Bandwidth required for FPGA mining. | Bitcointalksearch.org

TheSeven

hero member

Activity: 504

Merit: 500

FPGA Mining LLC

Quote from: wildemagic on May 21, 2012, 07:42:53 PM

IF latecy is the issue (i use ozco.in) then what are we talking?

My first hop pings are like 100-150ms most of the day (3G wireless internet in remote australia)

kind regards

More latency => more stales. But that order of magnitude should still be fine.
On a regular pool, 200ms of total roundtrip to the pool cause 0.03% stales on average.

wildemagic

member

Activity: 112

Merit: 10

Thanks mate, quite informative.

Just swapped out a nv550ti for an ati5850, still not generating much traffic.

But I have a couple of quad spartan boards on the way, looks like I should be all good.
Total hashing power will be under 2Gh so I wont be generating much traffic by the looks of it.

kind regards

pieppiep

hero member

Activity: 1596

Merit: 502

I just monitored a little bit with wireshark, it's about 1kB per request to the server.
So if you mine at 4GH/s you have about 1kB to get work and 1kB to report work each second.
If you mine at really lower speeds you probably get a little more data per hashing power because you don't try the whole block.
At 800 MH/s it is about 5 seconds per block so it probably won't matter, but my videocard is as slow as 60 MH/s and would take a little over a minute to complete a block and thus would generate a lot of stales.

Conclusion, at 800 MH/s I estimate it at 2 kB / 5 seconds, 86400 seconds a day, 17280 times 2 kB, 33.75 MB/day.

wildemagic

member

Activity: 112

Merit: 10

IF latecy is the issue (i use ozco.in) then what are we talking?

My first hop pings are like 100-150ms most of the day (3G wireless internet in remote australia)

kind regards

TheSeven

hero member

Activity: 504

Merit: 500

FPGA Mining LLC

This highly depends on which pool you're using.
For pools that support roll-ntime, the traffic is dominated by share submissions. Those occur every 4 gigahashes on average for a difficulty 1 pool, and usually cause around 500-1000 bytes of traffic, depending on what kinds of overhead you count.
So while latency is somewhat critical for mining, actual bandwidth isn't. We're talking like 1KB/s of traffic per 4GH/s here. If you're really constrained, you could try to find a pool with a higher difficulty. That could bring traffic down to almost zero.

mufa23

legendary

Activity: 1022

Merit: 1001

I'd fight Gandhi.

Quote from: P_Shep on May 21, 2012, 03:44:25 PM

This much:

--->| |<---

You are wrong. I know for sure that it is --->| |<---

wildemagic

member

Activity: 112

Merit: 10

Quote from: likuidxd on May 21, 2012, 05:38:13 PM

Just took a gander at one of my routers that is connected to 8GH/s
2492 TxB/s, uptime 2 days 5 hours

~6 B/s per GH I guess

Great news, thats tiny.

kind regards

likuidxd

sr. member

Activity: 476

Merit: 500

Just took a gander at one of my routers that is connected to 8GH/s
2492 TxB/s, uptime 2 days 5 hours

~6 B/s per GH I guess

edited for retardation, I need to double check my work after work. It's probably still wrong Roll Eyes

wildemagic

member

Activity: 112

Merit: 10

Ok, thanks for your input ppl.

I dont have a data limit, just an upstream limit.

Still, if anyone has some actual numbers, that would be great.

Can someone put DUMeter or Netmeter on one of their host boxes and log the data for me plz?

kind regards

Lethos

sr. member

Activity: 476

Merit: 250

Keep it Simple. Every Bit Matters.

It doesn't use huge amounts of bandwidth, but as I understand it, it will be in constant use.
So it will add up, I certainly wouldn't want to be on a bandwidth plan with a low cap, just to be on the safe side.
Speed is not really an issue, as what it sends is tiny bits of information, it just does it very regular.

BR0KK

hero member

Activity: 784

Merit: 500

I had my FPGA Cluster at a UMTS Router for a month .... The limit (1GB -5GB <-- not sure) was exceeded after that month. But it worked well with low speeds (56K Down and 14K up

)

aqrulesms

sr. member

Activity: 373

Merit: 250

Quote from: wildemagic on May 21, 2012, 03:50:07 PM

Well from my own 50M/hash miner I deduced it was at least a few megabytes per hour, but I do have other processes going.

Just wondered if anyone had some actual stats, as my upstream is limited to 40-50Kbytes a second, I dont want my upstream to limit the amount of work I can process.

kind regards

It's fine. Bitcoin mining does not use much bandwidth at all. The amount it uses is not a concern with those speeds, and especially if you are not charged for how much data you use or caps by the ISP.

BR0KK

hero member

Activity: 784

Merit: 500

The same as for GPUs?

wildemagic

member

Activity: 112

Merit: 10

Well from my own 50M/hash miner I deduced it was at least a few megabytes per hour, but I do have other processes going.

Just wondered if anyone had some actual stats, as my upstream is limited to 40-50Kbytes a second, I dont want my upstream to limit the amount of work I can process.

kind regards

P_Shep

legendary

Activity: 1795

Merit: 1208

This is not OK.

This much:

--->| |<---

Dalkore

legendary

Activity: 1330

Merit: 1026

Mining since 2010 & Hosting since 2012

Quote from: wildemagic on May 21, 2012, 03:42:15 PM

Ive poked around a bit, but cant really find any numbers. Im just wondering what kind of amount of data an 800M/hash miner generates.
(or any size if you have the stats)

Kind regards.

Barely anything from what I have read and asked a few people with knowledge of the process. Basically you are getting a math problem, crunching it locally and then sending out an answer. Correct me if I am wrong please.

Dal

wildemagic

member

Activity: 112

Merit: 10

Ive poked around a bit, but cant really find any numbers. Im just wondering what kind of amount of data an 800M/hash miner generates.
(or any size if you have the stats)

Kind regards.

Topic: Bandwidth required for FPGA mining. (Read 3626 times)