Bitcoin Forum

Hello. I'm really hoping someone can help me with this one!!

Have big problems with my thread-concurrency. Cant get it over 8192, no mather what I'm doing. If I have intensity higher then 13, I get _a lot_ of HW-errors.

* Sapphire Radeon R9 290. (Others specs for Computer: 16GB Ram, Intel I5, 120SSD HDD, Win7 64bit).
* Using CGMiner 3.7.2.
* Have Win7 64 bit.
* Have installed Catalyst Center 13.12.
* Have installed AMD SDK 2.9.
* Yes, I have a screen connected to the card, and its on.
* Yes, I'm using the commands "setx GPU_MAX_ALLOC_PERCENT 100" and "setx GPU_USE_SYNC_OBJECTS 1".
* Yes, I have tried using a .BAT file.

I can get it higher in GUI-miner Scrypt, but I can't see HW errors there, and its seems to be unstable.

I have tried using all the config files I have found on the Internet (Been working on this for days..).


SOMEBODY??

Thanks for all replies.

Go into the directory where cgminer is and delete all the *.bin files.
Run: cgminer.exe --scrypt -u username -p password -o stratum+tcp://Yourpool
Let it run for a little and hit "Q"
Go into your cgminer directory and you should now have a file named something like scrypt130511Hawaiiglg2tc30592w256l4.bin
The "tc" number is your "thread-concurrency".

I removed the "thread-concurrency : "x", from my .conf file and started it. Also removed all .bin files.

I then got a binfile with tc4032 in it.

What does this tell us?


..By the way, my error message is as following, when I put thread-concurrency in my .conf file higher then 8192:

"Maximum buffer memory device 0 supports says 536870912"
"Your scrypt settings come to ...xxxx..."
"Error -61: clCreateBuffer <padbuffer8>, decrease TC or increase LG"
"Failed to init GPU thread 0, disabling device 0"
"Restarting the GPU from the menu will not fix this"
"Try restarting CGMiner"

I see you on windows, so the following account of my suffering the same symptoms with my sapphire r9 290 card are likely of no help at all, sorry.

had a similar problem with my setup, any TC above 8192 would just give errors or refuse to load on gpu , quating insufficient space on gpu mem.

tried deleting the cl bins, freshly build still refused to load,
tried fresh pull an compile of miner softwares, no help.
I tried numerous times reinstalling fglrx drivers and opencl ask and openal headers, no difference.
Tried putting it in a windows box, and it worked fine, grrrr, thats usless to me, i dont have a windows install myself, and dislike having to work around issues when i do encounter windows.
My stuffing around iwth drivers (ati's support for non windows OS is realy very poor after all.
 
But finaly I tried running it from local terminal window, no ssh, no screen command, and IT BLOODY WELL WORK NOW.
My script i normaly ran would screen->ssh->screen, keeping termcaps/envvars where i thought relevent, worked fine on my 7970 (my last gpu).

Damn thought that was the first thing i illiminated out the picture when started  debugging the problem, evidently not.

random tenuously related story i know.

When you start your miner, does it say something along the lines of "SUCCESS: SAVED setx GPU_MAX_...?

If it's to fast for you to see, just add timeout /t 5 to your bin file before the cgminer.exe.

Also, try using SGminer.

I'm having the EXACT same problem. 3x Gigabyte R9 290, Win 7 x64, SGminer 4.0, CGWatcher. I can't get the TC above 8192 without a buffer error. I'm not sure what "setx GPU_MAX_ALLOC_PERCENT 100" and "setx GPU_USE_SYNC_OBJECTS 1" are. I'm certainly not executing them anywhere. If I raise the intensity over 13, I get over 90% hardware errors.

Could this problem be related to the RAM we use i.e. how many slots etc?

Show your full settings..

{
"pools" : [
 {
 "url" : "stratum+tcp://xxxxxxxxx",
 "user" : "xxxxxxx",
 "pass" : "xxxxxxx"
 }
],

 "intensity" : "13",
 "vectors" : "1",
 "worksize" : "512",
 "kernel" : "scrypt",
 "lookup-gap" : "2",
 "gpu-engine" : "1010",
 "gpu-fan" : "50-85",
 "gpu-memclock" : "1500",
 "gpu-threads" : "1",
 "thread-concurrency" : "8192",
 "gpu-powertune" : "20",
 "temp-cutoff" : "90",
 "temp-overheat" : "87",
 "temp-target" : "85",
 "auto-fan" : true,
 "api-port" : "4028",
 "expiry" : "120",
 "gpu-threads" : "1",
 "hotplug" : "5",
 "log" : "5",
 "no-pool-disable" : true,
 "queue" : "1",
 "scan-time" : "60",
 "scrypt" : true,
 "temp-hysteresis" : "3",
 "kernel-path" : "/usr/local/bin"
}

Yes, the commands seems to work fine. I get the message you say ("SUCCESS: Saved...") - So that shouldn't be the problem...  :-\

I can try SGminer... But it doesn't solve my problem even if it works.

First thing I would suggest for you is to do is to restore your clocks to stock settings, i.e. 947/1250 if you got the reference card or 1000/1300 if you got the tri-x version.
After that, try with TC of 20481 and see if it works.

I know you probably don't want to hear it, but you are not getting GPU_USE_SYNC_OBJECTS 1 and GPU_MAX_ALLOC_PERCENT 100 if TC 4032 is the highest Cgminer can pull on it's own.

You should be able to see them in the registry when they are active. They pop up here, though searching is easier...

HKEY_CURRENT_USER/Environment/

If you are absolutely sure that setx commands are taking, you might have something causing drivers to crash. For example, I found that using the environment variable "setx GPU_USE_SYNC_OBJECTS 1" while the GPU fan is unplugged (as it would be, watercooled) will NOT allow the driver to "unlock," the GPU_USE_SYNC_OBJECTS variable is immediately reset, in other words. Cgminer will default to TC4032 in this case.

Get rid of worksize 512, it makes you look like an exceptional noob as no such command exists. 256 is the maximum. It's hard to start mining directly at 1500 memory, at least in my experience with over 30 290's. 2d voltage levels are too low, and the card crashes trying to use overclocked memory before the voltage is raised to 3d levels. Leave that out for now, you'll see what I mean if you ever get it working right, artifacts immediately on starting mining, but not if you swap to 1500 after a minute. Not all 290's will do 1500, most won't, stable.

TC 8192 is for a 7970 with 2048 cores x 4, a low TC meant for dual threads which you also aren't using, another dead giveaway of your noobishness. If you are going to mine, it would behoove you to at least read the manual (SCRYPT-README) or take more than 5 minutes to find a config file  ::)





 

Go to start menu, right click on computer, go to properties, then advanced system settings. Click on environment variables.  Make sure GPU_MAX_ALLOC_PERCENT 100 is in ALL CAPS.  If it is in lower case then that is what is causing this.

Yes, the problem is definitively with this little environmental variable and nothing else (drivers, OS, etc).
I had countless issues with this myself.

Eventually, I had to create a brand new windows user and add the following to "System \ Advanced \ Environmental Variables..." for that user

gpu_max_alloc 100
GPU_MAX_ALLOC_PERCENT 100

Case is important! I don't have 'GPU_USE_SYNC_OBJECTS' defined, it does not look like it is crucial at all - everything works great on my R9 290 using following settings:

-g 1 -w 512 --lookup-gap 2 --thread-concurrency 24550 -I 19

Could you tell how many RAM sticks are you using?
Are they of same brand and same memory and same speed?

This is totally irrelevant. The only problem here is this crazy variable - as mentioned by the people before.

THANKS EVERYONE. PROBLEM SOLVED!  :)

I got PM's and also messages in this thread which helped me.

Problem was the commands "setx GPU_MAX_ALLOC_PERCENT 100" and "setx GPU_USE_SYNC_OBJECTS 1".

Even if I write them with upper case in CMD, they still was saved with small letters at My Computer -> Properties -> Advanced System Settings -> Environment Variables.

When I clicked edit there, and changed them to big letters (upper case), everything started working!!! THANKS EVERYONE!!  ;D ;D



Next question: I get around 820KH\s, and I know its possible to get 870+\-. Anyone got a good config?   ???

Have changed Thread-Concurrency to 20481. That helped a little bit.

Anyone got a good config file for my computer?   ::)

120GB SSD. 16GB ram. Intel I-5. Sapphire Radeon R9 290 (21227-00-50G). Win7 64-bit.

I want the BEST one obviously. Maybe I will buy 2 more Radeon R9 290 cards, so configs for 3 of them is also appreciated  8)

Try the one I posted before:

-g 1 -w 512 --lookup-gap 2 --thread-concurrency 24550 -I 19

I easily get on > 800kh/s this config on R9 290, with no overclocking.

Hello miners.

I have just installed new 290 miner and I had the same error -61. After some time I figured out that bins in cgminer folder are causing problems but that was not all.

I have copied cgwatcher from my 280x miner and it was messing with the config, so if you have some old watcher or cgminer I suggest you go with clean download to see if problem persists.

I found that under clocking the core increased my khs. My stock core clock is 1040 on my Gigabyte R9 290, but saw improvement by lowering it until it reached 967, then it started dropping.

Try MSI Afterburner for this. start your miner and Afterburner. Then make tiny changes in Afterburner and watch the result in real-time in your miner. Keep going until you find the best core clock.

On TC, I suggest you use a multiple of your # shaders, +-1. Not sure why the +-1 helps, but it does.

Here's my config if it helps, it does 845khs.

"intensity" : "20",
"vectors" : "1",
"worksize" : "256",
"lookup-gap" : "2",
"thread-concurrency" : "30719", 

"gpu-engine" : "967",
"gpu-memclock" : "1250",
"gpu-powertune" : "20",

"temp-cutoff" : "90",
"temp-overheat" : "85",
"temp-target" : "80",
“auto-fan” : true,

"api-port" : "4028",
"expiry" : "120",
"failover-only" : true,
"gpu-threads" : "1",

"log" : "5",
"queue" : "1",
"scan-time" : "60",
"temp-hysteresis" : "3",

"kernel-path" : "/usr/local/bin"

my 290 cant increase intensity due to hardware errors
and cant increase tc due to some thing extrange i dont know :S

i have 32 bits ... can be this the problem?????

i can  only get 520 kh

Add to beginning of your batch file:
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1

Better if you start your own thread next time so people can easily see what you did and what was offered for you to do.