Alienware Aurora AMD Ryzen Edition: Quick Look at Overclocking
To start with, we used Ryzen Master's manual mode to just give the processor an all-core speed of 4.4 GHz, and let the system figure out how best to deal with boost. We cranked the clocks up and set the maximum CPU voltage to 1.375 and let it run the built-in stress test, which it passed easily with a CPU temperature pushing 85 degrees Celsius. That's still a touch under the Ryzen Master-defined max of 95C, so we felt pretty good about that. Unfortunately, the system locked while running Cinebench.
We didn't really want to feed the CPU more power, so we dropped the clock to 4.3 GHz. That resulted in errors in Cinebench R20. Still hesitant to give it more juice, we dropped the all-core speed to 4.2 GHz, which passed with flying colors. In fact, the system successfully ran Cinebench R20 back to back three times before we turned to Prime95. It also survived around 15 minutes of stress testing, and the CPU temperature never climbed above 87 Celsius. It seemingly leveled out at that point, which was good enough for us.
In our stock-clocked tests, the Aurora scored 8926. After around an hour of tinkering, we managed to squeeze a whopping 9629 out of the aurora, which represents nearly an 8% increase in performance. There might be a little performance left to find in our Ryzen 9 3950X, but deadlines pressed us for time. Still, it's a minimal time investment for a good performance gain.
Next we turned our attention to the GeForce RTX 2080 Ti. For this we used MSI Afterburner's OC scanner to find the optimal clock and voltage curve. Afterburner settled on an average clock increase of 151 MHz, which is a nice boost just shy of the default clock of 1545 MHz. This one we stress tested with Furmark, and there were no issues out of the gate. Once we felt good about our graphics overclock we added back Prime95 to see the full load wattage. Now the system draws 515W under this incredibly taxing load. Let's see what our overclock got us in the two most demanding games in our test suite.
Control saw some mild gains at 3840x2160, but it's still not enough to be completely playable with high ray tracing. With the medium effects, the 99th percentile frame rate was actually down slightly, but that can vary from run to run. Since this game is almost entirely limited by the GeForce RT cores, we don't think the CPU bump did much here. The average overall was still around 6% higher, which is definitely measurable but not something we could really feel.
The differences are again measurable but have minimal impact to how the game feels. At ultra settings, we saw a 6% boost in overall frame rates and a similar bump to 99th percentile scenarios. When we dial the settings back to we see roughly the same percentage gain. Power usage wasn't increased all that much either. We went from consuming around 395 watts while gaming up to around 430 watts. That's a difference of around 7%, similar to the gains we saw in performance, despite a pretty hefty CPU voltage bump. Overclocking didn't really move the needle on efficiency, which makes us think there's still more in there.
The GPU overclock was nice, however we believe that Gears 5 benefited more from the CPU overclock. Remember how we said that approximately 8% of all frames were CPU limited in our previous runs? Let's see how a decent CPU overclock helps the situation. For gaming purposes, this really made us pine for a Ryzen 7 3800X which could potentially hit higher clocks since it has fewer cores to keep cool.
Now more than 99% of all frames were waiting on the GPU. What exactly that means is a little mysterious. "CPU limited" might just mean that it was only utilizing 99% of the graphics card instead of 100%, but now we were maxing out the GeForce RTX 2080 Ti all of the time, and gained around 7% performance. That's with the overclock, too. Still, for a gaming-focused system, 16 CPU cores is overkill for anybody who isn't streaming software-encoded to Twitch or uploading to YouTube for the very highest image quality. Most folks who dabble in such things might be better off with fewer cores at a higher clock.
We may be leaving performance on the table, too, especially with the Ryzen 9 3950X. The Aurora's UEFI doesn't give access to all the different latency clocks of the HyperX 2933 memory, but Ryzen Master exposes all of those settings. It's possible that we could have wrung more out of the system if we'd had more time, but that's the fun part about tweaking. There's always something left to get.
Now that we've driven the Aurora's performance to the edge, let's sum up what we've learned and read the final verdict.