Raven Ridge Isn't Yet Ready To Have GFXOFF & Stutter Mode Flipped On For Linux

[finalzone]

Unfortunately it is not only Lenovo with really bad AMD laptops. I have a 13" HP Envy Ryzen and out of box it is really unusable bad. Battery life could be better ( i think that affects most Ryzen notebooks but at least for Dell and i think Lenovo there were some recent BIOS updates, no idea if they helped) but thats okay. Also no touchscreen support in linux due to broken acpi-table (thats shared by all hp amd products?) but the worst thing is the fan control. It is so loud and always on, thats just unbearable. In windows you can at least change the profile but it will also cripple performance. Fortunately someone updated NBFC to support the Envy Ryzen notebooks, otherwise this piece of aluminum would just decay in some drawer until HP hopefully gets their shit together and release a decent BIOS.

Which HP Envy Ryzen version do you have? Standard or the x360. The x360 model I own is upgradable i.e. replaceable memory and storage.

About the touchscreen issue, yes it affects wall HP AMD products hence very shocking to be honest such problem exists for a long time. Some users lacking experience on kernel functionality managed a workaround as highlighted https://bugzilla.kernel.org/show_bug.cgi?id=199529. Currently, no upstream kernel maintainer available while bug report have users willing to test the suggestion.

About the fan control, check the BIOS settings and disable "Fan always On" to see if that resolve the problem.

Apology for the offtopic as the subject is about Riven Ridge.

[thxcv]

Which HP Envy Ryzen version do you have? Standard or the x360. The x360 model I own is upgradable i.e. replaceable memory and storage.

About the touchscreen issue, yes it affects wall HP AMD products hence very shocking to be honest such problem exists for a long time. Some users lacking experience on kernel functionality managed a workaround as highlighted https://bugzilla.kernel.org/show_bug.cgi?id=199529. Currently, no upstream kernel maintainer available while bug report have users willing to test the suggestion.

About the fan control, check the BIOS settings and disable "Fan always On" to see if that resolve the problem.

Apology for the offtopic as the subject is about Riven Ridge.

It is the 13" x360 model. Memory is not upgradable but storage should be, i think. That BIOS setting does not do anything. It is not only that the fan is always on, it also spins really fast all the time. It is just some crooked fan control by HP but i think only the 13" model is affected by this.
The out of box experience of this thing is really bad.

[finalzone]

It is the 13" x360 model. Memory is not upgradable but storage should be, i think. That BIOS setting does not do anything. It is not only that the fan is always on, it also spins really fast all the time. It is just some crooked fan control by HP but i think only the 13" model is affected by this.
The out of box experience of this thing is really bad.

It looks like an issue with last year model. Storage should not be the problem as I was able to upgrade to SATA SSD on the 2018 15" x360 version. I was impressed by the nearly great out of box experience. If only HP can properly handle the Linux firmware update via LVFS.

[finalzone]

My vulkan experience is really bad. Even not in VKcube i see this glitch. I thought it was performance-wise before i ran vkcube. Could play WoW or Overwatch on linux, you know.

I found vkmark, a Vulkan benchmark similar to glmark2, and ran it for testing. Try it on your side and see if there is an issue.

[chithanh]

At work we got new Dell Latitude 5495 (with Ryzen Pro 2700U). I'm running Gentoo with kernel 4.19.1 now on mine.

Positive:

• It boots Linux without any ivrs_ioapic workarounds (this was dealbreaker for the E485)
• BIOS/UEFI update possible via UEFI setup menu, I updated to 1.2.3 without any issues
• Kernel build time comparable to my old desktop (FX-8350)
• Two SO-DIMM slots, dual channel
• Battery life

Neutral:

• It gets rather loud during compile, but is inaudible when idle
• I get AMD-Vi page fault errors in dmesg with the wireless card (ath10k), but seems to have no ill effects
• Mine did not come with the optional finger print reader, but even if it did, it is unsupported in libfprint (Broadcom 5880, no specs)

Negative:

• Two-finger touchpad scrolling seems broken in Linux, still investigating but it seems that hid-alps.c needs the product ID added
• amdkfd doesn't load due to some IOMMU error (IOMMU is enabled in UEFI setup and detected by amd_iommu_v2). Will file bug report.
• No ECC memory support, the necessary traces are not connected on the mobo (verified this with Dell support).

1. E485/E585 don't support keyboard backlight, however E480, E580 do. I would by E485 if it had that feature.

There might be more to that story. Old Lenovo E485 datasheet from June 2018 lists keyboard backlight as an option, but that was removed in the new datasheet from July 2018. Also backlit keyboards are still listed as compatible FRUs for the E485 (someone tried, the keyboard works but the backlight does not light up).

Possibly Lenovo had planned to offer E485 with keyboard backlight but some oversight made it impossible to use.

3. T480 has WQHD (2560 x 1440) option, where AMD hasn't.

Yes, that is a real bummer.
The 2560x1440 screen can be bought individually as replacement part though, if you absolutely want that.

[agd5f]

[*]amdkfd doesn't load due to some IOMMU error (IOMMU is enabled in UEFI setup and detected by amd_iommu_v2). Will file bug report.

Bios bug. I think it should be fixed with this patch:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=44d8cc6f1a905e4bb1d4221a898abb0d7e9d100a

[Drago]

Bios bug. I think it should be fixed with this patch:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=44d8cc6f1a905e4bb1d4221a898abb0d7e9d100a

How is possible Windos to work on such vast variety of "broken" Laptops? Do M$ do the same like this patch, adding workarounds all over for fucked up BIOS/UEFI?

[chithanh]

Bios bug. I think it should be fixed with this patch:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=44d8cc6f1a905e4bb1d4221a898abb0d7e9d100a

Unfortunately, no. That patch is already part of 4.19.
But I found a bug report which describes my issue: https://bugs.freedesktop.org/show_bug.cgi?id=107898 and indeed iommu=pt is a workaround.

How is possible Windos to work on such vast variety of "broken" Laptops? Do M$ do the same like this patch, adding workarounds all over for fucked up BIOS/UEFI?

I think ROCm is Linux-first software anyway.

But yes, AMD needs some kind of certification program like with FreeSync monitors, that OEMs must submit to before they can use the Ryzen/Vega/Fusion/etc. brand in marketing.

[agd5f]

How is possible Windos to work on such vast variety of "broken" Laptops? Do M$ do the same like this patch, adding workarounds all over for fucked up BIOS/UEFI?

It's not that the BIOSes are broken per se, they are just designed and tested against a specific OS and only the things that that OS uses are tested. A lot of things that Linux enables by default are not used at all on windows so those ACPI/UEFI interfaces are not often tested on OEM platforms unless Linux is actually validated on the platform. Some examples: IOMMU, interrupt remapping, CRAT tables for ROCm, etc.

[Drago]

It's not that the BIOSes are broken per se, they are just designed and tested against a specific OS and only the things that that OS uses are tested. A lot of things that Linux enables by default are not used at all on windows so those ACPI/UEFI interfaces are not often tested on OEM platforms unless Linux is actually validated on the platform. Some examples: IOMMU, interrupt remapping, CRAT tables for ROCm, etc.

Fair enough, I am not hardware/uefi specialist. But then how come windos is the de-facto "standard" where everything is supposed to work. Till now I was thinking Linux is lagging behind Windos in terms of hardware support, but now you are saying that on the contrary, Linux supports more uefi features. Strange...