Topic: The Habanero - 650GH/s - OOS - page 48. (Read 96043 times)

Oh, oh, oh .... one thing I forgot to ask earlier.  Will the standard Hashfast MinePeon image work with these Habanero boards?

H@shKRaker

Hola Sidehack.  Yea, the KillaWatt says That Dell PSU is doing 950 at the wall.  As a side note I have the PSU's fan running at max speed.  That kinda leads me to believe the PSU's current trip is partially influenced by heat build inside the PSU case.  On that current sense resistor thing .... are those the two metallic looking resistors nearest the PSU exhaust fan?  If so I messes up the other day and decided to touch them while it was running.  D@MN they were hot.  Heck, I almost pissed myself they were so hot.  One thing 'd think about when doing kits for the habanero is maybe making the PCI-e wiring 14AWG as opposed to 16AWG.  This way any heat build due to resistance will NOT happen under that PCIe shroud OR along the wiring.  I've had that happen before on an OCZ PSU and it melty-melted and a certain amount of magic smoke got out.  In any case, you got a GREAT product.

H@shKRaker

Oh hey there H@shKraker. I'm actually this afternoon trying to find out more about the viability of using our D750 kits with these boards. Everything looks to me like one PSU kit and one habanero is about a perfect matchup. I haven't run these supplies to 950W myself but I'm not surprised they do it; yesterday I was messing with loads and found the max current tripped between 77 and 81A depending on the specific PSU model. Watch your current when you get into high powers though, the board is only rated for 80A sustained and past that you risk damaging the current sense resistors.

Make sure you slap some fans on that radiator that have a high static pressure and I think you're good to go.  I've got some Scythe Typhoons that I use on my radiators for PC Watercooling but if noise is not an issue some good server fans will work.

Hey guys, I was thinking of getting a couple of these and cooling them with Corsair H110 coolers and powering them with either Sidehack's break out board and Dell D750 PSU (which I know will successfully supply 950watts because I'm doing that right now on a fully populated BitFury M-Board gear PLUS a S1) *OR* one of the HP DPS-800 units (which will also do 1KW).  Any reason I should run away from this idea?  Many thank for your reply(s).

H@shKraker.

@PatMan:  Any news on this? Heard anything yet?

UK/EU need a reseller/official distributor........

Only the highest test point on their chart required more than 750W, which I know those 2950 supplies can actually source close to 80A before overcurrent protection cuts out and 783W should be no trouble for them. Have these boards been tested beyond the 783W point?

Yep, I get that.  My problem is density.  I've got really really cheap power as long as I can get my miners into 24U of space.  I was doing some back of the envelope calcs and figured I could get a couple of these boards into some old Dell 2950 servers I have lying around.  So nice and compact, plenty of cooling (80mm rads with liquid cooling) and 2 built-in 750W 90% eff psu's.  I thought that if I could cram 12 of these running at only 650gHs per board (1.3TH total per case/ 15.6TH in 24U) or so... I'd be a happy camper at about 1.5K CapEx (each box) less then a spondoolies sp10.

Hence my question about power consumption...

Or maybe I'm missing something obvious...which is very possible...


BTW...nice John Galt reference...huge Ayn Rand fan!

The last boards that MrTeal and ChipGeek produced were officially rated at 32GH/s but many of those boards would do 38GH/s consistently. Assuming you can effectively push it then it would kind of suck to be limited by your PSU. With 1000W you have plenty of headroom.

This question may show my ignorance...but given the <1watt /gHs, could you power this board with a 750W server psu (accepting that you are underclocking it)?  The web site says that a 1000W psu is recommended.  Why?

Any updates on power suggestions?  Recommended rad size, fan CFM?

Helpful info!  I placed an order.  So hopefully I can test VMC's board with the Habanero and post results (in another thread).

Bit of a small update, this week will be extremely busy so bare with us on any replies.

https://peppermining.com/busy-week-ahead/

I have a 110 the mounting bracket is kinda cheap and bends, but well anything you will be placing on the board when mounting the cooling head you will see the ends bend as you tighten it down to get proper cooling for each die.  One thing I have learned is the stock fans that come with these coolers do not push enough air though the coolers to keep the chip cool.  I'm hoping to do a bit of a write up about this today or tomorrow (monday), I'm also sure Mr. Teal will add some comments based upon his engineering background.

I could be wrong here since I haven't received my VMC versions yet (tomorrow) - but I think the VMC version also has milling for the caps to sit in. I will let you know.

Thanks Mr. Teal. That's what I was looking for. I'll place an order after I hear how the first batch is performing. I have a lot of antminers to replace.

That is interesting, there wasn't any reference to an included one on the VNC webpage.

The difference is that is cut out of a piece of plate and you use a pad to allow it to be pressed against the capacitors, with the majority of the force pushing down on those delicate ceramic caps (unless they've changed it recently).

https://www.dropbox.com/s/vddqqeyg0fx5kfr/heat%20sink%20back%20plate%20pic.htm

With our design, it is much larger (almost a full 120mm square) and is in contact with the board itself to help dissipate heat from the board. That's done by milling out pockets under the ASIC to allow the decoupling caps to sit in. This also puts the stress of the mounting pressure on the board instead of on the relatively fragile ceramic caps. The backplate itself is screwed to the board using 6-32 screws with a 105mm hole spacing, so you can mount a standoff and easily install a 120mm fan blowing on the bottom side if you don't have them in case with airflow around it.

What do you guys think?

http://www.corsair.com/en-us/hydro-series-h110-280mm-extreme-performance-liquid-cpu-cooler

or

http://us.coolermaster.com/product/Detail/cooling/glacer/glacer-240l.html

Ignore the crazyass aircooler, but the aluminium block is provided with the VMC board AFAIK. Is the Habanero's block similar or what are the differences?

Well, the VMC board is the Evo board.

There are a few software improvements that we're porting from the Chilis such as having the ability to auto clock the voltage and frequency to max out your thermal headroom. This is probably even more important with these boards since they tend to be thermally limited more than anything.

Hardware wise, most apparent difference other than the board being slightly larger will be our addition of a custom backplate to both sink heat from the VRMs and board in general as well as to support the chip. Make no mistake about it, dissipating 600W from a chip isn't easy, and you need a lot of pressure to get good enough contact to do that. The stock backplate that ships with most cpu coolers isn't up to the task of preventing board flex in really high pressure installs and you will get board warp. Our backplate bears right on the PCB under the chip to support it with minimal flex. The added bonus is that at high hashrates even an efficient VRM needs cooling. Having a heatsink added helps reduce the amount of air you need to move to do so, increasing efficiency.

Outside that, there's the intangibles. We'll be burning these in long enough to test that they function as expected, and then shipping them. They're not pulls from a personal farm and you don't have to hope that we're pulling a random one instead of just pulling a weak performer and sending it to you. Then there's customer service; we do screw up and miss an email or message now and then but our primary focus is on making, selling and shipping boards. We're not going to sit on our "in stock" boards for two+ weeks before shipping them because we're too busy trying to figure out how to run a mining farm. I won't lie on the specs; while we quote the DC power of the board since we're selling raw boards and the wall power will be installation dependent, those are measured results. We say 650GH/s because we're confident that given proper cooling the boards will all get there. This isn't authoritative as I've only had a few samples, but I think even 700 should be achievable for most people with good cooling. However 750GH/s will probably be very hit or miss, and 750GH/s at 0.833J/GH is a blatant lie.