Topic: Swedish ASIC miner company kncminer.com - page 61. (Read 3049521 times)

I could but it would be largely pointless. Being a tad above 40A is not as dangerous as running these things at temperatures for which they arent rated to handle 40A. So, I think temps are much more important then going over the 40A by a bit. Hence, I would recommend the max temp of my DCDC threshold code to be set to 80 or 85C.

Gentarkin, great job on the software btw, I am one of your donators you mention current above 40A being an issue can you make your software keep the settings in a "safe" range etc ?

At those temps Im surprised ur DCDC's have lasted as long as they have. Also, those arent DIE temps, there are no die temp measurements on these Titans, those are just internal temps of the DCDC's.
There is absolutely no way to detect internal temperature of the dies on Titan. Unless they have an undocumented thermistor somewhere in there LOL! ~ I doubt it.

Also, I have no idea what these units were originally supposed to be rated, but perhaps they were all supposed to be 5 cube units all along and they decided to take the cheap way out by shipping 4 cube units and telling everyone to OC. And yes u are right about early versions of firmware 300mhz was the max, the allowance of 325mhz is shown on github as a change to the spimux-titan.rbf file. Which is why I know the key to unlocking 350mhz is in that file =)

Heh ..not even gonna look ...I have one cube with 2 dies runs hot as hell always has 325 mh all out.(since march 2015 when I got my 2nd unit)
.....i always expect it to burn out ....looked yesterday 90c plus as per usual
that was when I got the used unit...so all the cube goes full out.....at max 325 setting any other cube i throttle it back (or titan says no way) looked today
100 to 101c on the 1 die chugging merrily along...I am befuddled as always ..so that is what 8 months of full out wtf is going on ..then again it has been
11 months wtf is going on with the dies I have set to OFF its only fair I be confused the other way around

so again confused on what witchery is involved in this tried to tone it down last march the die would  not come back and previous attempts..but ..put it back at 325 setting etc and viola
bat out of hell since (highest producing die too ..heh figures)

so i only have 2 weeks left on my warranty so it will blow up 2 days after

but man i'm clueless as hell on how just plain 'weird' these units act from my own experiences and the folk's on here saying stuff

So I'm trying to go with the 'idea' (likely) as best I can recall in them 'heady' months when the Titan was expected to arrive on time and such

It was designed for 250mh .....the dc's etc components (power jacks) etc etc with the 'appropriate' over engineering for that 250mh unit

They figured out before shipping (w/o telling us imho) that the units were probably gonna ship late....so upped the hash from 250 to 300mh to
compensate for the lateness i suspect they saw coming (although to be fair I doubt they thought they'd be as late as they were) Again still within
the likely 'over engineering' on a 250mh design

They shipped......1st batch had bad dies..the firmware sucked with lots of errors etc...Initially I was told to mine at 300mh anything above that was
my own risk...the firmware ...buggy as hell for 2 months plus was not helping hash .even with LTC ...the hardware had the bad die issues..thus suddenly
when I had 2 bad dies on 2 diff cubes ..I was told to overclock it to 325 speed on adv page (ie ok by knc techs) so I did and it worked ok (with 2 off dies)

Side note: By the way I seem to remember the 1st firmware you could not even SET the speed above 300mh or am I incorrect? Anyone?

But then they changed the rules and said because my orig titan got 306.1 mh with 2 dead dies anything over 300mh they would not take back equip and
nor any cube with only 1 die off.....amazing how rules got changed. Again this was against their previous advice of 300mh was max.

So essentially to cover their ass on shoddy firmware and equip and lateness of shipping you were 'required' to overclock.... the F*ck out of your Titan
to get around their failed design/promises and ship date..thus burning a lot of stuff out early as you can see from the 'wreckage' of this thread over the last
year or so.....all risk on us!

to get to my POINT:

 IT IS IMHO A FRIGGING MIRACLE THAT ANY OF THESE TITANS THE WAY THEY WERE DESIGNED 'MANAGED' TO GET ABOVE 300MH FOR A YEAR! I JUST NEED
 TO DEAL WITH THAT 'REALITY' (ie KNC lied on what the product should do vs what it could do)

So trying to go with the flow...they are frigging overclocked to the max at 325mh adv page settings..so it is easier to limit my expections to reality then to try to figure out how to tweak
these units above 300mh when they already are in 'flaky mode' at probably 300mh ONLY ..from what they imho were designed for!

Got to admire that KNC 'evil genius' chutzpah thou! We have all the issues/risks/etc and they limit their problems with these units by a lot!

By the By I have 2 weeks left on my warranty ...anyone want to use me to swap a cube out to knc for their 200 bucks I'm trustworthy enough to do so
(then again with 2 weeks left they may just never send it back claiming it did not arrive even thou on the shelf..ie must remember the 'evil' we are talking about)

But we should have a mechanism on here to move cubes back thru trusted parties for folk who's warranty has already run out....I mean its not like they
did not play such games with us..the old razzle dazzle misdirection tricks... (then again send cube 200 buck fee....in say a month the knc tech is fired (or quits) due to his impending loss of a job...as the 12 month fix them requirement goes away)

hell probably should just send them to Vegas would be safer I guess ..just frosts my cookies for those with cubes as door stops as we speak because
it is already past 12 months and they are stuck......whatever just wool gathering on above

Cheers man, much appreciated and you're most welcome! Indeed, I've no doubt that people have expanded on those ideas with even more improvements, which is what it's all about. I'd like to try some more things, but I'm still working my way outta the hole this Neptune put me in, lol. Speaking of bitter (cold), way up north from you, I'm already switching over to using my miners for heat, brrrr! Seems like we completely skipped Fall up here.

hey teslaman, good to see you -- you were putting out a lot of good info on modding these things back when there was very little if any other info to go on. Copper heatsinks, 2 part thermal epoxy, noctura fans, cutting the Eplate and probably a bunch of other stuff I forgot, were all originally his ideas. After doing and redoing a bunch of these thing I like to think I learned a few of my own tricks too, but he got the ball rolling. thx dude 

OK quick question.  I own one of these and one of my cubes is being a little finicky.  Any advise would be appreciated.  When I run the miner only one die seems to be working one cube 3, but as it mines, it switches dies which work???  So one minute die 1 will work and the other 3 wont and then after watching it for a couple of minutes die one goes off and die 4 works, etc, etc,...  Any clues, or anyone that has had this happen.  I have 3 titans and this is the only one that does that.  thanks in advance for any help.

Thanks for the mention TXSteve.

The one thing in that video that I wouldn't recommend doing more than once, is bending those ends down on the cube case. It got pretty flimsy the second time I did that and it completely broke off the third time! KnC used some really cheap metal, which is even more aggravating considering what we paid for them, heh.

As TXSteve mentioned, taking the stickers off the VRMs is a good idea, especially if you use the thermal tape, like they did in the video.

GenTarkin, here's my before and after temps with the heatsinks and thermal epoxy/adhesive. https://i.imgur.com/2aE6f7S.jpg This is on a Neptune cube, but they are laid out the same. That hot VRM must be messed up or something, because it always runs really hot compared to all the rest of my cubes.

These also compare the different fans, KnC's supplied fans and the Noctua upgrade:
Copper Heatsinks with thermal tape: https://i.imgur.com/fcX9cp7.jpg
VS
Copper Heatsinks with thermal epoxy: https://i.imgur.com/yNIsXby.jpg
As you can see, the epoxy worked quite a bit better than the 3M thermal tape that came on the heatsinks.

I didn't do a test going from the stock thick thermal pads, to the heatsinks with thermal tape. I was doing this upgrade on the Bonus Neptune, which came with NO thermal pads! By the time I modified that one regular cube (first link), I had already switched to using the epoxy.

Here's my full gallery if you want to browse through the rest: http://imgur.com/a/zQDSa

Yeah, I would say on dies which drive the DCDC's above 41A , I would first try to lower the voltage a notch, if thats not stable then put the die at 300mhz and then tweak the voltage down as desired.
Also, like I mentioned above, I would keep the DCDC temp threshold at 85C(if ur trying to achieve longest lifespan that should be the max)

So, safe bet would be to keep DCDC amperage at 40 or below if I wanted the cubes to last longer? I see at 325 some of them goes as high as 44.

Um, Im not trying to reverse engineer that file. I was saying if someone wanted 350mhz, the key to unlocking it is in that file. Im not willing to do it and either brick my fpga or burn up my titan LOL

I bet KNC kept it closed source cuz they didnt wanna have to service a bunch of peoples cubes who abused them by running them at 350mhz and frying everything.

I can test it of sent me firmware

BTW, if anyone feels like tackling it.... LOL

350mhz is doable, if ur freaking crazy and want to melt ur titan asap, but it is doable.
The key lies in reverse engineering the spimux-titan.rbf binary file. Its closed source, maybe, if ur lucky u can try to beg KNC to release the source =)

BE WARNED!, ur DCDC's will probably catch fire and ur PSU will melt eventually and power cables turn into burnt spaghetti noodles but what the hay!!! YOLO!! =P
If anyone loves fireworks, 350mhz+ is the way to go!

Ultimately its up to the user and their preferences of how long they want the machine to potentially last.
Its hard to answer ur question definitively due to so many environment variables & operating conditions between cubes.
What everyone needs is a laser temp reader to take temperature checks of the inductor on the DCDC modules while its hashing. If its true that the inductor is generally within 10C of the webgui stat then a SAFE MAX temp @ 40A is... between 70 & 75C, so webgui stat should no higher than 80-85C.

Whats interesting is with more airflow these DCDC's are rated for more current safely at any given temp, even if the temp is the same vs less/no airflow.

For example, temp of inductor 80C @ no airflow is safe up to 34A.  80C @ 2-3m/s of airflow is safe up to 40A
At 100C inductor temp: 20A safe for no airflow, 30A safe for 3m/s of airflow

Another facet of max life from DCDC's is keeping them at their rated 40A max or lower. I know most of mine in the titan @ 325mhz are at 40-41A ... which I imagine aint bad, but I think around 43A ur starting to really push it. The actual Current limiting function of the DCDC's kicks in at 46A

This is the life expectancy of the DCDC's given everything is within spec:
MTBF at 90% confidence level = 11.52 Mh

What would be the safest Maximum DCDC temp setting on advance page of your mod regardless of the ambient temp? I currently set it at 90 and all dies run at 325, should I lower it?

BTW, heres an interesting tidbit for yall, I think why many titans "dies end up dieing permanently" are infact because the DCDC modules burn out. After looking over the tech sheet, its easy to see why. If the ambient temp (surface temp of the inductor) reaches the 80-90's these are only rated for 25-30A to maintain their expected lifespan.
The fact some folks run them at these temps 24/7 and draw over 40A through them, its easy to see why the DCDC's would die ....

When vegas measured the temp of the inductor from the side, it measured about 55C , the webgui showed about just over 60C ... so there is a really small delta between the ambient temp & internal junction temp(webgui reading) of these DCDC's

Meaning, if ur webgui shows 90C area, the inductor is probably not far behind, most likely running in the low 80's ... which means nowhere near 40A should be pulled from the DCDC at that point.

I was able to run all 4 of my cubes at 325 MHz and gain a little over 60 MH by putting those small heatsinks on DC-DC modules and reapplying thermal compound to the big heat sink of 3 cubes. The 4th cube was able to run at 325 w/o any modification, it was running stable at 300 MHz and I didn't try 325 before since other cubes would not run at 300 for over 20 mins. Thanks for all the tips and link to YouTube mod video. If you are planning to do this, as TXSteve mentioned before don't forget to peel off the sticker on DC-DC, I forgot to take sticker off of 1 cube and it is running warmer than the other 2.

Of course I'd release it, eventually

But hey, I put a lot of work into these titans since last spring, mostly building on teslamans mods in this thread. The little copper heatsinks on the dc-dc and cutting the eplate that we're talking about today, started with teslaman many months ago, it's not new info. I didn't pay for his generous advice so I am not really looking for compensation. I had a head start because I knew it would be a real challenge getting them thru this south texas heat, and general tarkins mod was a huge help in automating things, like monitoring temps and die failures.

Today the high was a refreshing 82f, so I should be good for another 9 mos or so, this is when it's nice down here

Only if you actually release it!  

I dont have a titan, not sure if vegasguy is up to it, will see haha!