AMD’s Ryzen 3000 CPUs struggle to hit boost speeds

If you’re going to drop $500 or more on a processor, you want one that does what it promises on the box. With the Ryzen 3 3900X CPUs, though, that may not be the case. According to well-known overclocker Der8auer, though, AMD’s Ryzen 3000 series is having difficulty hitting the advertised boost clock speeds.

De8auer is best known in the PC community for delidding and overclocking CPUs, so he’s certainly going to be one to spot a processor acting out of character compared to its promised speeds. After noticing the discrepancy, Der8auer conducted a survey among Ryzen 3000 users and found that only 5.6% of respondents are seeing their Ryzen 9 3900X CPUs hitting rated boost speeds. Other AMD 3000-series CPUs are doing better job of hitting those boost speeds but are still struggling.

Tom’s Hardware confirmed through AMD that the only one core on a CPU is guaranteed to hit a a rated boost clock, but Der8auer’s results show that many users aren’t hitting it on any core, let alone a single one.

The survey includes performance data from 2,700 systems with AMD Ryzen 3000-series chips running Cinebench R15 and HWInfo, an AMD-recommended utility. Many got within 25MHz of the advertised boost speed, but most didn’t make it there.

A flawed survey

AMD Ryzen 9 3000-series

There are a few issues with this survey. Der8auer is an expert when it comes to CPU performance, so we’re definitely not questioning his interpretation of his data. However, Der8auer is also going to attract people who pay attention to the particulars of processor operation to his audience. This self-reported survey highlights a specific subset of computer users; this already biases the result.

Further⁠—and Der8auer acknowledges this⁠—users who have reason to be suspicious of their speeds and users who find that they’re getting sub-par speeds are more likely to respond to a survey like this, while other users might dismiss it.

5.6% is a dismal return, though, and it would be a wild skew to see the numbers differ significantly in a more scientific survey.

AMD hasn’t commented on the findings yet. The issue shows up on a variety of boards, too. Der8auer still recommends the 3000-series CPUs. He explains that AMD just shouldn’t advertise speeds that users can’t reasonably achieve.

If you’re in the market for a new AMD 3000-series Ryzen CPU, it’s something to keep an eye on.

avatar
5 Comment threads
17 Thread replies
0 Followers
 
Most reacted comment
Hottest comment thread
11 Comment authors
KrogothK-L-WasterSpunjjiWacoderFunkenstein Recent comment authors
  Subscribe  
newest oldest most voted
Notify of
derFunkenstein
Guest
derFunkenstein
Waco
Guest
Waco

Excellent! My 3900X under water hasn’t hit the peaks advertised under any workload ever since I updated the BIOS on my board. It’s still super stable and fast but I’m glad AMD is taking ownership of the problem (even if it’s a perceived problem).

joe9000
Guest
joe9000

It’s simple, the stock coolers don’t cut it. AMD needs to start offering CPUs without a cooler so we aren’t paying for something that goes straight in the trash bin. Those that buy the stock coolers just have to live with the results.

Krogoth
Guest
Krogoth

They are more than sufficient. The base clock is what matters. Boost clock is nice, but it is not guaranteed. Every company that has implemented turbo-clocking or similar has already put a disclaimer on it.

People just got too spoiled when AMD/Intel were too conservative with their binning now they are pushing their SKUs to their bleeding edge.

It would be much more alarming if silicon was having issues reaching base clock speed under stock cooling and kept throttling under a heavy load. It would be embarrassing if they were outright unstable (*cough* Pentium III 1.13Ghz *cough*)

Krogoth
Guest
Krogoth

I don’t the understand the controversy over this. Boost/Turbo clockspeeds are not guaranteed nor have they ever been. They are always have been opportunistic based on thermal and voltage settings (within stock specs). The same thing has been happening on the Intel front for years and it was mostly noticeable on their laptop and server SKUs but the intended customers didn’t make a fuss over it. The difference from yesteryear and current CPUs is that Intel and AMD didn’t push their silicon to their bleeding edge with stock voltage and settings. It is no surprise that majority of the current… Read more »

K-L-Waster
Guest
K-L-Waster

Putting the boost clock in big font then burying a “not all systems can hit this” disclaimer in the fine print isn’t exactly customer friendly.

If Intel or Nvidia tried this most of you would be calling for CEO jail time…

Krogoth
Guest
Krogoth

They have already done it. This is nothing more than outrage from a tiny, but vocal minority. It is being further fueled by fanboys and cheerleaders of their favorite brand(s).

It would much more alarming if it was happening with silicon falling to reach base clocks or suffer from general instability.

chuckula
Guest
chuckula

Cinebench exists, these arguments are all invalid*

* But if Intel pulled a stunt half as bad as this we’d be screaming about how it’s the darkest day in human history and Intel should be sued into oblivion for lying. But hey, it’s AMD so there’s always a free-pass involved.

derFunkenstein
Guest
derFunkenstein

Not to mention that the survey represents a self-selected sample and has tons of variables in play. But yeah, Cinebench is the reason to get upset.

chuckula
Guest
chuckula

No, Cinebench is the excuse as to why having close to 95% of the chips not reach the advertised turbo at stock clocks is perfectly fine and NBD. As for the doubt about this being true, Der8auer is late to the party on this one (and probably intentionally so) because the failure of these chips to hit full turbo has been pretty well-known since launch. Der8auer is late because he looked at a large population of chips to come to this conclusion, and that takes time. Go look at https://siliconlottery.com/collections/matisse where you see what they are selling (when they get… Read more »

derFunkenstein
Guest
derFunkenstein

This just in: Silicon Lottery charges crazy prices for CPUs that overclock better. I don’t think SL’s pricing means anything. How about this i9-9900K top overclocker for a near-90% markup? Does that indicate an issue?

https://siliconlottery.com/collections/all/products/9900k51g?variant=15392436093014

Who’s talking about Ice Lake besides you? I just don’t think anything is well known at this point because it hasn’t been investigated properly. There’s zero control (scientifically) in this report, including whether the report is legitimate. The maximum frequency has never been advertised as an all-core, or even mulit-core speed, either. Your target keeps moving.

derFunkenstein
Guest
derFunkenstein

Honestly, this is just outrage culture at its finest. Oh, my new CPU (maybe, but it’s completely unverifiable because software tools poll speeds too infrequently) up to three percent short of the advertised max, SOMEBODY CALL THE POLICE!!!!

crabjokeman
Guest
crabjokeman

I feel like this is AMD’s way of trolling Intel because they’re confident of forthcoming headroom in GloFo’s 7nm process. Advertise HUGE GHz and ignore IPC. Pentium 4, anyone?

DPete27
Guest
DPete27

For the record Der8auer was talking about this issue within the first week of launch if you watch his YouTube videos.
https://www.youtube.com/watch?v=WXbCdGENp5I
and
https://www.youtube.com/watch?v=tlUE8GlkbGA

Spunjji
Guest
Spunjji

A “shady” alteration that reduces operating voltage significantly in exchange for a statistically irrelevant shift in performance? Perish the thought.

I wasn’t aware that we had any representative testing of Ice Lake to go on either. Hitting 4.1Ghz with fans set at 100% is… well, I’m decidedly whelmed, and interested to see what the performance / power equation works out to vs. 14nm++.

Spunjji
Guest
Spunjji

Who’s we? Do you have a mouse in your pocket?

In seriousness, though, this issue appears to be getting roughly the level of coverage it deserves – just as issues with Intel’s notebook chips failing to hit rated speeds has done. I’m not really sure what would satisfy you.

derFunkenstein
Guest
derFunkenstein

So he’s just taking peoples’ words for it that they know how to test and how to observe clock speeds? Or that they even have the CPU that he’s looking for? YouTube as a platform is entirely too engrossed in generating outrage, so I just can’t take anything I find there seriously. Even the thumbnail and title for this video were designed to generate outrage: IT’S WORSE THAN I THOUGHT! SOMEBODY FIRE UP THE CAMERA! I don’t have a Ryzen 3000 CPU. In fact, Zen 2 having this kind of issue would probably make me feel a little better about… Read more »

willmore
Guest
willmore

Yeah, 3700X here’s hitting full speed on multiple cores running Prime95–until it gets to temp. Might just need to use a cooler other than the stock if I plan on running a workload like that for long–I am not.

Justin A
Guest
Justin A

One problem is that being involved in a business that sells binned chips gives him the image of having a conflict of interest. Odd how no one has mentioned that.

Justin A
Guest
Justin A

Clearly, the best solution is to have a fully-independent reviewer (one that receives no free samples and early-access hardware from AMD and Intel) to buy a bunch of each of the 3000 series chips and do some testing. One thing I would do if I were testing this is to cut back HWInfo so that it’s measuring only one thing, instead of the plethora of things it measures by default. All of that processing itself could impair the CPU’s ability to fully boost. There is also the problem of Windows 10 doing background nonsense so it would be better to… Read more »

derFunkenstein
Guest
derFunkenstein

While you’re correct it’s the only way to perform a scientific test, the investment in writing such an article is through the roof. It’ll never happen. “Hey go buy (I dunno) half a dozen top-end CPUs…”

Spunjji
Guest
Spunjji

Even a dozen wouldn’t really be a scientifically valid sample size.

One thing I would say here – while cutting out the background processes reduces the variables and makes the test more scientific, it also smells a lot like stacking the deck. If the processors can’t hit those speeds in normal use under Windows 10 then that’s A Thing; whether it’s A Thing Worth Caring About is definitely debatable though.

Pin It on Pinterest

Share This