Radeon RX Vega 11

[metaquarx]

Request to add Radeom RX Vega 11 to the whitelist of gpus.

This is the successor to the Vega 8, which is already whitelisted. It seems the Vega 11 is automatically "downgraded" to a Vega 8 in fahcontrol? huh.

Log.txt:

Code: Select all

*********************** Log Started 2020-03-18T00:22:02Z ***********************
00:22:02:************************* Folding@home Client *************************
00:22:02:        Website: https://foldingathome.org/
00:22:02:      Copyright: (c) 2009-2018 foldingathome.org
00:22:02:         Author: Joseph Coffland <joseph@cauldrondevelopment.com>
00:22:02:           Args: --config /opt/fah/config.xml --exec-directory=/opt/fah
00:22:02:                 --data-directory=/opt/fah
00:22:02:         Config: /opt/fah/config.xml
00:22:02:******************************** Build ********************************
00:22:02:        Version: 7.5.1
00:22:02:           Date: May 11 2018
00:22:02:           Time: 19:59:04
00:22:02:     Repository: Git
00:22:02:       Revision: 4705bf53c635f88b8fe85af7675557e15d491ff0
00:22:02:         Branch: master
00:22:02:       Compiler: GNU 6.3.0 20170516
00:22:02:        Options: -std=gnu++98 -O3 -funroll-loops
00:22:02:       Platform: linux2 4.14.0-3-amd64
00:22:02:           Bits: 64
00:22:02:           Mode: Release
00:22:02:******************************* System ********************************
00:22:02:            CPU: AMD Ryzen 5 2400G with Radeon Vega Graphics
00:22:02:         CPU ID: AuthenticAMD Family 23 Model 17 Stepping 0
00:22:02:           CPUs: 8
00:22:02:         Memory: 5.79GiB
00:22:02:    Free Memory: 3.54GiB
00:22:02:        Threads: POSIX_THREADS
00:22:02:     OS Version: 5.5
00:22:02:    Has Battery: false
00:22:02:     On Battery: false
00:22:02:     UTC Offset: 0
00:22:02:            PID: 3587
00:22:02:            CWD: /opt/fah
00:22:02:             OS: Linux 5.5.9-arch1-2 x86_64
00:22:02:        OS Arch: AMD64
00:22:02:           GPUs: 1
00:22:02:          GPU 0: Bus:7 Slot:0 Func:0 AMD:5 Raven [Ryzen vega 8 mobile]
00:22:02:           CUDA: Not detected: Failed to open dynamic library 'libcuda.so':
00:22:02:                 libcuda.so: cannot open shared object file: No such file or
00:22:02:                 directory
00:22:02:OpenCL Device 0: Platform:0 Device:0 Bus:NA Slot:NA Compute:1.1 Driver:19.3
00:22:02:***********************************************************************

VendorID: 0x1002
DeviceID: 0x15dd
(from lspci -nn)

Code: Select all

07:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc. [AMD/ATI] Raven Ridge [Radeon Vega Series / Radeon Vega Mobile Series] [1002:15dd] (rev c6)

Last time Vega 11 was requested, proof of DP support was requested. Here is a snippet of clinfo:

Code: Select all

    Double-precision Floating-point support         (cl_khr_fp64)
        Denormals                                   Yes
        Infinity and NANs                           Yes
        Round to nearest                            Yes
        Round to zero                               Yes
        Round to infinity                           Yes
        IEEE754-2008 fused multiply-add             Yes
        Support is emulated in software             No
    Address bits                                    64, Little-Endian 

[Joe_H]

AMD reuses device ID's a lot across different products. So if your Vega 11 is recognized by the client, that is based on the ID number read a tconfiguration time. The description part of the entry is purely cosmetic and not used by the client for more than display. The entry in the GPUs.txt file is usually based on the first report of a new device ID we get, sometimes when one of the maintainers gets a chance they will modify the description to reflect that more than one GPU is referred to by that ID.

[CyTeK1978]

aren`t the Vega RX 10 and RX 11 the Same (11 with more open shaders?)

[bruce]

The manufacturers generally assign a unique device identifier no matter what the hardwre differences ... more VRAM, more shaders, different manufacturing technology, Even different assigned clock rate ("binning") FAH always enables GPUs by their hardware ID.
On the other hand, both AMD and nVidia do give new public names to the exact same hardware. Here's an example:

AMD:5 Ellesmere XT [Radeon RX 470/480/570/580]

FACT: The human readable portion between the [...]s doesn't change how FAH works with the device.

[toTOW]

0x15dd / Raven [Ryzen vega 8 mobile] is already in the list of supported GPUs.