It's not that relevant since the Compac doesn't have that chip on it, but I guess the general overclocking concept does still apply.
I should probably have put an S1 underclocking writeup on there, since I spent a week generating a full chart of efficiencies at every 10mV interval between stock and the point the chips stopped working. I should probably have put a lot more stuff on there, really. The website's been pretty neglected lately.
that is true, this is what I get for googling the TPS53355.. but as you said, the Infineon is superior to it (100A over the 53355s 30A).. as you said, split the voltage planes, feed it 12V go OH SHI- and enjoy hot boxing the blue smoke  |
|
|
|
If you can separate the power drive rail from everything else on 5V (the buck chip, USB chip, LDOs, and LED circuit) so they don't explode, maybe you can do it. With a higher input voltage you'll be reducing the overall duty cycle quite a bit which will make the input caps' job easier because the high-side current burst they have to buffer is shorter. Honestly, the weak link in the buck circuit with a 5V input is the USB jack, and after that it's the inductor. The USB jack is rated for 1.5A, probably because for USB applications the standard only defines current up to that point. If you trust it to 3A that's 15W right there. At 0.85V the theoretical max current is a little over 17A. How about that, the inductor is rated for 17A RMS. So if you can jimmy the input to be comfortable with more than 3A still at 5V, the next thing to replace is the inductor. A lot of old gear, from S1 to RBox, have 0.47uH inductors on them probably good for over 30A. Now you're cooking with gas. The FETs are actually rated for close to 100A if you can keep them cool, and at the gate voltages that driver gives them they should have an Rdson between 2 and 3 milliohm - on par with the internal drivers in everyone's favorite TPS53355 (which has a 2mOhm low side and 5mOhm high side switch). A lot of 30A 53355 VRMs have three 100uF ceramic output caps; so does the Compac, but it also adds a 470uF tantalum-poly cap with 30mOhm ESR for extra buffering.
If you do separate the 5V chip supply from the main power (which techically could be done by drilling out a single via) you can circumvent the Compac's real high-load Achilles heel, which is its lack of ripple isolation at the buck driver. I have a 2.2uF cap at the buck chip to help decouple, but that's not so useful when you're looking at 15-20A current kicks on the input during a high-side switching. If that kick drops the input voltage below about 4V the buck resets, which means you don't get stable power on the output and the RMS voltage is lower than your setpoint says it should be. Isolating the chip supply's 5V line from the power swiching supply with something between 2.2 and 4.7 ohms resistance will cut down a lot of the brief input voltage tanks and keep your core voltage stable, which will allow you to push the chip even further. It's humorous to say, but with a very few changes, that little stick miner could push enough power through your ASIC to catch it on fire even with that giant-ass heatsink and still on 5V.
Im'a look to see what I can jimmy up from my dead NRBs I'm sure they are running with them TPS53355, nothing beats hashing out the magical blue smoke  just for lolz, reading your writeup on cube OC'ing on your website (all about the TPS53355 and feeding the BE100s with more power), relevant information is still relevant! |
|
|
|
Ive been waiting for this.. FWIW, BFGMiner can theoretically go up beyond 1000 MHz now... try at your own risk :p
Oh boy! time to play the "Who has the fastest and hottest stick!" looks like i may have to get a beefer Buck.. may have to get the stick to take 12V at 30A... /me hunts out old PoS 12V USB hub (for those people going "what the?!" http://www.poweredusb.org/pdf/PoweredUSB_v08g.pdf ) |
|
|
|
Thought I'd share my misery for the last 24 hours.
Tried playing with voltage and frequency (for the contest) on my two sticks and minera-cgminer pi.
Everything went to hell. Only this morning do I think I got things back to where they were- 24Gh at 225.
Also edonkey gave me this so I didn't have to wait for a 1K share to appear, which is what cgminer seems to require before down regulating my share difficulty. But it doesn't work. I did put a space and a number (80) after 'diff'
/var/www/minera/minera-bin/cgminerStartupScript: line 4: --suggest-diff: command not found
Any thoughts on this one?
Good job! Just a lil curious at what voltage your running. Seems incredible to reach 24ghs at 225 freq. 2 sticks not one stick. unless I misread his post well, "They" was mentioned, not "It" |
|
|
|
*Edited as long quote
2700CFM at 0.33HP, my sun server does more air then that and at 110dBa :>
since my little hub got 30A at 5V directly from the PSU..
They actually say 1062 CFM in video. It could be wrong though it seemed a little dated on video. It is a interesting designed fan's. All I use are more traditional round ones. I am impressed with your hub. I would love to see how high a stick goes on that. how many fans? are they 80mm radial design? whats the static pressure? how many rpms? what temps are they in? is 110 db a or c-weighted <= a/c weighing makes a difference in perceived sound in human ears due to the shape of the ear canal. animals like cats/dogs hear much different) https://en.wikipedia.org/wiki/A-weightingthese are 0.33 horse power i believe these are big and each one weighs 27lbs. I ran them around 2007 in construction. that have warnings not to move them while running and warnings to brace doors as they will close them with air presure and keep children and drapes away from them. gotta love it yup servers are LOUD and can move serious air indeed. 110db can hurt your ears long term ( i work at an ent medical practice with audiologists and run pro sound in a band ill grab a sound meter and check) I have some delta fans kickin around somewhere in the basement the video is from 2006 I think these are fairly quiet we ran them in residential houses. and we do have servers at work that need air conditioned environments. this one you can rent if needed the old thing is a Sunfire 280R, the 3x12cm fans (Fan Module 540-5088) are louder then a Vax vacuum, I have to yell to talk to the other person! So OK, its not 110 db, but it is rather loud. and they will push on your hand if you have it in the air stream (im thinking 250cfm each).. also the 2700 is what the factory is rating them blowers for.. I used to use a electric radiator fan out of a car as an exhaust fan, they did 2500cfm and were only 16". I'm half tempted to setup a room to do that again, although the house I'm in has fold out windows, not sliders, so a little harder to setup.. |
|
|
|
As enthusiastic as I am about a contest to maximize a stick, I don't know that I'm up to it now. I found these comments in the stick support thread- meech - steadily running @300mhz 7.2 v adj. @ 1.52a with 4 hashing about 16-17ghs edonkey's getting 16GH @306 Mhz chig reports running at 400 Mhz with no errors in 24hrs (no hash rate mentioned) phillipma1957- I did run at 400 and got 22gh but was pulling over 2.1 amps or just about ten watts a stick. (and i bet if phil unplugged the other 19 sticks in his stud(ly) hub, he could shoot lightning bolts into a stick and manage to keep it alive for 24 hrs.) and this- Did you see we jacked one up a bit yesterday and got it to 488MHz? Most of the high-power parts were actually still stock. Keep an eye on sidehack, he might just want to win one of his sticks back. I've got 5 sticks that's been going 5 days @ 350MHz, with a big 2HW, would have been a longer run if the power didn't go out for about 2 hrs last Friday. http://imgur.com/e8Xqr7hhttp://imgur.com/hhcItj3I'm waiting on some better fans to move more air, before going any higher. And if what I got planed works, should be able to hit 450+ MHz (but got a little testing to do yet to see if it's going to work). *cough* http://www.jondon.com/dri-eaz-sahara-pro-x3-turbodryer-air-mover.htmlwatch the video please and download the manual (pdf) *umm cough* i do have access to more powerful ones i used to do fire and flood restoration in construction thats just the 1st link i could google real quick @ phil is that cheating? HMM! I kind of want one... LOL 2700CFM at 0.33HP, my sun server does more air then that and at 110dBa :> since my little hub got 30A at 5V directly from the PSU.. |
|
|
|
Note --set cbm:clock=xNNNN will also work; you just need to abbreviate it "cbm" and can't spell out "compac".
(That's what is being fixed in 5.4 - being able to spell it out)
I was unable to get --set cbm:clock=xNNNN to work. I believe you said you used it on the pre-production stick that was sent to you, but I cannot duplicate your success on the production stick. I'll attempt to duplicate with mine, I believe there is little to no difference between the test and production versions, if anything, only the naming revision on the firmware. I could send you one of the production models Luke, if you think there is a difference? |
|
|
|
Kind of an aside because I already voted for the 2 boards +25 deal.
But in the spirit of voting often, and because I have 3 fully functional S2's (a whopping 30 boards, sorry no BKWhoppers),
I want to suggest an evensteven trade deal. Boards for Pods. No cash or coin.
You'd have to come up with numbers and mix e.g 2 S2 boards plus 1 S1 board missing no more than 2 capacitors, would get get you one pod.
Or six S2 boards, and 4 S1 to S3 upgrade boards would get you 4 pods. You get the point.
I don't have or expect to have any S5's or S7's and I wouldn't be put off if those who did got a little better deal as these appear to be more critical to the operation.
I hope you consider this and I hope you recognize that once you release any 'close to final' payment schedule the UPS man is going to be all over you and you'd be smart to find some extra carts and dolly's.
If they were S5 I don't think trading would be a problem. Unless a bunch of parts on it seems like S2's are a lot of old chips. So I don't think they will hold a even trade on S2's. But I could always be wrong. I was going to point out the same, board for pod, S5 has the required components, S1/2/3 all have the older chips, no good for the project, the only thing worth it is the component for power (buck ICs, caps, etc). The 4 odd New-R-Box boards would cover 1 pod, instead I would be more inclined to find the BM1384 chips off the S5. |
|
|
|
|
Thats probably my problem, bad wick, I have to flux it and hold an iron to it for about 5 minutes to warm it up... and its only basic 5mm wick too..
i might have to go back and remember what Ben Krasnow did when he was playing with QFN devices, i think he just pulled the spring pins out of a socket and soldered them to a footprint..
|
|
|
|
If belly pad contacts don't get into the VDD corners, I guess a 56QFN socket could work.
I never messed with the BM1382 but that package looks like it sucked quite a bit. BM1384 is fairly straightforward and BM1385 appears to be a stock footprint. Anything with perimeter pins like QFN is going to be a lot easier to work with than BGA or a bump die package because you can more easily verify alignment by looking at the solder rather than just hoping it was right. Belly pad floatation also helps a bit. I haven't worked directly with much BGA so I don't know how well those self-align, but they tend to be larger packages as well so the weight could prevent some of that movement.
I must ask, how does one align these QFN packages? every time I attempt to solder a replacement cp1202, the leads shift between the PCB pads, which requires me to reattempt it. In attempting to re-solder the package, cleaning the solder away (for the 2nd time), a pad or 2 just comes away, ruining he PCB. As for BGA, you really need to X-ray it to see if it is properly attached. 2nd hand airport baggage x-ray machine may help? Depending on who made the socket, some are a little round nail that touches the belly, others are pads, some even have simply a hole. |
|
|
|
Heatsink isn't part of the base cost. Someone wants 50 boards with 50 heatsinks, that someone will also be paying for 50 heatsinks. One S5 board only gets you the 8-chip PCB and nothing else. You provide your own cooler, power, computer running cgminer et cetera.
I do not have a socket to test the chips standalone. Since it's, as far as I know, a custom footprint (I don't know any other QFN with corner pads like this one) I'm not sure where to even look for a socket.
My assumption is at least 24 viable chips per 30-chip board. More than that is a bonus.
just to clarify, with the power, and passives, I was thinking the other components on the pod, fare enough for the rest, (sinks, power supply and mining controller) being up to the buyer to supply. Yeah, forgot about that customization of the QFN (Y U DO DIS BITMAIN?!), I've never really had a good hard look at the layout, I thought maybe the Vcc pads would extend to the edge of the package? I would think that would be just as painful as BGA and the unknown of it being properly soldered. |
|
|
|
As many people as are popping up with boards, I'll probably look to making more like 200 than 50.
I guess what I was thinking for a cap was limiting the number of free pods that could be gotten for S5 trades; say, past the cap an S5 board would pay for half a pod instead? That does seem kinda crappy though. I think I'll leave it wide open.
Regarding "extra pods" if folks don't buy them all up, that's not really a problem. Heck I'll just mine on them myself.
Well theres got to be a cap somewhere, you do need the other parts to make the pod: Power, sinks, other passives.. if someone comes in with 50 boards and say "50 pods plz" and you only had 20 heatsinks to go with them, who's paying for the other 30? Also, how many of these boards are going to have viable ASICs on them? how have they been treated after death? manhandled like some 500lb gorilla? and for them, are you willing to say to them "no sorry, chips are no good"? do you have a rig to test these chips off-board? (got one if them $100+ QFN sockets?) |
|
|
|
Also note that BFGMiner 5.4 will support setting Compac clocks as the MHz number directly.
Nice, i have one customer that was running BFG and switched to cgminer because somehow he couldn't attain more than 12 GHs with these sticks. Is this a known issue, or should i debug it further? I have no problem getting 15 Gh/s myself. AKA, debug further. Get the latest BFG, set just the MHz number, see if it works, don't forget to add a little more juice to compensate. I was getting 15GH/s out of the box myself. I was running in stock CG with a OC'ed U3, so I wasn't sure the numbers propagated to the GSC (both were running at 250MHz/0.875V according to CG) |
|
|
|
Also note that BFGMiner 5.4 will support setting Compac clocks as the MHz number directly.
Nicely done, making things easier! |
|
|
|
i was going to make a little tune of "packie packie, ima pack'n ma steeeks" but someone is going to call me racist..   |
|
|
|
Soooo kids, Aussieland has dem steeeks! Opening the package  (ohh that smell, that sweet smell of new stuff.. embarrassed sidehack? *throws on a Gordon Freeman voice* I can smell you... )  ooohhh yeah! tricass's, heslos and TheRealSteve's sticks!  And the rest!, Stringer Bells, BTC6000s and mine/replacements for any dead ones So, packing them up and getting them ready to post on to you guys |
|
|
|
sidehack, what do you thing a peltier would do? 22 watts of cooling capacity. cool the power section and chip. I would need frost precautions etc but I can do that easy.
stupid? oh yeah that has vapourminer written all over it. I mean anything worth doing is worth overdoing as the saying goes.
ive only known +50W peltiers to go below 0 degrees Celsius, i've been playing with a 40mmx40mm, that runs at 15v 7A for best results, cant get it lower then 7 degrees lower then ambient, off a sink.. what about novec, anyone wants to drown these babies? only if they are in a fish bowl, with an IP cam on them, sinkless |
|
|
|
now I know where my other screen went, PostmasterBlaster! time to visit Tradetown! |
|
|
|
Hooray! It's about time PostmasterBlaster got around to delivering.
Its probably been sitting at the shop for a few weeks, and they haven't got around to delivering it, I got it when I when there to RMA a LCD screen me: "hi, i have this to ship to china" her: "oh boy.." scales item, gets declaration paperwork out the back her:"Oh, I have a package here for you" me:"Oh lookie, it finally arrived!" her: "ah, yeah" grab it, wipe dust off the box.. AUSTRALIA POST IS BEST POST! so people, you'll get your sticks at Christmas, way a head of time! |
|
|
|
Guess what arrived at the doorstep of my work.. (TBA Photos yeah! ) now to pack them up for everyone |
|
|
|
|
Show Posts
