Profile cover photo
Profile photo
Marc Bevand
207 followers
207 followers
About
Posts

The Ordos Bitcoin mine (20% owned by Bitmain) was visited by journalists:

https://spectrum.ieee.org/computing/networks/why-the-biggest-bitcoin-mines-are-in-china
https://www.techinasia.com/inner-mongolia-bitcoin-mine
https://qz.com/1055126/photos-china-has-one-of-worlds-largest-bitcoin-mines/
https://qz.com/1054805/what-its-like-working-at-a-sprawling-bitcoin-mine-in-inner-mongolia/

The mine has 8 buildings and 25000 machines: 7 containing 21000 Bitcoin mining machines, and 1 containing 4000 Litecoin mining machines. IEEE Spectrum journalist Morgen E. Peck says the Bitcoin mining units were Antminer S9 units (each rated between 11 and 14 TH/s, let's say 12.5 TH/s on average). Therefore the mine operates at 21000 × 12.5 = 263 PH/s, which was about 4.4% of the global hashrate of approximately 6000 PH/s when Morgen visited on July 14th (https://twitter.com/morgenpeck/status/885981355817021446). This percentage is confirmed by the author: "By my own calculations, the hardware on the grounds—some 21,000 computers—accounted for about 4 percent of all the computing power in the Bitcoin network when I visited."

Tech in Asia claims the 8 buildings consume 800 megawatt-hours per day which is 33.3 megawatts (Quartz claims 40 megawatts.) Excluding the Litecoin machines, the Bitcoin machines presumably consume 7/8th of the total: 29.2 (Tech in Asia) or 35 (Quartz) megawatts.

A consumption of 29.2 (Tech in Asia) or 35 (Quartz) megawatts for 263 PH/s implies an efficiency of 0.111 (Tech in Asia) or 0.133 (Quartz) J/GH. For comparison, an S9 is rated 0.1 J/GH. Therefore this would make the aggregate PUE of the 7 Bitcoin mining buildings either good at 1.11 (Tech in Asia) or average at 1.33 (Quartz).
Add a comment...

Post has attachment

Post has attachment
The gPXE command "kernel <filename>" can't seem to boot the bootable binary images that memtest86+ provide (http://memtest.org/). It errors out with an obscure error 0x3d008001 and I have never been able to figure out why. I have tried all alternative combinations of imgfetch/imgload/imgexec/kernel. Anyway because of that error, for years I have always been recompiling memtest86+ to generate the ELF image, which boots perfectly fine using "kernel <filename>".

Today I was recompiling the latest memtest86+ (5.01) and the ELF image threw the same error when attempting to boot it!

/memtest5.01.elf: ok
. Error 0x3d008001
Could not fetch menu.c32: Error 0x3d008001
Could not boot from filename "/gpxe-1.0.1/script": Error 0x3d008001
No more network devices

An hour of debugging led me to find a bug in memtest86+'s Makefile. The linker line to generate the ELF image:

$(LD) ‐s ‐T memtest.lds ‐b binary memtest_shared.bin ‐o $@

Need to be changed to:

$(LD) ‐m elf_i386 ‐s ‐T memtest.lds ‐b binary memtest_shared.bin ‐o $@

(Beware, don't copy paste the command—I had to replace hyphens-minus (U+002D) with hyphen (U+2010) to prevent G+ from formatting the text as strikethrough.)

That is, add "‐m elf_i386". Or else the custom ELF loading address (specified as 0x10000 by memtest.lds) is ignored and set to 0x0. You can see that by running "readelf ‐a memtest" and checking the Program Headers section. Now I can boot the ELF image from PXE just fine.

The bug was reported to the developer.

Let's hope the details in this post will be indexed by Google and will save others from tedious debugging ;-)
Add a comment...

Post has attachment
BIOSTAR is also releasing a consumer motherboard with a very large number of PCIe slots: 12 of them. They have just announced the "TB250-BTC PRO" motherboard (not to be confused with the "TB250-BTC" that offers only 6 PCIe slots.) Strangely they announced it without showing any picture...

http://www.biostar.com.tw/app/en/mb/introduction.php?S_ID=884
Add a comment...

Post has attachment
ASRock released a consumer motherboard with 13 PCIe slots, for cryptocurrency miners. Impressive.

http://cryptomining-blog.com/8753-asrock-h110-pro-btc-mining-motherboard-with-13-pci-e-slots/
Add a comment...

Post has attachment
Bitcoin mining ASIC designer Canaan, who only has 2 competitors (Bitmain and Bitfury), just received a $43 million Series A investment:

http://www.coindesk.com/bitcoin-miner-canaan-raises-43-million-blockchain-ai-push/
Add a comment...

Post has attachment
I found this interview of Sam Cole (KnCMiner CEO), explaining that KnCMiner had to fold down because of the competition from Chinese miners and because their overheads costs of having data centers in Sweden is too high. Interesting. He says at least 2 times that had he had more capital, he would have built more data centers and been able to continue operating. But he never explicitly says this would have only been true if these extra data had much lower overhead costs.

https://www.youtube.com/watch?v=2BgjKOvTlRA
Add a comment...

Post has attachment
Add a comment...

Post has attachment
There are 2.9 to 5.8M users of cryptocurrencies worldwide. The field is estimated to employ more than 2000 persons.
Add a comment...

Post has attachment
Interesting account of the 30k BTC auction. Turns out the US Marshals forgot to include a fee when transferring the coins to Draper, so everyone was anxious until in took 1 hour to get a confirmation :)

http://www.coindesk.com/bitcoin-milestones-adam-draper-silk-road-auction/
Add a comment...
Wait while more posts are being loaded