[HN Gopher] New Intel Linux Graphics Driver Patches, Up to 10~15...
       ___________________________________________________________________
        
       New Intel Linux Graphics Driver Patches, Up to 10~15% Better
       Performance
        
       Author : mikece
       Score  : 154 points
       Date   : 2023-04-27 14:14 UTC (8 hours ago)
        
 (HTM) web link (www.phoronix.com)
 (TXT) w3m dump (www.phoronix.com)
        
       | jandrese wrote:
       | It seems the extra performance is entirely due to being more
       | aggressive about switching to higher power modes, not due to
       | efficiency improvements.
        
         | DesiLurker wrote:
         | Well if AMD can burn up their CPUs then We should be able to
         | atleast overclock our GPUs a bit. plus you get nice planned
         | obsolescence benefit on the side.
         | 
         | - some Product Mgr at Intel probably.
        
           | ip26 wrote:
           | _We cannot allow a chip-melt gap!_
        
           | cassepipe wrote:
           | context ?
        
             | mschuster91 wrote:
             | Probably AMD's 7000 issue: https://www.theverge.com/2023/4/
             | 24/23695829/amd-7000x3d-cpu-...
        
       | readyplayernull wrote:
       | How can we get these patches in Ubuntu?
        
         | wswope wrote:
         | Pull kernel source, apply patches, build kernel, run kernel.
         | 
         | https://wiki.ubuntu.com/Kernel/BuildYourOwnKernel
        
         | jaegrqualm wrote:
         | Assuming it's accepted right away, it'd be in the merge window
         | for kernel 6.4, and since Ubuntu 23.04 just came out with 6.2,
         | you'll have to wait for 23.10 at the earliest.
         | 
         | Or you could just compile your own kernel and test it out.
        
       | cramjabsyn wrote:
       | If when it lands I wonder how much it would help with video tasks
       | like plex transcoding
        
       | abdullahkhalids wrote:
       | What's the timeline for something like this landing into linux
       | distro repositories?
        
         | klooney wrote:
         | Years for Ubuntu/RedHat, six to twelve months for Fedora, it
         | varies for everything else
        
           | okdood64 wrote:
           | > Years for Ubuntu/RedHat, six to twelve months for Fedora
           | 
           | Is there some literature on why this is? I'm curious. Is
           | Fedora more "experimental"?
        
             | doodlesdev wrote:
             | Fedora does breaking point releases around every six
             | months, and yes it's kind of more "experimental", generally
             | speaking the repositories are as up to date as possible in
             | a point release distribution, kernel version are always up
             | to date, and Fedora is generally early adopter of breaking
             | changes in programs such as GNOME, Wayland, Pipewire, new
             | Wine versions, etc. Also Kernel updates are done mid-cycle
             | which means this update could land even in just a few weeks
             | from now.
             | 
             | Meanwhile Ubuntu does a new LTS release once every two
             | years, which are already somewhat outdated when they get
             | released.
        
           | TingPing wrote:
           | Fedora updates the kernel mid-cycle so could be weeks to
           | months depending on various factors. Mesa versions would be 6
           | months usually though.
        
       | bagels wrote:
       | Is this just for mobile devices? Not sure why the diver should
       | command low power states when plugged in and playing a game other
       | than for thermals.
        
         | dagmx wrote:
         | Why not? Desktop CPUs and other components do it all the time.
         | Why waste energy?
        
           | bagels wrote:
           | Maybe there is an argument that games are a waste of energy,
           | but I would prefer maximum performance for my games, makes
           | them more enjoyable, or on the margin, playable in the first
           | place. Low power states make sense when you're not actually
           | trying to use all the performance.
        
             | dagmx wrote:
             | Not all games or applications will use up all your
             | resources so lower power states make sense to save energy
             | and reduce thermals as well.
             | 
             | Alternatively , a game/app might trigger boosts that can't
             | be sustained, or some instructions like AVX that are
             | intensive that require dropping thermals to sustain.
             | 
             | Power state switching is very common and a very interesting
             | space to dial in.
        
         | brucethemoose2 wrote:
         | Sometimes you want to conserve every bit of power so the SoC
         | can boost hard when it really needs to. And you dont want the
         | CPU/GPU running faster than they have to, otherwise you sap TDP
         | the rest of the chip could use.
         | 
         | That being said, Intel IGPs arent (yet) fast enough to be
         | throttling down in games :P
        
         | nicktelford wrote:
         | It's for Intel's integrated graphics, which are most prevalent
         | on laptops.
        
           | wongarsu wrote:
           | I imagine NUCs/tiny-form-factor desktops and business
           | computers in general also make a significant fraction of the
           | install base
        
       | zokier wrote:
       | > Google engineers are interested in hooking this into Feral's
       | GameMode
       | 
       | Afaik ChromeOS gamemode is completely independent from Ferals
       | work?
        
       ___________________________________________________________________
       (page generated 2023-04-27 23:01 UTC)