When the first reviews emerged of AMD’s Ryzen 7 processors, benchmarks suggested solid multithreading performance in high-end applications, good multitasking performance, and surprisingly anemic Ryzen 7 gaming performance. But to understand Ryzen gaming benchmarks, you need to understand several things:
- Ryzen is the first really new x86 architecture variant in years.
- PC game benchmarks for testing CPUs run at low resolutions with low detail levels.
- Games developed with Ryzen in mind perform better.
That Ryzen is new, yet still manages to perform well in applications tests, suggest the new architecture meets its promise of increased instructions per clock (IPC). While single-threaded performance lags a bit behind the newest Intel Kaby Lake CPUs, Ryzen remains a breath of fresh air for AMD.
In part I of the Ryzen diary, I talked about the Ryzen PC as it came out of the box. Part II discussed changes I made to the system, plus a bit of on-the-fly troubleshooting. In this third installment, I take a brief sidetrip into the art of benchmarking but eventually get to a bit of subective gaming experience on the Ryzen 7 1800X PC discussed in the first two parts.
Gaming Benchmarks are Weird
Game benchmarks, whether using simple timedemos or more sophisticated frame interval counting, are among the most complex benchmarks. That’s because games consist of multiple subsystems with different goals: graphics output, AI, world loop management, database tracking, and more. Modern AAA studio games often have a heavy focus on graphics. Games like the original XCOM and Civilization used to be graphically simple even though the internal simulations may have been complex. But modern iterations such as Civilization 6 and XCOM2 make heavy use of graphics modern graphics. Open world games like Rise of the Tomb Raider, Ghost Recon: Wildlands, and Mass Effect Andromeda often result in the graphics hardware being the limiting factor in performance, not the CPU.
To counteract this, CPU game benchmarks tend to be run with all the graphics features dialed down as much as possible and as at low a resolution as possible. Game performance tests run in this manner show Ryzen lagging Intel CPUs, sometimes by a little, sometimes more substantially.
No one, of course, actually plays in this way, unless they’re forced to — users of Intel HD Graphics, for example. Anyone spending $500 for a CPU will likely want a damned good graphics card to enhance the game experience with that processor. Once you drop in a good graphics card and run at full resolution with lots of eye candy, performance differences become far less pronounced.
Tuning Ryzen
That Ryzen represents the first new x86 architecture in years means we’re at the start of the cycle of optimization. No current application has been compiled with Ryzen in mind, yet the new CPU runs most applications pretty well. Signs indicate that future applications may do better. AMD worked with Oxide Games, developers of the RTS Ashes of the Singularity, in an effort to optimize the game for Ryzen. Dave Altavilla at hothardware.com noted performance increases up to 20% after Oxide implemented Ryzen optimizations.
AMD also issued a new balanced power profile for Windows to work around an issue where Windows would put unused Ryzen cores into their deepest sleep state. This flushes any internal caches and affects performance due to the high performance cost of waking up the core. This can happen hundreds or thousands of times per second. PC Perspective tested the new power profile and saw modest, but measurable, performance gains with only an incremental increase in power consumption. Once Microsoft brings Windows up to speed on Ryzen power modes, this stopgap solution won’t be necessary.
But How Good is the Ryzen 7 Gaming Experience?
Let’s talk about subjective impressions. I installed Mass Effect Andromeda, Ghost Recon Wildlands, and XCOM2 on the Ryzen 7 system provided by AMD, connected to an Asus 1440p monitor and using an eVGA GTX 1080 SC GPU. I played all three games on my own for a number of hours. On top of my own game time, I had friends on the system during our Friday night LAN parties. Overall, we’ve logged 30 hours on the Ryzen system. During that time, we had one crash-to-the desktop during a Ghost Recon co-op session — and that was before the first Wildlands patch. One CTD in 30 hours of gaming is not a big deal; I’ve had that happen on Intel system, particularly with new games out-of-the-box.
More importantly, gameplay felt fluid and smooth. We cranked up the games detail levels to Ultra mode and always ran at 2560×1440 We never noticed any hiccups, stuttering, or low frame rates. Not only were games playable on Ryzen, but the gameplay experience was excellent. I can safely say without reservation that you can build a find gaming sysetm with Ryzen 7 and a good graphics card.
So much for Ryzen 7. What would you see with a lower-cost system using Ryzen 5 and a more modest GPU? I’ll tackle that in a future post, as I build up my $1,200 Ryzen 5 gaming PC.
8 comments
Skip to comment form
Thanks. BTW, first time seeing your writing in a very long time. I missed it quite a lot. You have been bookmarked and refreshing like a madman.
Author
I post when I have time, as my day job at tech analyst is pretty demanding. But I’m also having fun writing about what I enjoy most!
Did you do any benching with an amd rx-480 or a fury?
Author
Haven’t really done more than subjective testing yet. However, others have benchmarked RX 480 (and the new RX 580). Mostly, the GTX 1060 still outperforms the 480 on Ryzen, though the AMD GPU wins in about 25% of the tests. Most of the AMD wins come in DX12 games, but the 1060 does well across the board, including some DX12 titles.
A reviewer on YouTube, AdoredTV made a surprising find, once you didn’t use an Nvidia graphics card with a Ryzen CPU, the performance picked up quite nicely, matching intel level performance in most cases. More than likely a driver side issue on Nvidia’s Part. May be worth looking into.
https://www.youtube.com/watch?v=0tfTZjugDeg&t=1079s
Author
This might be interesting to test with midrange GPUs, but even a somewhat hobbled GTX 1080 or 1080 Ti would blow away any current AMD GPU. That’s why AMD really needs to get Vega out the door.
But you do understand what I’m saying right, if you test an amd gpu against an equivalent Nvidia gpu (rx480 vs 1060) on ryzen, ryzen seems to perform better with an amd gpu.
We are not looking at gpu performance, rather, cpu performance with different brands of gpu in a certain performance bracket.
I would have liked to have seen this with the fury.
Author
Yep. I plan on doing testing on a couple of midrange GPUs on the budget gaming rig I’m building using a Ryzen 1600X. I’ll use the best outcome in the final build.