The only trouble's going to be I need to warm up the robot room
You know a thing or 6 about warming up a room. And he'll have ample hardware to do it once my lil pallet of presents arrive for him ta play with  |
|
|
|
<snip> I mark the S5 controller and fan at 5W, and the board at a shade under 65W, which 350MHz for 8 chips should run about 154GH which puts this guy at a solid 0.42W/GH at the board alone (not counting fan and controller). Pretty decent. <snip>
Pretty impressive considering my first 2 miners were BFL 10GH/s Jala's bought through TigerDirect. They pulled 65w each + the 20w needed to run the Host PC. Them 2 lil' bastards are what got me hooked on BTCSweet. |
|
|
|
The freezer is useless for this. As others have said they are made to reach a low temp in a very small volume with NO additional heat input to the system. The math: BTU/hr=Watts x 3.4129 so 1 miner @ 1,200w = 4, 095.5 BTU/hr 1 Tonne=12,000 BTU/hr, large A/C units are rated in Tonne Bottom line is miners produce massive amounts of heat to be moved from point-a to point-b and typical air conditioners much less a refrigerator/freezer ain't gonna work  |
|
|
|
Well, I like to overbuild things. Who in their right mind would put a 16A-rated buck on a USB stick? But y'all are having endless fun with it. I don't want to scrimp on node-level buffering so I have a good full-voltage output cap bank and good node-level caps as well. The per-node polys are 470uF Compac leftovers (which will probably be 470 or 680 pulls from S1 and such on the final) and a 100uF S5 pull tossed in for good measure. You gotta remember, I'm designing this to use scrounge parts so I have to play with what's available and keep costs down.
Good node-level current availability is essential for ASICs to initialize at low node voltages, since there's a bit of a spike when the chip first fires up which might not be available from idling chips higher up the line, and if you're running at 600mV per node you really don't have much room for sag. Part of that will be overcome by good node-level caps, and part by dynamic voltage control - we'll probably have the controller init the string at a higher node voltage and baseline frequency, then ramp down to user setpoint voltage as the frequency ramps up to user setpoint. This should ensure minimal excess power draw and reliable startup.
Just tossing this info out there so folks can have an idea of what happens when power is getting from the regulators to the chips. http://powerelectronics.com/power-electronics-systems/five-things-every-engineer-should-know-about-pdnNot just a matter om having path from point-A to point-B. Needing to take into account lumped-component values and resonances they can cause, it can be more like designing power RF circuits. Edit: Add to that this rather graphic bit on what can easily happen to an ASIC's power... http://powerelectronics.com/community/why-pdn-measured-using-vna-and-not-oscilloscope?page=1resistive load step response:  Same load step with just 1 resonant node in the PDN:  |
|
|
|
fuck I bought 3x s7 batch 8, and tomorrow should send
My god we hope that work well
Odds are they should be fine. I have 3 of the earlier ones from Batch-8 and they are running perfectly. Never OC'd, fans turned down to 50-55%, all running the Bitmain PSU's. I've also placed an order for 2 more from batch-8 that I should get in a week or so. |
|
|
|
Does this work with all A1-based miners?
I have one of the few 1.2TH units that AMT finally delivered to some of its customers. Looks like it is really from LKtec. Been running solid for over 1.5 years but your upgrade looks far better than what it has.
I guess yes, but do not know for sure as i dont own one. Do you have a download link for your installed firmware? I will know after inspecting it. Feel free to PM me. Not sure if I have an image of the firmware. Once this unicorn of a miner (rare) arrived after >6mo of delays I just walked away from the AMT/Bitmine.ch fiasco. I was one of the lucky ones that actually got something from the AMT/Bitmine.ch scammers  I suppose I can always pull the card an make an image of it. Wow, you must be one of those "happy customers" on the pictures 😂 j/k If you provide an image, i promise to have a look asap. Would a zip of the card work? I use HDclone at work and not sure if its 'smart' .img files which do not copy the free-space bits are usable by other programs. 'Happy"? Not by a long shot. A lucky one, yes and never ever to be seen in any AMT promo. Ordered AMT's 510GH/s unit in Feb of 2014 for $4k, advertised as "ships within 4 weeks" unlike the 1.2TH/s rigs that were advertised and sold as pre-order only with to-be-determined delivery. I was sent a 1.2 TH/s rig only because they were all the scammer Joshoua Zipkin could get during his stay in China. That said, eventually actually getting 2x the hash I paid for vs folks that paid the full $8k for their TH/s rigs - several folks had waited since late 2013 - I was DAMN lucky. The events are best covered in the Skype chats linked in my sig. |
|
|
|
Does this work with all A1-based miners?
I have one of the few 1.2TH units that AMT finally delivered to some of its customers. Looks like it is really from LKtec. Been running solid for over 1.5 years but your upgrade looks far better than what it has.
I guess yes, but do not know for sure as i dont own one. Do you have a download link for your installed firmware? I will know after inspecting it. Feel free to PM me. Not sure if I have an image of the firmware. Once this unicorn of a miner (rare) arrived after >6mo of delays I just walked away from the AMT/Bitmine.ch fiasco. I was one of the lucky ones that actually got something from the AMT/Bitmine.ch scammers I suppose I can always pull the card an make an image of it. |
|
|
|
|
Does this work with all A1-based miners?
I have one of the few 1.2TH units that AMT finally delivered to some of its customers. Looks like it is really from LKtec. Been running solid for over 1.5 years but your upgrade looks far better than what it has.
|
|
|
|
|
In the US I know that under a claimed value of $2500 there is no Duty or Tax to be paid with goods shipped from China.
I always use UPS and only once was I charged a brokerage fee ($40). Why that one time charge I have no idea as over the past couple years I've gotten several dozen Ants with no fees charged. However, that was when the ordered equipment value was under $2k.
I just placed an order to Bitmain for 2 more s7's and with shipping total is $2545 so I'll soon know if that changes.
|
|
|
|
I am in... Same here. What wallet address do we send to to get a ticket? As for s5's... I have 3 that I'm willing to donate to the effort. They generally are fully functional but every week or so will drop one board, a simple hard reboot and they are happy again. |
|
|
|
|
oh well. Also have 15 s3's in the process of being retired.
|
|
|
|
Got it Philip (I frequent Data Centers, network engi)
Regarding the 2 USA not 4?
Is it a network delay issue or node saturation?
tee
eye
A
No, it's a matter of Customs Duty and fees. Believe they kick in at $2500 so you need to split the orders to stay under that. As for PSU's so far I'm very happy with the 6 I've gotten from Bitmain. At a rated 1600w they put a s7 load right around the 80% max load sweetspot. Also have scads of the IBM DPS200 PSU's. |
|
|
|
Spoiler alert - I'll probably have an update regarding the pod miner tomorrow. It may even include pictures.
Faboo!  Along those lines I currently have 1 s2 off-line and another soon to be replaced with a s7 so if ya need chips yer welcome to the blades.. Also have 4 s1 hash boards sans heatsinks from doing two s1>s3 upgrades. |
|
|
|
|
I doubt it. Pretty sure he thinks it will all go away some day.
|
|
|
|
14/16nm full custom will be the last major efficiency improvement - for a few years.
IBM has already announced they're researching 8nm (IIRC) for the next generation process.
Agreed. After several years of announcing 14/16nm is 'soon to be here' the foundries like TSMC, GloFo, Samsung et al are only now finally nailing down the production processes for 14/16nm chips with reasonable yields. Still have massive dead chip rates vs the larger nodes but getting 30-40% viable chips per run ain't bad. The current scrap rates are very acceptable for companies (cell phone/PC chip makers) that can afford it but still relate to barely acceptable for other chips users. Still, give it time and yield rates will continue to climb. As for the 10-8nm nodes...  Not going to see that fore at least 3-5 more years assuming they get the EUV light sources to be more powerful and stable. The current way being used (zapping droplets of tin with >30kw lasers and collecting the light produced) is working but IMHO not the final solution. re http://optics.org/news/6/6/31 and http://www.photonics.com/Article.aspx?AID=56802 |
|
|
|
I like that it's water cooled. Interested.
Just caught this thread. Per the requested wish list re: water/fluid cooling. Just make sure the setup can take real-world cooling system pressures. Preferably say 50 psi or more. Why? So the end user can implement a reliable system using a line powered industrial pump(s) and fans on the radiator(s). The 12-24vDC pumps & reservoirs folks seem to want to use may work for individual miners but.... put a rack full of them together and then one big central system is called for. Make them able to be plugged into a large dry-chiller system or other heat removal sink (possibly a spa?) and you have a winner. |
|
|
|
220v plug on a 30 amp breaker.
Need a good Pdu any suggestions?
<snip> IF your in the US you won't find one that goes over 24 amps, max code law here!!! for a 30 amp line. you can Polly push it to 30 amps max if you disable the breakers on it, but if it is working right the breakers on it should shut down @ 24 amps or tad bit over .each bank on the one i use will go to 15 amps per bank Actually, Tigerdirect sells a 220v 32 amp one from TripLite: Tripp Lite Single-Phase Monitored PDU - 7.4kW 32A 230V, 0U Vertical Rackmount, 36 C13 & 6 C19 Outlets - PDUMNV32HV2 http://www.tigerdirect.com/applications/searchtools/item-details.asp?EdpNo=9105661Ya is a heap'o unused outlets and sorta pricey but - Tigerdirect accepts BTC and at current exchange rate, well, |
|
|
|
If that PSU was having issues, it's a lemon. I think Phil was running on 120V back then too. 240V is easier on the PSUs, I hope you'll be using 240V with 6 S7s?
EVGA says you can pull 1300W DC from them, IE ~1450W at the wall for 10 years. That's good enough for me! Worse case they replace the PSU for free.
That's good to know as i was getting nervous after reading what Phil had to say. Looking at graphs, efficiency is decreasing at least 5% at or above 90% load. Yes, I was running at 110/120V, but I send them all out to hosting with sidehack-posted review in Hosting service thread. I like his arrangements and service. Aside from the decreasing efficiency which for ATX supplies actually largely is from the fans running near/at full speed and taking more power to do it, with ANYTHING dealing with power be it plug/circuit ratings and including PSU ratings always follow the 80% rule. For the feeders it keeps plugs/cords/wiring behind walls at safe temps when under a constant near-max load - which miners present - for the PSU's you need that margin to handle the random (hopefully) short-term events all powerlines have. For example of pushing ratings just talk to Bitmain and their experience with early s2 PSU's.... But as long as you have spares to cover ya while EVGA ships a replacement then your call. Und ja, as has been often said, any PSU is much happier on a 208-240v line because of the lower currents on the AC-line side of the power conversion. |
|
|
|
You sure you guys can't put a resistor between some of the pins to regulate the fan speed?
If the resistor is put in the +12v line feeding the fan, sure it will work as a brute-force way to do it. Question is it a 3 or 4 wire fan? With either 3 or 4 wire it will be reporting it's speed, 4-wire gives it built-in PWM control. With 4-wire most often no signal = fan ON (for safety and convenient way to let fan go full speed if wire 4 is open) and +5v = Fan min speed. Play with the signal duty cycle to set the speed between it's min and max speed.. Dunna know if the PSU's fan circuit does anything re: monitoring the speed except for maybe checking for a stalled fan. |
|
|
|
It's funny you mention poorly-designed power on the Monarch, because I remember talking with one of their guys on here a while ago and he was going over how difficult it was to design and test the regulators and that was one of the big holdups that delayed the product.
If it was very difficult for them to design and test the regulators I would say they were inexperienced. My guess is inexperienced and locked into a 'well it worked before' mentality. Just gloming together more VR modules in parallel to give needed very high current and low voltage is just a small part of the battle. Getting the desired power from the VR's cleanly to the chips is a whole 'nother ballgame and requires looking at the PDN like a radio circuit than a simple DC bus. Probing the power planes, the spiking rise/fall times and ringing you see on them even just an inch away from a different spot will be very different. To me at least, that is the biggest advantage to the string topology - currents are far lower making power plane considerations much easier to deal with. |
|
|
|
|
Show Posts
