A lot of AMD Ryzen 3000 processors do not touch the backup clock: report

Overclocking Der8auer released the results of a survey of more than 3,000 Ryzen 7nm owners who purchased AMD's new processors since they went on sale in July. Last month, reports revealed that the Ryzen 3000 family had not hit its start clock, as well as some expected hobbyists. Now we have data on exactly what these numbers look like.

There are, however, two confounding variables. First, Der8auer had no way of determining which AMD users had Windows 1903 installed and was using the latest version of the company's chipset drivers. AMD recommends both maximum performance and the desired supercharging behavior. Der8auer acknowledges this, but believes that it is incumbent on AMD to communicate with end users regarding the need to use certain versions of Windows for optimal performance.

Secondly, there is the fact that surveys like this one tend to self-select. It is possible that only the subset of end users that are not see the desired performance will respond to such an investigation. Der8auer also acknowledges this as a very valuable one, but believes that his community of viewers is broadly supportive of AMD and favorably focused on the smallest processor manufacturer. The full video can be seen below; we extracted some charts for discussion.

Der8auer carefully examined the survey data to find results that were meaningless or that were obviously made in bad faith. He has compiled data on the 3600, 3600X, 3700X, 3800X and 3900X models. The clock distributions were measured up to two deviations from the average. The maximum boost clock was tested using the Cinebench R15 single-line test, as recommended by AMD.

In the case of the Ryzen 7 3600, 49.8% of the processors reached their 4.2 GHz boost rate, as indicated above. However, as clocks increase, the number of processors that can reach their amplifiers decreases. Only 9.8% of 3600X processors reach their 4.4 GHz. The 3700X chart is presented below for comparison purposes:

The majority of 3700X processors are capable of reaching 4.375 GHz, but the 4.4 GHz amplified clock is a more difficult leap. The 3800X improves these numbers: 26.7% of the processors have reached the boosted clock. This seems to correspond to what we have heard from other sources, according to which the 3800X is a better overclocker than the 3700X. However, the 3900X is fighting more, with only 5.6% of processors reaching their full boost.

We can assume that at least some of the people who participated in this study did not install Windows 10 1903 or updated AMD drivers, but it was AMD users that had the most reason to install these updates. up to date, which should help limit the impact of the confusion variable.

The ambiguous meaning of "up to"

After analyzing the results, Der8auer explained that he still recommended AMD's 7 nm Ryzen processors with comments like "I absolutely recommend buying these processors". His statements are unambiguous, nor in our performance review. AMD's 7 nm Ryzen processors are excellent. But an excellent product can still have problems that need to be discussed. So let's talk about CPU clocks.

The reason why Intel (which made it debut) launched Turbo Boost as a product feature was to give itself room for maneuver with regard to processor clocks. At first, Turbo Boost processors seemed to simply treat the higher optional frequency as the actual target frequency, even when the load was less than 100%. This is no longer true for several reasons. AMD and Intel processors sometimes operate at lower clocks depending on the mix of AVX instructions. High-end processors such as the Core i9-9900K can slow down considerably if they are under full load for an extended period (20 to 30 seconds) if the motherboard is configured to use the power settings from default of Intel.

In other areas, like smartphones, it is not necessarily unusual that a device never works at the maximum clock. Smartphone vendors do not advertise base clocks at all and provide no information about a supported SoC clock. Often, it is the responsibility of the reviewers to define the behavior of the device based on the post-launch analysis. But the Intel and AMD processors have generally been considered, at least theoretically, to be able to hit the Boost Clock under certain circumstances.

The reason I say that this view is "theoretical" is that we observe many variations in the behavior of the processor, even during a single revision cycle. It is common for UEFI updates to arrive after the start of our tests. Often, these UEFI updated specifically to repair pointing problems. We correspond with various motherboard manufacturers to tell them what we have observed and we update the platforms throughout the review to ensure that the behavior of the diet is appropriate and that the cards work as planned. When checking overall performance, however, we tend to compare the performance test results to the manufacturer's expectations, instead of focusing strictly on clock speed (the performance, after all, is what we try to measure). If the performance is strangely low or high, the clocks on the CPU and RAM are the first place to check.

However, it is not unusual to be more or less 2-3% compared to the manufacturer or our fellow examiners, and occasional excursions of 5 to 7% may not be extraordinary if the reference is known to produce a largest range of scores. Some tests are also more sensitive than others to the timing of RAM, the speed of the SSD or a host of other factors.

Now, consider the Der8auer data on the Ryzen 9 3900X:

Only 5% of the batch processors are capable of reaching 4.6 GHz. But a processor clocked at 4.6 GHz is only 2% faster than a processor clocked at 4.5 GHz. A gap of 2% between two products is close enough that we call it an effective tie. If you evaluate CPUs strictly on the basis of performance, with a reasonable margin of 3%, you would get an "acceptable" clock range of 4662MHz to 4738MHz (assuming a 1: 1 relationship between clock and performance). And if you take into account this gap in the graphs above, a noticeably higher percentage – though not, not all – AMD processors "are eligible" and effectively reach their highest clock level.

On the other hand, 4.5 GHz or less is factually do not 4.6 GHz. There are at least two significantly different ways of interpreting the meaning of "up to" in this context. "Up to X.XGHz" does it mean that the processor will occasionally get boost boost under certain circumstances? Or does it mean that some Processors will be able to hit those boost frequencies, but you will not know if you have one or not? And how important is this distinction, if the whole performance of the piece corresponds to the expected performance that the end user will receive?

Keep in mind that one thing these results do not Tell us what overall performance looks like on all Ryzen 7 processors. Knowing the highest boost rate reached by the processor does not tell us how long it has been. A processor that maintains a constant 4.5 GHz clock from start to finish will outperform a processor that explodes at 4.6 GHz for one second and drops to 4.4 GHz to complete the job. These two behaviors are possible under a model "up to".

Manufacturers and consumers can see this problem differently

Although I do not want to rain on his parade or his next article, we spent the last weeks in ET troubleshooting a laptop that my colleague David Cardinal recently bought. More precisely, we try to understand its behavior under load when the CPU and the GPU are simultaneously used. Without revealing anything about this future story, let me say this: the process consisted of understanding how complex thermal management is between different components.

I think manufacturers are increasingly looking at power consumption and clock speed as a way of balancing the performance and power allocated to the components they need and limiting them everywhere else. The increased variability is on the agenda. What I suspect of AMD, in this case, is to set a performance standard that its processors should provide rather than a specific clock frequency target. If I had to guess the reason the company did this, I guess it's because of the inherent difficulties in keeping high clock speeds at lower process nodes. AMD probably chose to push the limits of its clock goals because the processors compared themselves better to their Intel counterparts with respect to maximum clock speeds. Any negative response from critics would be mitigated by the fact that these new processors offer significant advantages over previous-generation Ryzen processors and their equivalent Intel counterparts.

Was it the right call? I am not sure. This is a situation where I really see both sides of the issue. The Ryzen 3000 family offers excellent performance. But even after taking into account variations due to the Windows version, driver updates, or UEFI issues from the manufacturer, we do not see as many AMD processors reaching their optimized maximum frequency as we do. 'Expectations, nor high-end processors with higher boost clocks have more problems than low-end chips with lower clocks. AMD's claims that the output frequency of TSMC 7 nm would be higher than that of 12/14 nm GF seemed somewhat suspect at this stage. The company absolutely got the performance gains we wanted, and the power improvements on the X470 chipset are also very good, but the timing situation was not as detailed as it should have been at launch. .

According to rumors, AMD would have changed behavior with the latest versions of AGESA. An employee of Asus, Shamino wrote:

I have not tested a newer version of AGESA modifying the current state of 1003 boost, not even 1004. If I am aware of the changes, I will specify it more specifically. They were too aggressive with the boost previously, the current boost behavior is more consistent with their confidence in long term reliability and I have not heard of any change in this position, although I am not sure. have heard about a "more customizable" version in the future.

I do not have any specific knowledge of this situation, but it would surprise me. First, reliability models are usually developed long before production. Companies do not make major changes after launch, except in exceptional circumstances, because there is no way to ensure that the updated firmware will reach the products it needs to reach. When that happens, that's the big news. Remember when AMD had a TLB bug in Phenom? Secondly, AMD's use of the Adaptive Frequency and Voltage Scale is specifically designed to adjust the internal voltage of the processor to ensure that clock targets are met, thereby limiting the impact of variability and keeping the processor at the perfect place for the clock.

I'm not saying that AMD would never make any adjustments to AGESA that impact synchronization. But the idea that the company discovered a critical reliability problem that required it to make a subtle change that reduced the clock by a few MHz in order to protect long-term reliability does not immediately match my understanding of processor design, binned and tested. We contacted AMD for additional information.

I am always confident and comfortable recommending the Ryzen 3000 family because I spent a lot of time with these chips and I could see how fast they are. However, AMD's "up to" start clocks are also slimmer than we originally knew. This does not change our expectations of the overall performance of the piece, but the company seems to have decided to interpret this cycle "up to" differently from previous product launches. This change should have been communicated. In the future, we will take a closer look at the behavior of Intel and AMD clocks as part of our exam cover.

Now read: