[Work in progess] Burnins Avalon Chip to mining board service

[SebastianJu]

I did a number of overclocking testruns now. The normal Voltage i used was 1300mV.

I found its a bit strange when i sat the fan to 100% this setting became active only after 30seconds to 1 minutes after the start. I dont know why that is. Its the same with Temp 40°C but it seems to be a bit faster. Normally i would think a 100% setting means 100% from the start but it isnt.

One should keep in mind, when using windows, that cgminer seems not to keep windows from closing down automatically. Its not like showing a video that prevents windows from going to hibernation. So when you dont change that setting in windows before it will stop after some time. I had this over night.

Results:

Code:

MHz	% more	Accep-	HW-	HW to	Tempe-	Addit-	Result			Real Hash-	Note
	Hash-	ted	Errors	Accep-	rature	ional				rate gained
	rate	shares		ted		setting				compared to
				shares						previous
										hashrate
	
282
400	41.8%	3400	0	0%	49°C		Effect positive		+ 41.8%
410	2.5%	3000	1	0.033%	48°C		Effect positive	 	+ 2.467%
420	2.44%	3000	4	0.1333%	49°C		Effect positive		+ 2.3067%
430	2.38%	3040	12	0.3947%	48°C		Effect positive		+ 1.9853%
440	2.32%	3008	68	2.26%	49°C		Effect compensatory	+ 0.06%
440	2.32%	34000	677	1.99%	48°C	1310mV	Effect compensatory	+ 0.33%		Its not possible to set 1 more millivolt with Standard cgminer
440	2.32%	12000	248	2.066%	49°C	Temp40°	Effect compensatory	- 0.77%		
440	2.32%	23504	297	1.263%	46°C	Temp40°	Effect compensatory	+ 1.057%		
						1310mV
450	2.27%	3008	473	15.724%	48°C		Effect negative		- 13.454%
450	2.27%	10224	901	8.8125%	45°C	100%Fan	Effect negative		- 6.542%	
450	2.27%	13336	807	6,05%	49°C	1310mV	Effect negative		- 3.78%
450	2.27%	14024	895	6.38%	49°C	1310mV	Effect negative		- 4.11%
						Temp40°

It looks like a lower temperature doesnt have a big effect. The more effect has a higher voltage. Even 10mV is helping not bad. 1311mV is not possible but 1310 is the last possible value. When you want to use more you have to compile cgminer and change a constant. What i will do later.

I believe burnin wrote 1.5V is a sure way to kill the chips. I wonder what is the max. I want to find out but dont like to risk the chips too much too... im not sure if it could be found out safely. The only way i can think of is to use only one miner for the test but that doesnt say anything about longer runtimes.

@burnin... is there a way to only use 10 chips of one miner? Maybe with a jumper?

Is it possible to power a raspberry pi with 5v power and gnd from a atx-psu-cable connected to pin 2 and 6 or 4 and 6 at the rpi? See here for the pins: http://elinux.org/RPi_Low-level_peripherals#General_Purpose_Input.2FOutput_.28GPIO.29

Im not sure about the gain in hashrate with higher clockrate and the correlation between accepted shares and HW-Errors are a correct indicator since i think that HW-Errors have a bigger impact on the hashrate than my calculations show. At least the average hashrate drops by more percents than i calculate it. The best indicator probably would be to let it run until 10 or 20k accepted shares and check the second hashrate value (avg) until you find the best value.

[ProfMac]

I believe burnin wrote 1.5V is a sure way to kill the chips. I wonder what is the max. I want to find out but dont like to risk the chips too much too... im not sure if it could be found out safely. The only way i can think of is to use only one miner for the test but that doesnt say anything about longer runtimes.

The way to do this is to pay burnin to make a board with only 1 or 2 chips on it, and to ramp it up.  Keep it at each step an hour or two.  Keep a database of voltage, frequency, HW, and hashrate.

If I ran the test on my chips, I would set it as 400 MHz or whatever the max spec is, and run the voltage up and plot HW & hashrate.  I would stay inside the design spec on this test.

So ... the thing is to get a quote from burnin to run the test, set a pay-to address, and when enough money is there, those are the people who learn the results...

[SebastianJu]

I believe burnin wrote 1.5V is a sure way to kill the chips. I wonder what is the max. I want to find out but dont like to risk the chips too much too... im not sure if it could be found out safely. The only way i can think of is to use only one miner for the test but that doesnt say anything about longer runtimes.

The way to do this is to pay burnin to make a board with only 1 or 2 chips on it, and to ramp it up.  Keep it at each step an hour or two.  Keep a database of voltage, frequency, HW, and hashrate.

If I ran the test on my chips, I would set it as 400 MHz or whatever the max spec is, and run the voltage up and plot HW & hashrate.  I would stay inside the design spec on this test.

So ... the thing is to get a quote from burnin to run the test, set a pay-to address, and when enough money is there, those are the people who learn the results...

Thats an idea... i would contribute to such test since burnin has the ability to do this relatively cheap to find the result. When i want to go above 440MHz i need to go higher thatn 1310mV. Thats not possible with standard cgminer so a special cgminer is needed where the constants are changed to allow higher voltage.
Burnin? Can you do such test?

The only risk then remains is hot environments that could break the chips and that the chips then might only run for some weeks. But one could go a bit under he breaking values...

[MorePork]

Burnin did some OC tests already and it looks like you'll need 1.34V to run at 450Mhz.

https://bitcointalk.org/index.php?topic=179769.msg2745262#msg2745262

And i have some numbers to go with those from yesterday:
Slightly different air cooling setup therefore different temperatures with air cooling. (fan placement)
TL;DL : 450Mhz [9Ghash/s] - STABLE
But at the cost of 94Watts of power.

Air:
431 - 54, 48, 1.30V, 87W, stable
450 - 56, 48, 1.30V, 90W, HW Errors
450 - 57, 52, 1.34V, 94W, slightly increased error rate compared to what i normally call "stable" but close enough

Water:
450 - 54, 32, 1.34V, 94W, slightly less hw errors then with air

[iikun]

Burnin did some OC tests already and it looks like you'll need 1.34V to run at 450Mhz.

https://bitcointalk.org/index.php?topic=179769.msg2745262#msg2745262

And i have some numbers to go with those from yesterday:
Slightly different air cooling setup therefore different temperatures with air cooling. (fan placement)
TL;DL : 450Mhz [9Ghash/s] - STABLE
But at the cost of 94Watts of power.

Air:
431 - 54, 48, 1.30V, 87W, stable
450 - 56, 48, 1.30V, 90W, HW Errors
450 - 57, 52, 1.34V, 94W, slightly increased error rate compared to what i normally call "stable" but close enough

Water:
450 - 54, 32, 1.34V, 94W, slightly less hw errors then with air

Does cgminer allow this voltage?  It seems the best/only way to deal with the errors experienced at 400+ MHz.

[kano]

I set the limit to 1.31 since I was thinking that 1.28 was the normal limit to use ... I can't remember why

burnin what would you say people shouldn't go above?

(i.e. I'll set that limit in cgminer and if anyone wants to destroy their board they can just recompile with a higher limit)

[eraziel]

I set the limit to 1.31 since I was thinking that 1.28 was the normal limit to use ... I can't remember why

burnin what would you say people shouldn't go above?

(i.e. I'll set that limit in cgminer and if anyone wants to destroy their board they can just recompile with a higher limit)

Can you put it up to 1.35 instead and give out some kind of warning or make the default option not take values above 1.3 and have a separate option that takes values above that and stating higher risk of stuff being broken?

[kano]

I set the limit to 1.31 since I was thinking that 1.28 was the normal limit to use ... I can't remember why

burnin what would you say people shouldn't go above?

(i.e. I'll set that limit in cgminer and if anyone wants to destroy their board they can just recompile with a higher limit)

Can you put it up to 1.35 instead and give out some kind of warning or make the default option not take values above 1.3 and have a separate option that takes values above that and stating higher risk of stuff being broken?

No, I'm not adding another option.
The limits will be the expected, non-destructive limits, and, as I said, if you want to exceed them, simply recompile cgminer.
At the moment, if you do a java API "ascset|0,help" then it will tell you the limits also.
Having those defined means people will know what they can use.
... the more adventurous, or those who don't care if it destroys their hardware, simply edit and recompile, it's very simple to compile cgminer ... and we even include the details how to

[maxmint]

I set up team "BitBurner" on BTCGuild – feel free to join.

[maxmint]

I set up team "BitBurner" on BTCGuild – feel free to join.

Need PIN

Sorry, just removed the PIN. Please try again.

[SebastianJu]

Burnin did some OC tests already and it looks like you'll need 1.34V to run at 450Mhz.

https://bitcointalk.org/index.php?topic=179769.msg2745262#msg2745262

And i have some numbers to go with those from yesterday:
Slightly different air cooling setup therefore different temperatures with air cooling. (fan placement)
TL;DL : 450Mhz [9Ghash/s] - STABLE
But at the cost of 94Watts of power.

Air:
431 - 54, 48, 1.30V, 87W, stable
450 - 56, 48, 1.30V, 90W, HW Errors
450 - 57, 52, 1.34V, 94W, slightly increased error rate compared to what i normally call "stable" but close enough

Water:
450 - 54, 32, 1.34V, 94W, slightly less hw errors then with air

I know, but when i would test out the real limits i would lose not a few money. Burnin has the possibility to do this for a low price and when we donate for this test he maybe is willing to do it. Of course he would need a recompiled cgminer then.

[maxmint]

I'm mining at BTCGuild right now and noticed quite a difference between local hashrate (as reported by cgminer) and the estimated speed at BTCGuild.
cgminer reports 128.9 GH/s on average, while BtCGuild reports 106.7 GH/s.
Is this difference within usual ranges?

[Tehfiend]

Is there a "quick setup" guide posted somewhere? My miners are en route so I figured I better get the proper PSU(s) and setup my Raspberry Pi in advance but don't want to comb through the 118 pages in this thread for the specifics. I figured since they are shipping all the relevant info would be condensed somewhere but can't seem to find anything. If not I'll get back to reading this thread...

P.S. I love you Burnin <3

[BenTuras]

Is there a "quick setup" guide posted somewhere?

http://www.burninmining.com/news/

[Tehfiend]

Is there a "quick setup" guide posted somewhere?

http://www.burninmining.com/news/

Wow not sure how I missed that  . Thank you!

[SebastianJu]

I now run my miner from the rpi. It works so far but i want to know how and where i can change the const for the maximum voltage. I used ./autogen.sh --enable-avalon to compile but i dont know what file to edit for the const.

Then, regarding my previous question, i run the rpi through a smartphone loading cable but would like to run it from the atx psu. I found complicate tutorials for creating a usb-connector with power but i dont like it. So is it possible to simply power the rpi through pin 2 and 6 or 4 and 6 connected to a 5v cable from the psu?

I would like to have that the rpi, when it starts, automatically starts cgminer and that a crashed or broken cgminer is restarted, or the whole rpi. How to achieve this? It would be best to not having to see what the rpi does, only connect power and it mines and controls itself.

[Tursk]

I now run my miner from the rpi. It works so far but i want to know how and where i can change the const for the maximum voltage. I used ./autogen.sh --enable-avalon to compile but i dont know what file to edit for the const.

Probably these lines in driver-avalon.h

Code:

#define BITBURNER_DEFAULT_CORE_VOLTAGE 1200 /* in millivolts */
#define BITBURNER_MIN_COREMV 1000
/* change here if you want to risk killing it :)  */
#define BITBURNER_MAX_COREMV 1310

[roy7]

I use MinePeon with my Pi and it works great. Starts mining on bootup, web interface to configure pools.

[villex]

Then, regarding my previous question, i run the rpi through a smartphone loading cable but would like to run it from the atx psu. I found complicate tutorials for creating a usb-connector with power but i dont like it. So is it possible to simply power the rpi through pin 2 and 6 or 4 and 6 connected to a 5v cable from the psu?

Check this out: http://dangerousprototypes.com/docs/ATX_Breakout_Board

[SebastianJu]

I use MinePeon with my Pi and it works great. Starts mining on bootup, web interface to configure pools.

I think ill give it a try... can i edit cgminer consts there too somehow?

Then, regarding my previous question, i run the rpi through a smartphone loading cable but would like to run it from the atx psu. I found complicate tutorials for creating a usb-connector with power but i dont like it. So is it possible to simply power the rpi through pin 2 and 6 or 4 and 6 connected to a 5v cable from the psu?

Check this out: http://dangerousprototypes.com/docs/ATX_Breakout_Board

I know what voltages the cables from the atx have. The only thing i dont know is if i can safely take a 5V and a GND cable from the atx and power the rpi through its pins with it. I dont want to break it and for sure im not the only one coming up with that idea. Unfortunately i only found people that created a usb-connecter to power the rpi with a psu.