Linux 5.0 Release Includes FreeSync Support and Spectre Mitigation

AlphaAtlas

[H]ard|Gawd
Staff member
Joined
Mar 3, 2018
Messages
1,713
The Linux 5.0 kernel has been released, and among other things, it officially adds support for freesync displays on AMD GPUs. Phoronix notes that AMD previously supported FreeSync on Linux, " via their hybrid driver package with its DKMS module in Radeon Software," but posted a tutorial for enabling it and testing FreeSync support in Ubuntu. They note that Vulkan games (which presumably includes Valve's Proton renderer,) compositors, web browsers, media players and a few other apps aren't currently supported by FreeSync at this time.

Meanwhile, the publication also put the recent Spectre performance mitigation measures to the test, and found that performance on the Core i9 7980XE dropped by about 13% with the Spectre protections enabled. Core i7 8086K performance dropped by about 17%, while Ryzen 7 2700X performance only dropped by 3%. Unfortunately, running the same benchmarks on previous Linux kernels would be like comparing apples to oranges, so its hard to say exactly how much Linux 5.0 mitigates Spectre's performance hit, but it looks like certain workloads are still relatively sensitive to the security countermeasures.

To utilize FreeSync you need to be using the xf86-video-amdgpu DDX driver. You can verify so looking for "AMDGPU" in the Xorg.0.log. You also need the above-shown Xorg.conf snippet to enable the "VariableRefresh" AMDGPU DDX driver option. Using the xf86-video-modesetting DDX is unsupported at this time. Your xf86-video-amdgpu driver also has to be relatively new, but such supported X.Org driver can be found in the likes of the Padoka PPA... After enabling the VariableRefresh option and restarting the X.Org server, you can verify that the DDX is new enough and option is working by ensuring that VariableRefresh is successfully mentioned in your Xorg log.
 
Last edited:
Interesting, makes you wonder if whatever performance lead Intel has on AMD dies the moment these security holes are patched. I wouldn't be surprised if the new 10nm processors don't include hardware level fixes. AMD might actually start kicking ass and taking names... Too bad I switched to an Intel 9600... Lol:LOL:
 
Interesting, makes you wonder if whatever performance lead Intel has on AMD dies the moment these security holes are patched. I wouldn't be surprised if the new 10nm processors don't include hardware level fixes. AMD might actually start kicking ass and taking names... Too bad I switched to an Intel 9600... Lol:LOL:

Absolutely diminished, but Intel's arch is still better than Zen+.
Zen2 I think it's a very very different story :)
 
So glad to see more Linux news here! Yay!

By the time I get my Radeon VII Ubuntu should have this kernel easily available without compiling :D

Kind of a bummer about the Vulkan part of freesync. Hmm...
 
I look forward to testing this in RHEL the Intel patches have hurt my accounting servers. The newer EPYC one isn’t done validation Testing and I can’t do a full deploy on it yet.
 
all that really highlights the continuing issue that is linux. Way to much manually configuring files.. etc, that keep it out of the reach of the masses.
 
Since they are talking about software countermeasures slowing things, does Windows have these built in? Did we see as big a slowdown when the [Windows Update] patch hit?
 
Since they are talking about software countermeasures slowing things, does Windows have these built in? Did we see as big a slowdown when the [Windows Update] patch hit?
Yes we did, and in an upcoming patch much of that slow down is expected to be mitigated as well.
 
They note that Vulkan games (which presumably includes Valve's Proton renderer,) compositors, web browsers, media players and a few other apps aren't currently supported by FreeSync at this time.

Dammit.
 
Yes we did, and in an upcoming patch much of that slow down is expected to be mitigated as well.
Well then I'd take it that benchmarks at sites like this include that patch as well.
If AMD CPUs are almost not affected and Intel ones are by 15%, then this means that Zen+ single core performance is even worse than we think.
 
Well then I'd take it that benchmarks at sites like this include that patch as well.
If AMD CPUs are almost not affected and Intel ones are by 15%, then this means that Zen+ single core performance is even worse than we think.
Yes and no, if you are using an Intel series 7 or newer and running windows 10 and have updated your MoBo bios, then the change is negligible like 5% or something like that with AMD being slightly better at 3%, if you are running an older CPU or Windows 7 or 8 then yeah it hurt pretty bad 15% or so.

It was bad enough that my Gen 5 Xeon's had to get phased out a year early running Server 2012R2, so all got replaced with a single EPYC, Maintenance is loving it so much reduced load on the AC system with that upgrade accounting should love they decreased bills

EDIT: Intels' own reports put the decrease on the Gen 6 or older parts by as much as 21% depending on what the operation is.
 
Last edited:
already running

stats.png
 
Yes and no, if you are using an Intel series 7 or newer and running windows 10 and have updated your MoBo bios, then the change is negligible like 5% or something like that with AMD being slightly better at 3%, if you are running an older CPU or Windows 7 or 8 then yeah it hurt pretty bad 15% or so.

It was bad enough that my Gen 5 Xeon's had to get phased out a year early running Server 2012R2, so all got replaced with a single EPYC, Maintenance is loving it so much reduced load on the AC system with that upgrade accounting should love they decreased bills

EDIT: Intels' own reports put the decrease on the Gen 6 or older parts by as much as 21% depending on what the operation is.
Good to know. I currently run an Intel 3770. When I finally do upgrade, I might see a bigger performance increase than expected.
The only reason I say may is because my motherboard is from 2012 and there have not been any BIOS updates since this bug was found. I can assume it has not been patched for the hardware I run.
 
Good to know. I currently run an Intel 3770. When I finally do upgrade, I might see a bigger performance increase than expected.
The only reason I say may is because my motherboard is from 2012 and there have not been any BIOS updates since this bug was found. I can assume it has not been patched for the hardware I run.

I went from a 5Ghz all core 3770K (ddr3 2600 ram) to a 4.1Ghz 1600 (3200mhz ram) and then a 4.3Ghz 2700 and I have not regretted a single second of it. The performance hits to Ivy was there, you just do not realize how bad it is since you have not used a modern platform on a daily basis. Even giving up 1Ghz of clockspeed was a no brainer (since Zen was a bit above Haswell level with 3200 ram IPC wise) but the extra cores and lack of Intel/MS patches was such a blessing. I've owned 6c/12t systems before (4.2Ghz i970 with tri channel 12GB DDR 2133) but Zen was so "smooth" with everything from web use/HTPC/gaming/mining/you name it.

Zen 2 is going to be a wonderful upgrade, and I look forward to giving AMD even more money this summer to go along with the super low frame times and + FreeSync of my Radeon VIIs.
 
Well fudge. Thought I’d be cool and update the kernel on the Ubuntu box. Nothing like screwing it up on purpose. Now the video driver sucks and no second monitor plus it’s slow as balls in winter.....

I reverted back to 4.18 and it’s still borked..... Yay, go me.

I used ukuu, do I need to purge 5.0? I should have wrote down which revision of 4.18 Ubuntu 18.10 was running with.
 
Brand new Ubuntu Mate 18.04.2 install:

upload_2019-3-4_19-11-53.png


I'm looking forward to the 5.0 kernel myself, but I'm not in a hurry to get there :D
 
Since they are talking about software countermeasures slowing things, does Windows have these built in? Did we see as big a slowdown when the [Windows Update] patch hit?
The retpoline mitigation against Spectre was only rolled out in KB4482887 less than a week ago.
 
all that really highlights the continuing issue that is linux. Way to much manually configuring files.. etc, that keep it out of the reach of the masses.
Works fine without any of that, normal users need not concern themselves with bleeding-edge features (in fact, I recommend they do just that). And if you need those specific features, I would hold off on trying Linux unless you are comfortable with learning some basic linux administration skills (read a text file, learn a shell (bash, csh, etc), read a man page or wiki about xorg.conf, edit a config file). Those skills aren't generally required nowadays, but if you're going to be on the bleeding edge you'll want to be familiar with them.

Edit: That said, I do agree that it is a roadblock to adoption, especially since some bleeding edge features on Linux are stable features on Windows. But I've always been a "nothing ventured, nothing gained" sorta guy.
 
Last edited:
Well as a somewhat normal user, I can confirm that it is not needed to be cutting edge. This box hates 5.0, 4.20, 4.19 kernels. I managed to get back to 4.18 and now the system runs almost as it did before, with the exception of now Plex is unable to locate the server.... This is the Plex server, so I get to futz with that now.
I am so blessed..... I'm gonna need to do a beer run. (sigh)
 
OK, if I've updated my BIOS on my Ivy Bridge (BIOS hack with new microcode) and Haswell-E (BIOS update from manufacturer) rigs, should I downgrade BIOS on both and just rely on Windows microcode updates for the fixes and speed updates or will this speed fix also apply to updated BIOS?
 
OK, if I've updated my BIOS on my Ivy Bridge (BIOS hack with new microcode) and Haswell-E (BIOS update from manufacturer) rigs, should I downgrade BIOS on both and just rely on Windows microcode updates for the fixes and speed updates or will this speed fix also apply to updated BIOS?
Not sure Widows microcode affects this much......
 
Not sure Widows microcode affects this much......

The Intel microcode updates in Windows replaced BIOS updates for mobos that were never going to get a new BIOS. I'm wondering which takes precedence when both are in place?
 
I would think the BIOS would take precedence. Not entirely sure really. I didn't do any BIOS updates on my Intel boxes and only one of those now has Windows on it. I'd test that idea but the Gigabyte GA-Z77M-D3H hasn't had a BIOS update since December of 2013.

The GA-Z97N-Gaming 5 board I have doesn't have an update available either.
Maybe someone would take the time to expand on this idea.
 
I would not worry too much about this exploit.
It isn't going to blow things up. At worst, an attacker (and this would need to run on your system so you'd need to install infected software) can log memory dumps from other processes in your computer. With enough, maybe they can decipher important info. But as I said, you'd need to install infected software. This isn't something that can easily be exploited.

I feel it is not a big risk for home computers.

If you wonder if your computer now runs slower with the new BIOS, I guess you will have to run some CPU benchmarks and give it the old A B test.

I looked into this issue a long time ago and never could get a straight answer to whether software alone is a good enough measure however a Microsoft FAQ said that a fix required both a software update to the OS and a hardware (BIOS level) fix.
 
Back
Top