which leads to a fan that vibrates and won't spin at full speed (usually), and DEFINITELY decreases airflow for less cooling.

 Might not be the issue if your temps are close, but it IS a future "fail point" even if the card is working NOW.

I can't remember the last time I saw a Sapphire card that had anything but a ball bearing fan in them.

 I WISH they made NVidia cards.

As I recall, the fan on the S9 is rated for 5800 RPM at full speed.
It's NOT the high end of that Delta series (which is a 6800 RPM model as I recall rated for 250 CFM or so) but one step down from the top.

ANY computer-type equipment (specifically including cryptocoin mining gear) converts pretty much ALL of it's wattage consumed to heat.
The difference is measured in MILLIwatts at most (a little of the power the network interface uses gets pushed elsewhere - but it also takes IN a little from elsewhere, and we're talking MILLIwatts at most here).

 A typical "drier duct" is in the 4-6" range - which is NOT enough to handle the CFM of 4 miners like the S7 without a "booster" fan in the duct - and a 4" duct isn't going to be enough WITH a booster fan.

 I'd suggest upgrading the duct to 8" if possible AND adding a "booster" fan to it - then keep in mind you need air INTAKE as well somehow into the room.

 EVGA G2 series is my #2 option - and still available most of the time where my #1 option Seasonic X-Series is pretty much out of stock everywhere.

 I won't even look at the G3 series or the Seasonic "Prime" series (or that new lower-end series that replaces the Seasonic "G-series" models) since they moved to a sleeve-type fan instead of a double ball bearing fan.
 FAR too likely to die WAY too fast in high-load high-temp service like my mining environment.


 Corsair outsources their PS manufacturing to DIFFERENT makers depending on the model line - which is why their PS vary so widely from "very good" to "junk".

Low temperature isn't critical, though it can help keep cooling costs down.

 Dry and hot is BETTER for mining than wet and semi-hot, as evap cooling is CHEAP.

They've been having issues since August - and haven't fixed all of them.

 They haven't even TRIED to fix their Litecoin chart - it's still stuck in mid-August.

I'm getting close to just ignoring Bitcoinwisdom as a result.

 As I recall both EVGA and MSI make some with waterblocks, and there are plenty of aftermarket waterblock kits.

 I do find the "Seahawk" model name a bit irritating, but that's personal. 9-)


 I'm starting to move TOWARDS more riser usage for cooling reasons - my 3-card rigs that ran reasonably cool while folding are running the cards a LOT hotter since I moved them to ZEC mining - but if I put the "middle" card on a riser the cooling improves dramatically.

 Just built up my first rig with the ASUS B250 "Mining Pro" board (the one with 19 slots) as an experiment - more due to ease of multi-PS management than that I plan to USE anywhere near that many slots.
 Only major issue I had with the build turned out to be my own fault - I'd forgotten the Aorus 1080ti needed 2 x 8 pin power, not 8-pin + 6-pin - so now I've got a floating Aorus that needs a new home (I'll probably swap it for the Gigabyte Windforce in another machine).

 Due to power supply and infrastructure limits, each "part" is going to end up with 2 x 1080ti and either a 1080 *single-power connector" model or a 1070 on the motherboard for the 3'd card.
 Up side - EVGA "dual" power cables have wide enough spacing to be able to mount the riser boards far enough apart for good GPU cooling (4" centers NOT a problem, could go a bit more if I had the space).

 Looks like EVGA will be getting more of my GPU purchases in the future, as they have what I consider to be the best 1080ti model that does NOT need 2 x 8-pin power AND they love double ball-bearing fans.

 I just wish NewEgg/Rosewill would quit wasting money on those "SATA-PCI-E" adapters in their risers, and either "don't include the stupid dangerous adapter at all" or worse case switch to SATA-Molex.


 Time to start building out some of my older hardware and lower-performance GPUs as "gaming" systems and sell them on Craigslist....

 They want your PUBLIC IP address, not the local one (if you use a NAT router, they won't be the same).

 Try "myipis" as a google search to find various sites that will tell you your public IP address.

 Your setup is NOT LEGAL per your own description.
 BLATANT VIOLATION of the NEC, which most US jurisdictions have adopted as LAW (sometimes with additional restrictions) - and the very few areas that have not adopted the NEC have adopted an alternate standard that is very similar.

 Putting a splice in a box DOES NOT make it code compliant - there's more to it than that.

 It is very questionable about SAFE as well - still waiting to see where provision for strain relief is at, that IN AND OF ITSELF makes your setup unsafe AND a violation of Code.

 It's not defamation when FACTS are being stated.

 
 I'm not going to go into the "not covered wiring" issue, that's a point I personally think the Code is a bit excessively anal about - but it IS another violation.

 Be happy the guy I learned to be an Electrician under has not seen your setup or heard your "instructor" say is OK - that instructor would be getting DECERTIFIED FOR INCOMPETANCE in no time flat.
 Unless he WATCHED you making the splices, there is no way he has ANY CLUE if they are properly done or if they have any potential to be safe - and a LINEMAN should understand the need for stress relief better than I DO.


 I never said anything about "high voltage" - and by LINEMAN standards I agree that a 220v setup is low power - but by the standards of connections where electrical tape gets used, it's HIGH power.


 If you are in a building/home with a substantial space between it and other structures, then it might be justifiable to claim it's "your risk" - but if you are in space in a "shared" building/condo/townhouse/apartment/whatever or you are in a crowded area with only small spaces between structures, it is NOT just "your risk".

 My farm is not "huge" by any measure.
 
 I just don't insist on living "high on the hog" while I get it established and grow it.

 Personally, no - but Sidehack has mentioned having some chips, Hotmine has been selling Bitfury miners for a while, just for 2 rather well known examples.

 Raw chips would do "me" no good, I'm not a board designer or coder.

The temperatures of a GPU will NOT affect cold solder joints - whoever made that claim is totally clueless.

 100% speed on a fan DOES increase wear on it quite a bit over 80%, and is not generally NEEDED unless your cards are getting pushed fairly hard in a hot environment.
 A good ball-bearing fan is normally DESIGNED to run for years at 100% of design fan speed - used to be THE NORM back in the days before PWM fan speed control got common on some fan types.


 Dead fans CAN CAUSE CARD FAILURE due to overheating - but IME this is more common in power supplies (that don't tend to HAVE "thermal limiting" control) than on GPUs.

 I specifically looked for a space that already had the capability of good ventilation and LOTS of outlets - and cleared with the landlord (I think they're also the owner) ahead of time that I'd be installing evaporative cooling instead of using the existing window-style air conditioners.

 I'm fairly small scale though for a "pro" miner.


 Took me a bit to figure out why the space was as low cost as it is - then the first train rumbled by, there's a track less than 100 feet behind the place.

 Not always - my Sapphire HD7750 had a dual ball bearing fan, but the fan lasted less than 5 years.
 On the other hand, the HIS HD7750 cards I bought shortly AFTER I bought that Sapphire had all of their fans (except one) die in less than 6 months and that ONE didn't last a year - bloody junk "rifle bearing" junk.

 Dual ball bearing fans DO tend to last a lot longer in high-load high-temp environments than anything else though, which is why I try to avoid any other fan type on my gear (and refuse to even CONSIDER the new Seasonic "Prime" and EVGA "G3" power supply lines).


 Adding oil is at best a VERY SHORT term cure - if it leaked out in the first place, the seals are shot and adding more isn't going to last long.

 A PROPERLY done splice, with PROPER stress relief, won't get loose in less than decades - but they WILL corrode, causing increased resistance and a tendancy to overheat, in a LOT shorter time a lot of the time even when PROPERLY done.
 Most splices are NOT properly done though - Philopolymath's pictures show absolutely ZERO provision for stress relief, even if the splice itself was well done.

 Electrical tape is not intended for permanent installations because it DOES dry out and crack and get loose fairly quickly - sometimes in less than a year in a hot environment, rarely less than 10 even in an air-conditioned environment on a VERY LOW POWER splice.
 If it's tape on a high-power splice, the issue gets worse quickly as the splice ITSELF generates heat, even when WELL done, due to contact resistance.

 There is a reason insulated crimp connectors exist - though even THOSE can have issues long-term even when installed correctly, they're a LOT better than splices on a high-power or in a long-term installation.

 For reference, some smaller arc welders MAX OUT at 250 amps - like the Lincoln model my late father used.

 You CAN get the whole thing free of charge - just download each chapter seperately.
 The "pay for it" entire version is the print version, and the cost is to cover the cost of publication more than anything else.
 You can also generally get access to it enough to read it through most Union Electrician shops if you ask politely, and some LIBRARIES have a copy of it.

My Sapphire R9 290 cards (reference blower design) ran at 1050/1250 all day long at around 80 C when they were mining ETH (unless it's HOT in the room they are in) - thanks to being reflashed some time back with an undervolt BIOS from TheStilt.
 They COULD run at 1100/1250 but it didn't help hashrate noticeably while it kicked temp up quite a bit.

 Before that, they struggled to stay at STOCK 947 core clock before they thermal limited if the room was more than about 70F = and were hitting 93C or 94C (whatever the thermal limit point was) pretty much all the time.

The "Blades" actually were 2 40-chip units, not a single 80 chip - so chips=40 would be correct.

 They hashed in the rough ballpark of 2.6 Mhash/sec (a little over as I recall) at their "stock" 700 Mhz setting.