News AMD Announces BIOS Fix for Ryzen 3000 Boost Frequencies, Lands September 10th

Giroro

Splendid
"Our own investigation of AMD's new boost clock behavior also found that only one core on any given Ryzen 3000 CPU can hit the rated boost clock "

Except that's not what the data presented in that article shows. It's clear in that test that even the "best" core was not able to reach the advertised boost speed, at least not for a period of time that is observable.

So the question remains: has any review ever been able to prove that any core in any Ryzen 3000 CPU will reach advertised frequency for any amount of time?
 
Last edited:
  • Like
Reactions: dalef
"Our own investigation of AMD's new boost clock behavior also found that only one core on any given Ryzen 3000 CPU can hit the rated boost clock "

Except that's not what the data presented in that article shows. It's clear in that test that even the "best" core was not able to reach the advertised boost speed, at least not for a period of time that is observable.

So the question remains: has any review ever been able to prove that any core in any Ryzen 3000 CPU will reach advertised frequency on any core for any amount of time?

Steve from hardwareunboxed tested the same CPU over a bunch of x570 motherboards to see if the boost clock could be achieved... the funny thing is, MSI GODLIKE X570, the board that Toms used, was not able to achieve the Boost clock... but some others did.

Basically, Toms screwed up.... just accept it! (Just buy it!)
 
Last edited:
  • Like
Reactions: st379

st379

Distinguished
Aug 24, 2013
169
69
18,660
Steve from hardwareunboxed tested the same CPU over a bunch of x570 motherboards to see if the boost clock could be achieved... the funny thing is, MSI GODLIKE X570, the board that Toms used, was not able to achieve the Boost clock... but some others did.

Basically, Toms screwed up.... just accept it! (Just buy it!)

I want to believe they did not do it on purpose but after the "just buy it" article and the fact that they tested the athlon 200ge with gtx 1080 and did not test the igpu i am not sure it wasn't on purpose.
 

PaulAlcorn

Managing Editor: News and Emerging Technology
Editor
Feb 24, 2015
858
315
19,360
Steve from hardwareunboxed tested the same CPU over a bunch of x570 motherboards to see if the boost clock could be achieved... the funny thing is, MSI GODLIKE X570, the board that Toms used, was not able to achieve the Boost clock... but some others did.

Basically, Toms screwed up.... just accept it! (Just buy it!)

There is no screw up. Steve tested with one CPU. Silicon quality varies, and our sample was from a retail outlet, not one provided by AMD. Other Ryzen 3000 series processors that we have in-house have hit advertised boost clocks, or within 25mhz, with this very board.

Also, the central point of the article was NOT if the chip can hit boost clocks. It was to characterize boost behavior, particularly that there are slower cores that are simply unable to hit the boost speed, and that the scheduler targets cores.

You seem to be willingly refusing to accept what AMD confirmed directly to us: Not all cores can hit the boost clock, and the scheduler targets the faster cores first. Do you think AMD would admit to that if "Tom's messed up?"
 
Last edited:

TJ Hooker

Titan
Ambassador
Steve from hardwareunboxed tested the same CPU over a bunch of x570 motherboards to see if the boost clock could be achieved... the funny thing is, MSI GODLIKE X570, the board that Toms used, was not able to achieve the Boost clock... but some others did.

Basically, Toms screwed up.... just accept it! (Just buy it!)
According to the HWUB video, the X570 Godlike with a 3800X is in fact capable of hitting max advertised boost clocks. Assuming this is the video you're referring to:
View: https://www.youtube.com/watch?v=o2SzF3IiMaE


And even if the Godlike wasn't capable (and there's nothing about the board that would suggest that prior to testing), which boards can and cannot reach max boost seems semi random (in addition to varying by BIOS rev), so it wouldn't have been a matter of Toms screwing up so much as it would have been them simply being unlucky.

Edit: And of course the focus of that article was not on the max single core boost of the CPU, but rather the boost of the various individual cores. It remains to be seen how, if at all, the conclusions of that article will be affected by the new AGESA.
 
Last edited:

PaulAlcorn

Managing Editor: News and Emerging Technology
Editor
Feb 24, 2015
858
315
19,360
According to the HWUB video, the X570 Godlike with a 3800X is in fact capable of hitting max advertised boost clocks. Assuming this is the video you're referring to:
View: https://www.youtube.com/watch?v=o2SzF3IiMaE


And even if the Godlike wasn't capable (and there's nothing about the board that would suggest that prior to testing), which boards can and cannot reach max boost seems semi random (in addition to varying by BIOS rev), so it wouldn't have been a matter of Toms screwing up so much as it would have been them simply being unlucky.

It's also noteworthy that the article wasn't focused on whether or not the CPU can hit the boost clocks. It was centered on how many cores CAN boost to the clock, and how the scheduler is targeting the faster cores.

Regardless of the board, one fact remains, which we proved and AMD confirmed: All Ryzen processors come with a mix of fast and slower cores, and not all cores can hit the boost clock.
 
Last edited:

PaulAlcorn

Managing Editor: News and Emerging Technology
Editor
Feb 24, 2015
858
315
19,360
I want to believe they did not do it on purpose but after the "just buy it" article and the fact that they tested the athlon 200ge with gtx 1080 and did not test the igpu i am not sure it wasn't on purpose.

There is no "on purpose" to it. The article is factual, and the results therin are verified by AMD.
 
Do you think as the process matures, this might change with more cores hitting the boost clock?

I highly suspect the better chiplets are going into EPYC's right now so that is likely part of the picture. I'm sure TSMC can improve the yields quickly. Although I suspect we are going to be looking at 7nm+ next year for the really great yields to show up where we might have an all core 4.7Ghz 8 core CPU. Looking at silicon lottery stats with only 6% of the 3900x's hitting 4.2Ghz on all cores it's obvious we are dealing with a process in its infancy.
 
Honestly, this was (at least from my POV) to be expected. to some degree or so history would have one believe. This is a brand new process and on a smaller die. This isn't like Intel's 14 ++++++++ ad nauseum. It's pretty much new. It's also within the realm of margin for error, is it not? I say this as an AMD fan and FX 8350 owner who can't wait to move up to this iteration regardless of the current situation. Once things mature somewhat, I'm sure this will all be looked upon as a small hiccup rather than a lingering case of them.

It could have been worse. It can always be worse.
 
I got the advertised speed for my Ryzen 5 3600, and Im really happy with it. I know this is the lowest part on the line for the moment, and probably is the easiest one to reach advertised speed.

I know and understand people that spend way more money than me on higher tier Ryzrn 3xxx are mad, but just saying that any Ryzen 3xxx can't reach its advertising speed is wrong. Its was also wrong from AMD to advertise something they knew they couldn't deliver.

As for Roman's survey (der8auer), even though he explained what everyone has to do and how to do it, theres still the fact that a big portion of those who participated (I did) may or may not know how to properly setup BIOS, install Windows, and drivers, maybe some of them could even have background services eating part of thier CPU power and have no clue about it. So you have to take it with a grain of salt. Of course by the amount of participants for each CPU, the results don't look good, but that doesn't mean you can forget what I just wrote on the previous sentence.

Lets hope the new update will bring everyone closer to what they paid for.

Cheers
 
Actually, I would expect that to be the case. The yield learning curve is typically fast, and TSMC is one of the best in the business.

I want to see how it goes. Considering TSMCs normal products this is a bit of a newer one for them and I doubt they designed their 7nm for x86 based designs.

I think thats what is keeping Ryzens clock speeds lower than Intels. Even on 14nm they were not able to match Intels top speeds. But maybe as it sells and matures AMD will be able to get closer to Intel in top clock speeds.

What we need now though is for Intel to come out with a good competitive core matching product to get AMD to work even harder. Get us a nice price war going.
 
Anyone else considering the fact that the results Der8aeur got might be fairly skewed? To begin with the sample size is small, made up of people who either follow Derbaeur or noticed the problem and went looking for confirmation that it was a problem. Those who didn't have a problem never went looking and overwhelmingly would not have been the ones posting in the survey. I'm not denying that it is a thing, and a thing that REALLY needs to be fixed at that... it just might not be nearly the size of a thing that people are pointing out. Yes, these numbers are alarming, but if you consider the size of the community added to the people having the problem and as such paying attention to it, the problem could be much smaller than what is being reported. Just a thought I had while listening to people's heads explode over this all over the internet.
 
Anyone else considering the fact that the results Der8aeur got might be fairly skewed? To begin with the sample size is small, made up of people who either follow Derbaeur or noticed the problem and went looking for confirmation that it was a problem. Those who didn't have a problem never went looking and overwhelmingly would not have been the ones posting in the survey. I'm not denying that it is a thing, and a thing that REALLY needs to be fixed at that... it just might not be nearly the size of a thing that people are pointing out. Yes, these numbers are alarming, but if you consider the size of the community added to the people having the problem and as such paying attention to it, the problem could be much smaller than what is being reported. Just a thought I had while listening to people's heads explode over this all over the internet.

Sure. Minus the fact that AMD has confirmed it and are working on a BIOS fix for it. If AMD didn't confirm it and were not working on a fix for it then it might just be so puff piece but they did and they are so its obviously an issue that affects more than just the ones who partook in his survey.

And I don't think he is one to be biased. He is just an overclocker and will use the best chip for the job. I would think people would be rather glad this issue came to light and that AMD is working to fix it now instead of it not being reported by anyone major and AMD just going with the flow.