Topic: How reliable is the AntMiner S19 Pro? (Read 608 times)

Thank you for this information, I suspected it because I noticed that some failures occur after a restart following a long period of shutdown. Now as soon as I stop the miners, I try to store them in a dry place.

I also noticed that very often a miner crashes due to internet instability. I think that internet network instability must reveal bugs in the miner's software. Since the software also manages cooling, it can turn off the fans while the hashboard is still working... I think I've had cases like that on S9s.

I had an S17 but it died pretty quickly.

The S19s were a little better. out of 4 I still have 3 operational.   

In the case of the S19 Hydro, it disgusted me so much that I haven't looked at it yet, but it fails during operation, just after a network outage that lasted a few minutes. it doesn't come from humidity. I will keep you informed.

I've serviced and repaired an number of S17's over the past year or so, I have to say that their soldering quality leaves a whole lot to be desired.

If the S19 and later products are built to the same standards, its hardly a surprise that they fail.

Their boards are riddled with soldering issues, I guess they're built to the lowest cost possible.

As for regulator fails, I've only seen a couple and both were soldering issues again, most of the time its either transport damage and broken joints or ASIC's damaged due to the poor heatsink design on the 17's.

I would have expected the Hydro's to have been better, simply due to the arrangement of the heatsink assembly.

Its a shame really, their ASIC design is good with the exception of the IC package choice, they are subject to moisture ingress. I have to bake replacement parts for at least 48 hours before use to be sure they survive the soldering process.

If you say you're having issues in humid areas, this doesn't surprise me.

Hello,
I noticed it too, I've been mining in a sort of fairly humid cellar since 2015 and it's a disaster.

I have used at least one model from each family S1,S3,S5,S7,S9,S11,S17 and S19 and recently s19Hydro). If we ignore the S1 and S3 which were rather robust, the reliability of Bitmain products is generally quite catastrophic even if we can sometimes come across a less bad series.

There are, among other things, oxidation problems, but there are also unacceptable problems for a company that has been manufacturing miners for 10 years. I understand that the calculation chips are fragile, because they are pushed to the limit, but that auxiliary DC/DC converters fail on almost all models since the S7, that's not normal.

Until now I have always used Bitmain because the commercial service is good for me who is in Europe (nice delivery, payment in BTC...), but the reliability is not going up.

Personally, I work in electronics and in my job we have never had problems of this type. There are times when I wonder how they manage to have such low quality. The hardware seems designed by amateurs, the control of the fans which spend their time speeding up and slowing down, causing very harmful thermal variations to the welds, I also wonder about the quality of the software...
 
I recently bought an S19 hydro model to test, hoping that it would be better because there is no fan so no vibrations, no dust..., I already have a hashboard broken after two months. I'm fed up, I'm wasting my time fixing this shit. I have my free electricity (hydraulic) but I barely earn any money anymore, the earnings struggle to cover the purchase of the machines, I think I'm going to give up mining.

It really depends. On the version on the subversion on the area you are in (humid and hot or cold), etc.

S19 XP can work till next cycle.

hello to you my friend please stick 2 topic

Hello, everyone. 

From what i understand of the chip manufacturing process (since the CPU days) they are in fact all tested individually, and some pass at certain speeds while others don't but work with lower speeds (same voltage) and then are classified according to their "yield".

My guess is when they assemble the hashboards, they mix them to roughly achieve the desired overall speed rating per hashboard. Naturally you are going to have fewer chips of high grade than low grade. This is another reason they introduced the "auto"(tune) in 2017 when the option to manually set the speed disappeared from their fw, it makes more economical sense.

And when the low yields start piling up, lower speed models appear in the market.

No he mentioned silicon lottery.

every chip stands on its own.

So you could have a unicorn in the s19j that he mentioned.

I have been mining with tons of gear since 2011.

when gpu's mined btc 5 identical cards were shipped to me they had consecutive serial numbers (very rare to get that many in a row)

they all had different efficiency. 2 could over clock like mad. 1 could under clock really well and the other 2 were in the middle.

Even if you had 100 s19j pros all good made on the same day it only means that bath of chips were higher quality. Same production line 2 days later with chips from a different sheet you could be off by 5%.

The only  way to insure the best chips is to have them bin tested one by one before you assemble the gear.

That is not done for a miner on a regular basis.

helpful thanks - does that mean s19j pro is the best choice here?

The best efficiency record i have seen so far using Braiins OS+ is indeed with the S19j Pro 104T producing 64TH consuming 1343W. (20.87 W/TH) which is better than the XP using factory firmware.



It was set to Power Target 1616, HOT temp changed to 60C and Dangerous to 70C with dynamic scaling enabled with a step of 200W.

Theoretically with two of these you can surpass the XP:

Single XP 127T 2730W
Two jPros: 128.6T 2686W


FYI: S19 Pro has the chip BM1398 and the j uses the BM1362 while the XP uses BM1366. Silicon lottery varies the results of the individual chips the manufacturer decides to put in your hashboards, but the general trend is that their higher speed rated models are more efficient (better yield) than their lower ones. So as we have already seen for years, getting the fastest one running low gives the best efficiency results.

As for reliability, yeah its much better than the S17 disaster.


For comparison, the best results i have seen so far with the S19 95T doing 24 W/TH (80TH 1965W), and the S19 Pro 110T doing 26 W/TH (92TH 2415W).

Yes the two models do exist, true the J versions use 5nm and the non J use 7nm but that does not mean the 5nm version is more efficient, the number of chips and how they clock thems is what matters.

The J is said to be more efficient by a tiny bit, it seems like availability and price is what matters to most buyers.

Pro vs non Pro is indeed a huge difference, so try to get the Pro, if J and non J are the same price get the J if not get the Pro non j.

did not know that there is also an s19 'i' model then I will rather stay away from it

I still do not understand the differences between s19 s19j s19pro and s19j pro.Is there actually a model s19pro and s19j pro or is this wrong?

I read somewhere that the s19pro was manufactured on 7nm and the s19j on 5nm - can we conclude from this that the s19j is better because of the more modern manufacturing technology and thus better efficiency?

Well the s17 and s15 had glued heat sinks which were pretty big issue

my s19 seem better also heavy solid heat sink.

The S19 pros are pretty reliable when compared to the previous 17 series, there haven't been any significant issues reported, they have some funky PSUs and that's all about it, so make sure you get a spare PSU just in case you need it. The "i' version is the only exception, it sucks and it's the cheapest out there, I'd stay away from it.

When compared to M30s+ which are almost in the same efficiency group, the MicroBT beats the S19 pro by a slight difference in terms of reliability,  none of this is my own personal experience, but I dig into so many telegram groups and I know I few folks who own a dozen of each, so I am just relaying that information here for everyone's benefits.

i read bitmain s19 products are good. I do not understand which s19 model is worth to purchase.

despite of hashrate and energy specs do you suggest to buy s19 s19j s19pro or s19j pro? s19 xp is too expensive for me

Had to replace the control board in one of the S19A Pro's today    as only 2 of the hashboards were showing up.

That seems to have fixed it but that is after 6-7 months of 24/7 hashing...

Depends on the conditions. High humidity (but still below specs!) will ruin the temp sensor on the farthest board in a year or two. Applying protection on the temp chips (nail polish type) probably solves this. I guess air particles are collected here by airflow + the humidity eats away the temp chip's legs.

Otherwise, S19 reliability is way better than S17 and even S9 which was highly batch dependent.

For Bitmain only the s19 series are more reliable than the old units but if you spare time you can also check Canaan units they are also good and I think it would be an awesome choice if you have strong units from Canaan.

Read the post from this Most reliable SHA-256 Asic Miners

Thanks all.

Looks like their custom chips only run @ 45c compared to 89c with my current s9's!

I just read that the S19J was slightly underclocked and more stable/reliable but it does not seem the case.

Looks like i will upgrade then.