TL2 consumes about 1.6X power than TL

[Einst_Stein]

Advertisement (gone after registration)

The battery/power spec of TL2 according to CIPA is 250 while TL is 400. This is a ratio of 1.6 to 1.

The external V-finder seems also power hungry. Put both to work, power surge?

[Belle123]

Are the original T batteries the same for the TL2? They don't weigh me down too much, carrying few extra for the T. My other cameras have much heavier batteries by comparison, probably double to triple the weight. Last I read Leica knows what has caused the brick phenomenon and can be fixed via firmware IF your camera hasn't turned into a brick yet! So if you bought a TL 2, keep that EVF OFF!

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[Learner]

I believe the same battery works in in T, TL, and TL2

[fastfashn]

The battery/power spec of TL2 according to CIPA is 250 while TL is 400. This is a ratio of 1.6 to 1.

The external V-finder seems also power hungry. Put both to work, power surge?

 

The finder and the battery are the same.

[bencoyote]

In computing in general:

- the faster you run a processor the more power it consumes - if you want something more responsive you generally need to run faster; - the more bits you have to move the more power you consume - going from 16mp to 24mp means that you will need to move more bits;

- the more processing you do the more power you consume - adding more focus points over which you do contrast maximization to find focus lock

- the more blocks of storage that you write, the more power you consume - writing to SD consumes energy based upon the number of blocks you write so if you are writing a 24mp file rather than a 16mp file you will need to consume more energy.

 

If you just take the number of MP 24/16 is 1.5 which is darn close to 1.6x in the CIPA score and so I'd guess that most of that is due to the bump up of MP in the sensor.

 

However, as new semiconductor process technologies iterate you also get a reduction in power consumption for each and everything operation that you do. So they probably iterated the electronics one generation forward and got some benefits from that which were partially cancelled out by the MP bump in the sensor.

[TMorita]

The camera isn't writing to the SD card continuously, so that probably isn't the reason.

 

However, the camera is continuously reading the sensor to update the back LCD.

So it sounds like most of the extra power expenditure is reading the extra pixels from the sensor and processing them.

 

New semiconductor processes don't lead to lower power consumption anymore.

What you're referencing is called Dennard scaling, and basically it doesn't apply past about the 28 nm process node as I understand it.

Past about 28 nm, the wires become so thin that the static leakage current increases which leads to higher power consumption.

More info on Dennard scaling here:

 

https://cartesianproduct.wordpress.com/2013/04/15/the-end-of-dennard-scaling/

 

Wikipedia also has a good entry on Dennard scaling:

 

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

 

Toshi

[bencoyote]

Advertisement (gone after registration)

My professional job is doing high performance computing. I'm well aware of Dennard scaling (god am I aware of that). We can get down and dirty on this if you want but since this is camera forum it didn't seem like the place for that. ;-) Never the less the number of pico-joules per a fixed amount of computation has continued to go down over the years. You can see that in the reciprocal of the performance/Watt charts on: https://en.wikipedia.org/wiki/Performance_per_watt With higher resolution charts you can see that the each process technology somehow manages to squeeze a bit more computation out of each Joule used. It isn't the wonderful halcyon days before the end of Dennard scaling. Never the less Koomey's law https://en.wikipedia.org/wiki/Koomey%27s_law still is managing to hold for the time being  and process technology is part of that.

 

That being said, I have no idea if the TL2's embedded processor moved to the next semiconductor process.  I can however, tell from the specs that it will need to write more data to the SD card (actually a fairly expensive operation with regards to energy consumed) and it will have to read more data from the sensor and process it. I also think that increasing the number of AF focus points is likely to increase the processor's workload and so I think that the 1.5 increase in the resolution and the 1.6x decrease in CIPA score is more than coincidental. It is normal problem scaling physics. Try getting an exaflop machine to only consume 20MW. That's where you need to pull a few rabbits out of your hat. ;-) It is the same problem just in a different form.

[TMorita]

My professional job is doing high performance computing. I'm well aware of Dennard scaling (god am I aware of that). We can get down and dirty on this if you want but since this is camera forum it didn't seem like the place for that. ;-) Never the less the number of pico-joules per a fixed amount of computation has continued to go down over the years. You can see that in the reciprocal of the performance/Watt charts on: https://en.wikipedia.org/wiki/Performance_per_watt With higher resolution charts you can see that the each process technology somehow manages to squeeze a bit more computation out of each Joule used. It isn't the wonderful halcyon days before the end of Dennard scaling. Never the less Koomey's law https://en.wikipedia.org/wiki/Koomey%27s_law still is managing to hold for the time being  and process technology is part of that.

 

That being said, I have no idea if the TL2's embedded processor moved to the next semiconductor process.  I can however, tell from the specs that it will need to write more data to the SD card (actually a fairly expensive operation with regards to energy consumed) and it will have to read more data from the sensor and process it. I also think that increasing the number of AF focus points is likely to increase the processor's workload and so I think that the 1.5 increase in the resolution and the 1.6x decrease in CIPA score is more than coincidental. It is normal problem scaling physics. Try getting an exaflop machine to only consume 20MW. That's where you need to pull a few rabbits out of your hat. ;-) It is the same problem just in a different form.

 

You are assuming the performance per watt keeps increasing due to processors moving to smaller geometries.

This isn't fully correct as far as I know.

 

The end of Dennard scaling means chip designers can place more transistors on a die than they can actually power.

This is called "Dark Silicon" - it's silicon which is unused most of the time due to power constraints.

The way processors designers utilize dark silicon is to build custom hardware functions which are only used sporadically.

 

Intel has chosen to use the dark silicon to implement very wide vector instruction set extensions (AVX, AVX2, AVX512).

These instructions are more power-efficient because there's less power used in control logic as compared to individually-issued instructions, and also the data is fetched as a burst instead of individually.

 

http://www.informit.com/articles/article.aspx?p=2142913

 

It's unlikely that the TL2 processor has moved to the next process node IMHO.

The best price/performance is at the 28nm node, and has been for several years.

Past 28nm, the cost of chips goes up dramatically due to more complicated chip lithography, FinFETs, etc.

AFAIK most companies which build chips at lower volume are staying at the 28nm process node last I heard.

 

http://www.eetimes.com/author.asp?doc_id=1321536

 

Toshi

[TMorita]

Here's another more recent article where it mentions the 28nm process node is the sweet spot:

 

"28nm planar node has emerged as the sweet spot for many in the industry, as it combines the right balance of performance, power, area scaling and cost (PPASC) for application."

 

https://semiengineering.com/22nm-process-war-begins/

 

Toshi