AMD’s finalized fix for the controversial issue of lower than expected boost speeds on its Ryzen 3000 processors has been tested, but it might not magically solve all of the problems with the boost clocks.
The fix does indeed up those boost clocks, there’s a caveat in that a more fundamental problem still remains. Namely that the boosting activity isn’t always delivered in the right place (meaning the cores actively being used).
These are the findings made by Tom’s Hardware in an extensive examination of AMD’s BIOS fix, in which the Ryzen 9 3900X being tested boosted up to a maximum 4.65GHz, a 75MHz increase on the previously witnessed peak (with older BIOS versions) of 4.575GHz (and also above the rated 4.6GHz speed).
That uptick is obviously good news, but there’s a bigger picture here. As the tech site reminds us, what we have to remember is that Ryzen 3rd-gen CPUs don’t simply hit a uniform max boost speed across all cores, but rather have a mix of faster and slower cores as boost kicks in.
So, if the wrong cores are getting boosted – meaning the ones not actually being used by any given process – then obviously the chip is still missing beats in terms of overall performance levels, in spite of AMD’s improved boosting.
In short, if boosting is misdirected some of the time, then the impact made by any extra bit of boost gained by the BIOS change is going to be even more minimal (because we’re already talking about pretty minor performance increases anyway – although we shouldn’t lose sight of the fact that it is important that AMD’s chips hit their promised speeds, for obvious trust reasons).
Let’s dive into this in more detail: Tom’s performed its testing using a Ryzen 9 3900X processor in a Gigabyte X570 Aorus Master motherboard, as Gigabyte is one of only two motherboard vendors which has released the final version of the 18.104.22.168ABBA firmware that corrects the boost problem (Asus is the other, whereas the rest of the motherboard competition are still working with beta firmware).
Obviously that 22.214.171.124ABBA firmware was used, and the CPU was cooled by a Corsair H115i (Hydro series liquid cooler) and driven by a 1,500W PSU (to ensure there was no danger of the rig setup somehow contributing to not achieving max boost at any point).
A range of benchmarks were performed in the course of testing, including PCMark 10, GeekBench 4, VRMark, as well as Cinebench, the latter of which was used in overclocking expert Der8auer’s survey, the one which really lit a fire under this boost clock controversy.
An interesting side note here is that AMD has since come out and said that Cinebench isn’t the best benchmark to measure boost frequencies, as over an extended test such as this, “the operating frequencies may be less than the maximum throughout the run”. Although Der8auer noted that AMD actually recommended Cinebench R15 to be used for measuring boost.
Go figure, but at any rate, now AMD is recommending PCMark 10 as a good “bursty workload proxy” for testing the boost speeds of Ryzen chips, and indeed Tom’s highlights this benchmark suite as particularly interesting.
That’s because there’s much more boost activity during these tests with the new fixed BIOS compared to older BIOS versions, with greater numbers of boosts evident, and faster peak speeds reached.
So that’s all good news, and as we’ve already mentioned, Tom’s observed a 75MHz increase in boost on the tested 3900X with the fixed BIOS – but the sticking point is the aforementioned non-uniform boost behavior, and those differing faster and slower Ryzen cores.
The point is that the Windows 10 scheduler should be able to implement these boosts to the correct (in use) cores, but as Tom’s notes: “Unfortunately, after analyzing the test output in detail, we see that, in many cases, the processor simply boosts on the wrong core.”
It may be the case that one core is being driven fully to 100% use, and it gets no love from boost, whereas an idle core gets boosted instead. Hopefully this is a situation that will change when AMD and Microsoft get a better handle on targeting the scheduler more successfully.
We should clarify that this scheduler behavior is nothing new, and isn’t caused by this new BIOS update, which does indeed simply deliver the promised (slight) boost speed increase.
And, we should also bear in mind that the misdirected boost issue will vary in its significance depending on what you’re doing with your PC. Tom’s observes that boosts tend to occur on inactive cores more often than not with PCMark 10 and Geekbench, but other benchmarks saw more successful targeting of the boost.
Finally, we have to remember that when we first reviewed Ryzen 3000 processors out of the gate, they scored uniformly brilliantly across the range, and without a doubt offer fantastic value for money in terms of price/performance ratio, even with any boost speed shortfall.
And now those max boost speeds have, according to this report, been lifted slightly – with, AMD says, no chip longevity downsides – and hopefully, given time to work on the nuts and bolts in the background of Windows, they will be more appropriately applied before too long. And that should give us even better performance from Ryzen 3000 products.
As was the case with the whole boost controversy in the first place, it’s all a matter of perspective here, and the overriding viewpoint for us has to be what you’re getting with a Ryzen 3rd-gen processor – and that’s a hell of a lot of performance for your outlay.