Topic: GPU working at half speed... Opencl and ADL mismatch? (Read 7670 times)

Yeah, very true.


Well.. matter of taste I guess.
I personally see no reason why s1 would ever use 95/98/me/xp/vista/7/8 ; )
Especially on headless mining rig.

All unusual or specific situations, I guess. There are always exceptions. I don't see why anyone would mine on Server 2008, unless you actually had some nice cards in a real server that had hardware which truly required that OS.

Yes they do ; )

1) Windows Server 2008, clean install, any driver version - no GPUs work at all without dummy.
2) Asrock 970 extreme3 r2.0, Windows 7, my friend was having issues until he plugged dummy in one of the cards, can't provide details atm.
But display connected/dummy forces Windows to make that GPU primary/master and sometimes it matters.
3) nVidia + AMD cards, Windows 7, unless dumy is plugged into AMD card it won't work

Dummy plugs have no use as of the last few driver 1x. versions in windows. Not sure about other OS.

I'm having the same issue. The 3 cards were working fine when I had a monitor hooked up via HDMI cable inside my office. Now I had to move the rig outside my house b/c the A/C was competing with the heat and working nonstop so all I did was to install a wifi card and took the rig to my storage unit behind my house. Now I don't have a monitor to plug into it but I created 3 dummy plugs and still doesn't fix the issue with the 3rd card running half the speed. Without the dummy plug, I wouldn't even get the temperature to show up so that fixed that but why can't the 3rd card run full speed?

setx DISPLAY ":0"
- no need for dummy on w2k8 r2

Was getting ADL mismatch and GPU-Z "OpenCL bug detected" messages because Windows installed drivers from auto update, then even after ATI driver installation it used those 'old' libraries -> disable installing drivers from online and delete amdocl*, opencl* from System32/SysWOW64; I've also deleted all *amd*.* from sys32/wow64.

Yeah it's very irritating as you can imagine. I made two dummy plugs today, when I started the rig I saw in cgminer that the errant card seemed to go up to full speed for a moment however the whole rig shut down then. I've tried to start it again and one of the cards just shuts the system down with a dummy plug attached. the resistor wire seems a little thin I might try to get some more tomorrow.

Thanks for trying to help Kaz, I'm trying to set up linux on a USB stick to see if this helps..why does your setup need a dummy plug do you know? I think I read on a guide to setting up mining on linux that it wasn't needed sometimes... I have tried googling the issue believe me

I have the same issue when starting the rig without monitor connected or monitor connected to card which is not 1st to initialize in BIOS.
So.. when starting rigs I plug the monitor in.. this is a crappy solution, however it works.
I have not tested dummy plug yet.

Also I have the same issue on PC with nVidia and ATI GPUs installed, however it does not affect mining performance.

I dident think it was... I haven't set up a miner using win 7, all of mine are linux hosts... My setup requires Dummy plugs, but I can't say what you need to do in your situation...

Some times my 7970 would start off running at about 1/2 speed but after about 30mins they are back up to full performance..

Sorry I cant help you further, I have always found google to be my best friend in times like this.

Good luck!

It's not DOA, one of them is only running at half speed for some bloody reason. It only gave that error message once too.
I think that they all work fine in isolation. I will test them individually tomorrow again.
It's win7 on a clean install... I read somewhere about inserting dummy plugs, could that be the answer? 

Hi,

I have had the same issue:

"ADl found less devices than opencl!
There is possibly more than one display attached to a GPU
Use the GPU map feature reliably map Opencl to ADL
WARNING The number of Opencl and ADL devices did not match!
Hardware monitoring may not match up with devices!"

But my card was DOA... I am not sure what the issue with you card is, have you tried running the card by itself?

If the card is working as expected by itself I would try reinstalling the OS.

By the way what OS are you using?

Hi guys, I just set up my first small rig over the last couple of days, for fun more than profit really but I have a problem, I suspect/hope this may be a simple hiccup for someone who knows cgminer better than I but, anyway:

I have installed 3x 7950s in a small rig. One card is in a PCI-E x16 slot and two are in risers from PCI-E x1 slots. The two in risers are performing fine, mining at over 500 Mhash apiece. However the one in the 16x is stuck at half speed for some reason. I can't alter the clock speeds within cgminer (I can with the other two).
I tried shuffling the cards around to see what that did and I got an error message which I think preceded the x16 card going to half speed ( I think it had been working fine in isolation).

"ADl found less devices than opencl!
There is possibly more than one display attached to a GPU
Use the GPU map feature reliably map Opencl to ADL
WARNING The number of Opencl and ADL devices did not match!
Hardware monitoring may not match up with devices!"

Anyone have any idea how to get the card working at full speed?

I had logmein installed and thought it might be something to do with that but uninstalled it and no change.

I read this on another thread:

jake262144: "The root cause of the change is that AMD drivers are composed of loosely connected sets of technologies:
ADL - used for clocks, voltage temperature and fan control - being hardware-bound, sees only one device.
OpenCL - the language used for GPGPU ops - sees one device per each connected VDU.
Unfortunately, no one has been able to reliably tie ADL devices to their corresponding OpenCL virtual devices or in some scenarios (some systems using multiple double-GPU cards) even to the correct physical device!
I'm not exactly sure this matter requires any action on your part but if you wand to do something about it, one of the virtual openCL devices can be disabled with disable-gpu.
The number of threads per GPU core can be specified with gpu-threads.
It's all pretty well documented in the HOWTO."
 
I understand this is connected but I don't understand how to rectify it myself if anyone would be able to help out I'd very much appreciate it