Blender 3.3 AMD Radeon HIP vs. NVIDIA CUDA/OptiX Performance

[0xff]

$ blender -noaudio -b bmw27_gpu.blend -o /tmp/outputfile -f 1
Blender 3.3.1 (hash b292cfe5a936 built 2022-11-05 23:21:59)
Read prefs: /home/rnz/.config/blender/3.3/config/userpref.blend
Read blend: /home/rnz/Downloads/blender/benchmark/bmw27/bmw27_gpu.blend
Fra:1 Mem:62.66M (Peak 62.67M) | Time:00:00.23 | Mem:0.00M, Peak:0.00M | Scene, RenderLayer | Synchronizing object | Light
...
Fra:1 Mem:239.91M (Peak 259.79M) | Time:00:01.12 | Remaining:01:32.20 | Mem:386.99M, Peak:386.99M | Scene, RenderLayer | Sample 1/1225
Fra:1 Mem:239.97M (Peak 259.79M) | Time:00:20.62 | Remaining:00:27.16 | Mem:387.05M, Peak:387.05M | Scene, RenderLayer | Sample 513/1225
Fra:1 Mem:251.84M (Peak 271.62M) | Time:00:48.75 | Mem:387.05M, Peak:387.05M | Scene, RenderLayer | Sample 1225/1225
Fra:1 Mem:251.84M (Peak 271.62M) | Time:00:48.75 | Mem:387.05M, Peak:387.05M | Scene, RenderLayer | Finished
...
Fra:1 Mem:90.70M (Peak 271.62M) | Time:00:48.95 | Compositing | De-initializing execution
Saved: '/tmp/outputfile0001.png'
 Time: 00:49.53 (Saving: 00:00.58)

$ rocminfo 
ROCk module is loaded
=====================    
HSA System Attributes    
=====================    
Runtime Version:         1.1
System Timestamp Freq.:  1000.000000MHz
Sig. Max Wait Duration:  18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count)
Machine Model:           LARGE                              
System Endianness:       LITTLE                             

==========               
HSA Agents               
==========               
*******                  
Agent 1                  
*******                  
  Name:                    AMD Ryzen 9 5900X 12-Core Processor
  Uuid:                    CPU-XX                             
  Marketing Name:          AMD Ryzen 9 5900X 12-Core Processor
  Vendor Name:             CPU                                
  Feature:                 None specified                     
  Profile:                 FULL_PROFILE                       
  Float Round Mode:        NEAR                               
  Max Queue Number:        0(0x0)                             
  Queue Min Size:          0(0x0)                             
  Queue Max Size:          0(0x0)                             
  Queue Type:              MULTI                              
  Node:                    0                                  
  Device Type:             CPU                                
  Cache Info:              
    L1:                      32768(0x8000) KB                   
  Chip ID:                 0(0x0)                             
  ASIC Revision:           0(0x0)                             
  Cacheline Size:          64(0x40)                           
  Max Clock Freq. (MHz):   3700                               
  BDFID:                   0                                  
  Internal Node ID:        0                                  
  Compute Unit:            24                                 
  SIMDs per CU:            0                                  
  Shader Engines:          0                                  
  Shader Arrs. per Eng.:   0                                  
  WatchPts on Addr. Ranges:1                                  
  Features:                None
  Pool Info:               
    Pool 1                   
      Segment:                 GLOBAL; FLAGS: FINE GRAINED        
      Size:                    32785892(0x1f445e4) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
    Pool 2                   
      Segment:                 GLOBAL; FLAGS: KERNARG, FINE GRAINED
      Size:                    32785892(0x1f445e4) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
    Pool 3                   
      Segment:                 GLOBAL; FLAGS: COARSE GRAINED      
      Size:                    32785892(0x1f445e4) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
  ISA Info:                
*******                  
Agent 2                  
*******                  
  Name:                    gfx1010                            
  Uuid:                    GPU-XX                             
  Marketing Name:          AMD Radeon RX 5700 XT              
  Vendor Name:             AMD                                
  Feature:                 KERNEL_DISPATCH                    
  Profile:                 BASE_PROFILE                       
  Float Round Mode:        NEAR                               
  Max Queue Number:        128(0x80)                          
  Queue Min Size:          64(0x40)                           
  Queue Max Size:          131072(0x20000)                    
  Queue Type:              MULTI                              
  Node:                    1                                  
  Device Type:             GPU                                
  Cache Info:              
    L1:                      16(0x10) KB                        
    L2:                      4096(0x1000) KB                    
  Chip ID:                 29471(0x731f)                      
  ASIC Revision:           2(0x2)                             
  Cacheline Size:          64(0x40)                           
  Max Clock Freq. (MHz):   2100                               
  BDFID:                   3328                               
  Internal Node ID:        1                                  
  Compute Unit:            40                                 
  SIMDs per CU:            2                                  
  Shader Engines:          4                                  
  Shader Arrs. per Eng.:   2                                  
  WatchPts on Addr. Ranges:4                                  
  Features:                KERNEL_DISPATCH 
  Fast F16 Operation:      TRUE                               
  Wavefront Size:          32(0x20)                           
  Workgroup Max Size:      1024(0x400)                        
  Workgroup Max Size per Dimension:
    x                        1024(0x400)                        
    y                        1024(0x400)                        
    z                        1024(0x400)                        
  Max Waves Per CU:        40(0x28)                           
  Max Work-item Per CU:    1280(0x500)                        
  Grid Max Size:           4294967295(0xffffffff)             
  Grid Max Size per Dimension:
    x                        4294967295(0xffffffff)             
    y                        4294967295(0xffffffff)             
    z                        4294967295(0xffffffff)             
  Max fbarriers/Workgrp:   32                                 
  Pool Info:               
    Pool 1                   
      Segment:                 GLOBAL; FLAGS: COARSE GRAINED      
      Size:                    8372224(0x7fc000) KB               
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       FALSE                              
    Pool 2                   
      Segment:                 GROUP                              
      Size:                    64(0x40) KB                        
      Allocatable:             FALSE                              
      Alloc Granule:           0KB                                
      Alloc Alignment:         0KB                                
      Accessible by all:       FALSE                              
  ISA Info:                
    ISA 1                    
      Name:                    amdgcn-amd-amdhsa--gfx1010:xnack-  
      Machine Models:          HSA_MACHINE_MODEL_LARGE            
      Profiles:                HSA_PROFILE_BASE                   
      Default Rounding Mode:   NEAR                               
      Default Rounding Mode:   NEAR                               
      Fast f16:                TRUE                               
      Workgroup Max Size:      1024(0x400)                        
      Workgroup Max Size per Dimension:
        x                        1024(0x400)                        
        y                        1024(0x400)                        
        z                        1024(0x400)                        
      Grid Max Size:           4294967295(0xffffffff)             
      Grid Max Size per Dimension:
        x                        4294967295(0xffffffff)             
        y                        4294967295(0xffffffff)             
        z                        4294967295(0xffffffff)             
      FBarrier Max Size:       32                                 
*** Done ***     
​

$ yay -Ss rocm | grep -i installed
aur/rocm-core 5.3.3-1 (+2 0.17) (Installed)
aur/opencl-amd-dev 1:5.4.0-1 (+5 1.15) (Installed)
aur/rocm-llvm 5.4.0-1 (+11 0.26) (Installed)
​

[finalzone]

Better late never. After getting HIP working with ROCm-GFX8P repository from COPR, the benchmark is finally available running Sapphire Nitro+ Pure AMD Radeon RX 6950XT.

hipinfo

--------------------------------------------------------------------------------
device#                           0
Name:                             AMD Radeon RX 6950 XT
pciBusID:                         10
pciDeviceID:                      0
pciDomainID:                      0
multiProcessorCount:              40
maxThreadsPerMultiProcessor:      2048
isMultiGpuBoard:                  0
clockRate:                        2720 Mhz
memoryClockRate:                  1124 Mhz
memoryBusWidth:                   256
totalGlobalMem:                   15.98 GB
totalConstMem:                    2147483647
sharedMemPerBlock:                64.00 KB
canMapHostMemory:                 1
regsPerBlock:                     65536
warpSize:                         32
l2CacheSize:                      4194304
computeMode:                      0
maxThreadsPerBlock:               1024
maxThreadsDim.x:                  1024
maxThreadsDim.y:                  1024
maxThreadsDim.z:                  1024
maxGridSize.x:                    2147483647
maxGridSize.y:                    2147483647
maxGridSize.z:                    2147483647
major:                            10
minor:                            3
concurrentKernels:                1
cooperativeLaunch:                0
cooperativeMultiDeviceLaunch:     0
isIntegrated:                     0
maxTexture1D:                     16384
maxTexture2D.width:               16384
maxTexture2D.height:              16384
maxTexture3D.width:               16384
maxTexture3D.height:              16384
maxTexture3D.depth:               8192
isLargeBar:                       1
asicRevision:                     1
maxSharedMemoryPerMultiProcessor: 64.00 KB
clockInstructionRate:             1000.00 Mhz
arch.hasGlobalInt32Atomics:       1
arch.hasGlobalFloatAtomicExch:    1
arch.hasSharedInt32Atomics:       1
arch.hasSharedFloatAtomicExch:    1
arch.hasFloatAtomicAdd:           1
arch.hasGlobalInt64Atomics:       1
arch.hasSharedInt64Atomics:       1
arch.hasDoubles:                  1
arch.hasWarpVote:                 1
arch.hasWarpBallot:               1
arch.hasWarpShuffle:              1
arch.hasFunnelShift:              0
arch.hasThreadFenceSystem:        1
arch.hasSyncThreadsExt:           0
arch.hasSurfaceFuncs:             0
arch.has3dGrid:                   1
arch.hasDynamicParallelism:       0
gcnArchName:                      gfx1030
peers:                            
non-peers:                        device#0

memInfo.total:                    15.98 GB

​

[shanedav4]

I'm curious as to why the RX 6500 and RX 6400 were tested? My understanding was that all of the media creation parts of those two gpus are disabled.

[qarium]

I think if “GIMP” had a name that sounded better than a slur for someone with physical disabilities it may gain some traction.
Large companies don’t want to be seen supporting “gimped” software

there are multiple reasons why GIMP did not become important similar to blender.

there are many other opensource projects who did not go airborne yet ... Inkscape is another example.

i think we will soon see that in the next 5 years that many of these projects go the same route as blender.

[Eirikr1848]

this shows how opensource software become more and more important there is blender as battleground because it is important in the industry now.

Krita for example did get intel as sponsor because krita goes the same route as blender...

GIMP should be the same but is not yet important... but i think in the near future GIMP becomes a complete Adobe Photoshop replacement and because of this becomes important in the future.

as soon as such a software like blender becomes a battleground its like game over for the closed source ecosystem.

I think if “GIMP” had a name that sounded better than a slur for someone with physical disabilities it may gain some traction.

Large companies don’t want to be seen supporting “gimped” software

[qarium]

I don't know if this is a coincidence or a consequence, but looks like the AMD and Intel GPU stack development is — somehow — related with the Blender growth: the moment in which the latter became an Industry standard the two companies started to push on the accelerator... 🤔

this shows how opensource software become more and more important there is blender as battleground because it is important in the industry now.

Krita for example did get intel as sponsor because krita goes the same route as blender...

GIMP should be the same but is not yet important... but i think in the near future GIMP becomes a complete Adobe Photoshop replacement and because of this becomes important in the future.

as soon as such a software like blender becomes a battleground its like game over for the closed source ecosystem.

[Danielsan]

I don't know if this is a coincidence or a consequence, but looks like the AMD and Intel GPU stack development is — somehow — related with the Blender growth: the moment in which the latter became an Industry standard the two companies started to push on the accelerator... 🤔

[finalzone]

I just got Sapphire Nitro+ Pure Radeon RX 6950X for Ryzen 9 5900XT running on Fedora 37 Beta. Unfortunately, the HIP failed to get detected for some odd reasons but ROCM OpenCL got significant improvement, shame OpenCL support got dropped. I may try the benchmark this weekend.

[darkbasic]

maybe you can organise a growdfunding for the 6950xt

Maybe AMD could provide Michael a f*ck*ng card for free? AMD WTF? You're losing way more money every time he publishes a review with top of the line Nvidia GPUs and no AMD counterpart. Computing aside AMD blows Nvidia away in rasterization on Linux and yet you decide to not provide the highest end cards. I'll never understand.

[Mr.Elendig]

Wow, AMD looks really bad!

Specially for us rdna1 users. We were thrown under the buss.