我刚刚第一次将带有 eGPU 的 Razor Core X 连接到我的 Thinkpad。风扇在旋转,但nvidia-smi
没有显示 eGPU。
我能做些什么?
答案1
首先,检查dmesg | tail -n 200
。它可能会显示如下内容:
[ 74.959198] thunderbolt 0000:06:00.0: current switch config:
[ 74.959201] thunderbolt 0000:06:00.0: Switch: 8086:15da (Revision: 6, TB Version: 2)
[ 74.959202] thunderbolt 0000:06:00.0: Max Port Number: 11
[ 74.959203] thunderbolt 0000:06:00.0: Config:
[ 74.959204] thunderbolt 0000:06:00.0: Upstream Port Number: 1 Depth: 1 Route String: 0x3 Enabled: 1, PlugEventsDelay: 254ms
[ 74.959205] thunderbolt 0000:06:00.0: unknown1: 0x0 unknown4: 0x0
[ 74.999560] thunderbolt 0000:06:00.0: 3: reading drom (length: 0x56)
[ 75.301575] thunderbolt 0000:06:00.0: 3: uid: 0x1279cc9b0ba8400
[ 75.301686] thunderbolt 0000:06:00.0: Port 0: 8086:15d3 (Revision: 6, TB Version: 1, Type: Port (0x1))
[ 75.301689] thunderbolt 0000:06:00.0: Max hop id (in/out): 7/7
[ 75.301692] thunderbolt 0000:06:00.0: Max counters: 8
[ 75.301694] thunderbolt 0000:06:00.0: NFC Credits: 0x800000
[ 75.302174] thunderbolt 0000:06:00.0: Port 1: 8086:15d3 (Revision: 6, TB Version: 1, Type: Port (0x1))
[ 75.302178] thunderbolt 0000:06:00.0: Max hop id (in/out): 15/15
[ 75.302180] thunderbolt 0000:06:00.0: Max counters: 16
[ 75.302183] thunderbolt 0000:06:00.0: NFC Credits: 0x7800000
[ 75.302681] thunderbolt 0000:06:00.0: Port 2: 8086:15d3 (Revision: 6, TB Version: 1, Type: Port (0x1))
[ 75.302683] thunderbolt 0000:06:00.0: Max hop id (in/out): 15/15
[ 75.302685] thunderbolt 0000:06:00.0: Max counters: 16
[ 75.302687] thunderbolt 0000:06:00.0: NFC Credits: 0x0
[ 75.302689] thunderbolt 0000:06:00.0: 3:3: disabled by eeprom
[ 75.302691] thunderbolt 0000:06:00.0: 3:4: disabled by eeprom
[ 75.302692] thunderbolt 0000:06:00.0: 3:5: disabled by eeprom
[ 75.302806] thunderbolt 0000:06:00.0: Port 6: 8086:15d3 (Revision: 6, TB Version: 1, Type: PCIe (0x100102))
[ 75.302808] thunderbolt 0000:06:00.0: Max hop id (in/out): 8/8
[ 75.302809] thunderbolt 0000:06:00.0: Max counters: 2
[ 75.302811] thunderbolt 0000:06:00.0: NFC Credits: 0x800000
[ 75.302960] thunderbolt 0000:06:00.0: Port 7: 8086:15d3 (Revision: 6, TB Version: 1, Type: PCIe (0x100101))
[ 75.302962] thunderbolt 0000:06:00.0: Max hop id (in/out): 8/8
[ 75.302964] thunderbolt 0000:06:00.0: Max counters: 2
[ 75.302966] thunderbolt 0000:06:00.0: NFC Credits: 0x800000
[ 75.302967] thunderbolt 0000:06:00.0: 3:8: disabled by eeprom
[ 75.302969] thunderbolt 0000:06:00.0: 3:9: disabled by eeprom
[ 75.302971] thunderbolt 0000:06:00.0: 3:a: disabled by eeprom
[ 75.302973] thunderbolt 0000:06:00.0: 3:b: disabled by eeprom
这表明存在权限/安全问题。
让我们安装 Thunderbolt 管理工具以便我们可以修复它:
sudo apt install thunderbolt-tools
现在,让我们检查一下 Thunderbolt 是否能够看到该底座:
root@mymachine:~# tbtadm devices
0-4 Razer Core X non-authorized not in ACL
是的!
现在,让我们授权码头:
tbtadm approve 0-4
由此可见:
Authorizing "/sys/bus/thunderbolt/devices/0-4"
Already in ACL
system:5 Input/output error
拔下并重新插入底座并dmesg
再次查看显示:
[11187.232181] thunderbolt 0000:06:00.0: PCIe tunnel creation failed
因此我们再次看一下 Thunderbolt:
root@mymachine:~# tbtadm devices
0-4 Razer Core X non-authorized in ACL
事实上,我们可以看到基座已连接:
root@mymachine:~# tbtadm acl
0XXXXXb0-XXXX-XXXX-ffff-ffffffffffff Razer Core X connected
让我们尝试手动授权:
root@mymachine:~# echo '1' > /sys/bus/thunderbolt/devices/0-4/authorized
-bash: echo: write error: Input/output error
此时,我怀疑问题可能出在 BIOS 上。因此,重新启动,调出 BIOS 设置。它位于“用户授权”中,但让我们使用核“无安全性”选项(可能有助于弄清楚如何稍后再次锁定):
此时,重新启动机器。
在插入 GPU 之前,请确保已加载 Nvidia 驱动程序:
sudo modprobe nvidia-uvm
并尝试查找 GPU:
nvidia-smi
成功!
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.87.00 Driver Version: 418.87.00 CUDA Version: 10.1 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce RTX 208... Off | 00000000:3D:00.0 Off | N/A |
| 15% 36C P0 1W / 250W | 0MiB / 10989MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+
根据这个文件BIOS 设置含义:
- 没有安全性:允许 Thunderbolt 设备自动连接。
- 用户授权:允许用户授权后连接Thunderbolt设备。
- 安全连接:允许使用已由用户批准的已保存密钥连接 Thunderbolt 设备。
- 显示端口和 USB:仅允许连接显示输出和 USB 设备。不允许连接 Thunderbolt 设备