Swedish ASIC miner company kncminer.com

[GenTarkin]

Ive not uploaded any changes yet guys, just documenting my progress =)

[1713560085]

1713560085

[1713560085]

1713560085

[1713560085]

1713560085

[jelin1984]

Ok so try to put at new version and unlock at 350 MHz

Ask

Padrino
At forum

He want to make custom firmware for Titan miner

[GenTarkin]

Well I got a fake 350mhz working HAHAHA!
Its much more complex then simply modifying waas code =(, thats what I attempted and it just gives the appearance of 350mhz =/

So unless someone has clear documentation of modding it to 350mhz ... I aint willing to risk damaging my cubes.

[s1gs3gv]

IMO with the Titan the best strategy is 'If it (mostly) works, leave it alone', especially at current LTC prices.
I know how much fun it can be to tinker with firmware code but the risk of something going wrong and loss of warrantee isn't offset by the marginal increase in profitability running @ 350 instead of 325.

Too bad KNC never open sourced the FPGA code ...

[Tigggger]

So I have several sets of Titan Batch 2 cubes I got in june 2015, and they were brand new sealed in the box. 13 cubes have failed in different ways. Ive owned a few computer repairs shops over 20 years , so I have a ton of experience with thermal issues. So far I have 2 cubes with burned PCI-E connectors that I was able to restore and one cube that has 2 dies that read .8v (idle) that I tried the fix KNCminer kept saying would work, and that was to remove the single screw nearest the exhaust and remove the standoff too, and put a screw through a nut that is underneath the cube. The goal was to tighten the pressure on the heatsink. I had 8 cubes that had this issue and it did not work. All it did was start to bend the board in the middle. Ive applied my own fix after listening to KNCminer for over 45 days, and MY fix worked on the first cube I tried. So KNCminer now claim my warranty is void and refuse to help me. I will eventually repair all of these, and then file my class action suit to the tune of $70,000. Guys if you own ANY kncminer product, the deadline to file a lawsuit for the FULL amount YOU paid for them is August 20th 2015. I will be emailing everyone on this thread and letting them know. Im also optimizing my website for many keywords for google so others can see who KNCminer really are. My website bitcoinlasvegas.net is very high ranking on google as I get about 5000 visitors a week. Im already on page one for the search term "Kncminer Batch 2" , so Ill be posting most of the horror stories there. http://bitcoinlasvegas.net/kncminer-titan-batch-2-up-and-running/

Vegas

Awesome! What was the fix? I'd like to give it a shot!

Yes, please share your fix, I'm running a right bodge of a machine with 6 cubes (was a compromise to get my RMA handled quickly)

6 completely dead dies that never work, and 4 that sometimes decide to work if underclocked and overvolted.

[vegasguy]

Ive got loads of cubes to try this on, Ive only tried it one time for a cube that had .8V (IDLE) on dies #0 and dies #1. I am confused on something. KNCminer tell me that the dies are located under the main processor in the center of the board. Im quite sure its the vrms that are overheating. I see they use ericson vrms. My question is:  Is it the dies under the 8 square boards that go around the processor that has the ericson vrm on top,or are the dies located under the vrms. The 8 that go around the main processor on the board. Let me try my fix a few more times before I call it real fix.

Vegas

[GenTarkin]

Ive got loads of cubes to try this on, Ive only tried it one time for a cube that had .8V (IDLE) on dies #0 and dies #1. I am confused on something. KNCminer tell me that the dies are located under the main processor in the center of the board. Im quite sure its the vrms that are overheating. I see they use ericson vrms. My question is:  Is it the dies under the 8 square boards that go around the processor that has the ericson vrm on top,or are the dies located under the vrms. The 8 that go around the main processor on the board. Let me try my fix a few more times before I call it real fix.

Vegas

Keep in mind I dont have a physical titan so Im just guessing here, but based on my experience w/ everything else in PC land....
The heatsink / fan will be directly attached to the dies. VRM's usually are passively cooled.

Got a picture of the PCB? Ill tell ya.

[jelin1984]

Ask

Member

Padrino


He wait your message

[vegasguy]

I found these pics of the board: I think this is from a neptune, but it looks identical to the titan.

[GenTarkin]

I found these pics of the board: I think this is from a neptune, but it looks identical to the titan.

Really?! wow so the dies are protected then. They are underneath the heatspreader in the center there. Thats where the heatsink mounts directly.
For some reason I thought the dies would be bare like hashfast cards =P, guess not.
Anyways the VRM's are those smallish lifted boards around the ASIC

If you put ur finger on the yellow sticker and on the 2 little square things in front of it(that seem to have a metal center)
They will feel really hot when unit is running. Those are the VRM's for sure, putting tiny heatsinks on them may help cool them and the yellow sticker square heatsink would go well on that too =P
Theres probably other areas that could use little heatsinks. But that would be my initial guess off top of my head =)

[vegasguy]

Question if anyone knows :

Im going to remove the stickers and put thermal grease on top of that large black chip. Is this safe to put conductive grease there? How about the 2 smaller chips? I also have epoxy thermal grease I will use to bond the copper to the vrm chips, but I am hesitant to put any conductive material on top of those chips. It seems like the larger chips with the yellow sticker are ceramic, but Im not 100% sure. Any advice guys?

I found this pic on the net with the large processor cap removed and you can clearly see the dies underneath. Kncminer said 2 cores per die. So that makes sense.

click to enlarge to full screen:





Vegas

[GenTarkin]

Question if anyone knows :

Im going to remove the stickers and put thermal grease on top of that large black chip. Is this safe to put conductive grease there? How about the 2 smaller chips? I also have epoxy thermal grease I will use to bond the copper to the vrm chips, but I am hesitant to put any conductive material on top of those chips. It seems like the larger chips with the yellow sticker are ceramic, but Im not 100% sure. Any advice guys?

I found this pic on the net with the large processor cap removed and you can clearly see the dies underneath. Kncminer said 2 cores per die. So that makes sense.

click to enlarge to full screen:





Vegas

Whenever I put heatsinks on stuff that was questionable conductivity , I just used stickers thermal pads . Always worked but may have just been lucky lol
Ive had heatsinks touching one another that were on seperate VRM components on other devices like gridseed blades and there was never a short =)


Also, on another subject, for those of you who have dies that are no longer working, like permanently damaged but the board is not physically damaged, have any of you tried to reflow the boards?
They may have just stopped working from cold solder joints. I have brought back to life several laptop mobos w/ a good heat gun , Ive also done video cards in the oven w/ some success =)
Id say its worth a shot.

[pinhead666]

I have wondered too why there is no-one who provides re-solder service for dead boards.There are so many S5 and knc boards which propably could be saved with good bga resolder station.I would offer that service if I lived in US.

[vegasguy]

Well thats what Im looking to do, is a permanent fix that also brings the temps WAY down, and then  offer it to the public at an attractive price. There wont be any risk as I have a licensed computer repair shop for 20 years , so the cubes will be sent there.

Vegas

[vegasguy]

I read on one of the threads this:

I have removed the aluminum plate, so now I can see the Ericsson VRM's..... this is bad: they are the BMR4641002 (max 40A) instead of the BMR4641008 (max 50A).  :mad: ...... no OC possible !

The 1002 has a OverCurrentProtection of 48A and looking to my Advanced-page: nearly all the VRMs are already around 46-47A (and this is KnC's default setting).
They should have used the 1008 instead, which has 50A and an OCP of 62A.

So my question is how difficult would it be to replace the VRM on the board with BMR4641008 (max 50A)? Is this something that can be done with a soldering iron , or does it require a special machine like a reball machine? Is it JUST the top chip or the entire board underneath that VRM chip too?

Vegas
 

[GenTarkin]

I read on one of the threads this:

I have removed the aluminum plate, so now I can see the Ericsson VRM's..... this is bad: they are the BMR4641002 (max 40A) instead of the BMR4641008 (max 50A).  :mad: ...... no OC possible !

The 1002 has a OverCurrentProtection of 48A and looking to my Advanced-page: nearly all the VRMs are already around 46-47A (and this is KnC's default setting).
They should have used the 1008 instead, which has 50A and an OCP of 62A.

So my question is how difficult would it be to replace the VRM on the board with BMR4641008 (max 50A)? Is this something that can be done with a soldering iron , or does it require a special machine like a reball machine? Is it JUST the top chip or the entire board underneath that VRM chip too?

Vegas
 

Holy fuck KNC is cheap as all fucking shit! Yeah, you wouldnt want a 350mhz setting even if you had it. I wonder if this is why random dies drop... Constantly running over spec, I bet these VRM's could definitely use little heatsinks on them.
The question is what is the limit of the rest of the components in the VRM? sure you could swap out the main one but.... w/o knowledge or ratings of the other little ones... they could easily blow.

[Searing]

All these physical dead dies you guys mentions, I cant help but wonder if really bad shipping stresses(throwing boxes around etc...) caused the massive heatsinks to put insane pressure at times on the dies causing them physical damage.

My understanding is some of it was the first units out that first month.....too much thermal paste and other issues of shoddy workmanship.....was on the old knc thread now deleted of such problems ....likely true in my case as well

I should really balls up and look in my 1 dead die on 2 cubes each but again the titan(s) are getting 650.1 mh with 2 dead dies and 2 dies set down on a cube
to make it 312.5 from 325 ...so again as shoddy as I suspect these titans were made..I'm trying to refrain from even making eye contact when I go into the basement
in that they may take offense and brick themselves ..a common titan hobby imho..but again the titan(s) were only supposed to do 300mh and so most folk on here
say it is a miracle I'm getting what I'm getting so have been hands off on such so far

[jelin1984]

Any news from firmware with new bfg miner version???..

[vegasguy]

Yes. We are donating to the dev. I paid my share. I think he is asking .25 BTC.

Vegas

[GenTarkin]

Ill upload a new image today guys. I would really appreciate any tips. Worked hard on this.

Update on 350mhz: its not really worth it at this point. Ive looked at the way they clocked up neptune and I can say the eeprom on our devices seems completely different than neptune / jupiter. So using the same registers to set clock speed is a huge unknown to me at this point and I dont wanna risk killing any cubes. Id be willing to experiment on a cube if someone was willing to ship, also Padrino said he may be able to as well, but there are no gaurunteed results.
Also, as previously pointed out the VRM's are at their rated max. Running them beyond their rated max but under their OCP point is still risky, putting more stress on them is not a good idea. Especially 24/7, they will probably just pop sooner than later in such conditions, unless they were cooled really well.

What my firmware image will includin is all fixes Ive implimented to date for "soft" dropping dies + more stats in webgui + updated bfgminer(updated to latest).
I will continue to experiment w/ die drop handling. I read another way that may be more effecient. I will update if thats possible.

Once again donation addy is:
1Px71mWNQNKW19xuARqrmnbcem1dXqJ3At

The .5 bounty still has yet to fully be paid on, .3365 has been delivered thus far =)

Ill be uploading new image soon to github. For those of you who have my btsync share, its already updated in that, just wait for it to redownload.