Memory, Overclocking & Power Consumption
One of the biggest complaints with the previous generations of Ryzen processors was a lack of support for high-speed memory when every little bit higher than you could push the memory, the more overall performance you would see. AMD has used a new memory controller this generation to combat this and advertises fast memory is supported with ease. This comes with a small caveat though, anything over 3733 Mhz would shift gears and would drop the Infinity Fabric to Memory ration back to 2:1 instead of 1:1.
We did all of our testings with 3200MHz memory, something that itself was a struggle to achieve on the first generation.
3600Mhz works well at 69.9 ns latency.
We see a good bump from the 3200Mhz 1903cb score.
4000 MHz sees about the same overall bandwidth but jumps to 80.9 ns of latency.
The jump to 4000 MHz actually drops the score slightly from the jump in latency.
At 4266Mhz, we see the bandwidth largely unchanged, but the latency is starting to come back down at 77.8ns.
Moving the speed up a little more doesn’t really change the score.
Even at 4333MHz, we still see roughly the same bandwidth, but the latency has dropped again to 77ns.
Even a little higher we see no real change in score.
Next, we’ll drop the memory back to 3200MHz and try opening up the Precision Boost Overdrive window up, letting it grab another 200MHz. this gets us a good jump in score over the 1903cb stock.
Manually overclocking, we were unable to get stable above 4.4GHz all cores, but this does give us another 100 points or so in Cinebench.
Pushing the memory up to 4133MHz at the same time gives us just the slightest increase in score.
These settings give us a minute and a half faster 32M time in Super Pi as well.
Power consumption looks great and we see a drop from the previous 2700X under load thanks to the 65W TDP.