There is nothing "on the die" sensing temperature.
There is an LM75 component (not on the die, but on the board aka PCB) that senses temperature.
Everything else is semantics.
As there is nothing "on the die" sensing temperature, the data from the LM75 is the only information available to give an indication of the asic core temperature. It's obviously not the actual real core temperature, as tolip's IR gun figures show, but the data on the status/advanced page is coming from the LM75 and not from any other source. Which is what I have been saying.
If you need any proof of this - carefully - put your finger on the LM75 and watch the temperature value on the web interface change.
For people who don't know what a die is, here are some educational links
http://diephotos.blogspot.co.uk/
http://en.wikipedia.org/wiki/Integrated_circuit_packaging
http://electronics.stackexchange.com/questions/56649/what-is-a-die-package (see the second answer, which starts "A DIE is the actual silicon chip (IC) that would normally be inside a package/chip.")
Topic: Swedish ASIC miner company kncminer.com - page 294. (Read 3049515 times)
So you are saying the "Temperature" column here is coming from a circuit on the die, not an LM75 component on the board?
It is not either an LM75 ic OR a circuit on the die because it is BOTH of those things. The LM75A (if that's what's installed) is "a temperature-to-digital converter using an on-chip band gap temperature sensor"
"On-chip" in this sense generally implies the sensor is included in the IC package, not that it's reading an external sensor (read: on the SHA256 ASIC Die).
Further, the datasheet doesn't show any inputs for an external sensor. Thus, it is likely that it is reading the PCB/ambient around the place where it's mounted, not the ASIC chips core temperature.
Thanks tolip.
And maybe you could give your opinion on the statement "Core temperature sensing is performed by circuits on the die"
And maybe you could give your opinion on the statement "Core temperature sensing is performed by circuits on the die"
The LM75 measures it own case temperature.
There _IS_ a die in there though.
The original statement may have originated by someone who was wordlexic.
Die(s)/PCB/cube whatever.
I don't rely on that temp for much of anything.
I ALWAYS use the hottest temp I can find for cooling decisions.
As I type...
67C on bottom of PCB directly below ASIC
61C reported by LM75
I IR gunned the PCB temp.
I could probbably find hotter but bottom of PCB's is inside my pentagon with trapped V8 bottle.
It's tight in there and the bottle is not a cooperative captive.
https://i.imgur.com/9T8lRc2.jpg
YMMV
Could you please have a look and let me know what the part numbers printed on top of those two black 8 pin IC packages near the edge of the board and between the two cables in your photo are ?
Please forgive me for just answering the fucking question that was asked.
One is an
LM75
(closer to PCIe)
other is
24C32WP
k348k
It's the i2c EEprom
YMMV
Thanks tolip.
And maybe you could give your opinion on the statement "Core temperature sensing is performed by circuits on the die"
Could you please have a look and let me know what the part numbers printed on top of those two black 8 pin IC packages near the edge of the board and between the two cables in your photo are ?
Please forgive me for just answering the fucking question that was asked.
One is an
LM75
(closer to PCIe)
other is
24C32WP
k348k
It's the i2c EEprom
YMMV
So you are saying the "Temperature" column here is coming from a circuit on the die, not an LM75 component on the board?
It is not either an LM75 or a circuit on the die because it is both of those things. The LM75A (if that's what's installed) is "a temperature-to-digital converter using an on-chip band gap temperature sensor"; all it does is convert "temperature-to-digital" such that it can be presented as illustrated above.
How exactly do you think the LM75 would sense the temperature of the chip if it were not connected to a circuit on the die?
u27
So you are saying the "Temperature" column here is coming from a circuit on the die, not an LM75 component on the board?
It is not either an LM75 ic OR a circuit on the die because it is BOTH of those things. The LM75A (if that's what's installed) is "a temperature-to-digital converter using an on-chip band gap temperature sensor"; all it does is convert "temperature-to-digital" such that it can be presented as illustrated above.
How exactly do you think the LM75 would sense the temperature of the chip if it were not connected to a circuit on the die?
u27
"performed by circuits on the die" and "performed by a component on the board" are a bit different, no?
Yes, they are very different, and in this instance temperature sensing is "performed by circuits on the die".
"performed by circuits on the die" and "performed by a component on the board" are a bit different, no?
Yes, they are very different, and in this instance temperature sensing is "performed by circuits on the die".
Or do you still believe this?
Core temperature sensing is performed by circuits on the die.
Core temperature sensing is performed by circuits on the die; can't say it any clearer than that.
Likewise with the VRM temps; there is a circuit built in.
If it is an LM75A then it is "a temperature-to-digital converter using an on-chip band gap temperature sensor" i.e. "Core temperature sensing is performed by circuits on the die".
u27
"performed by circuits on the die" and "performed by a component on the board" are a bit different, no?
Or do you still believe this?
Core temperature sensing is performed by circuits on the die.
Core temperature sensing is performed by circuits on the die; can't say it any clearer than that.
Likewise with the VRM temps; there is a circuit built in.
If it is an LM75A then it is "a temperature-to-digital converter using an on-chip band gap temperature sensor" i.e. "Core temperature sensing is performed by circuits on the die".
u27
Could you please have a look and let me know what the part numbers printed on top of those two black 8 pin IC packages near the edge of the board and between the two cables in your photo are ?
One of them is an LM75, as I've told you a dozen times
Or do you still believe this?
Core temperature sensing is performed by circuits on the die.
user27, if you want to make the image smaller you can add a width or height attribute in the opening image tag. For example, i've added a width=256 statement inside the opening image tag.
Thanks, but I know really, was just messin'.
Could you please have a look and let me know what the part numbers printed on top of those two black 8 pin IC packages near the edge of the board and between the two cables in your photo are ?
Sorry, not my rig; it's all closed up now anyway as the cooling sucks without the box to direct the air flow/
What is the result in Gh/s gain?
It's more about cooling the VRM's properly so they don't burn out quite so quickly. However, with cooler ASIC temps he says his rig runs at about 3.5TH rather than 3.4TH (with one modified and four stock cubes). This could be down to reduced errors, I don't know/
If the same gain could be seen across the five cubes it could be upwards of 10% improvement. I should point out that this is based on a very subjective throw away comment and no proper testing.
u27
user27, if you want to make the image smaller you can add a width or height attribute in the opening image tag. For example, i've added a width=256 statement inside the opening image tag.
Could you please have a look and let me know what the part numbers printed on top of those two black 8 pin IC packages near the edge of the board and between the two cables in your photo are ?
You install the heatsink up to the plate or you removed the plate
As I said I can't claim that this is my handy work; but I think he removed the "E" shape plate yea.
You install the heatsink up to the plate or you removed the plate
As I said I can't claim that this is my handy work; but I think he removed the "E" shape plate yea.
You install the heatsink up to the plate or you removed the plate
Yes any photos please
So demanding...
(sorry, I couldn't figure out how to make the image take up any more space, please quote in full)
Yes any photos please
This is what you get if you put ram heat sinks on the VRM's and swap the fan out for a 3.3A 120mm one (fits inside the cube where the stock 140mm one was):
Connected to PSU via separate cable this also reduces load on that overworked PCI-e socket.
I can't claim that this is my handy work; but thought people might be interested to see.
u27
Connected to PSU via separate cable this also reduces load on that overworked PCI-e socket.
I can't claim that this is my handy work; but thought people might be interested to see.
u27
Jump to: