6 x 7970 in 3 PC's - Need help optimizing (middlecoin)

[koldtoft]

I have 3 miners with 2 X 7900 GPU's in each.

Large Fractal Design Define R4 Black Pearl / 5 x 140 mm fans in each / 2 x 7900 in each. (7970)


System and GPU info (Sorry some is in Danish, but I think it makes sense for english speaking readers)
Windows 7 - 64 Bit


I use GUI miner, but I think I should probably use CGwatcher or some other tool that?

I think my stale are way too high. Like 3 to 5 % (even higher some times) and my hash rate is only around 600 khash/s (and even around 550 for two of the cards)



http://middlecoin2.s3-website-us-west-2.amazonaws.com/reports/17HjaGvN64UD28NMigjBbu7Ru6kHy26TZi.html

I seem to have a very high rejected %


So basically, I think I have too many stale, to high rejected % and too low a hashrate.

Any help would be much appreciated.

[Blaater]

I had the same, I dont think you can do anything about it. I switched pools after trying everything.
The same gear on other pools did not have the rejects and stales, so I blaimed it on the pool.

ps. where you from? Might want to try the EU server, was a little better for me.

[koldtoft]

I had the same, I dont think you can do anything about it. I switched pools after trying everything.
The same gear on other pools did not have the rejects and stales, so I blaimed it on the pool.

ps. where you from? Might want to try the EU server, was a little better for me.

I am from Denmark (Europe)

I point my miners on middlecon.com (port 3333)

Where can I find the URL for middlecoin EU server?

Edit: Found it - eu.middlecoin.com:3333

[guitarplinker]

Did the EU server do anything about your stale share issue?

[pontiacg5]

Multipools are routing your connection through multiple other pools. Switching between those pools/coins will cause some losses. So, basically, you are seeing compounding stales and rejects.  

I do not think there is anything to be done about that. Hopefully the extra pay from pool switching covers the stales!

It'll take some working, but you should really quit guiminer and use cgminer. You will be able to tweak things so much more effectively once you learn the software. It's really pretty simple though, and the scrypt-readme file is a great place to start for 7970's 

[koldtoft]

Did the EU server do anything about your stale share issue?

Well yes and no...

I have some less rejected and some less stale now, but my Mhash/s seem to have dropped significantly over the last 6-7 hours (since i changed to the EU server). H2O wrote something about middlecoin being ddos attacked, but I don't see any drop in the over all Mhash. Just mine. And my miners seem to be running fine (maybe a 10 % drop). The pool report just looks crap now. (like 40% drop in Mhash/s)

I am going to let it run the rest of the night and then switch it back tomorrow if it does not pick up, but right now it looks bad for the EU server.

Multipools are routing your connection through multiple other pools. Switching between those pools/coins will cause some losses. So, basically, you are seeing compounding stales and rejects.  

I do not think there is anything to be done about that. Hopefully the extra pay from pool switching covers the stales!

It'll take some working, but you should really quit guiminer and use cgminer. You will be able to tweak things so much more effectively once you learn the software. It's really pretty simple though, and the scrypt-readme file is a great place to start for 7970's  

Thank you. Yes I will. I don't like Guiminer. I was thinking maybe CGwatcher to manage CG miner?


Ps. My 3 hour Mhash/s average have dropped from 3.6 to 2.26. I don't know if it's switching to EU that did it, but it looks like it right now.

[pontiacg5]

The pool calculates your hashrate based on the shares submitted, it has no direct way of knowing your cgminer display in other words.

Hashing to a set difficulty (the way pools are set up) is subject to variance. Switching pools likely reset your hashrate, so now you are subject to variance till it averages out to your real hashrate. Give it some more time.

[koldtoft]

The pool calculates your hashrate based on the shares submitted, it has no direct way of knowing your cgminer display in other words.

Hashing to a set difficulty (the way pools are set up) is subject to variance. Switching pools likely reset your hashrate, so now you are subject to variance till it averages out to your real hashrate. Give it some more time.

I get that, but it just seems like a radical drop. The graph bellow is my mhash/s from my middlecoin page http://middlecoin2.s3-website-us-west-2.amazonaws.com/reports/17HjaGvN64UD28NMigjBbu7Ru6kHy26TZi.html


As you can see it's running around 4 mhash/s and then when I switch to the EU pool it drops... now I switched it back, and it's back up.

I do have 3 times as many rejected on middlecoin.com:3333 as I do on eu.middlecoin.com:3333, but the fall is hash/s scared me back to using the non eu server.

I have 15+ hours on EU server, how long should it take for my hash/s stats to adjust themselves?

[Blaater]

it seems like we have the same problem.

many rejects, much stales on the US server.
low speed on the EU server but less rejects.

I never found the solution for it, and I blaimed the pool for it .

[GanjaFarmer23]

Try setting up the eu server as main and us as failover. The eu disconnects now and then and is the reason the hashrate drops. Should give you same overall hashrate and less rejects.

[koldtoft]

Try setting up the eu server as main and us as failover. The eu disconnects now and then and is the reason the hashrate drops. Should give you same overall hashrate and less rejects.

Can I do that in Guiminer?

I have been toying around with CGwatcher (trying to get CGminer to run better than it does when controlling it with guiminer) and I can get 650 khash/s with that, but I don't think it's actually mining. I can see the hashrate, but the cards stay well below 55c and ,y hash rate drops quickly on the pool.

This is the command I used

-o middlecoin.com:3333 -u 17HjaGvN64UD28NMigjBbu7Ru6kHy26TZi -p 12345 -I 19 --thread-concurrency 25600

But I think that's very wrong...

[slaveofmoney]

Try setting up the eu server as main and us as failover. The eu disconnects now and then and is the reason the hashrate drops. Should give you same overall hashrate and less rejects.

Can I do that in Guiminer?

I have been toying around with CGwatcher (trying to get CGminer to run better than it does when controlling it with guiminer) and I can get 650 khash/s with that, but I don't think it's actually mining. I can see the hashrate, but the cards stay well below 55c and ,y hash rate drops quickly on the pool.

This is the command I used

-o middlecoin.com:3333 -u 17HjaGvN64UD28NMigjBbu7Ru6kHy26TZi -p 12345 -I 19 --thread-concurrency 25600

But I think that's very wrong...

if you dont see HW errors, everething is ok, but you can add

Code:

--scan-time 1 --queue 4 --no-submit-stale

at the end of line
example

Code:

-o middlecoin.com:3333 -u 17HjaGvN64UD28NMigjBbu7Ru6kHy26TZi -p 12345 -I 19 --thread-concurrency 25600 --scan-time 1 --queue 4 --no-submit-stale

and see result

[koldtoft]

Ok my bat file looks like this now

-o middlecoin.com:3333 -u 17HjaGvN64UD28NMigjBbu7Ru6kHy26TZi -p 12345 -I 19 --thread-concurrency 25600 --scan-time 1 --queue 4 --no-submit-stale

CGminer looks like this..  But I don't think it's mining anything at all.  
A:0  R:0  HW:0  U:0


Very Strange. If I point it at another pool, like a feathercoin pool it mines at a lower khash/s rate (like 590 to 600) and Accepted starts counting right away. (also the card get's hotter, rather than staying the same low temperature)

[pontiacg5]

Ok my bat file looks like this now

-o middlecoin.com:3333 -u 17HjaGvN64UD28NMigjBbu7Ru6kHy26TZi -p 12345 -I 19 --thread-concurrency 25600 --scan-time 1 --queue 4 --no-submit-stale

CGminer looks like this..  But I don't think it's mining anything at all.  
A:0  R:0  HW:0  U:0


Very Strange. If I point it at another pool, like a feathercoin pool it mines at a lower khash/s rate (like 590 to 600) and Accepted starts counting right away. (also the card get's hotter, rather than staying the same low temperature)

That is a horribly managed pool. If I'm not mistaken, your pool difficulty is set higher than the network difficulty which makes absolutely no sense. All those "network difficulty set to ###" prompts are the pool switching coins. It's likely it's doing that so fast that you never get the chance to send out a ridiculously high diff share off your low hashrate...

Are you using one worker for multiple miners? I'd quit that if you are. Hopefully vardiff will scale back the pool diff setting, unless there is an option for you to change it?

[koldtoft]

That is a horribly managed pool. If I'm not mistaken, your pool difficulty is set higher than the network difficulty which makes absolutely no sense. All those "network difficulty set to ###" prompts are the pool switching coins. It's likely it's doing that so fast that you never get the chance to send out a ridiculously high diff share off your low hashrate...

Are you using one worker for multiple miners? I'd quit that if you are. Hopefully vardiff will scale back the pool diff setting, unless there is an option for you to change it?

Yes, I am using 6 GPU's spread out over 4 PC's on one worker. But the "worker" is is just BTC address. People say you can do that for this pool.

Some have hundreds of miners pointed at one worker like this one http://www.middlecoin.com/reports/1M3jtksp1upR33SX1VzeCfH5x9fc6zKykR.html

And no i can not change the difficulty.

[guitarplinker]

Try adding "--scrypt" into your batch file before the -o command.

So, after the change your batch file should look like this:

cgminer.exe --scrypt -o middlecoin.com:3333 -u 17HjaGvN64UD28NMigjBbu7Ru6kHy26TZi -p 12345 -I 19 --thread-concurrency 25600 --scan-time 1 --queue 4 --no-submit-stale

Try and see if that fixes your problem.

[afr33sl4ve]

Using no submit stale causes you to not profit on stales submitted to Middlecoin! Middlecoin pays you for submitted stales.

[MoneyMorpheus]

For tweaking your hashrate I would try adding the following:

--shaders 2048,2048 --lookup-gap 2,2 -I 13,13 -w 384,384

Also if I raise the thread-concurrency beyond 8192 I get lower hashrate; and the correct way for two cards would be:

--thread-concurrency 8192,8192

(or whatever value you are trying)

If you wanna mess a bit with oc you can also use:

--gpu-engine value,value --gpu-memclock value,value

I use 1070, 1070 and 1500, 1500, (which are stock values on my cards) but it gets up to 40 degrees outside and the cards overheat if I try to demand more during summertime...

For temp safety add something like this:

--temp-target 75 --temp-cutoff 95

If any of the cards reaches 95 degrees, cgminer will shutoff, and the fans should try to adjust themselves to keep the cards running at 75

Also if you want to look what other options people look in here: https://litecoin.info/Mining_hardware_comparison

PD:

For setting the us as failover add to the end

--failover-only -o x  -u x - p x

Where x is the address, user and password. the options after the failover should appear before it with different values