Radeon ROCm 4.3 Released With HMM Allocations, Many Other Improvements

[phoronix]

Phoronix: Radeon ROCm 4.3 Released With HMM Allocations, Many Other Improvements

AMD has released ROCm 4.3 as the newest version of their Radeon Open eCosystem stack for providing open-source GPU compute and CUDA portability for their supported graphics processors under Linux. ROCm 4.3 is the biggest update we've seen for this important enterprise piece to their enterprise GPU compute stack in a while...

https://www.phoronix.com/scan.php?page=news_item&px=Radeon-ROCm-4.3

[Andythe_great]

I heard about software layer that allow CUDA application to run on AMD hardware, is ROCm the thing?

[perpetually high]

If anyone runs into the following problem when running apt update:

Err:11 http://repo.radeon.com/rocm/apt/debian xenial InRelease
The following signatures were invalid: EXPKEYSIG 9386B48A1A693C5C James Adrian Edwards (ROCm Release Manager) <[email protected]>
Error: GDBus.Errorrg.freedesktop.systemd1.UnitMasked: Unit packagekit.service is masked.
Reading package lists... Done
W: GPG error: http://repo.radeon.com/rocm/apt/debian xenial InRelease: The following signatures were invalid: EXPKEYSIG 9386B48A1A693C5C James Adrian Edwards (ROCm Release Manager) <[email protected]>
E: The repository 'http://repo.radeon.com/rocm/apt/debian xenial InRelease' is not signed.
N: Updating from such a repository can't be done securely, and is therefore disabled by default.
N: See apt-secure(8) manpage for repository creation and user configuration details.

I had to do the following to get it going:

wget -q -O - https://repo.radeon.com/rocm/rocm.gpg.key | sudo apt-key add -

Then just apt update and apt upgrade and you should be good to go.

Still no rocm-smi included anymore. They've stopped since 4.0.0 and I haven't figured out why.

I had to do the following sym link to get it going again a few versions ago:

$ ls -al /usr/bin/rocm-smi
lrwxrwxrwx 1 root root 42 Mar 26 07:11 /usr/bin/rocm-smi -> /opt/rocm-4.0.0/bin/rocm_smi_deprecated.py

I use rocm-smi all the time to set the mem clocks so if this has been moved elsewhere, if someone could tell me where, i'd appreciate it.

[uid313]

I wonder what will happen when Intel launches their DG graphics cards for Linux. Will they eat the market share of Nvidia or AMD or both?

[tildearrow]

I heard about software layer that allow CUDA application to run on AMD hardware, is ROCm the thing?

Nope. ROCm only provides very limited tools to translate CUDA code to HIP (AMD's own thing), but otherwise it is not (unlike ZLUDA for Intel).

[tildearrow]

I wonder what will happen when Intel launches their DX graphics cards for Linux. Will they eat the market share of Nvidia or AMD or both?

DX graphics cards for Linux? What do you mean?
Since when was DirectX available on Linux?

I personally doubt they will eat the competition unless they optimize the GPU architecture...
Plus they are going the OEM/data center route as of now...

[uid313]

DX graphics cards for Linux? What do you mean?
Since when was DirectX available on Linux?

I personally doubt they will eat the competition unless they optimize the GPU architecture...
Plus they are going the OEM/data center route as of now...

Oops I meant DG not DX. Intel is bringing dedicated GPU cards, such as the upcoming Intel DG2 powered by Intel Xe technology.

Yeah but in the data center they will compete wit AMD and Nvidia too. They will compete with the AMD Radeon Pro and Nvidia Tesla offerings.

[Devorlon]

there still is not any official support for Radeon RX 6000/7000

Should be 5000/6000.

I got to correct Michael for the first time. nice

[bridgman]

Nope. ROCm only provides very limited tools to translate CUDA code to HIP (AMD's own thing), but otherwise it is not (unlike ZLUDA for Intel).

Just curious what you mean by "very limited" - in most cases the tools seem to take care of >90% of the porting work. The exceptions are when apps have been newly coded for CUDA features that do not yet have a HIP equivalent.

[tildearrow]

Just curious what you mean by "very limited" - in most cases the tools seem to take care of >90% of the porting work. The exceptions are when apps have been newly coded for CUDA features that do not yet have a HIP equivalent.

What I mean is that the tools are only for source-to-source porting, as opposed to being a translation layer like Wine.