Topic: [Review] Avalon 6 Miner - Winter Mining - Notlist3d - Also FAQ and Help - page 36. (Read 79347 times)

That is what he put.  If it's long term... not sure.  He has put they are working on optimizing on certain things I would guess this is in it as looking for best performance and efficiency.

I don't think you going to get all the answers you want today, but it's good to see we have a dev reading up.   Likely the key to alot including best operation speed and running temperature are in the "smart speed".   Which is still pretty new and were still learning about, and it sounds like it's still being optimized.

Sorry to be a pain but can I seek further clarification. Understand about not wanting 85-90C but are you saying that the chips are better at 65-70C than they would be at say 45-50C ?


Rich

Thank you for the explanation, this is now starting to make more sense.

For safty, we test it under ~ 85C, okey for some days. But ~90C is not good.

For performance, we find ~65-70C maybe better for both performance and efficiency.

ng

Thanks for the additional info. please feel free to share anything you are able to on the details. Many of us are as interested in the inner workings of a strate of the art miner like the Avalon6 as we are in the Bitcoin that they mine.

Just returning to a statement made in an earlier post.


From what you have said am I right in thinking that it is not correct that the "chip needs to run at 75C for optimal efficiency" but more that 75c is the Max temperature?


Rich

Thanks for the information it's great to hear that you are planning an update to down clock the A6!

I know it's a long shot but do you have any theorised, tested or project numbers around the down clock?

About temperate:

Next Cgminer software will adjust the default target tempature to ~68-70C. 75C is safe but a little high (too clost to the cut-off tempature).
Also, taget tempature can be set by user in next cgminer update.

About smartspeed:

This is a very complex system, partly built in the ASIC. Will add new algorithm in software side to optimize down clock and under voltege
condition later (may not finish in next update). But fully disable smartspeed is not recommend (will course a very high DH% and lost
20-50% real speed), so there will not a "DISABLE" switch.

ng

Thanks for joining the thread. Can you give a brief explanation of the temperate issue that is being discussed, does the miner really need to run at a minimum of75C ? Also an outline of the algorithm that SmartSpeed is using would be very instructive?


Thanks


Rich

Sounds great! We would love to hear more on smartspeed at some point.  It seems to handle overclock very well.  A lot of us are wanting to see what it can do as far as underclock, this might take a optimization as you mentioned.

Thanks for reading comments and trying to keep us up to date, it is much appreciated.

 

Hello guys, I'm reading the threads.

Right now the Cgminer software and MM firmware are preliminary versions, they "works " but not fully optimized of course.
Except for you guys metioned (fan control, voltage monitor, smartspeed, etc), there are also some other small bugs there,
We will try to release some updates soon.

ng

Most miners can run hotter then advertised.  It was not Avalon but another company I know of that gave a lower number to be safe on running temperature, but for big data centers it ran much hotter.  I will not suggest others do this as I don't want to break your warranty.

As far as asic's having a temp sensor on chip... I don't think that is likely.  From my readings and comparisons it seems temp sensor is close to exhaust side as temp out of exhaust and temp reading on GUI seem pretty close.  But this is speculation

I think we still have a decent amount to learn as far as "smart speed". It seems to do great as far as overclock, but we need to understand it better to get that underclock (which I am still hoping for).   The efficiency is still good for current gen as it is, so we have time to learn more.

Some have mentioned speed and fan percentage are directly linked with speed.  I will test in coming day's and see what speed change on fan minimum does.  Currently I have pretty cold air as ambient as I'm enjoying my window open, and the miner is still mining fine at 20 percent.    I will specifically try 20 percent see results and then 80 percent and see results, most likely a 30 min or hour test on each. 

I've ran over 5 day's at this point and no issues as far as rejects or other possible bad things with default settings.  And my temp in room with ambient outdoors has varied greatly.  So it seems to be able to handle a variety of ambient temps and still mine well.

Just trying to wrap my brain round this?.... A few thoughts.

So 75C chip temperature (Junction temperature?) is not that high and as has been said should not be confused with a temperature measured on the case of the chip or on  the circuit board (as in an Antminer). However I would regard ambient as being yet another temperature to get confused with?

So does this imply that the ASIC's have a built in temperature sensor? If so then yes you would expect to run at a higher temperature than if there is a separate sensor on the circuit board.

However what I do not understand is the concept that "the chip needs to run at 75C for optimal efficiency" I could understand if this meant not higher than 75C but why would it matter if the temperature was lower than this? I have always understood when it comes to chips that colder is better?

So putting all of that to one side for the moment we are next told "When it gets too cold you get a high error rate.  If you want to verify just set  your fan to min 80% and your hash rate will drop to about 3.0T" Now that would first imply that during warm up there is a high error rate and that the system would somehow have to allow for this?

So coming now to turning the fan to 80% and the hash rate dropping. Question for me is does the hash rate drop because the firmware just turns the frequency down when you turn the fan up or does the HW error rate actually go up and then the frequency is turned down to reduce the HW error rate?

Again I am finding this confusing. Surely if the chip gets a bit cold and the frequency is then turned down the chip will then run even colder because of the reduced frequency and is just going to keep going down and down?


All very strange and confusing to me, hopefully the full and true story will emerge at some point?


Rich

Thanks for posting.  It is interesting to read a lot is still not known on some of the smart stuff.  So nice to some info about it.

Alright transcript  below:


Sales Account
12:14 am (1 day ago)

to me
Most graphics cards run at 90 to 100 c. 75 c is the core temp of the chip.  Ambient temperature and core are very different


On Nov 28, 2015 12:04 AM, "Jerome Humphrey" <[email protected]> wrote:
75 degrees wow that seems quite high? Is their a design reason for this?

So what is the max temp range for the hash boards before they start to risk blowing up?

Cheers,
J

On Fri, Nov 27, 2015 at 11:57 PM, Sales Account <[email protected]> wrote:
The smart speed is very important.  The reason the fan control works the way it does is because the chip needs to run at 75C for optimal efficiency.  When it gets too cold you get a high error rate.  If you want to verify just set  your fan to min 80% and your hash rate will drop to about 3.0T

Thanks
yan

On Fri, Nov 27, 2015 at 11:30 PM, Jerome Humphrey <[email protected]> wrote:
Happy Thanks giving! You guys pulled my name out of a hat yet for the blog comp? Haha

Can you confirm that it is planned in a future firmware update that we will have control over the fanspeed ie be able to turn off smartspeed?

Thanks,
J

Note the email chain is longer but it's just me looking at getting stuff sent to me in NZ

Yes if you could copy/paste it.  I (and I think others) would be interested reading it in whole.   As there has not been a ton of info on "The smart speed" and were still really learning about it. 

With underclock I think it sounds like we need to find a way to get around it, or disable so the more we know the better.

This is the email I'm talking to Yan on Sales Account <[email protected]>  Or do you want a transcript?

Please post email Bonez so we can see exactly.  I'm still holding hope that someone will be able able to find a way to underclock.

I feel like were getting half of the info.  But I could be wrong.

That shouldn't prevent underclocking. It should be able to slow the fan down sufficiently - or even stop it - to reach 75C at any reasonable frequency/voltage. Am I missing something here?

well thats certainly interesting. i wonder why a chip performs better at a certain temp. Goes against what I've always thought.

Lol it's Sunday I'm a bit hungover

I'm talking to Yan in the sales dep Sales Account <[email protected]>

After asking if there is a firmware update in the works where we will have full control over the fanspeed he said "The smart speed is very important.  The reason the fan control works the way it does is because the chip needs to run at 75C for optimal efficiency.  When it gets too cold you get a high error rate.  If you want to verify just set  your fan to min 80% and your hash rate will drop to about 3.0T"

I also asked him about the ambient temp being so high and he said "Most graphics cards run at 90 to 100 c. 75 c is the core temp of the chip.  Ambient temperature and core are very different."