Intel Xe2 Lunar Lake Graphics Performance Disappoints On Linux

[phoronix]

Phoronix: Intel Xe2 Lunar Lake Graphics Performance Disappoints On Linux

While I have been very eager to test out the Core Ultra 200V Lunar Lake series on Linux in part due to the new Xe2 integrated graphics, after several days of pushing a new Lunar Lake laptop on Linux the results have been very disappointing. Besides needing a very leading-edge software stack to enjoy the Xe2 accelerated graphics out-of-the-box, the performance currently is poor. It's a fraction of the Windows performance and currently falls behind the Meteor Lake graphics performance and in turn also being well behind the AMD RDNA3.5 competition with the Ryzen AI 300 series laptops.

https://www.phoronix.com/review/lunar-lake-xe2

[Volta]

Intel never dissapoints. Everything is as usual.

[geerge]

So 6.12 shouldn't be considered the baseline supported kernel version if it's pissing half the iGPU performance away from bad drivers. 6.12 is likely the next LTS right, would that mean that major performance fixes might be backported? It would be a shame if it got working nicely by 6.14 say but most people don't update that often.

[avis]

There's no reason to be upset, I'm just a little annoyed.

Just like for AMD and NVIDIA, Linux is a second (third) class platform to support (properly) for Intel. And then they have had a ton of layoffs recently.

Performance will come later. Please retest in a year. I guess around 6.15/6.16 (unless Linus changes the numbering scheme again and he must) it will be just fine.

[Paradigm Shifter]

Linux support didn't used to be a third class citizen in Intel-land.

Although this just makes me want Strix Halo all the more.

Also, since a local shop has stock of both the Asus ProArt X870E and 9950X for close to the cheapest I've seen, I'm really tempted to not wait for the X3D chip for my new home pseudo-workstation.

[user1]

I remember when AMDGPU kernel driver was just released in early 2016, it already had full performance with the GPUs it supported back then.

This is why I don't trust Intel GPUs and will likely never buy one. Their Linux driver teams are just bad compared to AMDs Linux teams (+ also Valve).

And it's kinda sad because even though Nvidia has vastly improved its driver situation, especially now with explicit sync, as long as it's an external kernel module, I just don't want to deal with that because it's still a headache if I'm on the latest kernels. So if I want a driver that is part of upstream kernel and is also stable, mature and performant, AMD is currently the only option for me.

[avis]

I remember when AMDGPU kernel driver was just released in early 2016, it already had full performance with the GPUs it supported back then.

This is why I don't trust Intel GPUs and will likely never buy one. Their Linux driver teams are just bad compared to AMDs Linux teams (+ also Valve).

And it's kinda sad because even though Nvidia has vastly improved its driver situation, especially now with explicit sync, as long as it's an external kernel module, I just don't want to deal with that because it's still a headache if I'm on the latest kernels. So if I want a driver that is part of upstream kernel and is also stable, mature and performant, AMD is currently the only option for me.

I welcome you to https://gitlab.freedesktop.org/drm/amd/-/issues Currently Linux 6.11 is broken for multiple RDNA 2.0/3.0/3.5 users, and no patches have been submitted yet.

You may say "Use something stable, bro" and then I will have to remind you that RDNA 3.5 requires at the very least 6.10.x which will soon be EOL'ed (in less than 2 weeks) and it hasn't hit multiple distros yet.

We are back to the situation where Windows users enjoy day 1 support for new hardware and Linux users just suffer for months before they get their brand new HW properly working and fully supported.

So, there's no need to claim that AMD is somehow superior to Intel/NVIDIA as they have their own share of critical issues. Maybe just maybe the kernel has to drop all its drivers, create a semblance of stable API layers for them and develop them separately. Do you remember how the ALSA kernel drivers started? Yeah, they were developed exactly this way and no one objected.

[davidbepo]

Xe KMD has to be the worst thing to ever happen to intel linux graphics

[chuckula]

It's too bad the drivers aren't ready for launch day. Something tells me the main development efforts are for Battlemage, which is of course launching later, and the IGP here is still clearly a work in progress.

[user1]

I welcome you to https://gitlab.freedesktop.org/drm/amd/-/issues Currently Linux 6.11 is broken for multiple RDNA 2.0/3.0/3.5 users, and no patches have been submitted yet.

You may say "Use something stable, bro" and then I will have to remind you that RDNA 3.5 requires at the very least 6.10.x which will soon be EOL'ed (in less than 2 weeks) and it hasn't hit multiple distros yet.

We are back to the situation where Windows users enjoy day 1 support for new hardware and Linux users just suffer for months before they get their brand new HW properly working and fully supported.

So, there's no need to claim that AMD is somehow superior to Intel/NVIDIA as they have their own share of critical issues. Maybe just maybe the kernel has to drop all its drivers, create a semblance of stable API layers for them and develop them separately. Do you remember how the ALSA kernel drivers started? Yeah, they were developed exactly this way and no one objected.

Yes, I know AMDGPU isn't perfect and I've already seen that page dozens of times from your older comments here. It's still better than what I've seen on the Intel side. At least AMDs Linux drivers are also better optimized than Intel ones.

I'm not gonna get into the stable api argument, but I hope you realize that Linux GPU kernel drivers from any vendor has room for improved stability / qa testing (including Nvidia). I'd say among all GPU drivers on any OS, the most stable one is Nvidia's Windows driver.