[.05 BTC Bounty] CGMiner Hash Rate vs Pool Consistently 50% at Pool

bmoconno

sr. member

Activity: 280

Merit: 261

New In Town...

Quote from: Reckman on January 11, 2014, 11:20:37 PM

bmo and mikeyfinn had answer

it was thread-concurrency, but I had messed with it before and It caused crashes, I think removing the BINS helped

I'll send bounties

thanks

Thanks! Glad you're up and running without HW errors.

mikeyfinn

sr. member

Activity: 266

Merit: 250

Thanks much, and glad you are up and running!

Reckman

hero member

Activity: 711

Merit: 500

bmo and mikeyfinn had answer

it was thread-concurrency, but I had messed with it before and It caused crashes, I think removing the BINS helped

I'll send bounties

thanks

mikeyfinn

sr. member

Activity: 266

Merit: 250

"-I 17 -g 1 --thread-concurrency 12000 --lookup-gap 2" that should get you ~450 Kh/s,
hopefully with no HW errors.

mikeyfinn

sr. member

Activity: 266

Merit: 250

http://pastebin.com/1BfDXwS7

Reckman

hero member

Activity: 711

Merit: 500

Reckman

hero member

Activity: 711

Merit: 500

they are regular 270 cards not 270x

running stock still gives some errors and shitty WU

Code:

"intensity" : "13,13,13,13",
"vectors" : "1,1,1,1",
"worksize" : "256,256,256,256",
"kernel" : "scrypt,scrypt,scrypt,scrypt",
"lookup-gap" : "2,2,2,2",
"thread-concurrency" : "4032,4032,4032,4032",
"shaders" : "0,0,0,0",
"gpu-engine" : "0-0,0-0,0-0,0-0",
"gpu-fan" : "0-85,0-85,0-85,0-85",
"gpu-memclock" : "0,0,0,0",
"gpu-memdiff" : "0,0,0,0",
"gpu-powertune" : "0,0,0,0",
"gpu-vddc" : "0.000,0.000,0.000,0.000",
"temp-cutoff" : "95,95,95,95",
"temp-overheat" : "85,85,85,85",
"temp-target" : "75,75,75,75",
"api-mcast-port" : "4028",
"api-port" : "4028",
"expiry" : "120",
"gpu-dyninterval" : "7",
"gpu-platform" : "0",
"gpu-threads" : "2",
"hotplug" : "5",
"log" : "5",
"no-pool-disable" : true,
"queue" : "1",
"scan-time" : "30",
"scrypt" : true,
"temp-hysteresis" : "3",
"shares" : "0",
"kernel-path" : "/usr/local/bin"
}

mikeyfinn

sr. member

Activity: 266

Merit: 250

Did you say a gigabyte 270x? if so, try this for options:
-I 19 -g 1 -w 512 --lookup-gap 2 --thread-concurrency 15232 --gpu-engine 1130 --gpu-memclock 1500 --gpu-fan 85
this is what i use and i get ~480 with no HW errors

damiano

legendary

Activity: 1246

Merit: 1000

103 days, 21 hours and 10 minutes.

Im using a R9 270x as a backup atm and your config is way to aggressive.

I do about 390-414k

I 15

powertune 0 then 10

try it out

bmoconno

sr. member

Activity: 280

Merit: 261

New In Town...

The discrepancy you're seeing in cgminer's reported hashrate and the hashrate reported by the pool is a direct result of the HW (Hardware Errors) that your cards are producing. The number of HW should always be zero when scrypt mining.

I'd bet dollars to doughnuts that all those hardware errors are the result of running Powertune @ 20 on all four cards.

Here's how I would resolve this problem: (NOTE: IGNORE YOUR HASHRATE WHILE WE TRY TO REMOVE HW ERRORS)

At the command prompt enter the following commands:
- setx GPU_MAX_ALLOC_PERCENT 100
- setx GPU_USE_SYNC_OBJECTS 1
Edit your .conf file:
- Change "gpu-powertune" : "20,20,20,20", to "gpu-powertune" : "0,0,0,0",
- Remove the "shaders" : "1280,1280,1280,1280", line entirely.
- Change "gpu-engine" : "1050,1050,1050,1050", to "gpu-engine" : "900-1050,900-1050,900-1050,900-1050",
- Add this line: "auto-gpu" : true,
Delete any .bin files in your cgminer directory.
Restart your machine, then run cgminer with your newly edited .conf file

If cgminer is now running without HW errors, I would recommend (as others have suggested) increasing your thread-concurrency to 16000. If you previously had problems doing this, the variable commands (setx) I had you enter should have resolved that.

From there, I would either slowly iterate values for settings (gpu-engine/powertune) until you get your hashrate at an acceptable level with zero HW errors.

maardein

sr. member

Activity: 420

Merit: 250

I think your thread concurrency might be the problem. First, you don't need to set shaders and thread concurrency (tcc), since cgminer will automatically determine tcc when you set shaders. So, I would delete the tcc line, and keep the shader line.

Also lookup gap at 2, like suggested before, could be a good start.

ymer

sr. member

Activity: 308

Merit: 250

I'm pretty sure it's the pool, make a quick try on a different one.

_destro

newbie

Activity: 11

Merit: 0

Oh, how about this:

Set lookup-gap to 2.

Reckman

hero member

Activity: 711

Merit: 500

Reckman

hero member

Activity: 711

Merit: 500

Tried, running stock still having this issue, hmm wonder whats causing this o.O

_destro

newbie

Activity: 11

Merit: 0

Well, generally, HW errors means that the card is being pushed too much in some way. Lowering the intensity seemed to help a bit, but I'm suspecting the memory clock might be a little too high. Some of these R9s can be pushed quite a bit on the memory/core clocks, and some can't. I have a few R9 290s that are rock solid at 1500, and 2 others that don't like anything above stock.

Try this:

"gpu-memclock" : "1350"

If that seems to clear up the HW errors, and the WU is pretty close to the reported hashrate, I'd toy with bringing intensity back up to 19. I'd also try to slowly increase the thread concurrency. Many of these R9 270s can get up to 16000 TC.

Reckman

hero member

Activity: 711

Merit: 500

I have tried another pool,

look at the WU variance, the hash rate shouldn't be the same with WU of 50% more.

cozk

hero member

Activity: 812

Merit: 500

Try another pool.

The pool is estimating your hashrate.

Imho there is an issue on their side.

Cgminer is 100% accurate on your hashrate, pools are estimating it based on the shares you submit.

Change pool and it should be fixed.

Reckman

hero member

Activity: 711

Merit: 500

hmm this is bizarre I've never had anything like this

Reckman

hero member

Activity: 711

Merit: 500

Thanks for helping guys, I will tip for solution and help

Topic: [.05 BTC Bounty] CGMiner Hash Rate vs Pool Consistently 50% at Pool (Read 1494 times)