Announcement

Collapse
No announcement yet.

AMD APU vs. Radeon GPU Open-Source Comparison

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

  • #31
    Originally posted by chithanh View Post
    The folks from HardOCP know that the numbers are meaningless, but they include them anyway because people keep asking for them.

    The only thing that matters for a [H] review is the gameplay experience, something which can only be poorly expressed in terms of fps numbers.
    The value of a game benchmark is not in the results. Everybody knows what the desired results should be. The value of a game benchmark is in the configuration used to achieve the desired results. A game benchmark that runs at 3 frames doesnt even come anywhere near what is desired. That has no value.

    Comment


    • #32
      Originally posted by Michael View Post
      The testing isn't about letting you know if a game is playable or not per se but using it as a means of measuring/comparing performance improvements and differences, etc. I don't even play video games myself nor do a majority of the companies using PTS care about the game performance from that metric.
      As Duby and I have been saying for a long time now, your results are very much worthless. Take a look at the HardOCP link I posted earlier, note how they lay out their graphs with a a line at 30FPS as the minimum playable threshold and a line at 60FPS as the idealized threshold and how they test for minimum framerate as well as tweaking settings for maximum detail settings for a given card to still get a playable framerate.

      If cards start pushing past 100FPS in a game at max detail settings at 2560x1600 they retire the game from the list as at that point you have to go to 5670x1080 or higher which is uncommon for anyone not using a top of the line GPU. Why? Because at over 100FPS you are doing a CPU bound benchmark, not a GPU bound benchmark.

      Comment


      • #33
        Originally posted by curaga View Post
        Ouch. Why are they pretending to do a fair compare, if each playthrough is manual? Their numbers are meaningless.

        On the presentation, like others said, it's the exact same info, just presented differently. I really don't care if the min/max numbers were also in table form, because they alone are not that useful: what if it was a single anomalous spike? In the graph you can see how often the spikes appear.
        Take a closer look, they are actually testing the cards to their actual limits of maintaining playability by pushing the settings on each card to the point where they should just barely fall below the minimum playable threshold once or twice during the test.

        Those spikes are present in real demos and real gameplay because the GPU load in no game is a constant. As the number of things going on on screen increases and falls so to does the load on the GPU and thus the framerate also rises and falls with it. Theres a world of difference between the GPU load of sitting in a building with no firefight happening and going outside of it and having a war zone raging on.

        Comment


        • #34
          Originally posted by Michael View Post
          I was telling you what I care about. I know gamers care about achieving consistently at least 60 FPS.
          You run a gaming site moron.

          Comment


          • #35
            Originally posted by Kivada View Post
            If cards start pushing past 100FPS in a game at max detail settings at 2560x1600 they retire the game from the list as at that point you have to go to 5670x1080 or higher which is uncommon for anyone not using a top of the line GPU. Why? Because at over 100FPS you are doing a CPU bound benchmark, not a GPU bound benchmark.
            It is important to know if CPU or GPU is bottleneck.

            Comment


            • #36
              Originally posted by JS987 View Post
              It is important to know if CPU or GPU is bottleneck.
              Unless you are playing a heavy RTS or a simulator like Flightgear or MS Flight Sim then it's safe to say that the GPU is the bottleneck until you hit the levels of resolution and detail I mentioned before. Once those same titles are averaging 120+ FPS at 2560x1600, the highest resolution available on consumer monitors up until very recently.

              In 5 years when you can pick up a 3840x2160 for $150 then the goalpost will be moved. Just as it was when the average high res screen was only 1024x760 but most people ran 640x4080 or 800x800 to get acceptable framerates, this was upped to 1280x1024 then 1600x1200 then 1920x1200 then to 2560x1600. Such is technological progress.

              Comment


              • #37
                Originally posted by duby229 View Post
                I'm afraid you have me confused with another poster, I responded because I agreed that HardOCP does game benchmarks the best.
                I most certainly did mistake you for the original poster. Apologies.

                Comment


                • #38
                  Originally posted by Kivada View Post
                  You run a gaming site moron.
                  Does he?
                  According to the site creator/maintainer he doesn't.
                  IMHO, he doesn't as well: he runs a benchmarking site which focuses on linux/bsd.

                  Comment


                  • #39
                    Originally posted by liam View Post
                    Does he?
                    According to the site creator/maintainer he doesn't.
                    IMHO, he doesn't as well: he runs a benchmarking site which focuses on linux/bsd.
                    I'm not condoning name calling of course, that was out of line, but he is right. It doesnt matter what os the game runs on. What matters is that he's doing it wrong.

                    Comment


                    • #40
                      Originally posted by liam View Post
                      Does he?
                      According to the site creator/maintainer he doesn't.
                      IMHO, he doesn't as well: he runs a benchmarking site which focuses on linux/bsd.
                      May I point you to one of my previous pages and the homepage of this site:

                      From http://phoronix.com/forums/showthrea...836#post375836

                      So either he is confused as to what his stated job is or he is trying to weasel out of doing what is expected of him by keeping up with his peers doing the same job on other sites.

                      Comment


                      • #41
                        Originally posted by Kivada View Post
                        May I point you to one of my previous pages and the homepage of this site:

                        From http://phoronix.com/forums/showthrea...836#post375836

                        So either he is confused as to what his stated job is or he is trying to weasel out of doing what is expected of him by keeping up with his peers doing the same job on other sites.
                        I am not confused, but you must be... This site is provided as-is, if you don't want to read it or provide anything useful besides just spieling the same stuff, please leave. If anyone the people I am just trying to please the most are premium members and those that have a stake in some form to the site or the software, and certainly not those using AdBlock, etc.

                        Yes, Phoronix covers popular Linux game releases, other notable Linux / open-source gaming news, game benchmarks, etc. Just because another hardware review site publishes reviews one way, doesn't mean I need to do it that way.... Most of my "peers" also use Windows, should I go switch to being a Windows web-site too?

                        In fact, none of my commercial PTS customers at any of the major hardware vendors have even requested game testing like you mention. (Such support to easily calculate the maximum resolution/IQ settings, etc, could easily be added as a PTS module within a couple dozen lines of code to auto-determine the maximum values while staying above 60 FPS or another threshold, but any manual testing will not be accepted.)
                        Michael Larabel
                        http://www.michaellarabel.com/

                        Comment


                        • #42
                          Originally posted by Kivada View Post
                          May I point you to one of my previous pages and the homepage of this site:

                          From http://phoronix.com/forums/showthrea...836#post375836



                          So either he is confused as to what his stated job is or he is trying to weasel out of doing what is expected of him by keeping up with his peers doing the same job on other sites.
                          Do you think you can be a grown-up and simply point things out without petulance?
                          I hadn't noticed that. Of course, it doesn't change the facts that the operator of the site doesn't consider this a gaming site.
                          I suppose he talks about games a bit, but it seems like he's more interested in them from the perspective of them being of interest to the readers rather than following it b/c it is gaming.
                          The way he benchmarks games is fine, I think. It's not too dissimilar from how anandtech reviews video cards (they aren't a gaming site either but they'll make use of games to benchmark hardware). In fact, I think Michael does a better job in general since he always lists the frame timings. Of course their analysis is far better but they've got a much larger staff.
                          Certainly he could improve things greatly by presenting data in an easier to parse way, and benchmark a wider variety of games but, for what resources he has, I think he does a fine job.

                          Comment


                          • #43
                            Originally posted by Michael View Post
                            I am not confused, but you must be... This site is provided as-is, if you don't want to read it or provide anything useful besides just spieling the same stuff, please leave. If anyone the people I am just trying to please the most are premium members and those that have a stake in some form to the site or the software, and certainly not those using AdBlock, etc.

                            Yes, Phoronix covers popular Linux game releases, other notable Linux / open-source gaming news, game benchmarks, etc. Just because another hardware review site publishes reviews one way, doesn't mean I need to do it that way.... Most of my "peers" also use Windows, should I go switch to being a Windows web-site too?

                            In fact, none of my commercial PTS customers at any of the major hardware vendors have even requested game testing like you mention. (Such support to easily calculate the maximum resolution/IQ settings, etc, could easily be added as a PTS module within a couple dozen lines of code to auto-determine the maximum values while staying above 60 FPS or another threshold, but any manual testing will not be accepted.)
                            Your job is to be the PCPer/HardOCP/Hardware Secrets equivalent to Linux just as Barefeats is the Mac equivalent. Granted Barefeats focuses more on the performance in pro apps like Motion, After Effects and OpenCL calculations but at least thery stay current even if that software and hardware is quite expensive.

                            Now why would I do that and allow you to go on running the good name of Linux into the ground without saying anything back to you about it?

                            Thing is, you are leaving money on the table and chasing off all but the circle jerk of zealots that have signed up for the premium service, as I've stated many times, I'm keeping adblock on for the site and not paying for it until you learn to make the site better.

                            Define popular game. Nobody cares about half the games you've posted, really, Unvanquished and Euro Truck Simulator? Yet there are dozens that you have never posted about, if you remember about a year ago I posted a list of some 40 games that had been released over the previous 6 months that never once got a mention on Phoronix. These days there are hundreds you've missed, many quite good, I've got 120 or so games just on my Steam account, another 30+ on Desura, a good 60 in the wishlists waiting for when they come on sale or I've worked my way through my current library and have to buy more.

                            Take a look around at the Windows review sites that I've listed here numerous times. Theres a reason why they use minimum framerate and actual frametime at the output cable numbers. Because those are the things that will harm user experience the most.

                            What you test is garbage through and through and your "news" is always a regurgitation from other sources but without the common courtesy of a link back to the original article or mailing list you got it from which is standard practice on every other site on the internet.

                            Oh the big boastful talk about the mythical commercial PTS clients that there is no list of, who would they be again?

                            If you where serious about monetizing PTS you'd add FRAPS like functionality to get constant feedback on framerate and temps mid game as well as built in video capture since none of the desktop capturing software on Linux currently works all that good, even after much tweaking I can either have it crash after 5 mins or have no sound but not both unless I run the audio recording in Audacity via a male-to-male minijack cable from the line in to the line out and then have to sync it manually in editing. Not having to do that is something I'd pay for.

                            Comment


                            • #44
                              Originally posted by liam View Post
                              Do you think you can be a grown-up and simply point things out without petulance?
                              I hadn't noticed that. Of course, it doesn't change the facts that the operator of the site doesn't consider this a gaming site.
                              I suppose he talks about games a bit, but it seems like he's more interested in them from the perspective of them being of interest to the readers rather than following it b/c it is gaming.
                              The way he benchmarks games is fine, I think. It's not too dissimilar from how anandtech reviews video cards (they aren't a gaming site either but they'll make use of games to benchmark hardware). In fact, I think Michael does a better job in general since he always lists the frame timings. Of course their analysis is far better but they've got a much larger staff.
                              Certainly he could improve things greatly by presenting data in an easier to parse way, and benchmark a wider variety of games but, for what resources he has, I think he does a fine job.
                              I tried being nice years ago, I gave up on that as it's less energy intensive on myself to be blunt instead of dancing around the subject. That is why it's called being blunt. It takes to much effort to be eloquent or tactful.

                              Larabel's frame timings aren't being measured at the video output, go look up how PCPerspective does their frame time tests as this is the gold standard.

                              Larabel could have a staff as well, if he actually tried to build a good site from the start instead of putting so many ads on it I had no choice but to use adblock just to make the site even readable, or do you not remember when the site didn't just have text that was mostly links back to previous Proronix articles but also had in line text ads filling in the gaps between those so that every page was just a pile of links, half of which if you accidentally placed the cursor over would open up into windows that would cover half of the page. Or how about the full page ads that would takeover the site floating over the page and wouldn't go away?

                              It was crap like that of top of the bad reporting and bad netiquette(never citing sources of his articles and linking back to himself and only himself) that pisses me off about Larabel to this very day. When you question him about it you can see what he does, he complains about adblockers and says he only cares about the premium zealots. He should just shut the page down to all non premium members for a month then and see how well that goes for him.

                              I long for the day someone else starts a new Linux hardware benchmarking site. Anyone have the time for it? Make a kickstarter and I'll be among your first contributors.

                              Comment


                              • #45
                                Originally posted by Kivada View Post
                                Larabel's frame timings aren't being measured at the video output, go look up how PCPerspective does their frame time tests as this is the gold standard.
                                They are measured at the game engine, which is just as important. Frame latency at the engine level directly affects input latency, as the engine cannot process input for the next frame until the previous is out.

                                Measuring frames at the screen is also important, but it's measuring a different variable, that of the output chain. It is not as well correlated to input latency, but is able to expose driver/screen cheating. Apples and oranges.

                                TL;DR both measurements should be done, one is not better than the other.

                                Comment

                                Working...
                                X