Pages:
Author

Topic: Fury/Blizzard tuning and mods - page 40. (Read 115281 times)

full member
Activity: 140
Merit: 100
June 14, 2014, 01:27:15 PM
#81
Now that we have schematics, should we refer to these resistors as R10 and R9?

Texas Instruments pdf on top
Zeusminer PDF on bottom


Zeus schematic


J4bberwock - could you please confirm that this looks accurate.

If so we can update the formula to:
R10=0.591xR9/(desired voltage-0.591)

Zeus doc indicates R10=10.5k and R9=10k
We don't know if they have made any changes so maybe best to measure and plug into the formula I think.

TI doc seems to indicate R10 (TI R7) as being the resistor to change. Seems like R9 (TI R8) has additional function - Error Amplifier Compensation - I don't enough about this stuff to know for sure.  J4bberwock hope you can help me out here.

It's so great that we have the schematics to refer to now.
hero member
Activity: 840
Merit: 1000
June 14, 2014, 01:10:31 PM
#80
At clock 515, coinshift says i'm hashing at 4.8-5.0 MH including rejected for 2 Furys
I'll add 2 more overclocked ones and let them run all night long to confirm.
If it's real, I have created the first Zeus mini-blade.
hero member
Activity: 840
Merit: 1000
June 14, 2014, 12:54:52 PM
#79
Well I've been slowly increasing the clock speed every few hours and it seems that at least with Hashra, it stops displaying the correct hashrate above the 520Mhz mark.  Instead of 1.5mhs or whatever it should be, it shows as .51mhs.  Poolside is showing a range of 1.51 to 1.95mhs though.


Same thing here with hashra.
 I tried clock 600 and still the same hashrate poolside as clock 500. There must be something else that caps the real frequency in the chips. Maybe playing with  the crystal could help, but we need the datasheet for the chip.

Edit

600 clock gave me less than 515 poolside. Maybe related to cores disabling if I ask too much.
One Fury was reporting 1,5 in hashra and the second one was 0.5
I now have both reported as 1.5 and something like 4Mh/s poolside, so 2Mh each
sr. member
Activity: 252
Merit: 254
June 14, 2014, 12:34:30 PM
#78
Well I've been slowly increasing the clock speed every few hours and it seems that at least with Hashra, it stops displaying the correct hashrate above the 520Mhz mark.  Instead of 1.5mhs or whatever it should be, it shows as .51mhs.  Poolside is showing a range of 1.51 to 1.95mhs though.
full member
Activity: 140
Merit: 100
June 14, 2014, 10:39:59 AM
#77
That'll help!  Thanks ZiG.  http://zeusminer.com/shcematics/
sr. member
Activity: 252
Merit: 254
June 14, 2014, 10:29:37 AM
#76

Awesome!  2nd post updated.  Thanks Zig!
ZiG
sr. member
Activity: 406
Merit: 250
June 14, 2014, 09:58:33 AM
#75
full member
Activity: 140
Merit: 100
June 14, 2014, 07:05:45 AM
#74
Re: Fury/Blizzard tuning and mods

I am thinking about canceling my preorder of the fury because what needs to be done to get it stabilized, This thread 98% convinced me to cancel.

I've done nothing to stabilize two Fury's other than run BFGMiner.  Both are 1.4mh/s poolside right now.  340 clk.   

So no HW errors? No rejects?

Edit:
From my understanding. T.I.M. Needs to be reapplied. That's buying TIM. Heatsinks need to be applied. That requires more purchasing. And I really do not want to be a nuisance again having someone mod a device for me.

This thing should work out of the box without problems.

Hardware 5% & 6% errors.  Seems to get better with each new software build.

Poolside: 0.35% Stale (stale is same as my Gridseeds) 0.00% Invalid - that and poolside hashrate is what matters.   1,401 kH/s & 1,407 kH/s poolside avg now.  I'm pushing the clk too at 340 clk.  328 clk is recommended.

I'm not suggesting you buy one, just wanted to be clear that the ones I have at least exceed advertised performance of 1.3 mh/s.  I haven't taken them apart or anything like that.  So mine work "out of the box" with good results poolside.

full member
Activity: 140
Merit: 100
June 14, 2014, 06:45:53 AM
#73
I am thinking about canceling my preorder of the fury because what needs to be done to get it stabilized, This thread 98% convinced me to cancel.

I've done nothing to stabilize two Fury's other than run BFGMiner.  Both are 1.4mh/s poolside right now.  340 clk.   
full member
Activity: 140
Merit: 100
June 14, 2014, 06:42:45 AM
#72
Would using conductive thermal paste matter on the back of the board?  Does it ground to the enclosure anyway?  (Just trying to use up what I have on hand)

I have 10 3sq cm thermal pads.  It's not enough to cover the back of the board but could be used where the hot spots are.  Do you think thermal paste would perform better than leaving air gaps because of the shortage of pads?  It would take a lot of thermal paste to fill the gaps.

I touched a temperature probe to one Zeus chip at 340 clk.  (Fury was fully assembled) The reading was 50C.

legendary
Activity: 1288
Merit: 1004
June 14, 2014, 06:01:19 AM
#71
I think the thing we need the most is the ability to get these running on current versions of BFGMiner and cgminer.
So many of the dam errors are from the ancient cgminer 3.1.1
I would love to see support for it on the newer ones and setup on MinePeon or some such easy to use distro.  That would really help these little beasts out.
I am still not happy I had to add thermal paste to my Blizzard so I could get it stable enough for my review.
legendary
Activity: 1018
Merit: 1001
June 14, 2014, 05:36:48 AM
#70
Hi, is available a repository or a git source for compiling CGMINER in Linux with Zeus support, or better an hybrid Zeus and Gridseed support?

Thank in advance W_M
hero member
Activity: 840
Merit: 1000
June 14, 2014, 05:16:39 AM
#69
I did the mod...used a 50k trim pot adjusted down to 7k.

I had to leave but I was running it at 450mhz with a 3% error rate.  I just upped it to 500mhz and so far 0 errors in 10 minutes.  It's averaging 1.52mhs.  I'll let it run at 500 while I'm out and report back.  Temp probe only registered an average 3f increase with my heatsink setup.



Also....if anyone is running Hashra you have to edit the /var/www/index.php to allow those higher speeds

Super...

So...1 Fury/Blizzard/Zeus 1.3MH voltmod & overclocked to 500MHz = 3 Gridseeds voltmod & overclocked to 500M

NOT BAD at all...for $120...only 1 device/connections...vs 3...

Mine is coming...I don't know when... Angry

Cheers,

ZiG



I believe we can get 2.2 MH from the chips on the Fury so it will be like 4 Gridseed overclocked pods
Time for lunch, and after that, I'll set up the 2 Furys with server Psu and raspberry in the room with the other miners.
I'll report on the power draw too.
hero member
Activity: 840
Merit: 1000
June 14, 2014, 04:39:21 AM
#68
Some update to show you what I mean by serious cooling

The fan was from an HP DL580 G3 power suply, while I'm waiting for a new fan.
Makes same noise as a big vacuum cleaner, but it does the job very well.

Heatsinks are for VRM from arctic accelero L2 plus heatsinks I bent them to give a wider shape.








Still need to add some cardboard or something similar to close the sides where the fan was originally blowing So I'll have a nice airflow there too.


legendary
Activity: 3654
Merit: 8909
https://bpip.org
June 13, 2014, 11:30:01 PM
#67
Here is a formula to calculate actual hashrate without relying on pool numbers, which can be inaccurate:

difficulty_1_shares * 65536 / time_in_seconds

If you are using cgminer difficulty 1 shares are easy to obtain from the API ("Difficulty Accepted" in "POOLS" data, you can also add "Difficulty Rejected"). Time in seconds is the time elapsed since cgminer started. Newer cgminer versions also have this in the API ("Device Elapsed" in "DEVS") but unfortunately 3.1.1 doesn't so you'll need to get that from somewhere else (e.g. ps in Linux).

There are plenty of examples out there on how to query the API. Here is a quick one for Linux (red things are the ones you type in, green are the important bits):

Quote

such@moon:~$ telnet 127.0.0.1 4028
Trying 127.0.0.1...
Connected to 127.0.0.1.
Escape character is '^]'.
{"command": "pools"}
{"STATUS":[{"STATUS":"S","When":1402719573,"Code":7,"Msg":"3 Pool(s)","Description":"cgminer 3.1.1"}],"POOLS":[{"POOL":0,"URL":"stratum+tcp://useast.wafflepool.com:3333","Status":"Alive","Priority":0,"Long Poll":"N","Getworks":52,"Accepted":23,"Rejected":0,"Discarded":101,"Stale":0,"Get Failures":0,"Remote Failures":0,"User":"richbeyondimagination","Last Share Time":1402719533,"Diff1 Shares":12288,"Proxy Type":"","Proxy":"","Difficulty Accepted":11776.00000000,"Difficulty Rejected":0.00000000,"Difficulty Stale":0.00000000,"Last Share Difficulty":512.00000000,"Has Stratum":true,"Stratum Active":true,"Stratum URL":"useast.wafflepool.com","Has GBT":false,"Best Share":8956}],"id":1}Connection closed by foreign host.
such@moon:~$ ps -eo pid,cmd,etime | grep cgm
7287 /such/mine/cgminer-zeus -S        15:53
such@moon:~$

EDIT: Lower pool difficulty will give you more accurate numbers, and obviously a longer time period will too.
sr. member
Activity: 252
Merit: 254
June 13, 2014, 11:01:41 PM
#66
2nd post updated with J4bberwok's info that was posted. 

Also...my poolside hashrate is a LOT higher than the local reported hashrate. 
Local reported hashrate is around 1.52mh average.  Here's a pic of the poolside rates (under the nst6563.fury2 entry).  Single Fury. 
ZiG
sr. member
Activity: 406
Merit: 250
June 13, 2014, 10:28:47 PM
#65
I did the mod...used a 50k trim pot adjusted down to 7k.

I had to leave but I was running it at 450mhz with a 3% error rate.  I just upped it to 500mhz and so far 0 errors in 10 minutes.  It's averaging 1.52mhs.  I'll let it run at 500 while I'm out and report back.  Temp probe only registered an average 3f increase with my heatsink setup.



Also....if anyone is running Hashra you have to edit the /var/www/index.php to allow those higher speeds

Super...

So...1 Fury/Blizzard/Zeus 1.3MH voltmod & overclocked to 500MHz = 3 Gridseeds voltmod & overclocked to 500M

NOT BAD at all...for $120...only 1 device/connections...vs 3...

Mine is coming...I don't know when... Angry

Cheers,

ZiG

sr. member
Activity: 252
Merit: 254
June 13, 2014, 07:59:56 PM
#64
I did the mod...used a 50k trim pot adjusted down to 7k.

I had to leave but I was running it at 450mhz with a 3% error rate.  I just upped it to 500mhz and so far 0 errors in 10 minutes.  It's averaging 1.52mhs.  I'll let it run at 500 while I'm out and report back.  Temp probe only registered an average 3f increase with my heatsink setup.



Also....if anyone is running Hashra you have to edit the /var/www/index.php to allow those higher speeds
hero member
Activity: 826
Merit: 1000
°^°
June 13, 2014, 06:50:02 PM
#63
if you have a cnc mill,

you could mill a big heatsink down to touch chips only
hero member
Activity: 840
Merit: 1000
June 13, 2014, 05:29:36 PM
#62
Exactly what type of cooling are you referring to?  Something similar to the heatsinks on the chips I posted in my picture?  I've already got those + thermal paste on the back of the circuitboard and that let me clock up to 354 (~1.41mhs). 

I have a small 60mm fan blowing through over the circuitboard and heatsinks.  Thermal probe on the heatsink shows 76f-78f (ambient is anywhere from 69f-72f depending if the AC is on or not). 

If I didn't have plans tonight I'd be doing this mod....

with the small extra fan plus heatsinks, you should be ok.
I cut an old VGA heatsink in 3 parts that each goes over 2 chips with arctic ceramique applied for the chips and under the board.
Plus a spare Gridseed blade fan blowing inside the Zeus heatsink.
Not sexy at the moment, but efficient until I find what I'm looking for as custom cooling..
I"ll measure watts used tomorrow. I do expect something in the 70 watt range
Time to sleep for now.
Pages:
Jump to: