Ashes of the Singularity Scaling: The AMD Crossroads
Last week, the new game, “Ashes of Singularity” had a pretty comprehensive scaling review performed with both DirectX 11 and DirectX 12. The results were interesting to say the least. Multiple CPUs were used to test both the R9 390X as well as the GTX 980. While AMD enjoyed some impressive gains, NVIDIA had some fairly lackluster results that even prompted the company to release statements for damage control. It would be very easy to say that AMD is making a comeback and NVIDIA is gonna be in trouble, but that would be too easy. How can we come to the proper conclusions about these results?
Let me start off by saying that this is great news for AMD. It’s long been claimed that Radeon GPUs would be much better if the drivers could just utilize them properly. It seems this is almost true, but rather than drivers, it’s APIs that needed to take advantage of that hardware. However, I’m seeing some massive problems here that if AMD doesn’t quickly solve them, we can say goodbye to competition for a long time to come. I want to show you three conclusions that I saw from these results, and why I think there could be more bad news here than good if AMD doesn’t make some dramatic changes in the near future.
(Click for Larger View)
Let’s begin with the first big implication these AotS results are showing us. AMD needs to refocus their software development. This seems like something that is already in progress, but when we see what DirectX 12 is doing, it makes Mantle look like chump change. I’m glad AMD brought their low level API to fruition because it seems to have gotten the ball rolling on the other ones, but let the experts take it from here. Microsoft has a much better team (and budget) to handle this level of coding and it shows in the gaming results. AMD always seems to have issues with software. If they could focus their software teams to start churning out regular driver updates that make new game releases work smoothly at launch or quickly after, people would feel much more inclined to own AMD. This is still a major issue they face and even though they promised monthly driver updates a while ago, we’ve yet to see that level of commitment from AMD.
The next major issue I saw was the CPU bottlenecks. We’ve known for some time that INTEL was kicking AMDs butt all over the place in the CPU market. What I didn’t expect, was to see a dual Core i3 beating out 8 and 6 core AMD CPUs with DirectX 12. The reason why is, DirectX 12 is supposed to utilize multiple CPU cores more efficiently so that there is less of a demand on single-threaded performance. This sounded great for AMD, but a Core i3 is still offering better performance across the board than a 8370 does. Regardless of whether single-thread performance is still being a factor or the integrated GPU is being utilized, the fact remains that INTEL CPUs are better in DirectX 12 than AMD by a long shot. What I don’t understand is, why has AMD stuck with Bulldozer architecture for so long? I know many people have been asking this for some time anyways, but here’s where it really stumps me. While AMD was trying to push the industry to multi-core utilization to get past the eventual bottlenecks of single-thread performance, surely they realized long before now that their architecture was going to fail across the board. It looks like multi-threaded performance doesn’t benefit applications nearly enough if the single-threaded performance isn’t keeping up as well. I know this seems obvious, but I was holding out some hope that multi-threading would benefit more with Microsoft’s new API.
This brings me to my final conclusion about what I think is going to be a major concern for the future of AMD. Forward thinking is a good thing, but it can’t be used as an excuse for not taking the here and now into consideration. The disturbing trend I’m seeing is that AMD gets so caught up with trying to make the future happen, that what they offer in the here and now really isn’t worth it. The results for AotS in DirectX 12 are great for the R9 390X. We see huge improvements from the previous API, the aging Hawaii architecture is now exceding the performance of the more expensive GTX 980, and it seems to prove that AMD was right about simply needing better software. Unfortunately, this is only ONE game title at the moment. Right now, the gaming market is overwhelmingly saturated with DirectX 11 games. Unless a buyer only plans to play AotS, that person will have a much greater benefit from the NVIDIA card. What impresses me is how effective NVIDIA has been in utilizing DirectX 11. This tells me that there is an extremely good chance that when they release their next architecture, they’re going to address many of these issues because they are actually becoming relevant, currently. AMD can enjoy this victory, but unless they start showing some huge improvements in DirectX 12 with their hardware, they could just as easily be left in the dust again with future releases. It’s also important to note that this is only one game, which is far from being able to make any solid conclusions.
By now you might be a little worried about what I’m getting at here, and you should be. AMD just released the Fury X and even though HBM is great technology, what we ultimately have is a sub-par GPU. The performance is good, but overclocking hasn’t lived up to expectations created by AMD and it still struggles to compete with the GTX 980 Ti. This reiterates the trend I talked about earlier in that they were so focused on reaching a new plateau, that they missed the fact that buyers are looking to play the games they own today. The good news is, AMD can turn their fortunes around, but next year may very well be their last chance. Zen has a lot of promise, but it also has a TON of catching up to do with INTEL. The Arctic Islands GPUs are supposed to be an entirely new line up of architecture, but NVIDIA’s Pascal will be implementing HBM as well. I believe AMD can turn their fortunes around, but it will take some “today” thinking. Two things they can do are focusing on making sure the individual Zen cores are performing well and making sure the Arctic Islands GPUs not only overclock well, but scale well with both DirectX 11 and 12. They’re going to have a lot of work to do, and the financial situation certainly isn’t helping, but let’s hope AMD gets their nose to the grindstone and starts believing that they can raise the competition to a head-to-head level with there rivals.