Slashdot Mirror


AMD Ryzen Game Patch Optimizations Show Significant Gains On Zen Architecture (hothardware.com)

MojoKid writes: AMD got the attention of PC performance enthusiasts everywhere with the recent launch of its Ryzen 7 series processors. The trio of 8-core chips competitively take on Intel's Core i7 series at the high-end of its product stack. However, with the extra attention AMD garnered, came significant scrutiny as well. With any entirely new platform architecture, there are bound to be a few performance anomalies -- as was the case with the now infamous lower performance "1080p gaming" situation with Ryzen. In a recent status update, AMD noted they were already working with developers to help implement "simple changes" that can help a game engine's understanding of the AMD Zen core topology that would likely provide an additional performance uplift with Ryzen. Today, we have some early proof-positive of that, as Oxide Games, in concert with AMD, released a patch for its game title Ashes Of The Singularity. Ashes has been a "poster child" game engine of sorts for AMD Radeon graphics over the years (especially with respect to DX12) and it was one that ironically showed some of the worst variations in Ryzen CPU performance versus Intel. With this new patch that is now public for the game, however, AMD claims to have regained significant ground in benchmark results at all resolutions. In the 1080p benchmarks with powerful GPUs, a Ryzen 7 1800X shows an approximate 20% performance improvement with the latest version of the Ashes, closing the gap significantly versus Intel. This appears to be at least an early sign that AMD can indeed work with game and other app developers to tune for the Ryzen architecture and wring out additional performance.

3 of 121 comments (clear)

  1. Dishonest benchmark is dishonest by aliquis · · Score: 2, Informative

    To begin with the Ryzen 7 1800X doesn't end up giving the same performance as the i7 5960X even with this update as shown in the article.
    It's completely fair to to deal with them having different spec, 3.0-3.5 GHz vs 3.6-4.1 ghz on the Ryzen 7 1800X processor, but the 5960X is also one generation old, the 6900K would be the current top of the line 8 core one and the 6950X the best one. So if one is going to do a best 8 core vs best 8 core or best consumer line processor vs consumer/enthusiast one this test fail, the prices aren't the same though.

    However what really disturb me from a comparison stand-point is that they gave the i7 5960X 2133 MHz DDR4 vs 2933 MHz DDR4 for the Ryzen 7 1800X, that give the 1800X another opportunity to shine since infinity fabric run at the same clock as RAM but why wasn't the 5960X also given the same speed RAM? It can't run it?

    But it would be more honest with same speed ram and the 6900K or even 6950X it that's what they want to show, or for a similar price point just the 6800K even though that's just a 6 core processor. Which is the most relevant? Up to the reader I guess.

    If we go with OC RAM there would of course exist the opportunity to overclock the 5960X from the 3.0-3.5 GHz range up beyond 4.5 close to 5.0 GHz whereas the Ryzen 7 chip will do 4.0-4.1 GHz on OC. .. and as for gamers what most would rather compare it against is the i7 7700K anyway with just four cores but 4.2-5+ GHz clock-rate and a lower price.

    Anyway, the test manage to show the increase in FPS with the patch and also compare it against whatever other development could had happened to the game (the i7 run the game even slower than before now, the patch affecting it negatively or the game just having become more complex?), so if that was all that would be shown that could had been shown alone but since it's compared to an Intel processor and that Intel processor is a generation old and with slower RAM I don't really feel the test is honest. Also more relevant models would be the i7 7700K, the i7 6800K and the i7 6950X to compare against the "gaming king", the "same price Intel enthusiast processor" and "the best processor of the Intel enthusiast line."

    1. Re:Dishonest benchmark is dishonest by Billly+Gates · · Score: 1, Informative

      The reason is the infinity fabric with it's NUMA architecture sucks with Windows 10 and is server oriented.

      What's going on is Windows 10 CPU scheduling and power management LOVES randomly throwing processes and threads around cores during workloads. So imagine your a game busy working on something under Ryzen? You get interrupted and asked to move. Cache is now lost thanks to NUMA and needs to be reloaded from ram and the ram OC is bottlenecked to 2999 MHz. Now you continue. Think that would cause a stutter and fps loss? You bet!

      The Intel shares the same cache and no infinity fabric so no biggie and continues. So a Ryzen is not 8 core. It's really 4 2 cores attached by infinity fabric on a chip.

      MS can fix this with a patch but won't due to the small marketshare

  2. Re:Anemic on details? by Anonymous Coward · · Score: 2, Informative

    CFLAGS=-g0 -DTT_CONFIG_OPTION_BYTECODE_INTERPRETER -pipe -O3 -march=native -fweb -funswitch-loops -funroll-all-loops -funit-at-a-time -fsched2-use-traces -fsched2-use-superblocks -fsched-stalled-insns=12 -frename-registers -fprefetch-loop-arrays -fpeel-loops -fomit-frame-pointer -fmerge-all-constants -finline-limit=32768 -finline-functions -ffunction-sections -ffast-math -fdata-sections -fbranch-target-load-optimize2 CXXFLAGS=$CFLAGS