Topic: [GUIDE] GridSeed 5-Chip USB, Blade & Black Miner Support/Tuning - page 10. (Read 308626 times)

Sandor I tried replacing COM numbers with serials in my BAT file but it didn't seem to work, it just sets everything to 850. My config:

minerd-gc3355 --freq=850 --gc3355-detect --url=stratum+tcp://multi.ghash.io:3333 --userpass=user.1:pass --url=stratum+tcp://xxxxx.com:3333 --userpass=user.1:pass --gc3355-freq=\\.\8D6E30845355:850,\\.\8D6E30845355:875:3,\\.\8D6E30845355:875:4,\\.\6D91458E5748:850:0,\\.\6D91458E5748:825:1,\\.\6D91458E5748:825:2,\\.\6D91458E5748:850:3,\\.\6D91458E5748:825:4

pause

EDIT: oops maybe I should remove the \\.\ prior to each serial... Yes that worked.

Nice work Sandor loving the new features.

GC3355 device detection + serial string display has been added for Windows.
Binaries are updated, current version is 1.0c
Edit: You can now also set frequency by serial instead of COM port.

Hey, I've been having recent loss of comm with stratum and it doesn't try to reconnect after it loses comm.
I have to reset the usb hubs by re-powering them or resetting them each individually via device manager. After that they run fine for hours then just crap out again. This only started happening a day ago. I have each hub powered separately each with 2+Amp supply which well more than they need for 26 miners pulling maybe 100 or so mA each.
I have noticed some of the usb plugs are not keeping seated / connected properly and I have to wiggle one or two of them to get them to connect better again but even if one fails it causes only 10 or so of them to quit working due to the daisy chain setup I have going.
I have 3 hubs each daisy chained off each other.
The main hub feeds into my laptop USB port where it's controlled via cpuminer.
Anyone having similar issues?
It's not the internet connection causing it because if/when it does go out and comes back up, the system simply reconnects and starts accepting / submitting shares again.
Weird.
Any ideas?
Thanks

Sandor, I tried this but it didn't work. What is wrong with it?

--gc3355-freq=\\.\COM2:1163,\\.\COM3:1163,\\.\COM4:1163,\\.\COM5:1163,\\.\COM6:1163 --url=stratum+tcp://dosumining.com:3000 --userpass=xxxx:xxxx

This should work: --gc3355-freq=\\.\COM20:1163

I have not tried this , but I believe you need to specify the frequency for each chip? Like so  --gc3355-freq=\\.\COM20:1163:0,\\.\COM20:1163:1,\\.\COM20:1163:2,\\.\COM20:1163:3,\\.\COM20:1163:4

IDK thought I saw that in some readme somewhere that may be worth a try

Hey, give me some credit, eh? I mean, I'm no longer a newb 
Yes, the standard cpuminer command line, as such, is:
minerdoc --freq=1163 --gc3355=\\.\com2,\\.\com4,\\.\com5,\\.\com6,\\.\com7,\\.\com8,\\.\com19,\\.\com20,\\.\com21,\\.
\com22,\\.\com23,\\.\com24,\\.\com25,\\.\com26,\\.\com27,\\.\com28,\\.\com29,\\.\com30,\\.\com31,\\.\com32,\\.\com33,\\.
\com34,\\.\com35,\\.\com36,\\.\com37,\\.\com38 --url=stratum+tcp://dosumining.com:3000 --userpass=xxxx:xxxx
That's all there is to it.

So how did you make the screenshot then? You had to specify the device names and the frequencies somehow?

You can look in Device Manager to find the COM port numbers.

Well shyt, Suchmoon, I actually guessed and tried --gc3355-freq=\\.\COM20:1163,\\.\COM21:1163, and it didn't work.
When run, I get the response, Mp GC3355 devices specified, please use --gc3355-detect o-detection, or manually specifiy with --gc3355=DEV0,DEV1,,,,,DEVn
I don't know how to find out what each miner's dev name or number is...
Do you?

Thanks for trying to help!
Woofy

What's wrong with --gc3355-freq=\\.\COM20:1163,\\.\COM21:1163,...  ?

Or just --freq=1163 if you are running them all the same.

Right, I realize that, but I beg to differ. That is exactly what auto-tune IS supposed to accomplish. Optimum clock rates for each pod. Then one is supposed to 'rather left with' having to record each setting where you don't see more HW errors accumulating and set each pod manually to that specified frequency. Otherwise autotune is a waste of time IMO.

I just found the sweet spot for ALL of my pods, at least where 98% of them are still at 0 HW errors after an hour now, where before they would be up over 30 or even 50...some out the roof!
Anyway, after re-modding 6 of my original pods for 5V regulated power feed from the 12V rail, I tested each one separately and found that most of them really like running at 1163MHz! The chips settle at 492MHs to 498MHs. Pool side sees that as the same or more and some of them are peaking at over 800MHs..I"m looking at one right now that's been hovering at 790 for several minutes now...Pool side of course...My pool side total hash rate is hovering at over 14500KHs! Yeah baby!  

Before it would hover at around 10 to 12MHs...oops, it just hit 15211KHs . I see it hit over 17MHs a lot too! I want ZERO HW errors so those otherwise wasted hashes get put to good use = mo-coin!  

One thing that still bugs me though is not being able to set the whole set of chips (5) to one fixed frequency. I see how one can use com in the command line to set each chip to it's own fixed frequency, so why is there not a similar com option to set all 5 together? It's simple enough to do but of course if that were totally true, I would have done it already, wouldn't I? Anyway, you get the point. I just wanna set each pod to it's own hash rate and that's that.
They are doing wonderfully at 1163MHz though, so I might just leave it there. Still, I'd like to know how it's done....Just in case

Woofy

Pea Ess, just in case someone doesn't believe my claims....have a looksee  

 

Autotune does not set optimal clock speeds in all cores of all Gridseeds. You've got to take the values given from autotune and lower/raise them a bit. Not all cores from my devices were set to their optimal speeds. Autotune is VERY, very good, but not perfect.

As for the rest I don't know. Don't have as many Gridseeds as you, and no problems here.

I think serial numbers may only be supported in Linux. Please correct me if im wrong

I just noticed something very interesting that may be a key bug in the latest and previous versions of cpuminer.
When I was testing one of my 5 chip miners.
I usually run 26 miners at the same time. Hardware errors are minimal and sometimes they get ridiculously high right after I fire up cpuminer then settle out.
I've also noticed very high HW error counts after a 24 hour run after autotune has supposedly set all the pods to their optimal clock speeds.
I don't think I'm getting as much performance out of the set as I could.

The reason being, that when I was running just 20 of the miners, *6 of them were disconnected for cleaning and to change the 5V Fan power feed from USB to the 12V regulator's 5V output rail), after plugging in just one of them for testing then running the whole batch of 21 miners, I notice that at the clock setting of 1175, virtually all of them stay at 0 HW errors! Maybe one or two of them hit 1 HW error but that's it!
When running just 20 of them, HW errors go out the roof at first then seem to settle out after several minutes but wow! I mean HW errors go crazy high! Like 20 to 30 inside of a minute or two!
Again, when I plug in pod number 21 and start the mining program again, HW errors go away almost completely. I mean 19 out of 21 stay at ZERO HW error!
I am wondering why? I mean, is it my USB hubs acting up? Doesn't seem like that would be the case but what else could be reacting to the number of USB ports running but the USB drivers themselves?
Perhaps this is caused by a bug in the way cpuminer handles the USB ports, adding up and averaging all the pods hash counts....
I don't know.
Time for an expert to look into this? Sandor111? What do you think it may be?
Thanks ahead of time for any assistance or ideas anyone may have.
Woofy

Wow! I had no idea it would be this complex much less in a foreign language 
But thanks for your help!

I see a newish commit for Sandor's cpuminer allows you to reference Gridseeds by serial rather than COM assignment. How do you find out the serial for each Gridseed? Thanks in advance.

Try using a json file, example create a file called "gridseed.json". Use sample below for a start:


Above configuration is for after you've done autotuning. Replace the frequency for each individual chip.

When running the cpuminer, GSD 0 > COM2, GSD 1 > COM4... and so on.

To run with json file, execute: minerd-gc3355.exe -c gridseed.json

Okay, let me put it a different way....

-f, --gc3355-freq=DEV0:F0,DEV1:F1,...,DEVn:Fn

Okay, so since it seems that I cannot use com in the command line to identify which DEV (device) I want to set to fixed freq (F), how do I find out what the DEVice number of each pod is?
There has to be a command for that, right?

All I need then it to find that out and I think I'll be good to go.

Anyone?

Hi Phil,
From reading everything here and running the autotune feature in cpuminer I hit 838 as the best.  No errors and stable.
Here are the best set of cables I have used.
Psu's  I have many   .   not sure about cables but I have a lot.    so try freq = 838  sounds good.  thanks for info.
[/quote]

[/quote]

Psu's  I have many   .   not sure about cables but I have a lot.    so try freq = 838  sounds good.  thanks for info.