# PCIE65 issue - any development happened?

**URL:** <https://community.amperecomputing.com/t/pcie65-issue-any-development-happened/3016>\
**Category:** General Discussion\
**Created:** [June 30, 2025, 5:10pm UTC](https://community.amperecomputing.com/t/pcie65-issue-any-development-happened/3016 "2025-06-30T17:10:18Z")\
**Posts on this page:** 7\
**Page:** 1

<div class="post-metadata">

**Author:** ![hrw](https://yyz1.discourse-cdn.com/flex027/user_avatar/community.amperecomputing.com/hrw/32/31_2.png) [@hrw](https://community.amperecomputing.com/u/hrw)\
**Post date:** [June 30, 2025, 5:10pm UTC](https://community.amperecomputing.com/t/pcie65-issue-any-development-happened/3016/1 "2025-06-30T17:10:18Z")

</div>

I wonder did anyone spent some time on finding out where exactly problem is with pcie65 stuff.

GTK based apps have so many artifacts on text, that sometimes it is hard to read at all.

 ![obraz](https://canada1.discourse-cdn.com/flex027/uploads/amperedevelopers/original/1X/31b959f57ac9772deb58a268924c0edd2aee1c7b.png)

Qt/KDE ones display without any issues…

 ![obraz](https://canada1.discourse-cdn.com/flex027/uploads/amperedevelopers/original/1X/30d009c3ba5bec79d95dda08fd7aabdbf14d578d.png)

MPV (movie player) also shows text properly:

![obraz](https://canada1.discourse-cdn.com/flex027/uploads/amperedevelopers/original/1X/b0fe5495e6720bde127265fdc67554e4b3233027.png)

---

<div class="post-metadata">

**Author:** ![hrw](https://yyz1.discourse-cdn.com/flex027/user_avatar/community.amperecomputing.com/hrw/32/31_2.png) [@hrw](https://community.amperecomputing.com/u/hrw)\
**Post date:** [June 30, 2025, 6:31pm UTC](https://community.amperecomputing.com/t/pcie65-issue-any-development-happened/3016/2 "2025-06-30T18:31:52Z")

</div>

Also note that problem does not happen when llvmpipe is used instead of accelerated Mesa.

---

<div class="post-metadata">

**Author:** ![sevo](https://avatars.discourse-cdn.com/v4/letter/s/ed655f/32.png) [@sevo](https://community.amperecomputing.com/u/sevo)\
**Post date:** [July 1, 2025, 12:59am UTC](https://community.amperecomputing.com/t/pcie65-issue-any-development-happened/3016/3 "2025-07-01T00:59:26Z")

</div>

With llvmpipe, are you using the efi framebuffer or amdgpu drm kernel driver?

---

<div class="post-metadata">

**Author:** ![hrw](https://yyz1.discourse-cdn.com/flex027/user_avatar/community.amperecomputing.com/hrw/32/31_2.png) [@hrw](https://community.amperecomputing.com/u/hrw)\
**Post date:** [July 1, 2025, 6:59am UTC](https://community.amperecomputing.com/t/pcie65-issue-any-development-happened/3016/4 "2025-07-01T06:59:31Z")

</div>

llvmpipe jumps in when system has both ASPEED and Radeon graphics cards.

Nevermind is ASPEED as astdrmfb or as simple-framebuffer.

---

<div class="post-metadata">

**Author:** ![David.Zeng](https://avatars.discourse-cdn.com/v4/letter/d/9e8a1a/32.png) [@David.Zeng](https://community.amperecomputing.com/u/David.Zeng)\
**Post date:** [July 17, 2025, 3:35am UTC](https://community.amperecomputing.com/t/pcie65-issue-any-development-happened/3016/5 "2025-07-17T03:35:29Z")

</div>

Do you mean the PCIe-65 erratum for Altra family discussed in this github?

> <https://github.com/AmpereComputing/linux-ampere-altra-erratum-pcie-65/blob/main/README>

---

<div class="post-metadata">

**Author:** ![hrw](https://yyz1.discourse-cdn.com/flex027/user_avatar/community.amperecomputing.com/hrw/32/31_2.png) [@hrw](https://community.amperecomputing.com/u/hrw)\
**Post date:** [July 17, 2025, 5:56am UTC](https://community.amperecomputing.com/t/pcie65-issue-any-development-happened/3016/6 "2025-07-17T05:56:10Z")

</div>

Yes. That.

My machine has WX2100 and RX6700XT in slots. Booting without pcie-65 patches is waste of time as system misbehaves when sddm starts.

---

<div class="post-metadata">

**Author:** ![sevo](https://avatars.discourse-cdn.com/v4/letter/s/ed655f/32.png) [@sevo](https://community.amperecomputing.com/u/sevo)\
**Post date:** [July 18, 2025, 12:10am UTC](https://community.amperecomputing.com/t/pcie65-issue-any-development-happened/3016/7 "2025-07-18T00:10:57Z")

</div>

Well I suppose if the GPU drivers were rewritten to ensure aligned accesses and not assume X86 style write combining, the kernel patches wouldn’t be necessary. I assume the nvidia gpu drivers are written in such a way.