I heard it is solder, can you double-check? actually, removing the thermal is a lot harder than removing solder, and solder transmits heat a lot better than thermal which is probably a must for all these new miners.
Topic: Will the real Vnish please step up. (Read 360 times)
the whole issue was the bad quality unleaded tin used by Bitmain between the chips and the heatsinks, and replacing the whole of them does the trick.
I don't think that's true, the main issue "as explained above" is the solder between the chip and the PCB, wndsnb who fixed a truckload of hash boards can chime in and confirm/deny that.
another proof I have to support this is the hundreds of kernel logs all over the internet, if a heatsink is loose or completely unattached, the chip count will still pass and show full asic count, the miner will even start mining for a second or two and then give you that temp protection error and goes into sleep/mode for a while and then restarts.
In the event of a loose chip due to bad soldering, the chip count starts to act, it will show 0 chips or any other random number depending on the location of that chip.
If you search the internet or this forum you will see a lot more cases describing the latter problem, I'd say at least 8 out of 10 at least based on my own experience where many of 17 series gears died without any sign of a chip falling or losing contact to the chip, of course, that doesn't mean I disagree with the fact that the even the soldering between the chip and heatsink is terrible, I flew a handful of heatsinks on my S17pro with an air blower.
The modification might reduce the frequency of those issues, but I highly doubt it solves the main cause, my guess is it's more of a marketing scheme to convince people that "the problem has been solved" and now they can safely buy those mining gears that have over 50% failure rate.
It's a thermal. Would be a nightmare if they soldered it would require quiet a setup to remove it. Honestly wouldn't be surprised if they did it in the future.
I heard it is solder, can you double-check? actually, removing the thermal is a lot harder than removing solder, and solder transmits heat a lot better than thermal which is probably a must for all these new miners.
Of course it does, the whole issue was the bad quality unleaded tin used by Bitmain between the chips and the heatsinks, and replacing the whole of them does the trick. I cannot say if everyone doing the mod is using good tin (with lead), but in principle, resoldering all heatsinks would do the fix. In addition, this mod uses screws against the PCB. So far from the reports i have gotten, they work as good as the S19.
The factory flaw with the S17 family is the unleaded tin that develops something that makes it expand over time and loses adhesion causing chip overheat and silent death, unless its one of the only two chips connected to temp which is hardly ever the case.
Its no coincidence the S19 screws the plates against the pcb as well. Do you know if they still use soldering tin or are they back to thermal glue from chip to plate?
The factory flaw with the S17 family is the unleaded tin that develops something that makes it expand over time and loses adhesion causing chip overheat and silent death, unless its one of the only two chips connected to temp which is hardly ever the case.
Its no coincidence the S19 screws the plates against the pcb as well. Do you know if they still use soldering tin or are they back to thermal glue from chip to plate?
It's a thermal. Would be a nightmare if they soldered it would require quiet a setup to remove it. Honestly wouldn't be surprised if they did it in the future.
Also Vnish doesn't have a website. There are websites that offer the firmware and have a fee. Partner branded copies for different demographics and fess associated with them.
Of course it does, the whole issue was the bad quality unleaded tin used by Bitmain between the chips and the heatsinks, and replacing the whole of them does the trick. I cannot say if everyone doing the mod is using good tin (with lead), but in principle, resoldering all heatsinks would do the fix. In addition, this mod uses screws against the PCB. So far from the reports i have gotten, they work as good as the S19.
The factory flaw with the S17 family is the unleaded tin that develops something that makes it expand over time and loses adhesion causing chip overheat and silent death, unless its one of the only two chips connected to temp which is hardly ever the case.
Its no coincidence the S19 screws the plates against the pcb as well. Do you know if they still use soldering tin or are they back to thermal glue from chip to plate?
The factory flaw with the S17 family is the unleaded tin that develops something that makes it expand over time and loses adhesion causing chip overheat and silent death, unless its one of the only two chips connected to temp which is hardly ever the case.
Its no coincidence the S19 screws the plates against the pcb as well. Do you know if they still use soldering tin or are they back to thermal glue from chip to plate?
isnt Vnish asic.to? Been running asic.to on my s17s for over a year and love it.
Asic.to is a distributor for Vnish, just like AwesomeMiner, they all are legit and they all use the same firmware, you just get extra added benefits using them like more support from Asic.to or a great monitoring tool for AwesomeMiner.
The bad soldering issue to the heatsinks is no joke. There is now a popular mod going on where they replace those single heatsinks with large plates screwed to the pcb. This was a design/manufacturing mistake, but they just "quietly" killed this line.
The mod does not really solve the issue by any means, in most cases the issue is below the chip and not above it (bad pcb to chip solder, not chip to heatsink solder), however, using one block of heatsink on top of the chips makes "fixing" those miners a lot easier since it's pretty hard to troubleshoot the miner with all those heatsinks on, I see some people take out a dozen heatsinks just to find that one bad chip, so without a doubt, unscrewing 4 or 6 screws to get a perfect vision on all the chips is a smart idea nonetheless.
How about running original vnish on your device and also get free remote monitoring and management software?
https://mineitor.com/firmwares
https://mineitor.com/blog/tutorials-3/post/how-to-install-vnish-mineitor-firmware-59
https://mineitor.com/firmwares
https://mineitor.com/blog/tutorials-3/post/how-to-install-vnish-mineitor-firmware-59
Right now in northern SK I can climate control down to whatever I want easily and about 15 degrees C seems to work well and have been using the Vnish 60-62TH profile on about 20 machines and they have been rock solid.
I have brains a go and got greedy the first go and landed at about 65 yet temps seemed really reasonable with the low ambient so I let it eat. Made it about a week and one card went down, by the end of the next week another was down and limping on the last.
Figured well this is garbage gear I’ll give it another go. Lowered it to very similar hash rate and draw as my Vnish settings.
Ran rock solid for 2 weeks straight. I thought ok this might be the ticket. I wanted to like it! Then boom. Sum bitch lost a fan and was merrily tuning away for I’m not just sure how long loading up to 80 degrees then unloading it to a reasonable temp and then back up.
Still haven’t looked at the machine or thought about replicating why it didn’t shut down.
Nonetheless 3 more hash boards down.
I have brains a go and got greedy the first go and landed at about 65 yet temps seemed really reasonable with the low ambient so I let it eat. Made it about a week and one card went down, by the end of the next week another was down and limping on the last.
Figured well this is garbage gear I’ll give it another go. Lowered it to very similar hash rate and draw as my Vnish settings.
Ran rock solid for 2 weeks straight. I thought ok this might be the ticket. I wanted to like it! Then boom. Sum bitch lost a fan and was merrily tuning away for I’m not just sure how long loading up to 80 degrees then unloading it to a reasonable temp and then back up.
Still haven’t looked at the machine or thought about replicating why it didn’t shut down.
Nonetheless 3 more hash boards down.
With the x17 family, plugging a miner runs risk of failure, period. Only people doing immersion seem to be able mitigate this.
The bad soldering issue to the heatsinks is no joke. There is now a popular mod going on where they replace those single heatsinks with large plates screwed to the pcb. This was a design/manufacturing mistake, but they just "quietly" killed this line.
I remember someone in this forum recommending to use them horizontally with the larger heatsinks pointing up, in an attempt to use gravity to mitigate this somewhat. If you didn't know the reason these miners were so cheap compared to the others, this is why.
Used miners most likely have already damaged chips on them, it is a "silent killer", see, the manufacturer didn't bother to connect all the chip temp sensors, just two of them, so when any of the other chips start losing adhesion they start getting overheated without any warnings (maybe errors and chip retuning) until it finally dies, the fw may or may not compensate for this, being serial and "domain" based with the voltage regulated at a single point in the psu (not each hashboard or chip) makes this even more difficult.
But fear not, the manufacturer decided that putting chip sensors was a waste so the x19 no longer has any, yay? Just trust your now 4 pcb temp sensors there
Rather than more, think of less, in case you still want to gamble it and not do the fix (or hope that somehow your batch was better). A lower hashrate tends to improve efficiency anyway, and you want some longevity, tho if adhesion losing develops in one of your chips (doesn't necessarily needs for the heatsink to fall off right away, it can slowly develop and you would need something like an IR camera to catch it) you won't notice until its too late, and the repair will most likely require replacing that chip.
TL;DR: If you haven't fixed the bad soldering heatsink issue, do it before its too late.
Risk it? Horizontal placement with large heatsinks up, lower the power usage to improve efficiency and longevity.
The bad soldering issue to the heatsinks is no joke. There is now a popular mod going on where they replace those single heatsinks with large plates screwed to the pcb. This was a design/manufacturing mistake, but they just "quietly" killed this line.
I remember someone in this forum recommending to use them horizontally with the larger heatsinks pointing up, in an attempt to use gravity to mitigate this somewhat. If you didn't know the reason these miners were so cheap compared to the others, this is why.
Used miners most likely have already damaged chips on them, it is a "silent killer", see, the manufacturer didn't bother to connect all the chip temp sensors, just two of them, so when any of the other chips start losing adhesion they start getting overheated without any warnings (maybe errors and chip retuning) until it finally dies, the fw may or may not compensate for this, being serial and "domain" based with the voltage regulated at a single point in the psu (not each hashboard or chip) makes this even more difficult.
But fear not, the manufacturer decided that putting chip sensors was a waste so the x19 no longer has any, yay? Just trust your now 4 pcb temp sensors there
Rather than more, think of less, in case you still want to gamble it and not do the fix (or hope that somehow your batch was better). A lower hashrate tends to improve efficiency anyway, and you want some longevity, tho if adhesion losing develops in one of your chips (doesn't necessarily needs for the heatsink to fall off right away, it can slowly develop and you would need something like an IR camera to catch it) you won't notice until its too late, and the repair will most likely require replacing that chip.
TL;DR: If you haven't fixed the bad soldering heatsink issue, do it before its too late.
Risk it? Horizontal placement with large heatsinks up, lower the power usage to improve efficiency and longevity.
isnt Vnish asic.to? Been running asic.to on my s17s for over a year and love it.
How much extra hashpower are you actually getting out of the firmware upgrade and are you still using stock cooling?
I did early tests with HiveOS and remember getting 63-65th but that was dead of winter in a large warehouse. Due to the constants of where I'm mining now I don't OC, I use the custom firmware to fit my wattage and cooling limitations. I'm mining in a warm climate (SW desert) so heat is number 1 to deal with. With asic.to I can have thermal limits for swapping profiles. Stock firmware can't do 1/3 of the things the other 3rd party firmware can do. Current settings for the S17, 50.8th at 525Mhz for 1971 watts.
depending on the environment, for an S17/S17Pro, you can easily obtain +10th without risk for the miner with the stock cooling ...
All custom firmware will allow it, you just have to find the one that suits you best, functionality, dev costs, they range from 3% to 0.9% depending on the firmware and models)
All custom firmware will allow it, you just have to find the one that suits you best, functionality, dev costs, they range from 3% to 0.9% depending on the firmware and models)
isnt Vnish asic.to? Been running asic.to on my s17s for over a year and love it.
How much extra hashpower are you actually getting out of the firmware upgrade and are you still using stock cooling?
isnt Vnish asic.to? Been running asic.to on my s17s for over a year and love it.
So am I understanding it right Vnish.net is the real deal and Vnish-firmware.com is the knock off?
We put the Vnish-firmware on a dozen machines as a test and they are all on top of the leader board on my pool account.
But I might as well support the actual developer. Well my wife is a Russian national so biased a bit I suppose.
We put the Vnish-firmware on a dozen machines as a test and they are all on top of the leader board on my pool account.
But I might as well support the actual developer. Well my wife is a Russian national so biased a bit I suppose.
Jump to: