Announcement

Collapse
No announcement yet.

Linux 3.5 Can Massively Boost AMD Radeon Graphics

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #31
    Originally posted by Wyatt View Post
    Ah, I see. Thanks for clearing that up, though I really wish this was something that was addressed in the article in the first place.

    Thanks, I'm just speculating. My personal dream from these changes....a killer APU with openCL for my Gentoo. My Phenom II x6 eats emerge @system on a budget.

    ~Jux

    "CFLAGS=-march=native -Os -pipe -ggdb"

    Comment


    • #32
      Originally posted by crazycheese View Post
      Thats a circle, because if the driver is bad, no one will purchase gddr stacked card. Something should come first, I presume from development side.
      Those cards were not affected by these commits. There are certainly more tweaks that can enabled to improve performance on them as I mentioned in another reply, but in this case these commits will not affect them. Also, gddr cards will always perform better than ddr cards, so if you want better performance (regardless of driver improvements) from a 5450 or 6450, get the gddr version.

      Comment


      • #33
        Originally posted by evolution View Post
        Furthermore, I'm also expecting that the "default" profile will stop using always the maximum frequency of the GPU, because that kills the GPU lifetime. (Nouveau does the opposite, btw...)
        That's a fucking stupid thing, my HD5870 is in OpenCL full load 24/7. Maybe it will die in 15 years instead of 20, but who cares? Peoples keep saying the same thing for the cpu too, I have an old P4 overclocked since 12 years which I still use daily.
        ## VGA ##
        AMD: X1950XTX, HD3870, HD5870
        Intel: GMA45, HD3000 (Core i5 2500K)

        Comment


        • #34
          Originally posted by evolution View Post
          Furthermore, I'm also expecting that the "default" profile will stop using always the maximum frequency of the GPU, because that kills the GPU lifetime. (Nouveau does the opposite, btw...)
          The "default" profile uses the settings marked as "default" in the VBIOS. Some cards have high default settings in the VBIOS, others have low default settings.

          Comment


          • #35
            The default profile uses the clocks that are set by the vbios at boot; i.e., the clocks that are set when you boot the computer before the OS loads. As Bridgman noted, on some boards they are higher, on others they are lower. Whether or not they are the same as the high clocks varies from board to board. Nouveau does the same thing. The only difference is that nvidia tends to set the boot up clocks lower.

            Comment


            • #36
              By the way, just the usual reminder
              Micheal please stop benchmarking APUs with low clocks
              ## VGA ##
              AMD: X1950XTX, HD3870, HD5870
              Intel: GMA45, HD3000 (Core i5 2500K)

              Comment


              • #37
                Originally posted by evolution View Post
                Well, in opposition to the hype I've seen in the article, I'll take a more "conservative" approach: when 3.5 stable arrives, I'll give a new try to the FOSS ATI drivers (I went back to catalyst because of VAAPI and proper PM). From what I've seen, we'll have better 3D performance from now on (but does that apply to r600/r700 cards? I didn't see any card of that generation tested in the article...).

                Furthermore, I'm also expecting that the "default" profile will stop using always the maximum frequency of the GPU, because that kills the GPU lifetime. (Nouveau does the opposite, btw...)

                Finally, in the medium/long run, It'd be nice to have H.264 VDPAU/VAAPI/UVD acceleration. That wold be nice for those who still have weak CPUs (e.g. AMD E-350/Low-End Llanos/Nehalmen Core2Duos).

                Cheers
                unlike nvidia (google bumpgate), always maximum clock has no influence on the lifetime of the gpu.

                It might have an influence on the power circuitry... but you said gpu

                Comment


                • #38
                  Originally posted by agd5f View Post
                  Also, gddr cards will always perform better than ddr cards, so if you want better performance (regardless of driver improvements) from a 5450 or 6450, get the gddr version.
                  The gddr versions also use 5-15W more power, hint hint

                  Comment


                  • #39
                    Originally posted by energyman View Post
                    unlike nvidia (google bumpgate), always maximum clock has no influence on the lifetime of the gpu.

                    It might have an influence on the power circuitry... but you said gpu
                    At very least it produces lot of heat and enough of cooler noise from any more or less powerful GPUs. Which isn't good either. Why not enable smth like dynpm by default? CPU scales frequency by default almost anywhere for ages. Maybe it's time for GPUs as well, esp. powerful ones?

                    Comment


                    • #40
                      Originally posted by 0xBADCODE View Post
                      At very least it produces lot of heat and enough of cooler noise from any more or less powerful GPUs. Which isn't good either. Why not enable smth like dynpm by default? CPU scales frequency by default almost anywhere for ages. Maybe it's time for GPUs as well, esp. powerful ones?
                      Because it's still buggy, and someone has to fix it.

                      Comment


                      • #41
                        Originally posted by agd5f View Post
                        The default profile uses the clocks that are set by the vbios at boot; i.e., the clocks that are set when you boot the computer before the OS loads. As Bridgman noted, on some boards they are higher, on others they are lower. Whether or not they are the same as the high clocks varies from board to board. Nouveau does the same thing. The only difference is that nvidia tends to set the boot up clocks lower.
                        Thank you for your response... Now I'm a bit more clarified about the power profiles... But I'd like to have dynpm or a lower power consuption profile as a default for FOSS ATI drivers... without having to do "geeky" things, OC... (for me it's not a big deal, but for some "average Joe" Linux users...)

                        @darkbasic: Well, sometimes the time your electronics survive also depends from the "luck" you've with the components you bought. I also have a P4 for about 10y (OC from 2.8GHz to 3.2Ghz) and it's still working (using that PC right now btw ). Simultaneosly, I had a Radeon 9600 that only worked for 3y (also overclocked it with ATI Tool in the past )... But it's a fact that having your electronic components always working at maximum specs decreases their lifetime... And that's not a stupid thing as you're stating, is "common sense" in electronics...

                        @energyman: I know about some people that suffered with the "connectorgate" scandal... I even advised them to buy ATI next time (most of that people were using Windows instead of Linux...). But I also have some friends that use nVidia for a long time and never had problems... Now, I'm trying it for the first time (look at the "Goodbye ATI thread")... The "bumpgate scandal", if you're telling about the rebadging of nVidia cards, can also be applied to recent ATI card releases (except for the >77xx (aka GCN) series).

                        Cheers

                        p.s.: Still using a rv730 (4650AGP) card...
                        Last edited by evolution; 06-07-2012, 09:43 PM. Reason: Missing Info

                        Comment


                        • #42
                          Great! But how comparable is it with catalyst drivers?

                          Originally posted by e8hffff View Post
                          All the bad news about AMD Radeons, and now comes the strawberries and cream.
                          Good to know, but I'm still skeptic about the OSS drivers performance when compared with catalyst drivers. Would somebody please tell me how it'll stand in performance in comparison to the catalyst drivers? Also please somebody tell does it improves ATI HD 6250 (AMD C-50 APU) performance? If so, how much?


                          Note: Only yesterday I tried the kernel 3.4 with latest mesa stack. KMS was ok. But the performance (3d and hd video playback) was so damn bad that I wondered why do they release such bad, inefficient and feature-incomplete code at all.
                          Last edited by manmath; 06-08-2012, 01:09 AM.

                          Comment


                          • #43
                            Originally posted by manmath View Post
                            Good to know, but I'm still skeptic about the OSS drivers performance when compared with catalyst drivers. Would somebody please tell me how it'll stand in performance in comparison to the catalyst drivers? Also please somebody tell does it improves ATI HD 6250 (AMD C-50 APU) performance? If so, how much?
                            Note: Only yesterday I tried the kernel 3.4 with latest mesa stack. KMS was ok. But the performance (3d and hd video playback) was so damn bad that I wondered why do they release such bad, inefficient and feature-incomplete code at all.
                            you use a APU right? then your test is invalid because you don't do measure the performance you only do measure the bad and incomplete power-management because the default clock speed on AMD APUs are "LOW" because these APUs need power-management to set the right clock speed.
                            also you used the wrong kernel 3,5 is the one with the biggest speed improvements.

                            "Would somebody please tell me how it'll stand in performance in comparison to the catalyst drivers?"

                            you get 60-80% of the speed if you use a REAL GPU like my hd4770 and not a "APU-FAKE"

                            if you use a APU the slow-defaults hit you and you get 10-20% of the speed.

                            Comment


                            • #44
                              No, it's not a crappy gpu.

                              Qaridarium, thanks for posting the reply. But here are a few points.

                              #1. it's not an APU-Fake. On the same device Windows 7 handles 3d works and hd 1080p quite well. But linux with open source graphics simply fails to do that. I tried with catalyst, it works way better than the open source graphics, but it's too complicated in the sense i had to install catalyst drivers, vaapi-wrapper and then some workarounds in the media players. I thought open source graphics will be an easy way.
                              #2. Did you mean I'll get 10-20% speed boost if I used open source graphics on top of kernel 3.5 and proper xorg drivers? If it's so then it's very unfortunate to call it any improvement at all. Cos at present open source drivers on this c-50 apu don't offer even 10-20% performance of the catalyst drivers.

                              Just one puzzle, why does AMD releases such bad/shoddy open source drivers that are no where stand close to the proprietary ones.

                              Comment


                              • #45
                                Originally posted by manmath View Post
                                Qaridarium, thanks for posting the reply. But here are a few points.

                                #1. it's not an APU-Fake. On the same device Windows 7 handles 3d works and hd 1080p quite well.
                                then why you don't use Windows7 on your "WINDOWS/Closed-Source-only" hardware?
                                its a APU-FAKE if you use "Linux" because if you use a intel HD4000 graphic based INTEL-APU you get a much better "DEFAULT" result because the intel hardware is not "WINDOWS-ONLY/Closed-Source only"!
                                AMD fail on your hardware on (Opensource) Linux thats the real point.

                                Originally posted by manmath View Post
                                But linux with open source graphics simply fails to do that.
                                o well in other words AMD FAIL to do Opensource drivers for there APU-Hardware. but i know the name who is responsible for this: "Bridgman"

                                Bridgman ruin AMD's APU strategy.


                                Originally posted by manmath View Post
                                I tried with catalyst, it works way better than the open source graphics, but it's too complicated in the sense i had to install catalyst drivers, vaapi-wrapper and then some workarounds in the media players.
                                in other words AMD fail Double time on your hardware on linux with the closed source crap to!

                                wen Double fail FAILS:



                                Originally posted by manmath View Post
                                I thought open source graphics will be an easy way.
                                you thought? don't mess with Bridgman he will always beat you LOL !



                                LOL bridgman beat you so hard LOL


                                Originally posted by manmath View Post
                                #2. Did you mean I'll get 10-20% speed boost if I used open source graphics on top of kernel 3.5 and proper xorg drivers?
                                no. the defaults are broken you have to reclock the GPU on your APU to get the speed. if you clock it higher you get 50-60% of the speed.

                                Originally posted by manmath View Post
                                If it's so then it's very unfortunate to call it any improvement at all. Cos at present open source drivers on this c-50 apu don't offer even 10-20% performance of the catalyst drivers.
                                don't worry AMD got double damage rune on you



                                in other words you are fucked:



                                Originally posted by manmath View Post
                                Just one puzzle, why does AMD releases such bad/shoddy open source drivers that are no where stand close to the proprietary ones.
                                don'T worry this is just market-share politics from managers like "Bridgman"

                                you get what your market-share well-deserved

                                the market fucks you:

                                Comment

                                Working...
                                X