Announcement

Collapse
No announcement yet.

AMD Radeon R9 290 On Linux

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

  • #16
    Originally posted by brosis View Post
    At 95c, these cards are designed to fail, now at triple rate. Post High K Dielectric processes on the sub 45nm level aren't heat-friendly. From the past, we have learned the basics of the cooling - lower temperatures = lower wear, higher temperatures = higher wear.


    The regression isn' t the driver, the regression is the chip itself. Its like nvidias 4xx, but made even dumber.

    I wonder what temperatures do FirePro variants will run at.
    Exactly. And this has been done to compete with Nvidia.

    This is an overclocked card. Period. 95C is not acceptable. At this rate, gfx cards are going to consume more energy than an air conditioner.

    Comment


    • #17
      The R9 290 (especially the non-X variant that Michael owns) had a lot of improvements for clocking/fan speed/performance with Windows Beta 8 and Beta 9.2 which hadn't been released for Linux. The next Catalyst should hopefully catch up.

      Comment


      • #18
        Its is a safe benchmark. R290 has 5Tflops of 32bit SP. GTX680 has 3.2Tflops of 64bit SP "or" 6.4Tfops of 32bit SP (each 64bit core has 1ALU and 2FPUs). On the other side AMD has Drystone. Integer execution on Radeons 5000+ are close to DP/FP, its actually 2-3 times that of Nvidia, see GPU Coin Mining. Now a game its mostly FP+ASIC, but an inside Driver FX like Antialising can be optimized for Integer, and many Graphics Engines can do the same duo to consoles. Any way Nvidia cannot have big difference. The difference that you see on closed Linux vs closed Windowz drivers are probably that Linux Catalyst has Hacks/cheating on only for Source and Unigine. Any way all those things don't matter for Linux. We want good quality Libre "Gallium only" Drivers (not targeting PTX). And we want finally some Distro to have Mesa and Wine with the D3D9 State Tracker on their repositories. After that, we will find bugs on that State Tracker and we will correct them, but not before. We don't want to face a so delayed merge, because some of you you are not ready. What is the problem, you have a pain in the ass and we all waiting to pass???

        Comment


        • #19
          95 C is hot for microelectronics, not all solid state devices

          Originally posted by schmidtbag View Post
          AMD never said that wasn't hot, they said that temperature was safe. 95C is extremely hot for electronics, but depending on how they're made, some electronics can handle that temperature better than others.
          In fact, while 95 C would be extremely hot for a CPU and is hot for a GPU, there are other, much larger fab process devices for which this is only warm. Most of the large RF transistors used in radio transmitters are good to 150 C, which is a standard for silicon discrete transistors. Some ruggedized devices will even tolerate 200C, temperatures in both cases being limited by crystal regrowth that tends to destroy junctions over time. Radio amateurs often push them hard, as a ham needing to replace a $2 IRF 510 once a year could often care less. I've seen a mismatched transistor get hot enough to boil water on its case for short intervals with no damage.

          That being so, if AMD found a way to reduce electromigration and crystal regrowth, I could see them raising temperatures and reducing heatsink size/mass and fan noise. All the same, I don't like what hot metal masses due to circuit boards over time, they too must be ruggedized if high temperatures are to be run.

          Comment


          • #20
            Nice set of benchmarks

            That was a great set of benchmarks, Michael. Comprehensive, wide selection of applications, plus performance per watt.

            Thanks!

            Comment


            • #21
              It is more than obvious as most of the previous posts mentioned and Michael's article that the current Beta Catalyst Linux driver isn't ready for these cards in many critical areas. That's why no samples for Linux, they knew that embarrassment would come for the cards in tests. R9 290 is far more powerful than GTX 680 in hardware specs!

              Expected by me cause I have a feeling that AMD puts much more effort in making new hardware designs than coding for finetuned drivers and in my opinion this is obvious for windows Catalyst driver also! I believe it has problems there too, less but there are certainly bugs present.

              Obviously GCN differs very much than the older archs used in HD 5000/6000 cards and we are witnessing the same phenomenon like in Mesa with R600 and RadeonSI drivers, inside Catalyst's code base.
              Power consumption is a big problem... Heat is also if it passes through the card to the rest of the box components and raises the overall system's temperatures.

              Another small notice I want to make is the CPU bottleneck situation found on Source games here is perfectly clear that it is present also at the much more expensive and "more powerful" I7 4770K CPU than the shy and cheap in comparison AMD FX 8350. Paragraph conclusion FX 8350 worths much more than the overpriced 4770 I7 at the end of the day...

              AMD takes a bit of revenge in openCL tests where nvidia (driver and hardware) seems to be thrown out of the water but that is not enough for AMD to turn heads at their side in Linux rising gaming stuff...

              Comment


              • #22
                How you could have recorded the clocks fluctuating

                Hi Mike,

                You mentioned in your conclusion that you were not able to tell whether the clock speed down-clocked from heat due to the CCC not displaying it. You could have used the --odgc command to monitor the clocks.
                That is how I currently monitor the clocks on my R9-290X. Therefore it should also work with the R9-290.

                To record while a benchmark is happening, you can use a combination of the "watch" command and the controlFlow ">>" to redirect the output and append to a file per interval.

                Code:
                watch aticonfig --odgc >> clocksPer2Sec.txt
                With my R9-290X running on Uber mode, it tends to downclock once it reaches 92-94 C.

                With Metro LL with the slider-settings maxed to the right, I observe mostly FPS between 40+ to 60+. On the surface levels with the massive storm on the bridge and the last climax level; I observe FPS of 30 to 50+.
                I also observed that I could use CCC to overide the Anti Aliasing to maximum with EQ and using Edge Detect; however on levels with a lot happening, like the last climax level, the game frequently crashes. Switching to enhancing the applications AA instead of overriding it virtually eliminated the crashing.

                On another note, something I found peculiar; Serious Sam BFE with ULTRA everything for performance settings, would heat up the card faster and throttle down the clocks faster than Metro LL. It was like running Furmark.

                Comment


                • #23
                  Originally posted by DanglingPointer View Post
                  Hi Mike,

                  You mentioned in your conclusion that you were not able to tell whether the clock speed down-clocked from heat due to the CCC not displaying it. You could have used the --odgc command to monitor the clocks.
                  That is how I currently monitor the clocks on my R9-290X. Therefore it should also work with the R9-290.
                  No, the OverDrive clock information wasn't displayed correctly for the GPU on the driver.

                  PTS already has full support for realtime clock monitoring so I can just run:

                  MONITOR=gpu.freq phoronix-test-suite benchmark xxx and it will plot me real-time clock data on all the different Linux drivers, but the OverDrive information wasn't being reported for the R9 290.
                  Michael Larabel
                  http://www.michaellarabel.com/

                  Comment


                  • #24
                    I think potential buyers of Radeon R9 290 should just wait for models with manufacturer's alternative cooling solutions. AMD's reference cooler just sucks. My ASUS 7950 runs colder, at around 80C during torture tests like furmark.

                    Comment


                    • #25
                      Originally posted by AJSB View Post
                      If i read correctly what AMD said about these cards....


                      The 290 cards series DON'T run hot.

                      They were specially designed to run at 95C.

                      If for some reason , a 3rd party cooling solution for example tries to make it run cooler, the driver/card pushes performance till card reaches again that temperatures....OTOH, if a cooling solution is inefficient and card wants to push over 95C, performance of card is cut.
                      So it's all good for those of us that have the skill to build a liquid cooling setup out of a car radiator then.

                      I wonder how far the performance will increase if cooled by a phase change system? If only I had the skills to make one of these http://www.frozencpu.com/products/13...1G.html?tl=g49

                      Comment


                      • #26
                        custom better cooling

                        We're entering summer here down under in a couple of weeks. I've decided to customise the cooling of R9-290X so it can perform better under the Aussie heat.

                        Unfortunately all the after-market air-coolers take more than 2 slots. I can only afford two slots as all the other slots are taken up by cards.
                        I emailed Sapphire, Gigabyte, HIS and XFX if they were willing to sell me their custom coolers once they are released and so far only Sapphire has responded with a flat no advising I will have to buy another card.

                        So looks like I will have to go watercooling again. I've been trying to avoid watercooling as the last one I did (4yrs ago) was too high maintenance and restrictive (due to tubing and fittings all around the place, harder to just pull stuff out and replace/experiment/tweak). Also now I got barely any room within my case due to a Noctua D14. oh well, I suppose I'll end up enjoying the "tinkering-around" once I start it.

                        Another option, I just sell it now and recover as much as I can and get a 780 Ti. LoL

                        But, I'm betting AMD will come out in force with the cavalry from 1H of next year due to market forces from the ubiquitous SteamOS/SteamMachines. And eventually Mantle enabled and supported games on Linux...

                        Comment


                        • #27
                          Something tells me they're waiting for Mantle to be ready as well. For now Hawai are under testing for Windows so that we get a better software product, yeah i know wishfull thinking. They probably just forgot =(

                          Comment


                          • #28
                            I currently have a GTX 570. When I upgrade sometime in the future, I want to buy an AMD card and use the open source drivers. How soon do you guys think I can do so and see similar if not better performance from an AMD card that came after the GTX 570?

                            Comment


                            • #29
                              Originally posted by DanglingPointer View Post
                              So looks like I will have to go watercooling again. I've been trying to avoid watercooling as the last one I did (4yrs ago) was too high maintenance and restrictive (due to tubing and fittings all around the place, harder to just pull stuff out and replace/experiment/tweak).
                              As someone who has experimented A LOT with water cooling there are a couple of tips I can give you:
                              1. first and for most:USE QUICK COUPLING ADAPTERSThese things really make much more simple to play around, upgrade, plug/unplug, extend, etc.
                              2. Use large diameters, thick wall tubes of good quality. Large inner diameter (9mm and up. ~⅓" I think) make more flow and less resistance/pressure in the tubes. Large outer diameters (12mm and up. ~" ) makes sure that the tube is solid and won't tear. The quick-coupling adapters I've given as an example above use such wide tubing.
                              3. Use dual pumps in serie. That keeps more flow, and in case of 1 pump failing you still have the other one preventing the PC from melting. My setups used either 2x Lain DDC, or Innovatek HPPS + pump immersed in reservoir.
                              4. If you want to use a flow meter, use one with the least resistance (big internal diameter of the channel).


                              Also now I got barely any room within my case due to a Noctua D14. oh well, I suppose I'll end up enjoying the "tinkering-around" once I start it.
                              Well the good part with water cooling is that you can keep the bulk of the mass (radiator and reservoir) away from the heat source (away as in "out of the case", eventually with a long tube in between. If you go for the dual pumps as I'm suggesting, pressure and flow won't be a problem even with some distance).

                              I use radiator/reservoir hybrids, like the first serie of Zalman's reserators (The tall towers) or Kailon (a little bit less happy about the noise). That gives a lot of exchange surface without taking space inside the case (and as a bonus, these have a pump immersed inside, so you only need to think getting 1 extra pump).
                              I also put as many radiators on the out-take fan grills of the case as possible (Black Ice are my current favourite). If you choose big enough radiators of good enough quality, you may end up having way much more exchange surface than the typical in-case fan-cooled radiator. You can probably reach the same performance as your current monster, but without taking 99% of the free space. (Which in turn allows more air flow inside the case to help cooling the other components). I end-up with a rather spacious case, with the big honking radiators hanging on the outside.


                              Another option, I just sell it now and recover as much as I can and get a 780 Ti. LoL
                              Find someone into alt-coin mining. AMD GPUs are good on Scrypt-based alt-coins (Litecoin and Co).


                              But, I'm betting AMD will come out in force with the cavalry from 1H of next year due to market forces from the ubiquitous SteamOS/SteamMachines. And eventually Mantle enabled and supported games on Linux...
                              Well GCN 2.0 will probably more innovative and less incremental than GCN 1.1. Don't forget also that the other consoles too are using AMD hardware under the hood sa that's extra R&D that will eventually find its way into future Radeon products.

                              But I wouldn't bet on Mantle. 3D graphics are better served a completely cross-platform multi-vendor API, like OpenGL. Mantle would put is back into the 3dfx Glide days.

                              Comment


                              • #30
                                If anyone wants a little relief from the heat of one of these, you can remove the heatsink and do a careful application of thermal compound. It will still run at 95C, but it will be able to run a little faster. Note that this will probably void your warranty.

                                Like this. I'm surprised they picked that thermal compound, because they had run a big thermal compound comparison, and the best performance on their test GPU was with the Coollaboratory Liquid Metal Pad.

                                Comment

                                Working...
                                X