Topic: New Version (unoverclockable) ASICMiner Blade Overclock Kit - IN STOCK - page 2. (Read 15090 times)

I actually did some rework on my own blades today, built a new enclosure for them in a rack-mount case. I had problems in the new case, with keeping my original test blade cool; the polish and slight warp on the stock heatsink got the (full overclock) efficiency from 30% to 99% under the same fans. Atrocious thermal coupling beforehand.

Yeah, the fan you're getting to keep that blade cool should, with basic ducting, work for two or three of these. Good to hear everything's going well so far.

I am very happy with the upgrade you did to my blade.

One interesting tidbit. I already told you about my lack of proper cooling, hopefully to be remedied when my fans arrive.

But I have noticed that cooling the chip side heatsinks improved my hardware error  rate substantially.

Even without proper cooling I am getting better than designed output. I can't wait to see what I'll get out of this blade when fans are installed.

Art

I'll take 5 kits.  Incoming payment!

Got about 8 kits left. Gonna be doing a second round before long. In addition to upgrade, service, I'm also looking into Blade repair; if you're in the US and you receive a damaged Blade I might be able to help.

As in the OP, kits are 0.025BTC apiece and flat 0.01BTC shipping per order. I've only got around a dozen kits in stock - these have really been flying off the shelf.

whats the current price?

How much to upgrade 3 blades for me?

Blade upgrade kits are in stock and shipping. Parts for 20%, 30% and 40% speed increase in one kit, along with cooling components and parts to improve VRM stability. Got about 30 kits available at present.

I've now got my two blades running at ~14GH/s using 16.000MHz oscillators.  I've warped the heatsink to get better contact and on one blade I had to use thermal paste to get it stable.  Now they're running rock solid - nice jump from 12MHz to 16MHz.  ~33% speed increase, sweet!

Thanks again!

If you get these kits, do use a hot air station (they're not too expensive, or you can make one yourself).  The swaps are pretty easy with the right equipment - takes less than 30 mins each and that's if you're a first-time SMD modder.  The heatsink warping and application on the chips take longer than that!  ;-)

Blade Kits should begin shipping on Wednesday. I've already responded to anyone that PM'd me in the last week or so showing interest. I'll have enough parts for about 40 kits to go out, and if they sell well enough I'll put together more.

Kits this round are improved from the last round - better regulator components, and hardware for three different overclock speeds so you can run yours at the highest speed *you* can keep stable, not the highest speed *I* could keep stable - every board is a little different and they don't all behave at top speeds. Kits do still include extra chip heatsinks, but I do also recommend smoothing out the backside of the stock heatsink and maybe using a bit of thermal paste to improve heat transfer. This can be a huge part of keeping a blade running stable and effiient even at stock speeds.


Also, whatever you do, CHECK YOUR WORK. Always make sure that all parts have been installed correctly, and never EVER power up a blade with improperly installed or missing components. That's when disasters can happen, and it's easy to avoid.

Thanks man that is exactly what I was looking for!!!!

These guys on eBay have a decent selection of chip-sized heatsinks and ship from the US.

http://www.ebay.com/usr/diyfancylights

Hey where did you get those cool little heat spreaders on your chips? I have 1 blade with heat issues and Im thinking this may help.


Thanks

That sounds good. Please PM the snail mail address info I need to send the board to, as well as total BTC and a wallet address for your fees.

Art

The best luck I've ever had was with a three-point warp to the stock heatsink. Basically, laying it across two wood blocks and hitting it with a hammer through another wood block in the middle, and then halfway between the middle and the ends, to give the heatsink a fairly continuous slight curve - less than 1/8 inch total deviation, probably closer to 1/16 inch. When the board is screwed down over it, the backside of the board will flex slightly to press up against the heatsink, which will allow better heat transfer. I also used some thermal compound between the heatsink, pad and board and buffed the heatsink face smooth - the ones I've seen have been very rough, so not a good contact surface at all.
A three-point warp is good because it leaves a fairly continuous curve to the heatsink surface - just a single bend in the middle will make more like a triangle, so when the board flexes (into a continuous curve) it won't mate with the heatsink face evenly and could leave gaps under some rows of chips.

Between that and the simple chipside heatsinks in the kit, I've had a blade running at 14.5GH with good airflow, more stable than some of the stock boards I've seen under the same fan. It's pretty much the best bet without having to machine a plate to screw over the boards for heat spreading and pressure on the sink.

If you want to send the board in, I can see about improving stock cooling and, if that works well enough, try pushing some more hashes out of it for you. If it is a voltage issue I can look into that too, see if the VRM outputs are noisy and add some filtering to clean it up.

I have a question. I have one blade that consistently gives me a high hardware error rate. I have noticed that if I lower the p/s voltage the error rate also drops, but below a certain voltage, some chips stop running. If I send this blade for the overclocking job, would you be able to fix that?

I believe that this is a voltage noise issue.

As an example at 11.80 VDC power supplied, I have one chip not running, and I still get 7% hardware error rate. At the normal 12.8 VDC the normal p/s (not adjustable) puts out, this one blade gives me 20+% error rate.

I have plenty of cooling and the heatsing does not get very warm, even without fans. Maybe I need to jimmy the heatsink a bit.

What are your thoughts on this?

Art

Something bad happened, that is all that I know. I took the resistor back off after I could not get it working and checked the voltage and it shows 0 volts with no resistor and the vrm seems to work after putting the resistor back, I see ~1.03 volts or so with the 68k in place and had 1.20 with 110k. It is an odd one and has me trigger shy on doing  any more.

Huh, wow that is weird. I can see low numbers at high clock without adequate cooling, but not put back to stock. Never had that problem.

I did have one guy didn't connect a resistor properly and it blew out a buck regulator chip down around the control hardware at the end of the board, never came back to life. I wonder if the extreme overvolt on yours (from the miswired VRM likely outputting full power in open-loop) might have somehow affected a logic input to one of the interface or firmware controllers? Sucks that happened.

I think that something definty got fried when it was powered up without the resistor. Anyhow even after returning it to stock it would only hash at 500 mhash max with good cooling.  Sold it on eBay for parts so not a total loss , but it still sucks. Maybe someone will have luck fixxing it or harvest the asics.

Likely what happened is, since the VRM wasn't getting feedback from the output voltage, it basically locked onto the highest voltage output and possibly fried your chips.

As for running at 300-500MH/s, that's probably a cooling issue. All the ASICs on these guys are fairly independent, and one or a few dropping out shouldn't hinder the rest. The thing is at higher voltage and higher clock, more power is consumed in the chips so they get warm, and the V2 blades are legendarily bad at stock heatsink contact. Chipside heatsinks (as would be provided), and if possible a slight warp of the stock heatsink to pull the blade's board down firmly over it instead of floating with no pressure and very bad contact, will get these guys running well. I just the other day worked up a board to a 33% over-spec oscillator, that was running 14.5GH 99% efficient with nothing fancy but good airflow. Improving heatsink efficiency on these guys goes a LONG way. I've worked up about a dozen blades and had zero X'd chips that weren't from improper cooling.