Topic: [GUIDE] BitFury Miner Support/Tuning - page 22. (Read 148049 times)

so i spent some time rearranging my rig setup and got it up and running at 76.5GH (2 cards) before realising that i am pulling air through the tunnel, rather than pushing/blowing it. *facepalm*

Is this going to limit my cooling my any significant amount? It seems to handle very well at the moment, but im a little annoyed with myself for making the wrong assumption on airflow direction. Im even considering grabbing a second fan (230mm!) to double the airflow - would this be beneficial or simply cause turbulence?

Punin tested all boards in complete rigs for some time (i dont know how long). The notes at the H-Boards are a hint in which combination they are tested. You can use this notes as a suggestion to build up you rig. But its no warrany for optimal performance.

Ciao

Enrico

The biggest problem, it seems to me, is that these h-boards are produced with substandard quality, are noisy and picky, caps burned out so they were removed instead of being replaced with working ones, some chips cannot be seen if you don't disable one of them, errors spring out of nowhere in a working system and then disappear if you restart your system, h-boards connectors on m-board are too flaky.

I mean, august kits, ok, they were in a hurry, but on october kits, I'm disappointed by the fact that they still shipped EOL boards and that even new boards show all the problems of the old ones, so it means there was no improvement in hardware and software.

When you consider the 18K EUR price tag for the august kit and 9K EUR for the october one, this level of problems is barely acceptable.

spiccioli

Same here,

my october kit is made up of 12 new boards and four old end of line boards, each EOL board is the last of each bus, so A4,B4,C4,D4.

Everything was working at around 460 GH/s out of the box, but yesterday I decided to do a little pencil mod to the EOLs boards... hell broke loose!

I got tons of miso errors both on new and EOL boards, in the end I had to remove board B4, which, for unknown reasons was causing miso errors on all the cards in bus B and, from time to time, even to the other buses.

I'm using a 1000W single rail PSU, so it should not be a power issue.

I'd also like to know what is the max amperage that the  m-board can deliver but overall these kits remain extremely picky and instable systems which can stop working even for very small changes.

Here is my system now, as you can see I still have miso errors here and there, but they don't seem to affect hashing too much. Boards 3,B and F are EOL boards with caps, slightly pencil-modded to 1.20-1.22 on RF02.



spiccioli

August production cards ... with caps ...
October production cards ... without caps ....

but ...

august delivery ... with caps ....
october delivery .... without & with caps .... I got 1 out of 14 that had caps

a lot of cards have a signature like A1 ... or B4 ... B2 ... A3  ... it looks like they where
already used in a miner slot before ... so probably october delivery ... is build up out
of miner breakups ...

I'm unsure ... if that is the truth ... it's just a assumption ...

Just to be clear, are there actually some october H-cards that still have the caps, or is everyone refering to august cards when they're talking about cards with caps?

I separated my cards with caps and my cards without caps.  They've been hashing for quite some time now without issues.

I also had a couple occasions where hashrate plummeted and one card would disappear, after mixing 1 october and 2 august H-boards. That always required a total poweroff to fix. After getting 2 more october H-boards, this hasn't happened anymore and things have been stable for one week. Maybe it's because I changed the slot positions a little after getting those last H-boards. Or just pure luck, and it's going to crash today

I played swap-cards-around around until I finally pinned things down to one specific capless card that will not work in conjunction with my older cards.  

After taking that card out of the lineup things started working well with the mix.  If I put that card in any position in my setup things start going bad shortly thereafter, cards go missing, hashrate drops to zero, all sorts of peculiar things start happening...

Cards 8 and 9 are the original two cards with caps (and a pencil mod.)  The rest are stock capless cards, just like they came out of the box.  The older cards seem to work best as the last cards in the bunch.

Bank 1
1: 28.962GH/s
2: 29.564GH/s
3: 32.04GH/s
4: 31.496GH/s

Bank 2
5: 30.938GH/s
6: 28.647GH/s
7: 33.801GH/s
8: 35.09GH/s

Bank 3
9: 36.178GH/s
10: 0GH/s
11: 0GH/s
12: 0GH/s

Add me to the mix on those seeing incompatibility mixing caps and capless. I've done no modding.

I just added three new cards to my original. With the original in the first slot, it slowly died and dropped to 0, and all the cards seemed to get pretty hot. I moved it to the last slot, and everything seems to be running good (about 3 hours stable now). Although the original card is slower, it was always around that speed, so I don't think it is being negatively affected by the capless cards.

Nope (HDMI port facing upwards), it was running fine for a few minutes, and then it stopped working and I realized the component D17 was very hot.

the v2 m-board has the RPi flipped upwards rather than downwards like the v1. did you rotate it 180 when changing boards?

Hey my Raspberry PI just burnt out.
I upgraded to a new v2.3 M-board and it burnt my raspberry Pi. Component D17, a black diode near the micro-usb port had white smoke coming out of it.
I desoldered the component, and the PI powers up, but looks like the IO pins are shot as no Miners are detected anymore. Plugged the PI into a v1 board with same results.

   Did guys do RMA for Hboard drop to 0 hashrate and need to press stop then start?   I got my Hboards yesterday and 4 of them ran together hash good then drop to 0 hash in about 10 minutes and hboards get very hot.   My starter kit hash alone fine and no problem at all.


No one know why new Hboard is very very hot?

ps: Eu post also got problem with new Hboards too.

I have 1 of each style on a V2 m-board. both are overclocked into the 39-40GH range and in a new cooling arrangment so its hard to factor out some variables, but they seem to run well together, though every 4-6 hours one board will drop to 10-50% speeds until the miner start command is used. (crontab does it every 2hrs). The issue appears to be related to the overclock though. 40-41GH was stable for me with only a single card, but now im challenged to get it past 38.5GH due to the re-arranged cooling and extra card

Well, that looks good.  I've got 7 of the new capless cards hashing well, but have problems when I try to mix-and-match in either of my two older cards...




BTW - the hashrate improves slightly on the remaining cards as I remove some.  Maybe 4 is optimum

I just recieved my final 3 h-boards from Mbp today and they are capless. I have the original V1 M-board and original software, no changes from day 1 other than I lost 1 sd card and a pi. My observations so far on this 2 hour run are this.
My 1 original card with caps was pencil modded to 29-30 GH/s
I installed the new cards behind that one and they are all holding fairly steady.
I did not pencil mod the capless cards yet these are as received.



Card 1 is my pencil mod which never broke 30 before now. It hits 32 now.
The rest are holding steady at 30 to 31 with the occasional 29.
Again, These have only been online a couple of hours. I will update If they start doing stupid stuff.

*Very* good observation.  I'm dealing with something similar here, and I was in such a hurry that I didn't notice the missing caps on these boards.  My unit with mixed boards will hash fine for hours, and then out of nowhere, performance will tank and every chip kicks out miso errors.  Most of the time when this happens, an entire board just "disappears" from the stats (even though the boards further down the line are still recognized).

Has anyone noticed that the version 1.2 h-boards with and without capacitors don't mix well?  

I have two of the original cards with caps that run very happily at 32 Gh/s (with a little pencil-modding) until I add a third card without caps - then they both drop to sub-20 and the third card runs at 28.  Add another new card and all 4 drop in hashrate to sub-20.   Remove the third and fourth card, and the first two return to hashing at 32 Gh/s with no problem.  

I also seem to be able to run 8 non-cap boards by themselves with no trouble - they have a different resistor installed and they each come up and hash at 28-32 Gh/s. However, if I add one of the original boards in place of one of those eight, or in addition to those eight - they all seem to go to crap, regardless of the position the older card is installed in.  Hashrate drops from 220 Gh/s to zero over the course of an hour or so.   Lots of heat, no hashing.

Remove the board with bypass caps from the group without, and the newer ones hash fine on their own.   Remove the new cards, and the old two cards return to hashing just fine on their own.  Mix the two types, and strange, and invariably bad, things happen to hashrates on all of them....

There is some odd incompatibility here.  Did anyone ever figure out WHY the bypass caps were getting too hot and blowing up?

Yes, but just in case booth PSU work good. If one of them had a problem, e.g. bigger voltage variance, the m-board need to transport all the difference current. This is a bit risky especial because you dont get a fast replacement for the m-board.

Ciao

Enrico