Intel Haswell GT1 Graphics Have Been Busted The Past Half-Year On Linux

[re:fi.64]

Hmm this is weird, but I have a Coffee Lake i7 that was getting GPU hangs until I installed the patched kernel from the linked issue.

[JigglypuffSexDoll]

Hmm this is weird, but I have a Coffee Lake i7 that was getting GPU hangs until I installed the patched kernel from the linked issue.

Seems unlikely since all the code in the patch is in gen7_renderclear.c.

[MowKow]

since those security mitigations were mainlined I was getting GPU Hangs left & right, especially when combining vaapi de/encode with OpenGL offscreen rendering ( haswell, broadwell, skylake, coffeelake based XEONs )

since 5.7.x broadwell+ seem to be stable

what really bugs me:
there's no way to turn off those mitigations, even when, like in our case, the machines affected have no internet access at all

[chocolate]

Article states that the GFX chip is used in lower end processors that have "HD Graphics 2000" or "HD Graphics", none of the I3, I5, or i7 qualify, just the celerons and low end "pentiums" produced by intel based on the haswell-DT have these GFX on chip.

By direct experience, I believe the same problem affects the Iris Pro 5200 found in the i7-4750HQ.

[darkbasic]

Intel Graphics has always been a regression nightmare for me, that's why I would love to see an high end laptop with an AMD APU.

[ShFil]

HD4000 is Ivy Bridge, so it's not affected by this regression.

Haswell and ivy bridge have quite similar gpu (seven gen).

@topic
It's reason why so many people were angry about haswell, ivy, and sandy gpus don't getting support in new driver. They're first gpus of intel enough for light gaming, but without care (being in legacy driver) it's useless.

[Inopia]

Can't exploit the gpu if it hangs at boot. Checkmate hackers

[mattst88]

Haswell and ivy bridge have quite similar gpu (seven gen).

Thanks... I'm aware. I was on the Mesa team at Intel for 8 years.

[mattst88]

By direct experience, I believe the same problem affects the Iris Pro 5200 found in the i7-4750HQ.

The same commit caused the regression, but the fix isn't the same. Ville's patch fixed the GT3e regression I alluded to: https://git.kernel.org/pub/scm/linux...b198fae396d660

I haven't checked if it's been backported, but it has the right tags so I assume it has been.

[ShFil]

Thanks... I'm aware. I was on the Mesa team at Intel for 8 years.

Sorry, if you heard it this way.

I wanted to say that people probably asking about ivy, bc haswell and ivy bridge have quite similar gpu (seven gen).