When Ryzen launched, many reviewers noted the chip tended to compete extremely well against Intel in many tests, either in pure performance or performance-per-dollar. Ryzen gaming at lower resolutions like 1080p, however, was a weak spot in some tests. AMD promised that new BIOS (UEFI) updates and game patches would, over time, improve the performance of the platform. With AGESA 126.96.36.199 now rolling out to motherboard manufacturers, AMD has released some data on how Ryzen owners should tweak their platforms for optimal gains.
AMD tested the impact of the following settings:
- The impact of the new BankGroupSwap (BGS) BIOS option
- Single-rank DIMMs vs. dual-rank DIMMs
- Automatic sub-timings vs. manually-tweaked subtimings
- Max frequency vs. lower frequency at tighter timings
- Geardown Mode (GDM) on vs. off
Geardown Mode (GDM) isn’t an AMD invention; it’s part of the DDR4 specification. It allows DRAM to run at half of its specified frequency above DDR4-2667 when storing values on the memory module’s command or address buses (this is known as latching). According to AMD’s tests, the difference between GDM on versus off isn’t very significant. If you can hit your desired overclock with a 1T command rate, turning GDM off will improve your performance. If you can’t hit your overclock with a 1T command rate, leaving GDM on may boost performance.
Next up: BankGroupSwap. AMD describes this feature as follows:
[A] new memory mapping option in AGESA 188.8.131.52 [alters] how applications get assigned to physical locations within the memory modules; the goal of this knob is to optimize how memory requests are executed after taking DRAM architecture and your memory timings into account. The theory goes that toggling this setting can shift the balance of performance in favor of either games or synthetic apps.
Gains from using BGS appear to be in the 2-3 percent range, while synthetic benchmarks are slightly slower with BGS enabled. The largest single gains from any changes seem to be in the memory tuning department, when comparing manually tuned DDR4 with the defaults assigned to the motherboard by the UEFI.
The benchmarks above used single-rank, not double-rank, memory (double-ranked memory may not overclock as easily). But the gains here are significant — 8 percent in RotTR and 3DMark, while Hitman leaps by 14.5 percent. Does this mean that you’re going to always see those gains from every title? Absolutely not. Some games respond extremely well to memory overclocks, and some games don’t really respond to faster RAM at all. But at the very least, it shows that proper memory tuning can have a real impact on Ryzen performance.
Finally, there’s the argument over whether it’s better to have higher clocks and slower timings, or faster timings but slower clocks. In truth, it doesn’t really matter much either way — the largest gap measured was 4 percent. But if you’re looking to squeeze every last ounce of performance out of your Ryzen CPU, there’s some good data here.
Finally, here’s how it all shook out. Careful tuning and frequency testing boosted RotTR performance by 1.09x, Hitman by 1.16x, and the synthetic Sky Diver test by Futuremark by 1.09x. AMD makes some specific recommendations as to how you should tune a system based on your memory configurations and whether you have single or double-ranked DIMMs. If you want a fire-and-forget solution, dual-rank DIMMs offered the best solution. Everything else requires some degree of testing and evaluation to see whether these RAM settings match your own use cases. The total list of memory timing adjustment features added in Agesa 184.108.40.206 can be found here.
When AMD told reviewers Ryzen 7’s launch performance in certain games could be improved by better DRAM timing and in some cases, patch updates, we were skeptical. But based on results like this, it’s clear there’s at least some truth to the argument. Not every title is going to see these kinds of shifts, but clearly AMD left some performance on the table when Ryzen 7 first launched.