Linux Driver for Blockchain Compute 17.30.1029 - page 2.

heavyarms1912

full member

Activity: 729

Merit: 114

Quote from: rvlad on September 23, 2017, 08:16:25 PM

Quote from: beastman on September 22, 2017, 11:59:20 AM

@CryptoSquad, found this for Linux UV dwarfing-master here https://github.com/rumatoest/dwarfing.

I haven't tried it as I am running back on Win10 for now. If you try it please post your results, should bring your temps down.

@rvlad, I had compiled the Torvalds kernel but it is missing the AMD firmwares and I don't know where to find the patches. Kernel works but can't get the GPU running due to missing firmware.

Please see the message from laik2 proposing to copy AMD firmware from AMD driver.

But the question was to xmonkx which linux from Torvalds git was compiled: 4.12 or newer?

what's the reason to copy the firmware from the AMD driver?

CryptoSquad

newbie

Activity: 60

Merit: 0

Great let me know!

heavyarms1912

full member

Activity: 729

Merit: 114

Quote from: CryptoSquad on October 16, 2017, 05:12:41 PM

Think there’s a miss understanding

I’ve run the command you showed and it just shows my core and memory clock of each card

To be clear on my question

In Windows I get 30mhs per card with my custom bios

I’m Linux I am only getting 29.5 per card with the same card. Seems like the Linux driver has a slightly lower hash rate than the windows driver. I’ve heard others having that same discrepancy

The command you showed doesn’t really help? Unless I am missing something?

The command I provided gives you the current clock rate. If your clock rate is same as that on windows (monitor for several minutes) then it might be just the drivers. This was happening for me when dual mining. I was hitting the power limit (had a lower TDP and Max power in my custom BIOS) and card was throttling at lower core frequencies -> leading to lower hash rates.

I'll check how much I get on mine for solo mining at same clocks on windows & linux.

CryptoSquad

newbie

Activity: 60

Merit: 0

Think there’s a miss understanding

I’ve run the command you showed and it just shows my core and memory clock of each card

To be clear on my question

In Windows I get 30mhs per card with my custom bios

I’m Linux I am only getting 29.5 per card with the same card. Seems like the Linux driver has a slightly lower hash rate than the windows driver. I’ve heard others having that same discrepancy

The command you showed doesn’t really help? Unless I am missing something?

heavyarms1912

full member

Activity: 729

Merit: 114

Quote from: CryptoSquad on October 16, 2017, 05:06:35 PM

But is there a way to keep them in One state? What am I monitoring for?

you'll be checking for the active core/memory frequency of the gpu while mining. Open up the terminal in linux and run the commands.

CryptoSquad

newbie

Activity: 60

Merit: 0

But is there a way to keep them in One state? What am I monitoring for?

heavyarms1912

full member

Activity: 729

Merit: 114

Quote from: CryptoSquad on October 16, 2017, 03:01:09 PM

Ok but why the lower hashrate in Linux?

I mentioned it in my previous comment. Your cards might be jumping back and forth low-high clocks in windows (more frequently) while in linux might be stuck on 7th state clock or jumping between low-high clocks less frequently (There are 8 gpu core clock states for amd gpu)

Are the cards BIOS modded/undervolted?

CryptoSquad

newbie

Activity: 60

Merit: 0

Ok but why the lower hashrate in Linux?

heavyarms1912

full member

Activity: 729

Merit: 114

Quote from: CryptoSquad on October 16, 2017, 12:18:08 PM

Hi All

Having modded my AMD bios's using Windows to about 30mh per card (180mh on a 6 card rig) I then setup linux with this driver, and i am getting only 177mhs

Does this sound normal? Difference of 0.5 mh's per card for linux?

One thing i did notice is windows every now and then would have a dramatic drop, and then increase back to the expected speeds, where as this seems far more stable at one speed?

Thoughts on all of this?

mostly the windows has more frequent state changes whereas linux isn't doing that. monitor gpu clocks using below commands for active core and memory frequencies.
cat /sys/class/drm/card0/device/pp_dpm_sclk | grep '*'
cat /sys/class/drm/card0/device/pp_dpm_mclk | grep '*'

replace card0 with yours

CryptoSquad

newbie

Activity: 60

Merit: 0

Hi All

Having modded my AMD bios's using Windows to about 30mh per card (180mh on a 6 card rig) I then setup linux with this driver, and i am getting only 177mhs

Does this sound normal? Difference of 0.5 mh's per card for linux?

One thing i did notice is windows every now and then would have a dramatic drop, and then increase back to the expected speeds, where as this seems far more stable at one speed?

Thoughts on all of this?

heavyarms1912

full member

Activity: 729

Merit: 114

Anybody noticed very slow shutdowns/reboots with all amd gpus and driver 17.30?

sk1nt

newbie

Activity: 1

Merit: 0

I haven't been able to get the Vega's to work beyond 5.5 MH/s in Linux. I've tried ROMc kernel with amdgpu-pro drivers, compiling my own kernel, compiling the 4.14rc4 and attempting to add the patches listed (all fail). Maybe I'm missing something when trying to apply the patches, I'm using pwclient and can sort on all new patches, but as I mentioned above all fail.

Has anyone managed to get the Vega working in Linux and would be kind enough to share the steps, including specific patches against specific kernel. Would really appreciate any help before I break down and fire up win10.

MoneroCrusher

jr. member

Activity: 88

Merit: 1

Anyone got it to work with RX Vega 56/64 under Ubuntu 16.04 yet?

xmonkx

newbie

Activity: 8

Merit: 0

Quote from: rvlad on September 21, 2017, 08:57:58 PM

Quote from: xmonkx on September 14, 2017, 10:24:36 PM

Quote from: osnwt on September 01, 2017, 11:37:16 AM

Quote from: laik2 on August 31, 2017, 04:56:32 PM

1.Download latest linux 4.12.10 from kernel.org
2.Apply patches from here: https://patchwork.freedesktop.org/project/amd-xorg-ddx/series/?ordering=-last_updated( more advanced way)
3.Compile kernel -> the ubuntu way.
4.Install kernel using dpkg
5. Download amdgpu-pro-17.30 latest, unarchive and edit amdgpu-pro-install, under the --compute function remove amdgpu-pro-dkms.
6. Exec ./amdgpu-pro-install --compute to install only opencl and libdrm
7. dpkg -x amdgpu-pro-dkms.... /out; cd out/usr/src/amdgpu-pro/lib/firmware && rm /lib/firmware/amdgpu/* && cp * /lib/firmware/amdgpu
8. add amdgpu.vm_fragment_size=9 to /etc/default/grub as stated in previous posts.
8.1. add if you are using user to video group usermod -a -G video $user
9. update-grub && update-initramfs -u
10. Reboot and mine.

Is a real reason to build new kernel such as performance or stability issues?

Since I wasn't sure what patches to apply, I used a mix of both ways:
1. Installed not whole rocm package but only kernel image and headers 4.11 from rocm repository.
2. Installed compute-only 17.30 amd driver as recommended above (without dkms).
3. Added amdgpu module option.

Still there are issues with very slow boot (known issue if VBIOS was patched). But it runs.
Is 4.12 kernel somehow better?

Thanks laik2, I built linux from the git:linux-master and it works.

--
ROCm does not work for my rig with A + N on it.
Therefore, I did laik2 way to build a new kernel for my rig.
However, I found that linux-4.12.10 even linux-4.13.x from linux.org does not support vm_fragment.
There is no vm_fragment member in ./drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c.
And I don't know which patches to be applied from patchwork.freedesktop.org.
Finally, I google "linux amdgp vm_fragment_size", github.com/torvalds comes out.

What to do is just download linux from github.com/torvalds and buid and run laik2 steps.
And everything works perfect.

Thanks~

Which Linux have you downloaded from github.com/torvalds?
4.12?

It is 4.13.

bomberb17

hero member

Activity: 773

Merit: 528

I'm a bit hesitant to install this driver because I'm running my rig remotely, I only have access to Teamviewer and SSH and I can't play with grub.
Is there any 'safe' way to install these drivers?

rvlad

newbie

Activity: 8

Merit: 0

Quote from: beastman on September 22, 2017, 11:59:20 AM

@CryptoSquad, found this for Linux UV dwarfing-master here https://github.com/rumatoest/dwarfing.

I haven't tried it as I am running back on Win10 for now. If you try it please post your results, should bring your temps down.

@rvlad, I had compiled the Torvalds kernel but it is missing the AMD firmwares and I don't know where to find the patches. Kernel works but can't get the GPU running due to missing firmware.

Please see the message from laik2 proposing to copy AMD firmware from AMD driver.

But the question was to xmonkx which linux from Torvalds git was compiled: 4.12 or newer?

beastman

newbie

Activity: 9

Merit: 0

@CryptoSquad, found this for Linux UV dwarfing-master here https://github.com/rumatoest/dwarfing.

I haven't tried it as I am running back on Win10 for now. If you try it please post your results, should bring your temps down.

@rvlad, I had compiled the Torvalds kernel but it is missing the AMD firmwares and I don't know where to find the patches. Kernel works but can't get the GPU running due to missing firmware.

rvlad

newbie

Activity: 8

Merit: 0

Quote from: xmonkx on September 14, 2017, 10:24:36 PM

Quote from: osnwt on September 01, 2017, 11:37:16 AM

Quote from: laik2 on August 31, 2017, 04:56:32 PM

1.Download latest linux 4.12.10 from kernel.org
2.Apply patches from here: https://patchwork.freedesktop.org/project/amd-xorg-ddx/series/?ordering=-last_updated( more advanced way)
3.Compile kernel -> the ubuntu way.
4.Install kernel using dpkg
5. Download amdgpu-pro-17.30 latest, unarchive and edit amdgpu-pro-install, under the --compute function remove amdgpu-pro-dkms.
6. Exec ./amdgpu-pro-install --compute to install only opencl and libdrm
7. dpkg -x amdgpu-pro-dkms.... /out; cd out/usr/src/amdgpu-pro/lib/firmware && rm /lib/firmware/amdgpu/* && cp * /lib/firmware/amdgpu
8. add amdgpu.vm_fragment_size=9 to /etc/default/grub as stated in previous posts.
8.1. add if you are using user to video group usermod -a -G video $user
9. update-grub && update-initramfs -u
10. Reboot and mine.

Is a real reason to build new kernel such as performance or stability issues?

Since I wasn't sure what patches to apply, I used a mix of both ways:
1. Installed not whole rocm package but only kernel image and headers 4.11 from rocm repository.
2. Installed compute-only 17.30 amd driver as recommended above (without dkms).
3. Added amdgpu module option.

Still there are issues with very slow boot (known issue if VBIOS was patched). But it runs.
Is 4.12 kernel somehow better?

Thanks laik2, I built linux from the git:linux-master and it works.

--
ROCm does not work for my rig with A + N on it.
Therefore, I did laik2 way to build a new kernel for my rig.
However, I found that linux-4.12.10 even linux-4.13.x from linux.org does not support vm_fragment.
There is no vm_fragment member in ./drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c.
And I don't know which patches to be applied from patchwork.freedesktop.org.
Finally, I google "linux amdgp vm_fragment_size", github.com/torvalds comes out.

What to do is just download linux from github.com/torvalds and buid and run laik2 steps.
And everything works perfect.

Thanks~

Which Linux have you downloaded from github.com/torvalds?
4.12?

CryptoSquad

newbie

Activity: 60

Merit: 0

So i have been running this most of the day.

Whilst its great, dual mining for Siacoin is much better with the Windows Driver.

Not necessarily for speed, but my GPU's are getting to 80 degrees with my dcri as low as 10.

On Windows i run with dcri of 15, and my GPU's sit at 72 degrees max.

However, this linux driver is much more stable for mining just ETH than it is in Windows. In Windows i get random drops for no reason at all (Good temps, good volts etc), but on Linux i dont get that at all.

Anyone else had issues with temps dual mining on this driver?

(RX 580's 4GB OC Edition).

Let me know

Thanks

CryptoSquad

newbie

Activity: 60

Merit: 0

scrap the above. Got it working. Driver hadnt installed correctly first time

Thanks!

Topic: Linux Driver for Blockchain Compute 17.30.1029 - page 2. (Read 16017 times)