Random Results

Using “scene-referred” darktable workflow and recommended modules

CPU only: 

   darktable-cli setubal.orf setubal.orf.xmp test.jpg --core --disable-opencl -d perf
CPU + GPU, using OpenCL: 
   darktable-cli setubal.orf setubal.orf.xmp test.jpg --core -d perf -d opencl

GPUs tested:
Radeon RX 7800 XT 16GB
Radeon RX 7600 8GB
Radeon RX 6500 XT 4GB
Radeon RX 5700 XT 8GB
Radeon RX Vega64 8GB
Radeon RX 480 8GB 		M2 Ultra 76-core
M2 Pro 16-core
M1 8-core 		Arc A770 16GB
Arc A750 8GB
Iris Xe 4GB 		A100 80GB
A100 40GB
A40 45GB
GeForce RTX 2070 8GB
GeForce GTX 1080 8GB
GeForce GTX 1070 8GB
GeForce GTX 1050 2GB
TITAN X 12GB

Legend:   AMD  ,    Apple  ,    Intel  ,    Nvidia  ,   Open source — open source.

↧ Tests with 50 Mpx RAW setubal.orf and darktable 4.2+ history stack from setubal.orf.xmp, copies here. ↧


CPU / GPU System info “pixel pipeline processing” time
(average of 3 runs), s		 `darktable-bench` with mire1.cr2, s
darktable 4.8.1
CPU only ARM64 Neoverse-N1 64C @3GHz 128GiB RAM, Linux 5.15, Ubuntu 22.04 (gauss) 
Architecture:             aarch64
CPU(s):                   64
Vendor ID:                ARM
  Model name:             Neoverse-N1
Manufacturer: Ampere(R)
Version: Ampere(R) Altra(R) Processor
3.9
darktable 4.6.1
CPU only 2 × AMD EPYC 9474F 48C/96T @4.1GHz 1.5TiB @4.8GT/s RAM, Linux 6.2.0, Ubuntu 23.04 (doob) 3.4
+ GPU GA100 [A100 PCIe 80GB] (rev a1) 
[dt_opencl_device_init]
   DEVICE:                   0: 'NVIDIA A100 80GB PCIe'
   CANONICAL NAME:           nvidiacudanvidiaa10080gbpcie
   DRIVER VERSION:           550.54.15
   DEVICE VERSION:           OpenCL 3.0 CUDA, SM_20 SUPPORT
   GLOBAL MEM SIZE:          81038 MB
   MAX MEM ALLOC:            20259 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   USE HEADROOM:             600Mb
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 0.7
+ GPU GA102GL [A40] 45GiB (rev a1) 

[dt_opencl_device_init]
   DEVICE:                   1: 'NVIDIA A40'
   CANONICAL NAME:           nvidiacudanvidiaa40
   DRIVER VERSION:           550.54.15
   DEVICE VERSION:           OpenCL 3.0 CUDA, SM_20 SUPPORT
   GLOBAL MEM SIZE:          45403 MB
   MAX MEM ALLOC:            11351 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   USE HEADROOM:             600Mb
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 1.2
CPU only Ryzen 9 7900 12C/24T @5.2GHz, 32GB @7.2GT/s RAM, NVMe SSD, X670 PCIe 5.0, Linux 6.8.0, Ubuntu 24.04 (hilbert) 6.4 6.8
+ GPU open source Arc A750 8GB 
[dt_opencl_device_init]
   DEVICE:                   0: 'Intel(R) Arc(TM) A750 Graphics'
   CANONICAL NAME:           intelropenclgraphicsintelrarctma750graphics
   DRIVER VERSION:           23.43.027642
   DEVICE VERSION:           OpenCL 3.0 NEO 
   GLOBAL MEM SIZE:          7722 MB
   MAX MEM ALLOC:            3861 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   USE HEADROOM:             600Mb
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 4.1 5
+ GPU open source Arc A770 16GB 
[dt_opencl_device_init]
   DEVICE:                   0: 'Intel(R) Arc(TM) A770 Graphics'
   CANONICAL NAME:           intelropenclgraphicsintelrarctma770graphics
   DRIVER VERSION:           23.43.027642
   DEVICE VERSION:           OpenCL 3.0 NEO 
   GLOBAL MEM SIZE:          15474 MB
   MAX MEM ALLOC:            4096 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   USE HEADROOM:             600Mb
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 1.9 4.5
darktable 4.6
CPU only Ryzen 5 7600X 6C/12T 5.4GHz, 32GB RAM, Linux 6.6, Debian Sid (2024.02.06) (thelio-r3) 10.5
+ GPU Radeon RX 6500 XT 4GB
ROCm 6.0.2 
[dt_opencl_device_init]
   DEVICE:                   0: 'gfx1034'
   CANONICAL NAME:           amdacceleratedparallelprocessinggfx1034
   DRIVER VERSION:           3602.0 (HSA1.1,LC)
   DEVICE VERSION:           OpenCL 2.0 
   GLOBAL MEM SIZE:          4080 MB
   MAX MEM ALLOC:            3468 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 4.9
darktable 4.4.2
CPU only Ryzen 7 2700X 8C/16T 3.7GHz, 16GB RAM, Linux 6.5, Debian Sid (2023.12.18) (molly) 13.2
+ GPU Radeon RX 7600 8GB
ROCm 6.0 
[dt_opencl_device_init]
   DEVICE:                   0: 'gfx1102'
   CANONICAL NAME:           amdacceleratedparallelprocessinggfx1102
   DRIVER VERSION:           3602.0 (HSA1.1,LC)
   DEVICE VERSION:           OpenCL 2.0 
   GLOBAL MEM SIZE:          8176 MB
   MAX MEM ALLOC:            6950 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            WANTED
   MICRO NAP:                250
 2.8 
   MEMORY TUNING:            NO
5.2
CPU only Ryzen 9 3900 12C/24T, Debian Bookworm (data by Sebastian Bachmann) 8.8
+ GPU Radeon RX 7800 XT 16GB
ROCm 5.7 		0.95
CPU only 13th Gen Core i7-1360P, 16GiB RAM, NVMe SSD, Linux 6.0, Ubuntu 23.04 (Dell XPS 9320) 12.5
+ GPU open source Intel Raptor Lake-P [Iris Xe Graphics] 
[dt_opencl_device_init]
   DEVICE:                   0: 'Intel(R) Graphics [0xa7a0]'
   CANONICAL NAME:           intelropenclhdgraphicsintelrgraphics0xa7a0
   DRIVER VERSION:           22.43.24595
   DEVICE VERSION:           OpenCL 3.0 NEO 
   GLOBAL MEM SIZE:          12489 MB
   MAX MEM ALLOC:            4096 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            WANTED
   FORCED HEADROOM:          400
   MICRO NAP:                250
 7.7
CPU only Ryzen 9 3900X 12C/24T @4.2GHz, 32GB @3.6GT/s RAM, NVMe SSD, PCIe 4.0, Linux 6.2, Ubuntu 23.04 (storas) 8.8 8.9
+ GPU GeForce RTX 2070 8GB (TU106) 
[dt_opencl_device_init]
   CANONICAL NAME:           nvidiacudanvidiageforcertx2070
   DRIVER VERSION:           535.86.05
   DEVICE VERSION:           OpenCL 3.0 CUDA, SM_20 SUPPORT
   GLOBAL MEM SIZE:          7967 MB
   MAX MEM ALLOC:            1992 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            NO
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 2.2 3.9
CPU only Ryzen 9 7900 12C/24T @5GHz, 32GB @7.2GT/s RAM, NVMe SSD, X670 PCIe 5.0, Linux 6.4.11, Ubuntu 23.04 (hilbert) (≤300W) 6.3 6.7
+ GPU open source Arc A750 8GB PCIe 4.0 
[dt_opencl_device_init]
   DEVICE:                   0: 'Intel(R) Graphics [0x56a1]'
   CANONICAL NAME:           intelropenclhdgraphicsintelrgraphics0x56a1
   DRIVER VERSION:           22.43.24595
   DEVICE VERSION:           OpenCL 3.0 NEO 
   GLOBAL MEM SIZE:          7722 MB
   MAX MEM ALLOC:            3861 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            WANTED
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 (≤205W) 3 4.6
CPU only M2 Ultra 24CPU/76GPU (Mac14,14), 192GiB RAM, macOS 13.5, Darwin 22.6.0 (≤100W) 5.5
+ GPU M2 Ultra
[dt_opencl_device_init]
   DEVICE:                   0: 'Apple M2 Ultra'
   PLATFORM NAME & VENDOR:   Apple, Apple
   CANONICAL NAME:           appleapplem2ultra
   DRIVER VERSION:           1.2 1.0
   DEVICE VERSION:           OpenCL 1.2 
   GLOBAL MEM SIZE:          147456 MB
   MAX MEM ALLOC:            27648 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            WANTED
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 (≤70W) 1.4
CPU only M2 Pro 10CPU/16GPU (Mac14,12), 32GiB RAM, macOS 13.5, Darwin 22.6.0 (gegenbauer2) (≤38W) 11
+ GPU M2 Pro
[dt_opencl_device_init]
   DEVICE:                   0: 'Apple M2 Pro'
   PLATFORM NAME & VENDOR:   Apple, Apple
   CANONICAL NAME:           appleapplem2pro
   DRIVER VERSION:           1.2 1.0
   DEVICE VERSION:           OpenCL 1.2
   GLOBAL MEM SIZE:          21845 MB
   MAX MEM ALLOC:            4096 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            WANTED
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 (≤32W) 3
CPU only open source M1 8CPU/8GPU (iMac21,1), 16GiB RAM, Linux 6.3, asahi 13.1 (Arch) (102e-imac) 16.5
CPU only M1 8CPU/8GPU (iMac21,1), 16GiB RAM, macOS 13.4.1, Darwin 22.5.0 (102e-imac) 16.5
+ GPU M1
[dt_opencl_device_init]
   DEVICE:                   0: 'Apple M1'
   PLATFORM NAME & VENDOR:   Apple, Apple
   CANONICAL NAME:           appleapplem1
   DRIVER VERSION:           1.2 1.0
   DEVICE VERSION:           OpenCL 1.2 
   GLOBAL MEM SIZE:          10923 MB
   MAX MEM ALLOC:            2048 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            WANTED
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 5.8
darktable 4.4.0
CPU only Ryzen 5 5600X @4.6GHz 6C/12T, 32GiB RAM @3.2GT/s, NVMe SSD, Linux 6.2, Ubuntu 23.04 (storas) 12.5
+ GPU GeForce RTX 2070 8GB (TU106) 
[dt_opencl_device_init]
   CANONICAL NAME:           nvidiacudanvidiageforcertx2070
   DRIVER VERSION:           535.54.03
   DEVICE VERSION:           OpenCL 3.0 CUDA, SM_20 SUPPORT
   GLOBAL MEM SIZE:          7941 MB
   MAX MEM ALLOC:            1985 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            NO
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 2.3

CPU / GPUSystem info“pixel pipeline processing” time
(average of 3 runs), s
darktable:4.2.14.4
CPU only Ryzen 9 7900 12C/24T @5.2GHz, 32GB @6.4GT/s RAM, NVMe SSD, PCIe 5.0, Linux 6.4.0-060400rc6, Ubuntu 23.04 (hilbert) 8.5 5.9
+ GPU open source Arc A750 8GB PCIe 4.0 
[dt_opencl_device_init]
   DEVICE:                   0: 'Intel(R) Graphics [0x56a1]'
   CANONICAL NAME:           intelrgraphics0x56a1
   DRIVER VERSION:           22.43.24595
   DEVICE VERSION:           OpenCL 3.0 NEO 
   GLOBAL MEM SIZE:          7722 MB
   MAX MEM ALLOC:            3861 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            WANTED
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 3.2 2.9

CPU / GPUSystem info“pixel pipeline processing” time
(average of 3 runs), s
darktable 4.3.0+488~g59ff35fcf (adds OM System support)
CPU only Ryzen 9 3900X 12C/24T @4.2GHz, 32GB @3.6GT/s RAM, NVMe SSD, PCIe 4.0, Linux 6.2-rc5, Ubuntu 22.10 (hilbert) 11
+ GPU open source Radeon RX 5700 XT 8GB (Navi 10)
ROCm 5.4 
[dt_opencl_device_init]
   CANONICAL NAME:           gfx1010xnack
   DRIVER VERSION:           3513.0 (HSA1.1,LC)
   DEVICE VERSION:           OpenCL 2.0 
   GLOBAL MEM SIZE:          8176 MB
   MAX MEM ALLOC:            6950 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            WANTED
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 2.4
+ GPU open source Radeon RX Vega64 8GB
ROCm 5.4 
[dt_opencl_device_init]
   CANONICAL NAME:           gfx900xnack
   DRIVER VERSION:           3513.0 (HSA1.1,LC)
   DEVICE VERSION:           OpenCL 2.0 
   GLOBAL MEM SIZE:          8176 MB
   MAX MEM ALLOC:            6950 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            WANTED
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 5.5
+ GPU open source Arc A750 8GB PCIe 4.0 
lspci: 0b:00.0 VGA compatible controller: Intel Corporation DG2 [Arc A750] (rev 08)

`apt install intel-opencl-icd`

[dt_opencl_device_init]
   CANONICAL NAME:           intelrgraphics0x56a1
   DRIVER VERSION:           1.0.0
   DEVICE VERSION:           OpenCL 3.0 NEO 
   GLOBAL MEM SIZE:          7722 MB
   MAX MEM ALLOC:            3861 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            WANTED
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 3.5
+ GPU GeForce GTX 1080 8GB (GP104) 
[dt_opencl_device_init]
   CANONICAL NAME:           nvidiageforcegtx1080
   DRIVER VERSION:           525.78.01
   DEVICE VERSION:           OpenCL 3.0 CUDA, SM_20 SUPPORT
   GLOBAL MEM SIZE:          8080 MB
   MAX MEM ALLOC:            2020 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            NO
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 4.3
CPU only Ryzen 5 5600X @4.6GHz 6C/12T, 32GiB RAM @3.2GT/s, NVMe SSD, Linux 5.15, Ubuntu 22.04 (storas) 17.6
+ GPU GeForce RTX 2070 8GB (TU106) 
[dt_opencl_device_init]
   CANONICAL NAME:           nvidiageforcertx2070
   DRIVER VERSION:           520.61.05
   DEVICE VERSION:           OpenCL 3.0 CUDA, SM_20 SUPPORT
   GLOBAL MEM SIZE:          7949 MB
   MAX MEM ALLOC:            1987 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            NO
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 3

↧ Tests with Olympus RAW arecibo.orf and darktable history stack from arecibo.orf.xmp, copies here. ↧

darktable 4.2.0
CPU only Ryzen 5 5600X @4.6GHz 6C/12T, 32GiB RAM @3.2GT/s, NVMe SSD, Linux 5.15, Ubuntu 22.04 (storas) 6.7
+ GPU GeForce RTX 2070 8GB (TU106) 
[dt_opencl_device_init]
   CANONICAL NAME:           nvidiageforcertx2070
   DRIVER VERSION:           520.61.05
   DEVICE VERSION:           OpenCL 3.0 CUDA, SM_20 SUPPORT
   GLOBAL MEM SIZE:          7949 MB
   MAX MEM ALLOC:            1987 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            WANTED
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 1.2
CPU only Ryzen 9 3900X 12C/24T @4.2GHz, 32GB @3.6GT/s RAM, NVMe SSD, PCIe 4.0, Linux 6.2.0-060200rc5-generic, Ubuntu 22.10 (hilbert) 3.9
+ GPU open source Radeon RX 5700 XT (Navi 10) Compute units: 40, stream processors: 2560
ROCm 5.4 
[dt_opencl_device_init]
   CANONICAL NAME:           gfx1010xnack
   DRIVER VERSION:           3513.0 (HSA1.1,LC)
   DEVICE VERSION:           OpenCL 2.0 
   GLOBAL MEM SIZE:          8176 MB
   MAX MEM ALLOC:            6950 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            WANTED
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 0.85
+ GPU open source Arc A750 8GB PCIe 4.0
lspci: 0b:00.0 VGA compatible controller: Intel Corporation DG2 [Arc A750] (rev 08)

`apt install intel-opencl-icd`

[dt_opencl_device_init]
   CANONICAL NAME:           intelrgraphics0x56a1
   PLATFORM NAME & VENDOR:   Intel(R) OpenCL HD Graphics, Intel(R) Corporation
   DRIVER VERSION:           1.0.0
   DEVICE VERSION:           OpenCL 3.0 NEO 
   GLOBAL MEM SIZE:          7722 MB
   MAX MEM ALLOC:            3861 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            WANTED
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
 1
darktable 4.0
CPU only Ryzen 9 3900X 12C/24T @4.2GHz, 16GB @4GT/s RAM, NVMe SSD, Linux 5.15, Ubuntu 22.04.1 (hilbert) 4.3
+ GPU open source Radeon RX 5700 XT (Navi 10) Compute units: 40, stream processors: 2560
ROCm 5.2 
[dt_opencl_device_init]
   CANONICAL NAME:           gfx1010xnack
   DRIVER VERSION:           3452.0 (HSA1.1,LC)
   DEVICE VERSION:           OpenCL 2.0 
   GLOBAL MEM SIZE:          8176 MB
   MAX MEM ALLOC:            6950 MB
   ASYNC PIXELPIPE:          NO
   PINNED MEMORY TRANSFER:   NO
   MEMORY TUNING:            NO
   FORCED HEADROOM:          400
   AVOID ATOMICS:            NO
   MICRO NAP:                250
   ROUNDUP WIDTH:            16
   ROUNDUP HEIGHT:           16
   CHECK EVENT HANDLES:      128
   PERFORMANCE:              0.020894 (CPU 0.039102)
 1.3
darktable 3.8.0
CPU only Epyc 7713 2×64 @2GHz, 2TiB RAM @3.2GT/s, NVMe SSD, Linux 5.4, Ubuntu 20.04 (kontiki) 2.7
+ GPU A100 40GB 
[opencl_init] device 0 `A100-PCIE-40GB' supports image sizes of 32768 x 32768
     GLOBAL_MEM_SIZE:          40536MB
     DRIVER_VERSION:           495.29.05
     DEVICE_VERSION:           OpenCL 3.0 CUDA
 1
CPU only Ryzen 5 5600X @4.6GHz 6C/12T, 32GiB RAM @3.2GT/s, NVMe SSD, Linux 5.13, Ubuntu 21.10 (storas) 6.8
+ GPU GeForce RTX 2070 8GB (TU106) 
[opencl_init]device 0 `NVIDIA GeForce RTX 2070' allows GPU memory allocations of up to 1987MB
     GLOBAL_MEM_SIZE:          7949MB
     DRIVER_VERSION:           495.44
     DEVICE_VERSION:           OpenCL 3.0 CUDA
 1.1
darktable 3.7.0~git800.7b9f576d28-1
CPU only Ryzen 9 3900X 12C/24T @4.2GHz, 16GB @4GT/s RAM, NVMe SSD, Linux 5.11, Ubuntu 21.04, ROCm 4.3 (hilbert) 4.2
+ GPU open source Radeon RX 5700 XT (Navi 10) Compute units: 40, stream processors: 2560
amdgpu driver from stock kernel + ROCm utilities `apt install rocm-dev` 
[opencl_init] device 0 `gfx1010:xnack-' allows GPU memory allocations of up to 6949MB
     CANONICAL_NAME:           gfx1010
     GLOBAL_MEM_SIZE:          8176MB
     DRIVER_VERSION:           3305.0 (HSA1.1,LC)
     DEVICE_VERSION:           OpenCL 2.0
 1.4
darktable 3.6.1
CPU only Ryzen 5 5600X @3.7GHz 6C/12T, 32GiB RAM @2666MT/s, NVMe SSD, Linux 5.13, Ubuntu 21.10 (storas) 8.4
CPU only Ryzen 5 5600X @4.6GHz 6C/12T, 32GiB RAM @3.2GT/s, NVMe SSD, Linux 5.13, Ubuntu 21.10 (storas) 7.4
+ GPU GeForce RTX 2070 8GB (TU106) 
[opencl_init]device 0 `NVIDIA GeForce RTX 2070' allows GPU memory allocations of up to 1987MB
     GLOBAL_MEM_SIZE:          7949MB
     DRIVER_VERSION:           495.44
     DEVICE_VERSION:           OpenCL 3.0 CUDA
 1.2
darktable 3.6
CPU only Ryzen 5 1600 @3.6GHz 6C/12T, 32GiB RAM @3GT/s, NVMe SSD, Linux 5.11, Ubuntu 21.04 (storas) 7.9
+ GPU GeForce RTX 2070 8GB (TU106) 
[opencl_init] device 0 `NVIDIA GeForce RTX 2070' allows GPU memory allocations of up to 1993MB
     GLOBAL_MEM_SIZE:          7974MB
     DRIVER_VERSION:           470.57.02
     DEVICE_VERSION:           OpenCL 3.0 CUDA
 1.4
darktable 3.4.1
CPU only Ryzen 5 1600 @3.6GHz 6C/12T, 32GiB RAM @3GT/s, NVMe SSD, Linux 5.11, Ubuntu 21.04 (storas) 8.6
+ GPU GeForce RTX 2070 8GB (TU106) 
[opencl_init] device 0 `NVIDIA GeForce RTX 2070' allows GPU memory allocations of up to 1993MB
     GLOBAL_MEM_SIZE:          7974MB
     DRIVER_VERSION:           465.27
     DEVICE_VERSION:           OpenCL 3.0 CUDA
 1.5
darktable 3.4
CPU only Core i5-1135G7 4C/8T, 16GiB RAM, NVMe SSD, Linux 5.6, Ubuntu 20.04.2 (gimli) 12.3
+ GPU open source Iris Xe 4GB (Gen12LP)
`apt install intel-opencl-icd` 
[opencl_init] device 0 `Intel(R) Gen12LP HD Graphics NEO' allows GPU memory allocations of up to 4095MB
     GLOBAL_MEM_SIZE:          12590MB
     DRIVER_VERSION:           1.0.0
     DEVICE_VERSION:           OpenCL 2.1 NEO
5.3
darktable 3.2.1
CPU only Ryzen 9 3900X 12C/24T @4.2GHz, 16GB @4GT/s RAM, NVMe SSD, Linux 5.4, Ubuntu 20.04, ROCm 3.10 (hilbert) 5.3
+ GPU Radeon RX 5700 XT (Navi 10) Compute units: 40, stream processors: 2560
ROCm Package from a Debian Repository, `apt install rocm-dkms` 
[opencl_init] device 0 `gfx1010' allows GPU memory allocations of up to 6949MB
     GLOBAL_MEM_SIZE:          8176MB
     DRIVER_VERSION:           3212.0 (HSA1.1,LC)
     DEVICE_VERSION:           OpenCL 2.0
 1.2
CPU only Ryzen 5 1600 @3.6GHz 6C/12T, 32GiB RAM @3GT/s, NVMe SSD, Linux 5.8, Ubuntu 20.10 (storas) 10.2
+ GPU GeForce GTX 1070 8GB (GP104) 
[opencl_init] device 0 `GeForce GTX 1070' allows GPU memory allocations of up to 2029MB
     GLOBAL_MEM_SIZE:          8116MB
     DRIVER_VERSION:           455.45.01
     DEVICE_VERSION:           OpenCL 1.2 CUDA
 2.2
CPU only Ryzen 7 2700X 8C/16T @4GHz, 16GB @3GT/s RAM, NVMe SSD, Linux 5.4, Ubuntu 20.04.1 (geom) 6.8
+ GPU Radeon RX Vega64 8GB. Compute Units: 64, Stream Processors: 4096.
AMDGPU-PRO 20.45-1164792, `amdgpu-install --opencl=pal --headless` 
[opencl_init] device 0 `gfx900' allows GPU memory allocations of up to 6949MB
     GLOBAL_MEM_SIZE:          8176MB
     DRIVER_VERSION:           3188.4 (HSA1.1,LC)
     DEVICE_VERSION:           OpenCL 2.0
4.0
CPU only Xeon E5-2643 8C/16T 3.3GHz, 128GB RAM, Linux 5.4, Ubuntu 20.04.1 (molly) 10.5
+ GPU open source Radeon RX 480 (Ellesmere)
AMDGPU-PRO 20.45-1164792, `amdgpu-install --opencl=legacy --headless --no-dkms` 
[opencl_init] device 0 `Ellesmere' allows GPU memory allocations of up to 6743MB
     GLOBAL_MEM_SIZE:          8185MB
     DRIVER_VERSION:           3188.4
     DEVICE_VERSION:           OpenCL 1.2 AMD-APP (3188.4)
3.8
+ GPU GeForce RTX 2070 8GB (TU106) 
[opencl_init] device 0 `GeForce RTX 2070' allows GPU memory allocations of up to 1995MB
     GLOBAL_MEM_SIZE:          7982MB
     DRIVER_VERSION:           460.27.04
     DEVICE_VERSION:           OpenCL 1.2 CUDA
 2.1
CPU only Ryzen 5 2600 6C/12T 3.4GHz, 32GB RAM, SATA-II SSD, Ubuntu 20.04 (glab2) 9.2
+ GPU GeForce GTX 1080 8GB (GP104) 
[opencl_init] device 0 `GeForce GTX 1080' allows GPU memory allocations of up to 2029MB
     GLOBAL_MEM_SIZE:          8116MB
     DRIVER_VERSION:           460.27.04
     DEVICE_VERSION:           OpenCL 1.2 CUDA
 1.8
CPU only Xeon E5-1650 v4 3.6GHz 6C/12T, 64GB RAM, SSD, Ubuntu 20.04 (abacus) 8.8
+ GPU TITAN X 12GB (Pascal; GP102) 
[opencl_init] device 1 `TITAN X (Pascal)' allows GPU memory allocations of up to 3049MB
     GLOBAL_MEM_SIZE:          12196MB
     DRIVER_VERSION:           460.27.04
     DEVICE_VERSION:           OpenCL 1.2 CUDA
 1.8
CPU only Ryzen Threadripper 2970WX 24C/48T 4.2GHz, 64GB @3GHz RAM, NVMe SSD, Ubuntu 19.10 (toby) 3.7
+ GPU GeForce GTX 1050 2GB (GP107) CUDA cores: 640 
[opencl_init] device 0 `GeForce GTX 1050' allows GPU memory allocations of up to 498MB
     GLOBAL_MEM_SIZE:          1993MB
     DRIVER_VERSION:           460.27.04
     DEVICE_VERSION:           OpenCL 1.2 CUDA
 3.6

Using legacy darktable workflow

Processing RAW image file bench.srw with darktable history stack from supplied bench.srw.xmp. Both posted on darktable-user mailing list sometime in 2016, copies here (20 Mpx RAW from Samsung NX300). Most tests on current Ubuntu (though sometimes with the latest mainline kernel) and mostly stable versions of darktable.

CPU only: 

   darktable-cli bench.SRW bench.SRW.xmp test.jpg --core --disable-opencl -d perf
CPU + GPU, using OpenCL: 
   darktable-cli bench.SRW bench.SRW.xmp test.jpg --core -d perf -d opencl

CPU / GPUSystem info“pixel pipeline processing” time
(average of 3 runs), s
Legend:  AMD  ,    Intel  ,    Nvidia  ,   Open source — open source.
GPUs tested
Radeon RX 5700 XT 8GB
Radeon RX Vega64 8GB
Radeon RX 480 8GB
Radeon R9 390X 8GB
Radeon R9 290X 4GB 		GeForce RTX 2070 8GB
TITAN X 12GB
GeForce GTX 1080 8GB
GeForce GTX 1070 8GB
GeForce GTX 1050 2GB
GeForce GT 1030 2GB
darktable 3.2.1
CPU only Ryzen 5 1600 @3.6GHz 6C/12T, 32GiB RAM @3GT/s, NVMe SSD, Linux 5.8, Ubuntu 20.10 (storas) 12.3
+ GPU GeForce GTX 1070 8GB
[opencl_init] device 0 `GeForce GTX 1070' allows GPU memory allocations of up to 2029MB
     GLOBAL_MEM_SIZE:          8116MB
     DRIVER_VERSION:           455.32.00
     DEVICE_VERSION:           OpenCL 1.2 CUDA
 3.0
CPU only Ryzen 9 3900X 12C/24T @4.2GHz, 16GB @4GT/s RAM, NVMe SSD, PCIe4, Ubuntu 20.04.1 (hilbert) 11
+ GPU open source Radeon RX 5700 XT 8GB / Linux 5.9.0-050900rc6 / ROCm 3.8
Compute units: 40, stream processors: 2560
ROCm with Upstream Kernel Drivers, `apt install rocm-dev`
modinfo amdgpu:
filename:       /lib/modules/5.9.0-050900rc6-generic/kernel/drivers/gpu/drm/amd/amdgpu/amdgpu.ko

darktable-cltest:
[opencl_init] device 0 `gfx1010' allows GPU memory allocations of up to 6949MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           3186.0 (HSA1.1,LC)
DEVICE_VERSION:           OpenCL 2.0
 1.9
CPU only Intel 4790K@4.4GHz, 64GiB RAM, NVMe SSD, Windows 10. Contributed by Jan Müller. 20
+ GPU GeForce GTX 1070 8GB
     DRIVER_VERSION:           451.67
     DEVICE_VERSION:           OpenCL 1.2 CUDA
 4.37
darktable 3.0.1
CPU only Ryzen 9 3900X 12C/24T @4.2GHz, 16GB @4GT/s RAM, NVMe SSD, PCIe4, Ubuntu 20.04 (hilbert) 10.9
+ GPU Radeon RX 5700 XT 8GB / Linux 5.4 / AMDGPU-PRO 20.30-1109583
Compute units: 40, stream processors: 2560
./amdgpu-install --pro --opencl=pal --no-32 
modinfo amdgpu:
filename:       /lib/modules/5.4.0-42-generic/updates/dkms/amdgpu.ko
version:        5.6.5.20.30
license:        GPL and additional rights
description:    AMD GPU
author:         AMD linux driver team
darktable-cltest:
[opencl_init] device 0 `gfx1010' allows GPU memory allocations of up to 6732MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           3143.9 (PAL,LC)
DEVICE_VERSION:           OpenCL 2.0 AMD-APP (3143.9)
 2.0
+ GPU Radeon RX 5700 XT 8GB / Linux 5.4 amdgpu + OpenCL from AMDGPU-PRO 20.30-1109583
Compute units: 40, stream processors: 2560
./amdgpu-install --opencl=pal --headless --no-dkms --no-32 
modinfo amdgpu:
filename:       /lib/modules/5.4.0-42-generic/kernel/drivers/gpu/drm/amd/amdgpu/amdgpu.ko
license:        GPL and additional rights
description:    AMD GPU
author:         AMD linux driver team
darktable-cltest:
[opencl_init] device 0 `gfx1010' allows GPU memory allocations of up to 6732MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           3143.9 (PAL,LC)
DEVICE_VERSION:           OpenCL 2.0 AMD-APP (3143.9)
 2.0
+ GPU Radeon RX 5700 XT 8GB / Linux 5.4 / ROCm 3.5.0
Compute units: 40, stream processors: 2560
ROCm Package from a Debian Repository, `apt install rocm-dkms` 
modinfo amdgpu:
filename:       /lib/modules/5.4.0-42-generic/updates/dkms/amdgpu.ko
version:        5.6.0
license:        GPL and additional rights
description:    AMD GPU
author:         AMD linux driver team
[opencl_init]device 0 `gfx1010' allows GPU memory allocations of up to 6949MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           3137.0 (HSA1.1,LC)
DEVICE_VERSION:           OpenCL 2.0
 <2
+ GPU open source Radeon RX 5700 XT 8GB / Linux 5.8.0-050800-generic amdgpu / ROCm 3.5.0 with upstream kernel driver
Compute units: 40, stream processors: 2560
ROCm with Upstream Kernel Drivers, `apt install rocm-dev`
modinfo amdgpu:
filename:       /lib/modules/5.8.0-050800-generic/kernel/drivers/gpu/drm/amd/amdgpu/amdgpu.ko
license:        GPL and additional rights
description:    AMD GPU
author:         AMD linux driver team

[opencl_init] device 0 `gfx1010' allows GPU memory allocations of up to 6949MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           3137.0 (HSA1.1,LC)
DEVICE_VERSION:           OpenCL 2.0
 <2
darktable 2.6.2
CPU only Ryzen 7 2700X 8C/16T @4GHz, 32GB @3GT/s RAM, NVMe SSD, Ubuntu 18.04.4 (geom) 11.7
+ GPU Radeon RX Vega64 8GB / Linux 5.0, AMDGPU-PRO 19.50-967956
Compute Units: 64
Stream Processors: 4096
[opencl_init] device 0 `gfx900' allows GPU memory allocations of up to 4048MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           3004.6 (PAL,HSAIL)
DEVICE_VERSION:           OpenCL 2.0 AMD-APP (3004.6)
2.6
+ GPU Radeon RX 5700 XT 8GB / Linux 5.0, AMDGPU-PRO 19.50-967956
Compute Units: 40
Stream Processors: 2560
[opencl_init] device 0 `gfx1010' allows GPU memory allocations of up to 4048MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           3004.6 (PAL,LC)
DEVICE_VERSION:           OpenCL 2.0 AMD-APP (3004.6)
2.3
CPU only Ryzen 9 3900X 12C/24T @4.2GHz, 16GB @4GT/s RAM, NVMe SSD, PCIe4, Ubuntu 19.10 (hilbert) 11.1
+ GPU open source Radeon RX Vega64 8GB / Linux 5.3.0-40-generic amdgpu, ROCm 3.0.6 (rocm-dev)
(github.com/RadeonOpenCompute/ROCm) 
Compute Units: 64
Stream Processors: 4096
[opencl_init]device 0 `gfx900' allows GPU memory allocations of up to 6949MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           3052.0 (HSA1.1,LC)
DEVICE_VERSION:           OpenCL 2.0
3.4
+ GPU Radeon RX 5700 XT 8GB / Linux 5.6-rc1 amdgpu + OpenCL from AMDGPU-PRO 19.50-967956
(amdgpu-install --opencl=pal --headless --no-dkms) 
Compute Units: 40
Stream Processors: 2560
[opencl_init] device 0 `gfx1010' allows GPU memory allocations of up to 4048MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           3004.6 (PAL,LC)
DEVICE_VERSION:           OpenCL 2.0 AMD-APP (3004.6)
 2.1
CPU only Ryzen 7 2700X 8C/16T @4.0GHz, 32GB @3GHz RAM, NVMe SSD, Ubuntu 19.04 (hilbert) 11.6
+ GPU open source Radeon RX Vega64 8GB / Linux 5.1.0 amdgpu, ROCm 2.4 (rocm-dev)
(github.com/RadeonOpenCompute/ROCm) 
Compute Units: 64
Stream Processors: 4096
[opencl_init] device 0 `gfx900' allows GPU memory allocations of up to 6949MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           2874.0 (HSA1.1,LC)
DEVICE_VERSION:           OpenCL 2.0
3.9
+ GPU open source Radeon RX Vega64 8GB / Linux 5.2.8 amdgpu, ROCm 2.7 (rocm-dev)
(github.com/RadeonOpenCompute/ROCm) 
Compute Units: 64
Stream Processors: 4096
[opencl_init] device 0 `gfx900' allows GPU memory allocations of up to 6949MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           2949.0 (HSA1.1,LC)
DEVICE_VERSION:           OpenCL 2.0
3.8
CPU only Ryzen 7 2700X 8C/16T @4.0GHz, 32GB @3GHz RAM, NVMe SSD, Ubuntu 19.10 (hilbert) 11.7
+ GPU open source Radeon RX Vega64 8GB / Linux 5.3.0-19-generic amdgpu, ROCm 2.9 (rocm-dev)
(github.com/RadeonOpenCompute/ROCm) 
Compute Units: 64
Stream Processors: 4096
[opencl_init]device 0 `gfx900' allows GPU memory allocations of up to 6949MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           2982.0 (HSA1.1,LC)
DEVICE_VERSION:           OpenCL 2.0
3.8
darktable 2.6.0
CPU only Ryzen 7 2700X 8C/16T @3.7GHz, 32GB @2.133GHz RAM, NVMe SSD, Ubuntu 18.04 (hilbert)
Ryzen 7 2700X 8C/16T @4.0GHz, 32GB @3GHz RAM, NVMe SSD, Ubuntu 18.04 (hilbert) 		15.1
11.8
+ GPU Radeon RX Vega64 8GB / Linux 4.15, AMDGPU-PRO 18.40-697810
Compute Units: 64
Stream Processors: 4096
[opencl_init] device 0 `gfx900' allows GPU memory allocations of up to 4048MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           2686.5 (PAL,HSAIL)
DEVICE_VERSION:           OpenCL 2.0 AMD-APP (2686.5)
2.7
CPU only Ryzen 7 2700X 8C/16T @4.0GHz, 32GB @3GHz RAM, NVMe SSD, Ubuntu 18.10 (hilbert) 11.8
+ GPU open source Radeon RX Vega64 8GB / Linux 4.18 amdgpu, ROCm 2.0 (rocm-dev)
(github.com/RadeonOpenCompute/ROCm) 
Compute Units: 64
Stream Processors: 4096
[opencl_init] device 0 `gfx900' allows GPU memory allocations of up to 6949MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           2783.0 (HSA1.1,LC)
DEVICE_VERSION:           OpenCL 1.2
4.2
+ GPU open source Radeon RX Vega64 8GB / Linux 4.20 amdgpu, ROCm 2.1 (rocm-dev)
(github.com/RadeonOpenCompute/ROCm) 
Compute Units: 64
Stream Processors: 4096
[opencl_init] device 0 `gfx900' allows GPU memory allocations of up to 6949MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           2814.0 (HSA1.1,LC)
DEVICE_VERSION:           OpenCL 1.2
3.7
CPU only Ryzen Threadripper 2970WX 24C/48T 4.2GHz, 64GB @3GHz RAM, NVMe SSD, Ubuntu 19.10 (toby) 7
+ GPU Nvidia GP107, GeForce GTX 1050 2GB
CUDA cores: 640
[opencl_init] device 0: GeForce GTX 1050 allows GPU memory allocations of up to 498MB
GLOBAL_MEM_SIZE:          1993MB
DRIVER_VERSION:           440.33.01
DEVICE_VERSION:           OpenCL 1.2 CUDA
6
darktable 2.4.4
CPU only Ryzen 7 2700X 8C/16T @4.0GHz, 32GB @3GHz RAM, NVMe SSD, Ubuntu 18.04 (hilbert) 11.8
+ GPU Radeon RX Vega64 8GB / Linux 4.15, AMDGPU-PRO 18.20
Compute Units: 64
Stream Processors: 4096
[opencl_init] device 0 `gfx900' allows GPU memory allocations of up to 4048MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           2639.3 (PAL,HSAIL)
DEVICE_VERSION:           OpenCL 1.2 AMD-APP (2639.3)
 2.5
darktable 2.4.3
CPU only Ryzen 7 2700X 8C/16T @4.0GHz, 32GB @3GHz RAM, NVMe SSD, Ubuntu 18.04 (hilbert) 11.8
+ GPU open source Radeon RX 480 8GB (Ellesmere) / Linux 4.13 amdgpu, ROCm 1.7
(github.com/RadeonOpenCompute/ROCm) 
Compute Units: 36
Stream Processors: 2304
[opencl_init] device 0 `gfx803' allows GPU memory allocations of up to 6963MB
GLOBAL_MEM_SIZE:          8192MB
DRIVER_VERSION:           2576.0 (HSA1.1,LC)
DEVICE_VERSION:           OpenCL 1.2
 3.7
+ GPU open source Radeon RX Vega64 8GB / Linux 4.13 amdgpu, ROCm 1.8
(github.com/RadeonOpenCompute/ROCm) 
Compute Units: 64
Stream Processors: 4096
[opencl_init] device 0 `gfx900' allows GPU memory allocations of up to 6949MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           2617.0 (HSA1.1,LC)
DEVICE_VERSION:           OpenCL 1.2
 3.2
+ GPU Radeon RX Vega64 8GB / Linux 4.13, AMDGPU-PRO 18.10
Compute Units: 64
Stream Processors: 4096
[opencl_init] device 0 `gfx900' allows GPU memory allocations of up to 4048MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           2580.4 (PAL,HSAIL)
DEVICE_VERSION:           OpenCL 1.2 AMD-APP (2580.4)
 2.5
darktable 2.4.2
CPU only Ryzen 7 2700X 8C/16T 4.0GHz, 32GB @2.1GHz RAM, NVMe SSD, Ubuntu 18.04 (hilbert) 15.4
+ GPU open source Radeon RX 480 8GB (Ellesmere) / Linux 4.13 amdgpu, ROCm 1.7
(github.com/RadeonOpenCompute/ROCm) 
[opencl_init] device 0 `gfx803' allows GPU memory allocations of up to 6963MB
GLOBAL_MEM_SIZE:          8192MB
DRIVER_VERSION:           2576.0 (HSA1.1,LC)
DEVICE_VERSION:           OpenCL 1.2
 3.8
CPU only Ryzen 5 2600 6C/12T 3.9GHz, 8GB 2666MHz RAM, Barracuda ST2000DM001 HDD
Linux 4.15, Xubuntu 18.04.1 LTS. Contributed by Přemek Brada.		 21.9
+ GPU GeForce GT 1030 (384 CUDA cores), 2GB GDDR5, passive cooling
[opencl_init] device 0: GeForce GT 1030 allows GPU memory allocations of up to 498MB
GLOBAL_MEM_SIZE:          1992MB
DRIVER_VERSION:           390.77
DEVICE_VERSION:           OpenCL 1.2 CUDA
 16.3
CPU only Ryzen 7 1700 8C/16T 3GHz, 32GB RAM, SATA-II SSD, Ubuntu 18.04.01 (glab2) 14.6
+ GPU NVIDIA GP104, GeForce GTX 1080 8GB 
[opencl_init] device 0 `GeForce GTX 1080' allows GPU memory allocations of up to 2029MB
GLOBAL_MEM_SIZE:          8116MB
DRIVER_VERSION:           390.48
DEVICE_VERSION:           OpenCL 1.2 CUDA
 3.1
CPU only Xeon E5-2643 8C/16T 3.3GHz, 128GB RAM, Ubuntu 18.04 (molly) 10.6
+ GPU Radeon R9 390X 8GB (Hawaii XT) / Linux 4.18 / AMDGPU-PRO 19.10 (--opencl=legacy --headless) 
[opencl_init] device 0 `Hawaii' allows GPU memory allocations of up to 4048MB
GLOBAL_MEM_SIZE:          8187MB
DRIVER_VERSION:           2841.4
DEVICE_VERSION:           OpenCL 1.2 AMD-APP (2841.4)
 4.1
+ GPU NVIDIA TU106, GeForce RTX 2070 8GB 
[opencl_init] device 1 `GeForce RTX 2070' allows GPU memory allocations of up to 1988MB
GLOBAL_MEM_SIZE:          7952MB
DRIVER_VERSION:           418.67
DEVICE_VERSION:           OpenCL 1.2 CUDA
 2.7
darktable 2.4.1 and earlier
CPU only Ryzen 5 1600 6C/12T 3.2GHz, 32GB RAM, NVMe SSD, Ubuntu 17.10 (hilbert) 14.8
+ GPU Radeon RX 480 8GB (Ellesmere) / Linux 4.10 amdgpu / AMDGPU-PRO 17.50 (--opencl=legacy --headless) 
[opencl_init] device 0 `Ellesmere' allows GPU memory allocations of up to 4048MB
GLOBAL_MEM_SIZE:          8003MB
DRIVER_VERSION:           2527.3
DEVICE_VERSION:           OpenCL 1.2 AMD-APP (2527.3)
 3.8
CPU only Core i7-6700 3.4GHz 4C/8T, 32GB RAM, NVMe SSD (hilbert) 29.0
+ GPU Radeon RX 480 8GB (Ellesmere)
[opencl_init]
device 0 `Ellesmere' allows GPU memory allocations of up to 4048MB
GLOBAL_MEM_SIZE:          7963MB
DRIVER_VERSION:           2348.3
DEVICE_VERSION:           OpenCL 1.2 AMD-APP (2348.3)
 4.4
CPU only Ryzen 7 1700 3GHz 8C/16T, 32GB RAM, NVMe SSD (hilbert) 16.8
+ GPU Radeon RX 480 8GB (Ellesmere)
[opencl_init]
device 0 `Ellesmere' allows GPU memory allocations of up to 4048MB
GLOBAL_MEM_SIZE:          7851MB
DRIVER_VERSION:           2348.3
DEVICE_VERSION:           OpenCL 1.2 AMD-APP (2348.3)
 4.2
CPU only Ryzen 5 1600 3.5GHz 6C/12T, 16GB RAM, NVMe SSD (storas) 14.4
+ GPU GeForce GTX 1070 8GB
[opencl_init]
device 0 `GeForce GTX 1070' allows GPU memory allocations of up to 2027MB
GLOBAL_MEM_SIZE:          8111MB
DRIVER_VERSION:           375.39
DEVICE_VERSION:           OpenCL 1.2 CUDA
 3.2
CPU only Ryzen 5 1600 @3.6GHz 6C/12T, 16GB RAM @3GT/s, NVMe SSD (storas) 11.8
+ GPU GeForce GTX 1070 8GB
[opencl_init]
device 0 `GeForce GTX 1070' allows GPU memory allocations of up to 2029MB
GLOBAL_MEM_SIZE:          8116MB
DRIVER_VERSION:           440.33.01
DEVICE_VERSION:           OpenCL 1.2 CUDA
 3.1
CPU only Xeon E5-1650 v4 3.6GHz 6C/12T, 64GB RAM, SSD (abacus) 13.3
+ GPU TITAN X (rev a1) 12GB
[opencl_init]
device 0 `TITAN X (Pascal)' allows GPU memory allocations of up to 3047MB
GLOBAL_MEM_SIZE:          12190MB
DRIVER_VERSION:           375.88
DEVICE_VERSION:           OpenCL 1.2 CUDA
 2.5
CPU only Core i7-6700 CPU 3.40GHz 4C/8T, 16GB RAM, HDD (geom) 25.6
+ GPU Radeon R9 290X 4GB (Hawaii XT) 
[opencl_init]
device 0 `Hawaii' allows GPU memory allocations of up to 2879MB
GLOBAL_MEM_SIZE:          4083MB
DRIVER_VERSION:           2442.7
DEVICE_VERSION:           OpenCL 1.2 AMD-APP (2442.7)
 4.0
+ GPU Radeon R9 390X 8GB (Hawaii XT) 
[opencl_init]
device 0 `Hawaii' allows GPU memory allocations of up to 2879MB
GLOBAL_MEM_SIZE:          4083MB
DRIVER_VERSION:           2442.7
DEVICE_VERSION:           OpenCL 1.2 AMD-APP (2442.7)
 3.8
CPU only Ryzen 5 1600 6C/12T 3.2GHz, 16GB RAM, NVMe SSD, Ubuntu 17.10β (hilbert) 16.1
+ GPU Radeon RX 480 8GB (Ellesmere) / AMDGPU-PRO 17.30 / Linux 4.10 
[opencl_init]
device 0 `Ellesmere' allows GPU memory allocations of up to 4048MB
GLOBAL_MEM_SIZE:          7991MB
DRIVER_VERSION:           2442.7
DEVICE_VERSION:           OpenCL 1.2 AMD-APP (2442.7)
 4.1
+ GPU open source Radeon RX 480 8GB (Ellesmere) / Linux 4.11 amdgpu, ROCm 1.6
(github.com/RadeonOpenCompute/ROCm) 
[opencl_init]
device 0 `gfx803' allows GPU memory allocations of up to 6963MB
GLOBAL_MEM_SIZE:          8192MB
DRIVER_VERSION:           1.1 (HSA,LC)
DEVICE_VERSION:           OpenCL 1.2
 4.6
CPU only Ryzen 5 1600 6C/12T 3.2GHz, 16GB RAM, NVMe SSD, Ubuntu 17.10 (hilbert) 16.1
+ GPU open source Radeon R9 390X 8GB (Hawaii XT) / Linux 4.11 amdgpu, ROCm 1.6
(github.com/RadeonOpenCompute/ROCm) 
[opencl_init]
evice 0 `gfx701' allows GPU memory allocations of up to 6963MB
GLOBAL_MEM_SIZE:          8192MB
DRIVER_VERSION:           1.1 (HSA,LC)
DEVICE_VERSION:           OpenCL 1.2
 4.0
+ GPU open source Radeon RX Vega 64 8GB / Linux 4.11 amdgpu, ROCm 1.6
(github.com/RadeonOpenCompute/ROCm) 
[opencl_init]
device 0 `gfx900' allows GPU memory allocations of up to 6949MB
GLOBAL_MEM_SIZE:          8176MB
DRIVER_VERSION:           1.1 (HSA,LC)
DEVICE_VERSION:           OpenCL 1.2
 3.4
CPU only Core i7-3770 4C/8T 3.4GHz, 24GB RAM, SSD (glab2) 31
+ GPU GeForce GTX 1080 8GB
[opencl_init]
device 0 `GeForce GTX 1080' allows GPU memory allocations of up to 2028MB
GLOBAL_MEM_SIZE:          8112MB
DRIVER_VERSION:           387.34
DEVICE_VERSION:           OpenCL 1.2 CUDA
 3.8

/~sarunas