AMD Announces BIOS Fix for Ryzen 3000 Boost Clocks - hits September 10th

Status
Not open for further replies.
Sabrewulf...

Sabrewulf...

Gawd
Joined
Jan 18, 2008
Messages
1,019
Source

upload_2019-9-3_12-22-25.png


https://www.tomshardware.com/news/amd-ryzen-3000-boost-frequency-bios-fix,40308.html

"The statement comes amidst a growing chorus of complaints from Ryzen 3000 owners on forums, reddit, and social media that their chips aren't reaching the advertised boost clocks. In response to the growing number of complaints, YouTuber Der8aeur recently conducted a survey that garnered 2,700 respondents, of which only 5.6% were able to reach the advertised boost clocks for AMD's flagship Ryzen 9 3900X processor.

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, which AMD confirmed. That means the Ryzen 3000-series processors contain a mix of faster and slower cores, which is a departure from how AMD and its competitors have traditionally spec'd processors. Unfortunately, users must have the latest version of Windows 10 to use the Ryzen-aware scheduler, which targets the fastest cores with lightly-threaded applications, further complicating matters for frustrated customers trying to attain the advertised boost frequencies.

In either case, even with the presence of the necessary BIOS, driver, and Windows 10 scheduler, most customers have been unable to attain Ryzen 3000's advertised speeds with any of the models"

Hopefully it helps!
 
Stuff like this makes me glad that my 3900X Hasn't even shipped yet..... Too busy ATM to troubleshoot this kinda stuff so hopefully some kinks are worked out by the time it gets to me.
 
odditory said:
They'll shift *reported* clockspeed upward slightly, but benchmarks won't move with it.
Click to expand...
Nobody is that stupid. :D If that route is taken, there *will* be a slight boost in clocks, and a larger boost in reported clocks (and/or more hysteresis). That way, there is a performance gain that can be discerned, away from the noise.
 
jeremyshaw said:
Nobody is that stupid. :D If that route is taken, there *will* be a slight boost in clocks, and a larger boost in reported clocks (and/or more hysteresis). That way, there is a performance gain that can be discerned, away from the noise.
Click to expand...

Actually, historically, when previous boost clocks were raised for some CPU's, the benchmark scores either didn't improve or went down. The frequencies going up isn't necessarily a guarantee that the actual performance will improve. If we were talking about a few hundred MHz, then naturally we would see consistent improvements in performance. However, when we are talking about clock speeds under 100MHz, I wouldn't bet on it. What really matters is how long they hold the boost clocks. That too is often a problem so unless the new BIOS addresses this, I wouldn't bet on any benchmark scores improving substantially.
 
Good. I hope they fix the issue as I'm really sick of hearing about how everyone isn't seeing max boost speeds... I get 4375mhz and my 3700x should be hitting 4400. When is the class action going to be filed?
 
elite.mafia said:
Good. I hope they fix the issue as I'm really sick of hearing about how everyone isn't seeing max boost speeds... I get 4375mhz and my 3700x should be hitting 4400. When is the class action going to be filed?
Click to expand...

I would consider 4375 good enough too, but there are people who do not get beyond 4200 or so. Something is clearly wrong with those.
 
Dan_D said:
Actually, historically, when previous boost clocks were raised for some CPU's, the benchmark scores either didn't improve or went down. The frequencies going up isn't necessarily a guarantee that the actual performance will improve. If we were talking about a few hundred MHz, then naturally we would see consistent improvements in performance. However, when we are talking about clock speeds under 100MHz, I wouldn't bet on it. What really matters is how long they hold the boost clocks. That too is often a problem so unless the new BIOS addresses this, I wouldn't bet on any benchmark scores improving substantially.
Click to expand...

What about if I have an all core OC of 4100 on my R5 2600? Is that reallly all that much better than 4000 on the same processor with gaming?
 
Last edited:
MaZa said:
I would consider 4375 good enough too, but there are people who do not get beyond 4200 or so. Something is clearly wrong with those.
Click to expand...

aMD has said that VRM setups are important, and looking back I wish they would have clarified things to the point that even a simpleton could understand.

If you are using a lower tier board (b350/some x370 setups) you get 3.6Ghz all core all the time. You have the possibility of higher boosts but they are not guaranteed due to VRM design differences.

If you have a mid to upper tier board from the 4xx series, you get a guaranteed 3.6Ghz boost and the CPU will spend most of it's time clocked much higher.

Finally, if you are buying a x570 based board, you are guaranteed for the CPU to spend it's time between 80-100% of the rated boost clocks.

All these things are assuming the case has adequate air flow and the CPU is cooled by the wraith prism or a higher performing HSF/AIO which is installed correctly.

Now most of us that have a brain realize these things should be common sense but if AMD had assumed the worst and stated all this then this would be a big nothingburger.
 
ccityinstaller said:
aMD has said that VRM setups are important, and looking back I wish they would have clarified things to the point that even a simpleton could understand.

If you are using a lower tier board (b350/some x370 setups) you get 3.6Ghz all core all the time. You have the possibility of higher boosts but they are not guaranteed due to VRM design differences.

If you have a mid to upper tier board from the 4xx series, you get a guaranteed 3.6Ghz boost and the CPU will spend most of it's time clocked much higher.

Finally, if you are buying a x570 based board, you are guaranteed for the CPU to spend it's time between 80-100% of the rated boost clocks.

All these things are assuming the case has adequate air flow and the CPU is cooled by the wraith prism or a higher performing HSF/AIO which is installed correctly.

Now most of us that have a brain realize these things should be common sense but if AMD had assumed the worst and stated all this then this would be a big nothingburger.
Click to expand...

Adding some more, you would think that the fact that AMD has closed the gap to defeat Intel CPUs, despite a fairly large clock speed deficit, would have been received better.

If AmD would have just come out and said that the last combination of AGESA and chipset drivers has improved performance despite the fact that users may notice a lower boost clocks in single core/lightly threaded applications we could all move on. They can say "We gave our users a free performance boost" and leave it at that.
 
I just hope they also include something to get better idle voltage down clock, my 3900 wont idle less than 1.4 volts no matter what I do. one or two cores running the rest sleeping, Ryzen master still shows 1.4 volts or more.
 
I would think this will end up being more about the fact that pbo doesn't appear to work where as pb seems to be working fine on 3800x and 3900x cpus.

But we'll see.

I only hope that the next bios update also corrects a bunch of issues with improper acpi implementations. An additional hundred MHz isn't going to make any difference at all to performance. But bad acpi tables cause issues across all kinds of power management subsystems and it makes life annoying for OS's that make the silly presumption that manufacturers will implement specs correctly.
 
jimbo4822 said:
I just hope they also include something to get better idle voltage down clock, my 3900 wont idle less than 1.4 volts no matter what I do. one or two cores running the rest sleeping, Ryzen master still shows 1.4 volts or more.
Click to expand...

Honestly don't be surprised if it uses more voltage for low core count usage.
 
purple_monster said:
this is why you always buy intel/nvidia. this kind of stuff never happens.
Click to expand...
Lol, do people really beleive this? No, Intel never had a snafu or lied in marketing... And nvidia either.. never been caught cheating on benchmarks (like ATI was)... Sure, you're right, stick with the herd and ignore reality it it do any fit your mold. They all do it, AMD, Intel, Nvidia, Apple, just about every tech company. Some get caught out more than, some do it more often, etc, but don't kid yourself.


That being said, hopefully they can get a fix in that helps a little bit but I doubt you'll see much in the real world anyways.
 
jimbo4822 said:
I just hope they also include something to get better idle voltage down clock, my 3900 wont idle less than 1.4 volts no matter what I do. one or two cores running the rest sleeping, Ryzen master still shows 1.4 volts or more.
Click to expand...

it's not actually idling at 1.4v, the problem is that with the monitoring apps, it never lets the cores park even though windows tries to park them anyways so what ends up happening is that the app reports the highest voltage every time which happens to be 1.4v every time a core comes out of idle state into max boost state. the processor actually idles at ~.991v.
 
jimbo4822 said:
I just hope they also include something to get better idle voltage down clock, my 3900 wont idle less than 1.4 volts no matter what I do. one or two cores running the rest sleeping, Ryzen master still shows 1.4 volts or more.
Click to expand...
Yeah, I've noticed that with my 3600 is always at 1.4. Cpu z reports the boost running at 4.1-4.3 Ghz though:)
 
Tittle of the topic should be changed.. as I read "we will provide an update on September 10 to the community regarding the availability of the BIOS" I can understand as they will just make an announcement that day to say when the real date of the BIOS will arrive to the end user. So no hopes of that anytime soon, we already know how the waiting game works for AMD.
 
Araxie said:
Tittle of the topic should be changed.. as I read "we will provide an update on September 10 to the community regarding the availability of the BIOS" I can understand as they will just make an announcement that day to say when the real date of the BIOS will arrive to the end user. So no hopes of that anytime soon, we already know how the waiting game works for AMD.
Click to expand...

We have a timeframe for when you can expect a timeframe.
 
I have had my 3600 report between .7436 .7323 /48 to the "normalized" it seems with current Agesa etc (far more stable I might add) .82-.945 to ~1.023 - 1.015 type ranges ...

so .. likely AMD is really not all that far off seeing as THEY made the damn things, including x570 and AGESA is also made by them ofc... they stated for the voltages .. the LOWEST volt you will ever see is what the individual core lowest volt can be so, if a core magically able to reduce volt to X below all the others, it deep suspend/sleep at this rate .. i.e could be .94v, could be 1.24v, could even be .72v thereabouts.

I am certain quality of the silicon for the cpu is the most major factor, cooling likely is next, motherboard likely also has a very real place in there (many reasons) power clean/stable/enough in reserce . not enough knowledge other than, very likely ^.^
 
ccityinstaller said:
Adding some more, you would think that the fact that AMD has closed the gap to defeat Intel CPUs, despite a fairly large clock speed deficit, would have been received better.

If AmD would have just come out and said that the last combination of AGESA and chipset drivers has improved performance despite the fact that users may notice a lower boost clocks in single core/lightly threaded applications we could all move on. They can say "We gave our users a free performance boost" and leave it at that.
Click to expand...

The Zen2 release was received with universal acclaim and recomendations across the board. This is a seperate issue that is really becoming more about fanaticism and delusion.
 
Lakados said:
Stuff like this makes me glad that my 3900X Hasn't even shipped yet..... Too busy ATM to troubleshoot this kinda stuff so hopefully some kinks are worked out by the time it gets to me.
Click to expand...
Of course dick move from AMD on undershooting the clocks. But it's not something that is noticeable in real world application. So if you just want to use your cpu it's fine as it is. I don't even care as I run all clock OC anyway. The voltage AMD pushes to the cpu for boost is kind of scary as it is. I'd not be surprised if stock 3000s started to die while the overclocked ones will be fine.
 
M76 said:
Of course dick move from AMD on undershooting the clocks. But it's not something that is noticeable in real world application. So if you just want to use your cpu it's fine as it is. I don't even care as I run all clock OC anyway. The voltage AMD pushes to the cpu for boost is kind of scary as it is. I'd not be surprised if stock 3000s started to die while the overclocked ones will be fine.
Click to expand...

I really doubt that the high voltages are an issue. They designed them to run like that. It's a new 7nm node and maybe it just needs more voltage.
 
Amd knows what is safe to run on the CPUs more than we do. They know what the voltages actually are considering they change by the microsecond and all the monitoring methods available to consumers are not able to sample the sensors that fast to get a true picture of what is going on.

I think there is probably a glitch in how precision boost overdrive adjusts frequency on the 105w tdp chips....

There is also a very real acpi issue with some mb manufacturers, which would also impact power management functionality, which includes pstate management that could be impacting behavior of chips on those boards with hoopty tables.
 
Status
Not open for further replies.
Back
Top