Confirming that I received my batch 2 today.
I agree with the $10/Day per B3 estimation.
No announcements for batch 3 yet, right?

That would be in line with the 90-100 Day ROI we will likely see on these units.
A3 ROIed pretty quick if you HEDL, with recent market push.

Ive been watching the Bytom master repo and they've been busy crushing bugs lately.
If they can deliver on their plan to introduce an API for specific timeless Tensor functions, this could be a big enablement tool for ML/AI.
Understanding the BYTOM project requires a bit of understanding around the domestic growth of ML/AI tech in China.

Also, did anyone else catch the bit about Tensority being in compliance with gov. encryption RAs?
Interesting.

Unfortunately we know nothing about this version of BMMiner and Bitmain has yet to publish the source used in this firmware.
I'd be curious to know if the parameters work when passed to BMMiner directly.

PS the running cgminer, copy the launch syntax, kill the process, and then launch the miner from the CLI using the fan, fan PWM, and freq flags in line. Can't hurt to try.

I can archive the filesystem and host the gzip if anyone's interested in poking around.

I ABSOLUTELY AGREE WITH YOU!  

Do you have your shipment information from them?
What does your tracking say?

I can't be the ONLY ONE who got alerts for batch 2.
I do use DHL and perhaps that makes a difference. As does the final destination.
My shipping destination is usually pre-cleared with customs before any hardware hits the shore.
Products usually arrive at my door within 48 hours of leaving Bitmain's facilities.

I believe the breakdown was 2500 for the Global market and 2500 for the Chinese domestic market. (Remember, BYTOM is huge in its domestic market. Their Global outreach is just now beginning to take shape)

Also, Batch two units are beginning to arrive. Mine is set to be delivered early (Tomorrow) for some reason.
Expect B1/B2 A3 ROI times here. I’d be surprised if we’re mining 25 BTM a day by next week.

It’s also worth mentioning that just because one hasn’t been made public, doesn’t mean a GPU/TPU solver (Tensor Processing Unit like Sophon/Movidius/Google TPU) isn’t out in the wild. The source and C implementation of Tensority has been published for a while now.  The Nvidia TPU cores on their Volta GPUs are far more powerful than even the Sophon units, so it’s possible cloud provider mining is profitable atm.

All,
Curious to know if anyone else is seeing the same behaviors as me.

1) No matter the pool configuration order, 0 and 2 are always marked dead.
Bmminer always chooses pool priority 1, regardless of the port.
Same behavior for 443, 6666 and 25.

2) Rejected share rates seem to be very high. Roughly 45% in my case. It doesn’t appear that my profitability has been affected at the pool, but I’d like some anecdotal feedback that others are seeing the same reject rates.

Does anyone know if these are all getwork ports at the moment?
The only pool source released by BYTOM is getwork-based, and no formal stratum protocol reference has been published.
I suspect Antpool plans on deprecating GETWORK support on ports 6666 and 443.



Cheers,
9600

My only concern with this, is any potential reason Bitmain shipped them without PWM control enabled. It’s clear that there shouldn’t be a need for this volume of air when dissipating 330W, but I’d be curious to see if the rejected share rate gets higher as the temperatures creep closer to 70+

(I don’t think BMMiner is reading HW errors in the traditional sense in this Tensority build.)

The miner seems overengineered.
Has anyone tried pushing the ASICs past 750 yet?
(The default config shows “bitmain-freq" : "750" in bmminer.conf.factory)

Either there’s headroom both ways, or the blocks on these ASICs start to bug out at higher frequencies/temperatures.

Who’s willing to figure out? 

PS - Anyone else get shipping confirmation for batch 2? Seems early...

I have a significant QTY of Sophon SC1 TPUs in deployment for our research customers.
It’s worth noting that the device presents itself to the OS as a Xilinx memory controller with 16GB DDR4.
Bitmain has detailed FP16/32 INT functions available to the Metal API with a theoretical max throughout of 2TFLOPS.

If anyone is interested in getting schedule SSH access to a host with an SC-1, please feel free to PM me.

I have a significant QTY of Sophon SC1 TPUs in deployment for our research customers.
It’s worth noting that the device presents itself to the OS as a Xilinx memory controller with 16GB DDR4.
Bitmain has detailed FP16/32 INT functions available to the Metal API with a theoretical max throughout of 2TFLOPS.

If anyone is interested in getting schedule SSH access to a host with an SC-1, please feel free to PM me.

I have a significant QTY of Sophon SC1 TPUs in deployment for our research customers.
It’s worth noting that the device presents itself to the OS as a Xilinx memory controller with 16GB DDR4.
Bitmain has detailed FP16/32 INT functions available to the Metal API with a theoretical max throughout of 2TFLOPS.

If anyone is interested in getting schedule SSH access to a host with an SC-1, please feel free to PM me.

I have a significant QTY of Sophon SC1 TPUs in deployment for our research customers.
It’s worth noting that the device presents itself to the OS as a Xilinx memory controller with 16GB DDR4.
Bitmain has detailed FP16/32 INT functions available to the Metal API with a theoretical max throughout of 2TFLOPS.

If anyone is interested in getting schedule SSH access to a host with an SC-1, please feel free to PM me.

+1

I have a significant deployment of Sophon SC1 and other commercial and non-commodity TPUs.
More than willing to provide access to a few units for testing.

9600

First off, thanks to carlosmnk for the RPi image.
After pulling it apart and dumping the redisdb file for credentials (wiibox:wiibox), I've got a working miner (with some quirks)

I've also to compiled the miner from github in an attempt to run USB straight from the serial breakout, but received the same "cannot open sfards port" error on each OS. This is different from the "cannot find sfards port" you will receive when trying to launch a nonexistant device.

I believe this is something related to the CP210x drivers, as the serial board uses an uncommon quad port PN.
They really are just giant versions of the USB Dualminers and Gridseeds of yesteryear.

Also worth noting, I've got two different hash board revisions across my three blades.

I'm traveling all week but will post detailed pics and a sanitized image with DHCP enabled when I'm back.

While these chips are too cool to completely ignore, i do see why they had to abandon the project.
The sheer number of parts it takes to build a hash board is staggering, especially considering that this miner was launched when both BTC/LTC prices were much lower, and had recently been through a nosedive to sub $400.

Damn shame.
Had to buy mine broken from Russia, they just never made all that many.

The miner in the demo video looks suspiciously like an old Zeus miner chassis.
Their setup documentation mentions the device as a network appliance, however the demo shows some type of USB interface. Not to mention the video shows no signs of actual hashing or pool interaction. Just an instant initialization and immediately available 75th.

The supposed pool credentials are shown. I don't think its outside of the realm of reality to contact slush to see if there's any share data from this user to corroborate the claim.

Thinking about this obvious scam way too much.

After trying for what seems like years, I finally got my hands on an SF100. (Yay...?)
The miner was shipped to me without the RPI, so I dont have any SD cards to image onto a new Pi.
If I load up a new Pi with Rasbian, and compile the miner source from the Git repo, can someone confirm that's enough to get one of these moving in the right direction?
I'm not sure if there's any software/firmware related to the breakout boards and their GPIO interface.
Thanks!
9600

Just wanted to note that I just purchased a few Quads from them over email this past weekend.
Already have tracking, and its stating delivery time from CHN -> USA is three business days.

I didn't even know about the group buy, but was afforded the 10% discount and an allowance for customs.
Pretty good CS off the line!

I've seen the posts on Reddit re: 28nm Baikal ASICs, but delivery looks to be Q3/Q4 2017 if that.
https://www.reddit.com/r/dashpay/comments/5rsv6c/baikalminer_new_28nm_chip_do_not_buy_the_40nm/

I'm okay with 280 Day ROI given we might see a run on Dash.
20% Returns are +1 in my book.

*facepalm*

That is all.

So this one's got me stumped.

setx GPU_FORCE_64BIT_PTR 0
setx GPU_MAX_HEAP_SIZE 100
setx GPU_USE_SYNC_OBJECTS 1
setx GPU_SINGLE_ALLOC_PERCENT 100
EthDcrMiner64.exe -epool coinotron.com:3334 -ewal user.worker -esm 2 -epsw pass -dpool stratum+tcp://dcr.suprnova.cc:9112 -dwal user.worker -dpsw pass -allpools 1

Hey Everyone,
So I've been deploying quite a few used Neptunes over the past few months and I've never run into a problem like this...

Picked up a used batch 1 Neptune with an extra comp cube thrown in to make up for some dead/missing cores.  
With the best tuning, I can get about 3.3-3.4TH out of all cubes, which is lower than the seller advertised but still isn't the major issue.

What absolutely kills me is the controller. It hashes for about ten or fifteen minutes until, quite suddenly, all cores are disabled and it ceases to hash. CGminer/BFGMiner is still running but the cores never re-enable and the hash-rate settles at 0 with no device activity.  

A cgminer restart will "fix" the issue, but I'm basically sending a remote SSH command to the Neptune every fifteen minutes telling it to restart the cgminer service which, for some reason, also stops the Web interface from being able to talk to the miner via the loopback interface, therefore losing the ability to tune/read any stats until rebooting.

Definitely not a solution.
So, has anyone seen this controller behavior?

Thanks in advance.

This is a clever April Fools...I hope.

Ended up being a ground wire that had worked its way loose! Good to go after a tear down.