国家超算中心 昆山站 异构加速卡1 显存16GB详细配置， 海光 Z100SM HCU

zhangfeng11332026-06-16 11:03

一、国家超算中心 昆山站 异构加速卡1 显存16GB详细配置， 海光 Z100SM HCU整体环境概览

c 复制代码 
rocminfo
amdgpu 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                             
Mwaitx:                  DISABLED
DMAbuf Support:          NO

==========               
HSA Agents               
==========               
*******                  
Agent 1                  
*******                  
  Name:                    Hygon C86 7185 32-core Processor   
  Uuid:                    CPU-XX                             
  Marketing Name:          Hygon C86 7185 32-core Processor   
  Vendor Name:             HygonGenuine                       
  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):   2000                               
  BDFID:                   0                                  
  Internal Node ID:        0                                  
  Compute Unit:            8                                  
  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:                    32606196(0x1f187f4) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
    Pool 2                   
      Segment:                 GLOBAL; FLAGS: KERNARG, FINE GRAINED
      Size:                    32606196(0x1f187f4) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
    Pool 3                   
      Segment:                 GLOBAL; FLAGS: COARSE GRAINED      
      Size:                    32606196(0x1f187f4) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
  ISA Info:                
*******                  
Agent 2                  
*******                  
  Name:                    Hygon C86 7185 32-core Processor   
  Uuid:                    CPU-XX                             
  Marketing Name:          Hygon C86 7185 32-core Processor   
  Vendor Name:             HygonGenuine                       
  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:                    1                                  
  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):   2000                               
  BDFID:                   0                                  
  Internal Node ID:        1                                  
  Compute Unit:            8                                  
  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:                    33028140(0x1f7f82c) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
    Pool 2                   
      Segment:                 GLOBAL; FLAGS: KERNARG, FINE GRAINED
      Size:                    33028140(0x1f7f82c) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
    Pool 3                   
      Segment:                 GLOBAL; FLAGS: COARSE GRAINED      
      Size:                    33028140(0x1f7f82c) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
  ISA Info:                
*******                  
Agent 3                  
*******                  
  Name:                    Hygon C86 7185 32-core Processor   
  Uuid:                    CPU-XX                             
  Marketing Name:          Hygon C86 7185 32-core Processor   
  Vendor Name:             HygonGenuine                       
  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:                    2                                  
  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):   2000                               
  BDFID:                   0                                  
  Internal Node ID:        2                                  
  Compute Unit:            8                                  
  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:                    33028144(0x1f7f830) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
    Pool 2                   
      Segment:                 GLOBAL; FLAGS: KERNARG, FINE GRAINED
      Size:                    33028144(0x1f7f830) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
    Pool 3                   
      Segment:                 GLOBAL; FLAGS: COARSE GRAINED      
      Size:                    33028144(0x1f7f830) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
  ISA Info:                
*******                  
Agent 4                  
*******                  
  Name:                    Hygon C86 7185 32-core Processor   
  Uuid:                    CPU-XX                             
  Marketing Name:          Hygon C86 7185 32-core Processor   
  Vendor Name:             HygonGenuine                       
  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:                    3                                  
  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):   2000                               
  BDFID:                   0                                  
  Internal Node ID:        3                                  
  Compute Unit:            8                                  
  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:                    30914732(0x1d7b8ac) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
    Pool 2                   
      Segment:                 GLOBAL; FLAGS: KERNARG, FINE GRAINED
      Size:                    30914732(0x1d7b8ac) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
    Pool 3                   
      Segment:                 GLOBAL; FLAGS: COARSE GRAINED      
      Size:                    30914732(0x1d7b8ac) KB             
      Allocatable:             TRUE                               
      Alloc Granule:           4KB                                
      Alloc Alignment:         4KB                                
      Accessible by all:       TRUE                               
  ISA Info:                
*******                  
Agent 5                  
*******                  
  Name:                    gfx906                             
  Uuid:                    GPU-d420786172dc76e6               
  Marketing Name:          Vega 20 [Radeon Pro VII/Radeon Instinct MI50 32GB]
  Vendor Name:             HYGON                              
  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:                    4                                  
  Device Type:             HCU                                
  Cache Info:              
    L1:                      16(0x10) KB                        
  Chip ID:                 26273(0x66a1)                      
  ASIC Revision:           1(0x1)                             
  Cacheline Size:          64(0x40)                           
  Max Clock Freq. (MHz):   1700                               
  BDFID:                   1024                               
  Internal Node ID:        4                                  
  Compute Unit:            64                                 
  SIMDs per CU:            4                                  
  Shader Engines:          4                                  
  Shader Arrs. per Eng.:   1                                  
  WatchPts on Addr. Ranges:4                                  
  Features:                KERNEL_DISPATCH 
  Fast F16 Operation:      TRUE                               
  Wavefront Size:          64(0x40)                           
  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:    2560(0xa00)                        
  Grid Max Size:           4294967295(0xffffffff)             
  Grid Max Size per Dimension:
    x                        4294967295(0xffffffff)             
    y                        4294967295(0xffffffff)             
    z                        4294967295(0xffffffff)             
  Max fbarriers/Workgrp:   32                                 
  Packet Processor uCode:: 0                                  
  SDMA engine uCode::      0                                  
  IOMMU Support::          None                               
  Pool Info:               
    Pool 1                   
      Segment:                 GLOBAL; FLAGS: COARSE GRAINED      
      Size:                    16760832(0xffc000) 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--gfx906:sramecc-: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                                 
*******                  
Agent 6                  
*******                  
  Name:                    gfx906                             
  Uuid:                    GPU-2caa688172da5f17               
  Marketing Name:          Vega 20 [Radeon Pro VII/Radeon Instinct MI50 32GB]
  Vendor Name:             HYGON                              
  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:                    5                                  
  Device Type:             HCU                                
  Cache Info:              
    L1:                      16(0x10) KB                        
  Chip ID:                 26273(0x66a1)                      
  ASIC Revision:           1(0x1)                             
  Cacheline Size:          64(0x40)                           
  Max Clock Freq. (MHz):   1700                               
  BDFID:                   9728                               
  Internal Node ID:        5                                  
  Compute Unit:            64                                 
  SIMDs per CU:            4                                  
  Shader Engines:          4                                  
  Shader Arrs. per Eng.:   1                                  
  WatchPts on Addr. Ranges:4                                  
  Features:                KERNEL_DISPATCH 
  Fast F16 Operation:      TRUE                               
  Wavefront Size:          64(0x40)                           
  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:    2560(0xa00)                        
  Grid Max Size:           4294967295(0xffffffff)             
  Grid Max Size per Dimension:
    x                        4294967295(0xffffffff)             
    y                        4294967295(0xffffffff)             
    z                        4294967295(0xffffffff)             
  Max fbarriers/Workgrp:   32                                 
  Packet Processor uCode:: 0                                  
  SDMA engine uCode::      0                                  
  IOMMU Support::          None                               
  Pool Info:               
    Pool 1                   
      Segment:                 GLOBAL; FLAGS: COARSE GRAINED      
      Size:                    16760832(0xffc000) 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--gfx906:sramecc-: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                                 
*******                  
Agent 7                  
*******                  
  Name:                    gfx906                             
  Uuid:                    GPU-7c96396172fc1a8d               
  Marketing Name:          Vega 20 [Radeon Pro VII/Radeon Instinct MI50 32GB]
  Vendor Name:             HYGON                              
  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:                    6                                  
  Device Type:             HCU                                
  Cache Info:              
    L1:                      16(0x10) KB                        
  Chip ID:                 26273(0x66a1)                      
  ASIC Revision:           1(0x1)                             
  Cacheline Size:          64(0x40)                           
  Max Clock Freq. (MHz):   1700                               
  BDFID:                   17152                              
  Internal Node ID:        6                                  
  Compute Unit:            64                                 
  SIMDs per CU:            4                                  
  Shader Engines:          4                                  
  Shader Arrs. per Eng.:   1                                  
  WatchPts on Addr. Ranges:4                                  
  Features:                KERNEL_DISPATCH 
  Fast F16 Operation:      TRUE                               
  Wavefront Size:          64(0x40)                           
  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:    2560(0xa00)                        
  Grid Max Size:           4294967295(0xffffffff)             
  Grid Max Size per Dimension:
    x                        4294967295(0xffffffff)             
    y                        4294967295(0xffffffff)             
    z                        4294967295(0xffffffff)             
  Max fbarriers/Workgrp:   32                                 
  Packet Processor uCode:: 0                                  
  SDMA engine uCode::      0                                  
  IOMMU Support::          None                               
  Pool Info:               
    Pool 1                   
      Segment:                 GLOBAL; FLAGS: COARSE GRAINED      
      Size:                    16760832(0xffc000) 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--gfx906:sramecc-: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                                 
*******                  
Agent 8                  
*******                  
  Name:                    gfx906                             
  Uuid:                    GPU-d6dc514172da5f15               
  Marketing Name:          Vega 20 [Radeon Pro VII/Radeon Instinct MI50 32GB]
  Vendor Name:             HYGON                              
  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:                    7                                  
  Device Type:             HCU                                
  Cache Info:              
    L1:                      16(0x10) KB                        
  Chip ID:                 26273(0x66a1)                      
  ASIC Revision:           1(0x1)                             
  Cacheline Size:          64(0x40)                           
  Max Clock Freq. (MHz):   1700                               
  BDFID:                   25344                              
  Internal Node ID:        7                                  
  Compute Unit:            64                                 
  SIMDs per CU:            4                                  
  Shader Engines:          4                                  
  Shader Arrs. per Eng.:   1                                  
  WatchPts on Addr. Ranges:4                                  
  Features:                KERNEL_DISPATCH 
  Fast F16 Operation:      TRUE                               
  Wavefront Size:          64(0x40)                           
  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:    2560(0xa00)                        
  Grid Max Size:           4294967295(0xffffffff)             
  Grid Max Size per Dimension:
    x                        4294967295(0xffffffff)             
    y                        4294967295(0xffffffff)             
    z                        4294967295(0xffffffff)             
  Max fbarriers/Workgrp:   32                                 
  Packet Processor uCode:: 0                                  
  SDMA engine uCode::      0                                  
  IOMMU Support::          None                               
  Pool Info:               
    Pool 1                   
      Segment:                 GLOBAL; FLAGS: COARSE GRAINED      
      Size:                    16760832(0xffc000) 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--gfx906:sramecc-: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 ***

rocminfo 输出正常，amdgpu 驱动模块已加载 ，ROCm HSA 运行时版本 1.1，整机识别到 4 颗 CPU + 4 张 GPU。

二、CPU 节点（Agent1~Agent4）

  • 型号：海光 C86 7185 32核处理器，共 4 个 NUMA 节点
  • 主频：最高 2000 MHz，L1 缓存单节点 32768 KB
  • 类型：纯 CPU 计算节点，无 GPU 计算单元，仅提供主机内存
  • 内存：各节点主机内存容量略有差异，支持细粒度/粗粒度内存分配

三、GPU 节点（Agent5~Agent8，共 4 卡）

基础参数

  • 架构：gfx906（Vega20） ，对应型号 Radeon Instinct MI50 32GB / Radeon Pro VII
  • 厂商：HYGON 海光，ASIC 版本 1
  • 主频：最高 1700 MHz
  • 显存：单卡显存 16760832 KB = 16368 MB ≈ 32GB，和规格一致
  • 指令集：amdgcn-amd-amdhsa--gfx906:sramecc-:xnack-（无 SRAM-ECC、无 XNACK）

计算核心规格

  • 计算单元 CU：64 个/卡
  • 每个 CU 包含 4 个 SIMD，4 个着色引擎
  • 波前尺寸 Wavefront：64
  • 工作组最大尺寸：单维 1024，三维上限均为 1024
  • 支持 FP16 快速运算，适合推理、AI 计算场景

四、关键适配&使用建议

  1. ISA 标识重点

    所有 GPU 统一为 gfx906:sramecc-:xnack-，编译 ROCm 程序、PyTorch/TensorFlow、HIP 代码时，架构指定 gfx906 ，关闭 sramecc/xnack。

    示例环境变量：

    bash 复制代码 
    export ROCM_ARCH=gfx906
export HIP_ARCH=gfx906

  2. 显存与内存

    单卡 32GB 大显存，适合大模型推理、深度学习训练、HPC 并行计算；GPU 显存为独立显存，不可跨卡直接共享。

  3. ROCm 版本兼容

    当前 HSA Runtime 1.1，gfx906 对主流 ROCm 4.x ~ 5.x 全系兼容，不建议使用过新的 6.x 大版本（易出现适配问题）。

  4. 调度与任务分配

    4 张 GPU 分别对应 Node4~Node7，可结合 NUMA 绑定：CPU 节点与本地 GPU 亲和绑定，降低跨 NUMA 内存延迟。

  5. 功能限制

    输出显示 IOMMU Support: None、DMAbuf Support: NO，常规 HIP/ROCm 计算、AI 框架不受影响；如需容器直通、DMA 零拷贝场景会受限。

五、小结

整机为 4 路海光 C86 CPU + 4 卡 MI50 32GB(gfx906) 异构计算集群，ROCm 驱动、amdgpu 模块加载正常，硬件识别完整，可直接运行基于 ROCm/HIP 的深度学习、HPC、并行计算任务 ，编译时务必指定目标架构 gfx906。

相关推荐
青瓦梦滋1 小时前
Linux：TCP协议的socket套接字
网络·网络协议·tcp/ip
z落落1 小时前
C# WinForm TreeView 树形控件+ListView控件+菜单栏
开发语言·c#
烂白菜1 小时前
码道启辰：定时任务自由编排
运维·服务器·网络
梁辰兴1 小时前
计算机网络基础：对称加密密码体制
网络·计算机网络·计算机·对称加密·计算机网络基础·梁辰兴
Zhan8611241 小时前
WebSocket心跳与断线重连实战：芬兰赫尔辛基指数行情数据接口接入记录
网络·websocket·网络协议
悠悠121381 小时前
Linux 7.1 来了：新 NTFS 驱动、干掉 i486、FRED 默认开启，这次更新有点东西
linux·运维·服务器
用户805533698032 小时前
Linux 工作队列：把中断里做不了的事推迟到进程上下文
linux·嵌入式
大奎帝国2 小时前
Segearth-R2-03
深度学习·机器学习·计算机视觉
pingglala2 小时前
winscp连接linux失败解决方法
java·linux·服务器
热门推荐
012026年6月AI行业全景：从百模大战到Agent元年，这30天发生了什么？022026 AI 编程工具终极实战指南：Cursor vs Claude Code vs Copilot，开发者该怎么选？032026 年 AI 编程工具终极横评：Cursor vs Claude Code vs Copilot vs Windsurf04HTTP 与 HTTPS 的区别：从原理到实战详解05【AI】2026 年具身智能模型和世界模型总结06GitHub 镜像站点07AI科技热点日报 | 2026年6月1日08上线仅72小时被强制下架：Claude Fable 5 的短命09Claude Code、Codex、Cursor三分天下：2026年AI编程Agent生态全景剖析10《置身钉内》原文-可播放阅读