计算模式禁止GRID M60

亲爱的大家，
我已经虚拟化了NVIDIA Tesla M60，以便从多个虚拟机中使用它。
当我为虚拟机应用GRID M60-8Q GPU配置文件时，我可以通过“nvidia-smi.exe”看到计算模式为“默认”：
C： Program Files  NVIDIA Corporation  NVSMI> nvidia-smi
星期四07月18日18:11:22
+ -------------------------------------------------
---------------------------- +
|
NVIDIA-SMI 370.12驱动程序版本：370.12 |
| ------------------------------- + -----------------
----- + ---------------------- +
|
GPU名称TCC / WDDM |
Bus-Id Disp.A |
挥发性的Uncorr。
ECC |
|
Fan Temp Perf Pwr：用法/上限|
内存使用|
GPU-Util Compute M. |
| =============================== + =================
===== + ====================== |
|
0 GRID M60-8Q WDDM |
0000：02：04.0 On |
N / A |
|
N / A N / A P0 N / A / N / A |
755MiB / 8192MiB |
4％违约|
+ ------------------------------- + -----------------
----- + ---------------------- +
+ -------------------------------------------------
---------------------------- +
|
进程：GPU内存|
|
GPU PID类型进程名称用法|
| =================================================
============================ |
|
0 1600 C + G权限不足N / A |
+ -------------------------------------------------
---------------------------- +
但是当我应用任何其他GPU配置文件（例如，GRID M60-2Q）时，计算模式是“禁止的”，我无法将其更改为“默认”（尽管结果来自“nvidia-smi -c 0
“是”将GPU 0000：02：04.0的计算模式设置为DEFAULT。全部完成“，”nvidia-smi.exe“和”nvidia-smi -q -d compute“继续报告”禁止“：
C： Program Files  NVIDIA Corporation  NVSMI> nvidia-smi
星期四07月17日17:51:57
+ -------------------------------------------------
---------------------------- +
|
NVIDIA-SMI 370.12驱动程序版本：370.12 |
| ------------------------------- + -----------------
----- + ---------------------- +
|
GPU名称TCC / WDDM |
Bus-Id Disp.A |
挥发性的Uncorr。
ECC |
|
Fan Temp Perf Pwr：用法/上限|
内存使用|
GPU-Util Compute M. |
| =============================== + =================
===== + ====================== |
|
0 GRID M60-2Q WDDM |
0000：02：04.0 On |
N / A |
|
N / A N / A P0 N / A / N / A |
355MiB / 2048MiB |
0％禁止|
+ ------------------------------- + -----------------
----- + ---------------------- +
+ -------------------------------------------------
---------------------------- +
|
进程：GPU内存|
|
GPU PID类型进程名称用法|
| =================================================
============================ |
|
0 4432 C + G权限不足N / A |
+ -------------------------------------------------
---------------------------- +
C： Program Files  NVIDIA Corporation  NVSMI> nvidia-smi -c 0
将GPU 0000：02：04.0的计算模式设置为DEFAULT。
全做完了。
C： Program Files  NVIDIA Corporation  NVSMI> nvidia-smi
星期四07月17日17:52:26
+ -------------------------------------------------
---------------------------- +
|
NVIDIA-SMI 370.12驱动程序版本：370.12 |
| ------------------------------- + -----------------
----- + ---------------------- +
|
GPU名称TCC / WDDM |
Bus-Id Disp.A |
挥发性的Uncorr。
ECC |
|
Fan Temp Perf Pwr：用法/上限|
内存使用|
GPU-Util Compute M. |
| =============================== + =================
===== + ====================== |
|
0 GRID M60-2Q WDDM |
0000：02：04.0 On |
N / A |
|
N / A N / A P0 N / A / N / A |
355MiB / 2048MiB |
0％禁止|
+ ------------------------------- + -----------------
----- + ---------------------- +
+ -------------------------------------------------
---------------------------- +
|
进程：GPU内存|
|
GPU PID类型进程名称用法|
| =================================================
============================ |
|
0 4432 C + G权限不足N / A |
+ -------------------------------------------------
---------------------------- +
C： Program Files  NVIDIA Corporation  NVSMI> nvidia-smi -q -d compute
============== NVSMI LOG ==============
时间戳：星期四07年7月17日17:53:08
驱动程序版本：370.12
附加的GPU：1
GPU 0000：02：04.0
计算模式：禁止
那么，有人可以告诉我如何为这个GRID M60-2Q配置文件设置计算模式“默认”吗？
非常感谢你！

以上来自于谷歌翻译

以下为原文

Dear all,
I´ve virtualized a NVIDIA Tesla M60 in order to use it from several virtual machines.
When I apply the GRID M60-8Q GPU Profile for a virtual machine I can see, with the "nvidia-smi.exe", that the Compute Mode is "Default":

C:Program FilesNVIDIA CorporationNVSMI>nvidia-smi
Thu Sep 07 18:11:22 2017
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 370.12                Driver Version: 370.12                   |
|-------------------------------+----------------------+----------------------+
| GPU  Name          TCC/WDDM | Bus-Id       Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|       Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
| 0  GRID M60-8Q       WDDM  | 0000:02:04.0    On |                N/A |
| N/A N/A P0 N/A /  N/A | 755MiB /  8192MiB |    4%    Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                    GPU Memory |
|  GPU    PID  Type  Process name                            Usage    |
|=============================================================================|
| 0    1600  C+G Insufficient Permissions                   N/A    |
+-----------------------------------------------------------------------------+

But when I apply any other GPU Profile (for instance, GRID M60-2Q), the Compute Mode is "Prohibited" and I can´t manage to change it to "Default" (although the result from "nvidia-smi -c 0" is "Set compute mode to DEFAULT for GPU 0000:02:04.0. All done", "nvidia-smi.exe" and "nvidia-smi -q -d compute" continue to report "Prohibited":

C:Program FilesNVIDIA CorporationNVSMI>nvidia-smi
Thu Sep 07 17:51:57 2017
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 370.12                Driver Version: 370.12                   |
|-------------------------------+----------------------+----------------------+
| GPU  Name          TCC/WDDM | Bus-Id       Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|       Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
| 0  GRID M60-2Q       WDDM  | 0000:02:04.0    On |                N/A |
| N/A N/A P0 N/A /  N/A | 355MiB /  2048MiB |    0% Prohibited |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                    GPU Memory |
|  GPU    PID  Type  Process name                            Usage    |
|=============================================================================|
| 0    4432  C+G Insufficient Permissions                   N/A    |
+-----------------------------------------------------------------------------+

C:Program FilesNVIDIA CorporationNVSMI>nvidia-smi -c 0
Set compute mode to DEFAULT for GPU 0000:02:04.0.
All done.

C:Program FilesNVIDIA CorporationNVSMI>nvidia-smi
Thu Sep 07 17:52:26 2017
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 370.12                Driver Version: 370.12                   |
|-------------------------------+----------------------+----------------------+
| GPU  Name          TCC/WDDM | Bus-Id       Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|       Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
| 0  GRID M60-2Q       WDDM  | 0000:02:04.0    On |                N/A |
| N/A N/A P0 N/A /  N/A | 355MiB /  2048MiB |    0% Prohibited |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                    GPU Memory |
|  GPU    PID  Type  Process name                            Usage    |
|=============================================================================|
| 0    4432  C+G Insufficient Permissions                   N/A    |
+-----------------------------------------------------------------------------+

C:Program FilesNVIDIA CorporationNVSMI>nvidia-smi -q -d compute

==============NVSMI LOG==============

Timestamp                         : Thu Sep 07 17:53:08 2017
Driver Version                   : 370.12

Attached GPUs                      : 1
GPU 0000:02:04.0
   Compute Mode                   : Prohibited

So, could somebody tell me how to set Compute Mode "Default" for this GRID M60-2Q Profile?
Thank you very much!

华意仆

2018-9-19 17:14:00

你好
与这篇文章相同的答案：https：//gridforums.nvidia.com/default/topic/1603/nvidia-grid-vgpu/compute-m-prohibited-in-tesla-vgrid-m60-2q/
你们相隔3个小时发布同样的事情:-)
如果要在8Q以外的vGPU配置文件上运行CUDA，则需要Pascal GPU。
我正在询问NVIDIA工程公司，以了解这是否是Maxwell / Pascal GPU之间的技术限制
问候

以上来自于谷歌翻译

以下为原文

Hi

Same answer as this post: https://gridforums.nvidia.com/default/topic/1603/nvidia-grid-vgpu/compute-m-prohibited-in-tesla-vgrid-m60-2q/

You guys posted the same thing 3 hours apart :-)

If you want to run CUDA on vGPU profiles other than 8Q, you need a Pascal GPU.

I'm inquiring with NVIDIA Engineering to understand whether this is a technical limitation between Maxwell / Pascal GPUs

Regards

宫昊

2018-9-19 17:25:53

NVIDIA工程团队很快得到快速反应，说明为什么CUDA只能在Pascal GPU上使用多个vGPU配置文件：
“... Pascal有一个名为Preemption的新硬件功能，允许在vGPU配置文件上进行计算.Preemption是一个允许任务上下文切换的功能。它使GPU能够基本上暂停和恢复任务......”
Maxwell上不存在此功能，这就是为什么您只能使用带有8Q vGPU配置文件或Passthrough的CUDA。
问候

以上来自于谷歌翻译

以下为原文

Nice quick response from the NVIDIA Engineering Team as to why CUDA is only available with multiple vGPU profiles on Pascal GPUs:

"... Pascal has a new hardware feature called Preemption that allows Compute on vGPU profiles. Preemption is a feature that allows task Context switching. It gives the GPU the ability to essentially pause and resume a task ..."

This feature doesn't exist on Maxwell, which is why you can only use CUDA with an 8Q vGPU profile or Passthrough.

Regards

陈欢兆

2018-9-19 17:32:12

是的，我们错误地发布了相同的任务！
我们一起工作，但似乎我们有点“断开连接”:-)
非常感谢你们BJones和mcerveny的大力帮助。
这对我们非常有用！
亲切的问候

以上来自于谷歌翻译

以下为原文

Yes, we posted the same task by mistake! We work together but it seems we are a little "disconnected" :-)

Thank you very much both of you BJones and mcerveny for your great help. It was very useful for us!

Kind regards

乔丽娜

回帖（3）

华意仆

宫昊

陈欢兆

相关问答

计算M.禁止在特斯拉vGRID M60-2Q

使用Vsphere 6.5在Compute模式下使用2个M60卡遇到VM问题

m60是否支持OpenGL

带特斯拉M60的vSGA需要额外许可吗

特斯拉M60 Freeze，100％负载问题

M60设置为graphicsmode但仍然显示计算

在直通中使用NVIDIA M60时Windows闪烁是什么原因？

Grid 2.0

M60可以用于深度学习

一个esxi主机中有2个M60

20万+工程师都在用，免费PCB检查工具