GekkoScience Terminus R808 Miner Official Support Thread

[meckanic]

I'm assuming 480GH is the desired aggregated speed from at least 4 units? That's about 275MHz, which is pushing pretty hard. Some people reported taller feet and under-board airflow helped increase stability. The actual limit on this board is thermal derating of the main regulator, which is the rectangular add-on board in the corner by the pot. Keeping that cool will also help.

Yes the fans on these are pretty crappy and I won't be using them again.

They say max rate is 480 gh/s per each unit. I have multiple units, it would be nice to operate at around one terrahash per pair. I noticed as well a lot of heating in the power regulation modules. I guess I am going to have to make a plan as I don't really like the thermal cooling system.

Cheers

[sidehack]

You were lied to, or you seriously misunderstood. Who did you buy from?

These are tested at 200MHz, with a practical max of 275MHz. With 55 cores per ASIC and 8 ASICs per board, that's GH=0.44*MHz for 275MHz clocking in at a theoretical peak of 121GH.

Under pretty ideal cooling and power conditions, the BM1384 ASIC is good to around 500MHz which means the absolute best case liquid-cooled-miracle performance for this device could only be 220GH, still less than half what you want.

(in case you're wondering, I designed and built the R808 so I know more than "them")

[meckanic]

You were lied to, or you seriously misunderstood. Who did you buy from?

These are tested at 200MHz, with a practical max of 275MHz. With 55 cores per ASIC and 8 ASICs per board, that's GH=0.44*MHz for 275MHz clocking in at a theoretical peak of 121GH.

Under pretty ideal cooling and power conditions, the BM1384 ASIC is good to around 500MHz which means the absolute best case liquid-cooled-miracle performance for this device could only be 220GH, still less than half what you want.

(in case you're wondering, I designed and built the R808 so I know more than "them")

Well, then you'd know ;-) I am working on something, will get back to you shortly.

As an aside, at 200mhz, I generally run around 80-100 gh/s per unit on CKPool, which seems pretty fair. Only problem I have ever really had with the unit, apart from not being able to cycle it up to higher rates, over heat constraints, is the regulator heating (which is why I use an additional fan).

Cheers

[meckanic]

Ok, I designed an oil/glycol/distilled water cooling system (I prefer oil as it's electrically non-conductive), dual 1/4" NPT inputs with a single 1/4" NPT output. I only had 1/4" aluminum hanging around for the contact plate, but I have 1" PVC and Acrylic that I could throw into the CNC quickly. Looks like by tonight I will have a working solution to seriously cool these units. I have a flow switch around here somewhere to shut down the units if the pump stops, and I have a spare radiator, fans, reservoir, pump, fittings, etc.

Will post pictures later today when I get the units ready. One question, how does this effect warranty, using a oil cooled system with a no-flow automatic shutdown cutoff switch?

Cheers

[sidehack]

Okay, but don't forget that the limiter is heat in the main regulator. Depending on heat, it could cut off at 15A or so, which is only about 6W per chip.

Also, that's a lot of finagling. I'd say once you're taking stuff apart it becomes "roast at your own risk".

[lok8nusa]

Would adding heat sinks to the regulator help?

[meckanic]

Okay, but don't forget that the limiter is heat in the main regulator. Depending on heat, it could cut off at 15A or so, which is only about 6W per chip.

Also, that's a lot of finagling. I'd say once you're taking stuff apart it becomes "roast at your own risk".

I hear you ;-) I was thinking of adding a side-sink just to cool the regulator area. It will be an afterthought though as I am dealing with the main 8 chips first, I can add-in on the side for the regulator. It would be nice if I could oil cool that section as well, but lot of hills and valleys to deal with. I have the cnc and a laser scanner I can use for 3D profiling, so maybe. The main parts are almost milled, one is done, but I just used a ball mill for now, doesn't have to be perfect and pretty for prototyping, just functional.

I didn't get a lot of setup instruction with these. I can use command line management to control many functions on the R808, but I didn't get anything on jumper settings or the pot function? Would you have that handy?

Cheers

[sidehack]

I'm gonna bet most everything you need to know is in the first post.

[meckanic]

I'm gonna bet most everything you need to know is in the first post.

I went through the first post when I got the r808's. I see the core voltage is controlled by the pot, that's fine. What do the jumpers do? I see no reference for them?

Command line control was easy for changing freq speeds and step up times, I just wrote a config file. Like I said initially, nice little unit, easy to manage, I just want to push the envelope a bit, get 275mhz down to an operating temp of around 35/40 degrees C and then run up from there. I can monitor actual core temp via the USB I believe. I would have to look up the command.

Is there any way to increase clock frequency above 275mhz other than command line or possibly changing hardware? Is there a firmware "governor" on the unit, so the frequency cannot be increased past 275mhz?

Cheers

[sidehack]

As mentioned before, the limit is thermal derating of the main regulator. If you can keep that cold enough to not trip out, you can set the speed as high as you want.

Temperature is not available on USB. Nothing is. USB is wired straight to a UART adapter wired straight to the ASICs. If you can tag into the fan header (or the PWM testpad) it'll give you a rough idea of temperatures, within limits - it goes from 50% to 100% duty cycle between 30C and 60C.

The jumpers are a header used to program the onboard micro which handles string lockup resetting, overtemp shutdown and throttling fans based on heat. You won't need to know about it unless you are planning to write a custom firmware.

If you've got more questions, I'd try searching first. This thread and the dev thread can tell you pretty much everything, including the answers to all of today's questions so far.

[meckanic]

If you've got more questions, I'd try searching first. This thread and the dev thread can tell you pretty much everything, including the answers to all of today's questions so far.

I have been searching the forum for a dev thread for the terminus R808, unfortunately haven't found one specifically. I can figure it all out, I have a lot of toys ;-)

Anyway, I figured I would use medium velocity liquid transfer (less than 0.25 seconds/cm) across the highest surface area I could get, I wanted very little pooling time. So far so good. Figured I'd throw them up on one of my sites that I just set up last night for other projects I am doing (you can see site's not complete, have a lot of text to add lol). Two different flavours, Acrylic and PVC, I think I'll cut the thickness in half. Might change to 1/4" copper plate rather than aluminum as well for oil usage, water creates too much oxidization with copper.

http://dal-tek.com/hs2.JPG
http://dal-tek.com/hs4.JPG

I think this should work pretty well with a 300mm reservoir, 100 gph high flow/low pressure pump and a 14" x 7" radiator with dual 6" fans. It would be nice to have the regulator board cooled as well, but I'll look at that tomorrow.

Cheers

[lok8nusa]

I would like to run mine in a tank of this:
https://www.engineeredfluids.com/bitcool

The 2Pacs as well.  I just can't justify it at this time   

[meckanic]

I would like to run mine in a tank of this:
https://www.engineeredfluids.com/bitcool

The 2Pacs as well.  I just can't justify it at this time   

Had a buddy with a micro-cray about 15 years ago. It was a little bit bigger than a standard desktop, it was filled with oil. Of course it was built that way, so all of the electrical connections were apparently sealed or soldered, it had it's own recirculating system. I never did take it apart though, too messy lol. You could possibly do the same thing with the r808's, but you will have to make sure that any connections are sealed or soldered. Just a secondary note, you'll still need a pump, radiator and fans since if you leave the oil in one place without cooling it will eventually exceed the temperature working limits of the chips.

Cheers

[dutchlincoln]

now what i wonder: if someone designs something like this; why not design a new wort of S9, built with wate/liquid cooled blocks, optimised to run silent? its the biggest problem for the most people: noise...

If we could fabricate these with watercooling we could stack many and have decent hashrates and silent operation...

[lok8nusa]

@meckanic: The coolant I linked to is dielectric as well as non-corrosive. Basically just drop them in the tank.

@dutchlincoln: I don't think the home-miner's biggest issue is noise; I believe it's heat dissipation. I can get noise levels down with a bit of engineering around the miner. The only ways to deal with heat are removal (fans/ducting, liquids & radiators, etc.), or turn the miner down (downclock). It's practically impossible to run S7 and above at normal settings in the living room/bed room/den with household A/C because of the heat during the warmer months.

[sidehack]

The heat and noise are both a factor of power density, which also has an adverse effect on longevity. Home mining's biggest issue is manufacturers who think it's appropriate to stuff multiple kilowatts into a cracker box.

[dutchlincoln]

well, i think that the heat dissipation is the base of the noise, and the noise is what renders it unusable in home applications.
Agreed, one is tied to another, but i do know that its definately not easy to get rid of the noise... Other than immersion cooling or water cooled blocks. So...
Those fans all together consume a huge amount of power as well, which can be used to cool or pump as well.

[meckanic]

The heat and noise are both a factor of power density, which also has an adverse effect on longevity. Home mining's biggest issue is manufacturers who think it's appropriate to stuff multiple kilowatts into a cracker box.

I am testing for noise levels today. One simple fix for pumps is to put them into a soundproof box with a pair of passive air circulation baffles, the foam is only around a $1 a square foot and the box is easy to make. I used the wall foam, 2.5" egg carton style, in my recording studio and at 2,400wrms, you couldn't hear anything in the control room, except through the control room monitor speakers. If anyone needs the design, just ask and I will draw it up and provide a public link.

There are ways to get around everything ;-) I think I am going to put these oil/water cooled heat sinks up on ebay, might as well let everyone get some since I did all the modelling and production of my own units.

Cheers

[sunk818]

I'd be interested in the box design. Would we need two fans to pull and push air out of the box?

[stefanh]

Just a heads up.  Do not underestimate the importance of a quality power supply source.

My two Terminus were just working fine for many weeks with the plug packs provided with them.  I was always concerned though that the plug packs were running rather hot.  I had attempted overclocking, however I could not get them to be stable above 200MHz.  Then just last week I noticed they were not hashing away any more.  The devices were constantly being restarted.  Inspired by one of the other members on here, I went and found myself a cheap PC power supply with dual 6pin PCIe power connectors.  The two terminus are now hashing away at 275MHz and look to be very stable, without one restart.

ps.  Also looking forward to the new design Sidehack is working on.