Exceeding Limits: The 621 Million Transistor Chip

  • Thread starter Thread starter wolram
  • Start date Start date
  • Tags Tags
    Transistor
AI Thread Summary
The discussion centers on the KiloCore chip, which features 1,000 independent programmable processors and operates at a maximum computation rate of 1.78 trillion instructions per second with 621 million transistors. While some participants initially believed this transistor count indicated a limit in density, it was clarified that modern chips often exceed this number, with many desktop CPUs containing over 1 billion transistors. The KiloCore chip is noted for its unique massively parallel architecture and energy efficiency, consuming only 0.7W compared to traditional GPUs like the NVIDIA 970, which requires 145W. The conversation also highlights the chip's significance as the first of its kind designed at a university, emphasizing its innovative processing capabilities. Overall, the KiloCore chip represents a notable advancement in microchip technology despite its transistor count.
wolram
Gold Member
Dearly Missed
Messages
4,410
Reaction score
555
I thought we were on the limit for transistor density but this chip looks like it is way over that limit.

From science daily.
https://www.sciencedaily.com/releases/2016/06/160617215802.htm

A microchip containing 1,000 independent programmable processors has been designed. The energy-efficient 'KiloCore' chip has a maximum computation rate of 1.78 trillion instructions per second and contains 621 million transistors.
 
Engineering news on Phys.org
Ho hum. Modern off the shelf CPU chips routinely have over 1,000,000,000 transistors so 621,000,000 isn't particularly impressive.

The one in my desktop has 1,400,000,000 and it's not particularly high end.
 
Last edited:
@phinds - the chip described in the article is massively parallel, and is unique. IMO. It is fundamentally different from a GPU-like cpu as you cited. And apparently very efficient - energy consumption ~0.7W as opposed to an NVIDIA 970 which is 145W minimum. It runs on a single 6V battery.

"To the best of our knowledge, it is the world's first 1,000-processor chip and it is the highest clock-rate processor ever designed in a university," said Bevan Baas, professor of electrical and computer engineering, who led the team that designed the chip architecture
- UC Davis
 
jim mcnamara said:
@phinds - the chip described in the article is massively parallel, and is unique. IMO. It is fundamentally different from a GPU-like cpu as you cited. And apparently very efficient - energy consumption ~0.7W as opposed to an NVIDIA 970 which is 145W minimum. It runs on a single 6V battery.

- UC Davis
Agreed, but the OP was commenting on his thought that this had hit a limit in transistor density, which is what I was responding to.

EDIT: Hm ... I see that actually I was not looking at the DENSITY of the chip, just the number of transistors so my point may be off.
 
Last edited:
wolram said:
I thought we were on the limit for transistor density but this chip looks like it is way over that limit.

32nm technology is not as dense as it gets these days. Go here, and look to the right towards the top of the page for a table that puts 32nm in perspective.

https://en.wikipedia.org/wiki/International_Technology_Roadmap_for_Semiconductors

The big deal about the chip you referenced is the amount of parallel processing it can do - imo quite impressive.
 
Millions, trillions of transistors.
Wouldn't it be fun to scroll the circuit diagram of such a beast?

Ah... well... maybe not ; )

John
 
D_Arsonval said:
Wouldn't it be fun to scroll the circuit diagram of such a beast?
Actually, I'm pretty sure these things are built with magic ... no diagram required :smile:
 
Clarke's Third Law
"Any sufficiently advanced technology is indistinguishable from magic." (Arthur C. Clarke)
 
I doubt there ever was a circuit diagram. It went:

verilog --> netlist --> gds --> chip
 
Back
Top