Nvidia Quadro RTX 6000 on Dl380 - gen10 not work

[MohammadAhmadi]

Hi

Hpe dl380 gen 10

CPU: Intel Zenon Gold 5220

RAM:128GB

Power supply:2x 800W 

Graphic card: Nvidia Quadro RTX 6000 24GB

i install graphic card to Riser card(and connect 8pin mini to 2x-8pin cable), but server not started and Health LED in front is red and blinking.

[pchops]

---

i install graphic card to Riser card(and connect 8pin mini to 2x-8pin cable), but server not started and Health LED in front is red and blinking.

---

This GPU requires an 8pinx6pin Y-power cable, not an 8x8pin:

P/N: 869805-001 (cable kit P/N: 871830-B21, spares P/N: 875097-001)

Check that the card is properly seated in the riser and also check the Integrated Management Log for errors.

Also, make sure your server meets all of the requirements for this (R0Z45C) GPU.

references:

HPE ProLiant DL380 Gen10 Server quickspecs

HPE ProLiant Gen10 DL380 Cabling Diagrams

https://partsurfer.hpe.com/ShowPhoto.aspx?partnumber=875097-001

[MohammadAhmadi]

No gpu power cable has a 6+2 pin(6 pin and 8 pin).

please check this attachment link contain my gpu cable and riser card.

gpu power cable 

gpu power cable 

gpu power cable 

Riser card with gpu 

Riser card 

[pchops]

That is not the correct cable for the DL380 Gen10 -- from what I can find, that 879289-001 cable is for an ML350 Gen10. Technically, the GPU end should be wired the same as the 869805-001, but I don't know if the riser end of those two cables are wired the same. If they are not wired the same, hopefully nothing got damaged. If they are wired the same, then you've got a different problem (please check the IML for any related entries). Either way, you would need to use the correct cable if you want to achieve a supported solution.

NOTE that the correct cable kit (871830-B21) also includes the retainer clips that attach to the fan baffle to secure and support the rear extender bracket on the GPU.

Regarding the DL380 Gen10 quickspecs link I previously posted, it looks like you'll need to go back a revision of those quickspecs to be able to see the requirements I mentioned to support the R0Z45C (RTX6000) -- for example:

https://www.hpe.com/psnow/doc/a00008180enw.html?ver=54

NVIDIA Quadro RTX 6000 Graphics Accelerator for HPE R0Z45C

Notes:

­ Max Qty=2 per server

­ This option requires the High Performance Fan Kit (867810-B21), and the High Performance Heatsink (826706-B21).

­ This GPU requires Pwr Cable Kit (871830-B21) to also be selected.

­ System memory should be 2x GPU memory.

­ One Pwr Cable Kit can support multiple (max=3) GPUs.

Apologies for that...

[MohammadAhmadi]

Thanks for your response, But when I bought this cable, the seller asked me what the model of the graphics card is, and I said PNY, and the seller said  this cable is for the same graphics model and the same server.

[pchops]

It is quite possible that the cable you purchased is wired compatibly and may work, but it will not be a supported configuration, not to mention using a third-party (PNY) GPU card. If the cable is OK, then the issue lies elsewhere, but you will need to provide some additional information, such as:

• are there any related entries in the IML (Integrated Management Log)?
• does the RTX6000's fan spin when you attempt to power on the server?
• would you be able to "second box verify" that the GPU is good by trying it in another system?
• do other PCIe options function correctly in that same slot on the PCIe riser?

[MohammadAhmadi]

@pchops 

Im very sory for my late answer, but i checked IML and GPU status:

* in IML logs when try to turn on this errors shows up(IML Log in power on fault

       1.Server Critical Fault(Service Information: Power On Fault, System Board, AUX/Main EFUSE(10h))

       2.Server Critical Fault(Service Information: Power On Fault, System Board, P12V Main/AUX Regulators(04h))

*GPU fan not spin in power on fault

*Yes GPU is OK and worked in two other PC

*I tested in another Rizer card and same error happend

I hopefully you can help me to solve this problem