Dell R740 & Nvidia Tesla M60

Question

Hello, I try to install three GPU Nvidia Tesla M60 in a Dell R740. I use GPU install kit, that come with a cable that have one connector with 6 pins, and the other connector have 6 pins + 2 pins. The Tesla M60 have a power connector with 8 pins. So...I connect 6+2 pins in Tesla GPU Power connector, and the other connector side in the mother board power supply. When i have the three GPU connected, the server don't power up. The PSUs are alerted (orange blinked). When disconnected the GPUs, the server power up.  The questions is: Its necesary and extra power cable to connect the GPUs to the power? I find in the Tesla manual, that is not compatible connect directly an PCIe cable power to GPU. Tesla M60 Cable Attachment Support Notes: CPU-8-pin auxiliary power cable attached: Supported; required. CPU 8-pin cable must be attached. CPU 8-pin cable must carry up to 240 W. PCIe 8-pin cable: Not supported. PCIe 8-pin cable is not compatible with CPU 8-pin board connector. Plugging in forcibly can cause permanent damage to the board and the system. In Dell manuals and docs, i not find any issue with cable power that come in install kit. Any sugestions? Thanks in advanced!

DELL-Joey C · Accepted Answer

Hi,

The GPU install kit, are you referring to this: https://dell.to/38Lv2mV, the GPU enablement kit. The power cable should connect to the riser power connector, located at the end of the riser. Try install 1 GPU at a time.

What PSU watt are you using? The PSU settings must be set to non-redundant.

Dell-DylanJ · Accepted Answer

Looks like it shipped with risers 1A, 2A, and 3A, which is good. This would suggest that power is good, riser config is good, and we have 3x of the same GPU (good). I pulled this list of restrictions out of the documentation. Could you review your setup and see if any of this applies to you?

The only other thing that jumps to mind at the moment is potentially the volume of PCI lanes, but I doubt that's the issue here.

GPU Restrictions

R740/R740xd GPGPU restrictions (enforced by order Validator): 

•Requires 2 CPUs with low-profile heatsinks. 

•Ambient inlet temperature is restricted to 30°C for CPU 150 W/8 C, 165 W/12 C, 200 W, 205 W. 

•Max of three 300W double wide GPGPU (note that double wide takes up 2 PCIe slots). 

•Max of six 105W single wide GPGPU. 

•GPU requires redundant 1100W or higher PSUs and GPU enablement kit. 

•The GPU enablement kit includes: 
◦GPU air shroud 
◦Mylar 
◦Riser 3A 
◦GPU power cable for riser 3A 
◦Riser 2A 
◦GPU power cable for riser 2A 
◦Riser 1A or 1D 
◦GPU power cable for riser 1A or 1D 
◦Two processors with 1U heat sink modules and two CPU clips 
◦ Six high performance cooling fans 

• All GPUs must be same type / model. 

• The filler bracket on the GPU air shroud must be removed before installing the GPU. 

• Ensure that high performance fans, GPU air shroud and solid PCIe blanks are installed. 

• Install PSUs rated 1100 W or higher and set the PSU configuration to non-redundant mode. 

• Max of four single-wide and two double-wide GPUs with no GPU on riser 2A when NVDIMM-N battery is installed on GPU shroud. 

• GPU is not supported in expanded operating temperature of 5°C to 40°C.

DELL-Joey C · Accepted Answer

Hi,

I tried to check on some documentations on installation cables from various sources.

The TR5TP is supposed to be connecting to another power cable, which is from the GPU power cable doggle that should come with the GPU. Page 11: https://dell.to/317WxVp

Arti_DNO · Answer

Hi Joey, thanks for response.It´s correct the GPU enablement kit that i use to install. In raiser 1, attachted the GPU power cable in the raiser connector. But, in the others GPU, i attach the GPU power cable to the connector on the mother board. I tryed to put only one GPU in raiser 1 and power up the server, but the server not power on.. So I dettached all GPUs and power on the server, modifieded  the PSU setting to put in non-redundant. Power off the server, install all GPUs and try to power up server. The server failed...not power on. I use two PSU, each one with 2000w .   The GPU cable power that use. https://ibb.co/Wt2RzC7 Thanks!

Dell-DylanJ · Answer

If you could direct message either Joey or I the service tag of the R730, we can check the hardware config (at least at time of purchase). The reason that I ask for this is so that I can look at your riser and processor configuration. Considering you have 2x 2000w PSUs, I'd be willing to bet that power isn't the problem.

Arti_DNO · Answer

Hi, I sending a private message with the service tag of one server. Thanks you!

Arti_DNO · Answer

Thanks you for your time to reply my questions. It's possible that the customer have received a different parts code, so the cable it's not a correct. Regards.

DELL-Joey C · Answer

Hi,

The picture url that you gave, wasn't loading up on my side.

Can you let me know if you do have PN#TR5TP and #4VPD3.

I am not able to find any documentation on M60's power cable requirement, therefore I'm not able to advice if the card need which cable type.

Arti_DNO · Answer

Hi,At these moment only have one cable in my hand. The PN is TR5TP. Only these cable they come with the installation kit. With the GPU no cable come. Thanks!

Arti_DNO · Answer

I reached the same conclusion. Now I am waiting for dell to confirm if the installation service kit is correct and that NVDIA will confirm if a power cable should have come along with the GPU board.

Thank you very much for the help and your time.

Greetings and thanks

PowerEdge Hardware General

Was this post helpful?