip. Sanan is top maker
 
Status
Not open for further replies.

Top