From: sjunhyuk1@gmail.com
To: linux-usb@vger.kernel.org,
	platform-driver-x86@vger.kernel.org
Cc: Heikki Krogerus <heikki.krogerus@linux.intel.com>,
	Hans de Goede <hansg@kernel.org>,
	=?UTF-8?q?Ilpo=20J=C3=A4rvinen?= <ilpo.jarvinen@linux.intel.com>
Subject: Re: [BUG] HP OMEN Transcend 14 (board 8E41): GPU power ceiling on 65W USB-C charger is non-deterministic across sessions (25W vs ~65W)
Date: Tue, 22 Sep 2026 17:32:18 -0400
Message-ID: <20260922213218.5334-1-sjunhyuk1@gmail.com>
In-Reply-To: <20260922192312.40607-1-sjunhyuk1@gmail.com>
References: <20260922192312.40607-1-sjunhyuk1@gmail.com>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

Important correction to my earlier report: I ran the same charger
through HP's OMEN Gaming Hub on Windows (same laptop, dual-boot) and
found this is very likely *intended* HP behavior, not a bug:

  "Performance settings are overridden by system defaults while using
  battery or a low-power AC adapter. Connect a compatible AC adapter
  to restore normal system performance."

HP's own first-party software explicitly classifies the 65W charger
as a "low-power AC adapter" and documents that it intentionally
overrides performance down. So the 25W ceiling itself is apparently
by design, not a fault.

More importantly: **it also reproduces on Windows.** Same laptop,
same 65W charger, Windows test session: 25W ceiling from cold boot,
OMEN Gaming Hub's presence/absence made no difference, and the full
poweroff+physical-unplug/replug workaround that reliably restored
65W on Linux did *not* recover it on Windows in this test (stayed at
25W for a 60-second FurMark run, 60/60 samples). So my claim that "I
don't see this at all under Windows" was wrong - apologies for the
bad data point.

What's still odd, and still possibly Linux/kernel-relevant: on Linux,
the same 65W charger sometimes lands on ~65W instead of the
(apparently intended) 25W, non-deterministically across boots/
reconnects, and stays fixed at whichever value for the rest of that
session. If 25W is the intended default for this adapter class, the
~65W outcome looks like an unintended bypass of HP's low-power-
adapter policy rather than a "fix" - I just don't yet know whether
that bypass happens on Windows too (only tested one Windows boot
session so far, which stayed at 25W throughout).

So the interesting question is probably no longer "why is this
laptop sometimes clamped to 25W" (looks intended) but "what makes
the negotiated result non-deterministic instead of consistently
matching HP's stated policy" - which could still be something in the
UCSI/PD negotiation path. Sorry for the noise, and thanks for reading
this far.