[Question/Bug?] Why does the sysfs current_link_speed report a different value than pp_dpm_pcie

[Question/Bug?] Why does the sysfs current_link_speed report a different value than pp_dpm_pcie

[Maxime Schmitt]

[-- Attachment #1: Type: text/plain, Size: 546 bytes --]

Hello,

On my system the sysfs file "current_link_speed" is reporting "16.0 GT/s
PCIe".
The GPU has support for PCIe gen 4 (AMD Radeon RX 6800 XT).
However neither the CPU (AMD Ryzen 7 1700X) nor the motherboard have
support for PCIe gen 4.

The file "pp_dpm_pcie" returns the following:
0: 2.5GT/s, x1 310Mhz
1: 8.0GT/s, x16 619Mhz *
Which is correct.

Is this the expected behavior?

I am asking because I recently added AMDGPU support to a monitoring tool I
maintain (nvtop) and I don't know where to get the information from.

Cheers,
Maxime

[-- Attachment #2: Type: text/html, Size: 814 bytes --]

RE: [Question/Bug?] Why does the sysfs current_link_speed report a different value than pp_dpm_pcie

[Quan, Evan]

[-- Attachment #1: Type: text/plain, Size: 1206 bytes --]

[AMD Official Use Only - General]



From: amd-gfx <amd-gfx-bounces@lists.freedesktop.org> On Behalf Of Maxime Schmitt
Sent: Wednesday, April 6, 2022 1:40 AM
To: amd-gfx@lists.freedesktop.org
Subject: [Question/Bug?] Why does the sysfs current_link_speed report a different value than pp_dpm_pcie

Hello,

On my system the sysfs file "current_link_speed" is reporting "16.0 GT/s PCIe".
The GPU has support for PCIe gen 4 (AMD Radeon RX 6800 XT).
However neither the CPU (AMD Ryzen 7 1700X) nor the motherboard have support for PCIe gen 4.

The file "pp_dpm_pcie" returns the following:
0: 2.5GT/s, x1 310Mhz
1: 8.0GT/s, x16 619Mhz *
Which is correct.

Is this the expected behavior?
[Quan, Evan] "pp_dpm_pcie" is reporting the correct link speed. You can trust that.
While "current_link_speed" seems reporting the link speed the card is capable of. Not sure whether that fits design.
@Deucher, Alexander<mailto:Alexander.Deucher@amd.com> Any comments for the output of sysfs file "current_link_speed"?

BR
Evan

I am asking because I recently added AMDGPU support to a monitoring tool I maintain (nvtop) and I don't know where to get the information from.

Cheers,
Maxime


[-- Attachment #2: Type: text/html, Size: 5044 bytes --]

Re: [Question/Bug?] Why does the sysfs current_link_speed report a different value than pp_dpm_pcie

[Lazar, Lijo]

+Mike

On 5/9/2022 8:17 AM, Quan, Evan wrote:
> [AMD Official Use Only - General]
> 
> *From:* amd-gfx <amd-gfx-bounces@lists.freedesktop.org> *On Behalf Of 
> *Maxime Schmitt
> *Sent:* Wednesday, April 6, 2022 1:40 AM
> *To:* amd-gfx@lists.freedesktop.org
> *Subject:* [Question/Bug?] Why does the sysfs current_link_speed report 
> a different value than pp_dpm_pcie
> 
> Hello,
> 
> On my system the sysfs file "current_link_speed" is reporting "16.0 GT/s 
> PCIe".
> 
> The GPU has support for PCIe gen 4 (AMD Radeon RX 6800 XT).
> 
> However neither the CPU (AMD Ryzen 7 1700X) nor the motherboard have 
> support for PCIe gen 4.
> 
> The file "pp_dpm_pcie" returns the following:
> 
> 0: 2.5GT/s, x1 310Mhz
> 1: 8.0GT/s, x16 619Mhz *
> 
> Which is correct.
> 
> Is this the expected behavior?
> 
> */[Quan, Evan] “pp_dpm_pcie” is reporting the correct link speed. You 
> can trust that./*
> 
> */While “current_link_speed” seems reporting the link speed the card is 
> capable of. Not sure whether that fits design./*
> 
> */@Deucher, Alexander <mailto:Alexander.Deucher@amd.com> Any comments 
> for the output of sysfs file “current_link_speed”?/*
> 


In our architecture, GFX endpoint is not directly connected to 
CPU/Chipset PCIE. We have upstream/downstream bridges in between

CPU/Chipset <-> || US Port <-> DS Port <-> GFX EP||

The actual speed will be reflected in the US port connected to the 
CPU/chipset port or pp_dpm_pcie node as it shows the speed reported by 
power management firmware.

Thanks,
Lijo

> *//*
> 
> */BR/*
> 
> */Evan/*
> 
> I am asking because I recently added AMDGPU support to a monitoring tool 
> I maintain (nvtop) and I don't know where to get the information from.
> 
> Cheers,
> 
> Maxime
>