News AMD's Ryzen 3000 Boost Fix Works, But Hits Wrong Cores

[Admin]

AMD's BIOS fix boosts our Ryzen 9 3900X up to 4.65 GHz, but often on the wrong cores.

AMD's Ryzen 3000 Boost Fix Works, But Hits Wrong Cores : Read more

 

[redgarl]

With all the stupidity you guys posted in the last couple of weeks, I would take a break... no but seriously, it is almost pathetic. You have been used by Intel to fight their war without having them to lift a finger. And now you just keep going...

So how much could this impact my cinebench multi-thread score? 0.05%? Also, you were the one saying that missing 25MHz of boost was a problem and now you prefer the old BIOSes because the average boosting time was a little bit longer...

No but seriously? WTH? You need to create issues to feel relevant?

Toms: It works.... BUT!!!... (insert something to throw dirt at AMD)

 

[Arbie]

Why do you continually slant your headlines to be as anti-AMD as possible? Besides being click-bait? We just got through with this:

https://forums.tomshardware.com/thr...eads-to-false-ryzen-benchmark-scores.3520916/

The present article should have been entitled "AMD's Ryzen Boost Fix Works (but Win Scheduler still weak)". Instead, you give the impression that the AMD fix is actually worthless.

It's a shame that you do all this careful technical work only to have some biased or ignorant editor slap on misleading headlines. You come off looking like Intel tools.

 

[Iczeman]

Another perfect analyze and review so big thanks for it.
Pity that is not mentioned if WHEA errors are finally fixed with ABBA Agesa on all X570 mainboards and when updates will be released to x470 and x370. Still waiting and getting impatient as from AMD still silence, oh I'm not impatient but more fed up by waiting as this is critical issue and improved boost is just nice to have as expected out of the box already.

 

[st379]

Where is the "this article is sponsored by intel"?
Latest and greatest in "favour" of Amd:

1. Just buy it nvidia
2. 95% of Amd cpu are defective.
3. Athlon 200ge paired with 10,0000$ gpu and no review on the igpu.
4. Amd cpu will die on you tomorrow (this article Intel wrote and of course published here)
5. You can't really benchmark Amd cpu

But......

Thank you for exposing the weak cores on ryzen that was the only decent article followed by an onslaught on Amd.

I like you exposed the weak cores the other articles are utter....

 

[logainofhades]

For my usage, the bios update was useless. I didn't see any performance increases. Instead of all my cores sitting at 4.275ghz, they are all bouncing all over the place, between 3.6 and 4.4ghz. You fanboys can cry foul, all you want, but until Microsoft gets their act together, I don't expect to see any performance gains. AMD has stated that their chips are a mix of faster and slower cores. If windows is not properly prioritizing those fast cores, you are not going to get the kind of performance, your CPU is capable. of.

 

[joeblowsmynose]

" ... but the increased peak frequencies often don't equate to noticeable differences in performance "

Re: Expectations around performance increases
Let's be serious for a moment ... how much increase in performance were we really expecting to see in a 50mhz improvement on ONE core - assuming that you can even keep windows/the chip focusing all work on a single core?

50-100mhz of one core on a 3900x is (Edit: I had the math wrong forgot to account for 12 cores, lol) 0.125-0.25% of its processing power, just to throw that out there. (rough napkin math)

Steve from GN did a good video on this bios update, and published gaming comparisons and there actually was an increase in FPS in most games, although not a lot - as expected.

You could easily OC your RAM, tighten up timings and increase gaming performance by up to 10-20% (Steve at GN has a great video on this also), but let's instead of looking at things like that, focus all our awareness on this topic, and create a false expectation that performance should improve by 'X' amount with this fix, then, be disappointed with AMD that this hasn't increased performance by "enough". Then maybe Intel will hire us for marketing.

It was a minor issue (but a valid one) that pretty much didn't impact performance to begin with.

 

[Arbie]

You fanboys can cry foul, all you want, but until Microsoft gets their act together ... If windows is not properly prioritizing those fast cores...

So calling foul because AMD is being blamed for what is primarily a Windows weakness makes people fanboys??

 

[logainofhades]

You all scream bias anytime anything negative is brought up, that involves AMD. If there was such an anti AMD stance here, how come they dominate the Tom's best CPU lists? I fully remember the days when you only saw AMD at that low end price point, now they are top in nearly every category.

https://www.tomshardware.com/reviews/best-performance-cpus,5683.html

https://www.tomshardware.com/reviews/best-cpus,3986.html

 

[kinggremlin]

So calling foul because AMD is being blamed for what is primarily a Windows weakness makes people fanboys??

Intel's development of Turbo Boost Max:

-Every core tested, cores' ranking from fastest to slowest stored on CPU
-BIOS support development
-Write custom driver themselves for Windows to funnel single (Broadwell)/ dual (Skylake+) core loads to fastest cores
-Work with Microsoft to get driver built into Windows 10
-Write patch themselves for Linux support, because no OS scheduler supports this feature natively

AMD's Chinese knockoff version of Turbo Boost Max development:

-screw up so bad that almost no CPU's hit advertised boost clocks even when using the program AMD recommends for testing this feature
-release "fix" that boosts random core which may or may not be the fastest core, then prays that Windows will coincidentally send work to that core.
-sit back and wait for AMD fanboys to blame everyone except AMD for this fiasco, especially Microsoft, even though this isn't a Microsoft OS exclusive limitation.
-mission accomplished

Yes, this is AMD's fault. Watching the AMD faithful defending AMD is like watching religious people defend their God. Nothing is ever AMD's fault, they can do no wrong. And if they are caught time and time again over hyping their product? Ah well.... AMD works in mysterious ways...

 

[DMAN999]

LOL,
The comments here are at least very entertaining.
I always get a good chuckle over all the "Fanboy" name calling.
I personally am a Fan of whatever CPU gives me the best performance for the price and meets my needs at the time.
I don't care who made it.
I even built a Cyrix based PC once (486DX iirc).

I currently have a Ryzen 7 3700x and I am very happy with it, BUT I have owned way more Intel CPU's over the last 30 or so years.
In fact this is only my second AMD CPU ever, the first being the Ryzen 5 2600 I initially bought for this build last November.

And just FYI,
My Ryzen 7 3700x does hit 4.4 GHz+ single core even without the newest AGESA (1.0.0.3ABBA) and with a few BIOS settings changed.
Look here :

Discussion - Ryzen 3700x Max Single Core Clock Speed ?

What is the Maximum single core clock speed your Ryzen 7 3700x hits ? I was just wondering what everyone is getting for a max clock speed because of all the recent press about these chips not hitting the Max clock speeds (4.4 GHz ). Mine hit 4375 MHz (4.375 GHz) on one core while running...

forums.tomshardware.com

 

[justin.m.beauvais]

Well, I'm glad to see we are all being reasonable here.

Yup, AMD screwed up. They sort of over promised. I mean, the CPUs do have cores capable of hitting the advertised speed, they might even do it in real world tasks from time to time. That said, it isn't like any CPU with a turbo mode will sit at that frequency all day long. CPU work naturally comes in bursts. Heck, I've never seen my 7700HQ even hit its full turbo speed of 3.4GHz despite sufficient power and cooling and my 4590 RARELY hits its max of 3.7GHz. But that can't have anything to do with this. Probably not related in the slightest.

In the real world aggregate performance is the most important thing, and AMD's CPUs don't seem to be doing too bad in that regard. I mean unless they are so broken they aren't running the system clock correctly. It is just shocking the level that people are analyzing the Ryzen 3000 series compared to the two previous generations. It is as if AMD suddenly releasing a better CPU has people in such disbelief that they are willing to look past actual performance and are just out to find everything they can to bring AMD down a notch... but that can't be, right? I mean, people don't just start to take shots at other people just because they have become more popular, right? That never happens. AMD must really have serious issues here.

I guess I'll just stop recommending them and I'll put off my plans to buy one. I'm sure a good solid i5 that costs more for the same performance will work just fine. At least I know it is coming from a reputable company with a huge R&D budget that can just throw money at an issue until they get it right. I mean, trying to take the process node lead from Intel? AMD must have been over extending themselves. Their chips are probably riddled with major flaws just waiting to be found. Intel never has that problem.

Disclaimer: If you can't see the sarcasm dripping from this post, I can assure you it is there.

Despite these hiccups, I think AMD has released a solid CPU. I will probably even buy one, but they are a smaller company that designed Ryzen on a shoestring budget (compared to Intel) and continue to operate with far fewer resources than their competition. Small problems will exist, but those shouldn't overshadow their solid product. Just as Specter and Meltdown didn't destroy people's faith in Intel, this won't destroy people's faith in AMD. AMD came back from the brink of death to give us CPUs that do a lot of things really well, and they were really enthusiastic about that. Heck, I would be. They just took that enthusiasm a little far, and now every little mistake they make is being tossed in their face and hyped to outrageous proportions. Seriously, the tech media has made this just about as big as Specter and Meltdown, and it isn't.

I like Intel, but there is a special place in my heart for AMD. My first computer was an AMD K6. Both companies have made mistakes, and I've seen most of them. In the grand scheme of things, and given the disastrous mistakes of the past, this problem is pretty small. Stop making a mountain out of a mole hill.

 

[mattkiss]

LOL,
The comments here are at least very entertaining.
I always get a good chuckle over all the "Fanboy" name calling.
I personally am a Fan of whatever CPU gives me the best performance for the price and meets my needs at the time.
I don't care who made it.
I even built a Cyrix based PC once (486DX iirc).

I currently have a Ryzen 7 3700x and I am very happy with it, BUT I have owned way more Intel CPU's over the last 30 or so years.
In fact this is only my second AMD CPU ever, the first being the Ryzen 5 2600 I initially bought for this build last November.

And just FYI,
My Ryzen 7 3700x does hit 4.4 GHz+ single core even without the newest AGESA (1.0.0.3ABBA) and with a few BIOS settings changed.
Look here :

Discussion - Ryzen 3700x Max Single Core Clock Speed ?

What is the Maximum single core clock speed your Ryzen 7 3700x hits ? I was just wondering what everyone is getting for a max clock speed because of all the recent press about these chips not hitting the Max clock speeds (4.4 GHz ). Mine hit 4375 MHz (4.375 GHz) on one core while running...

forums.tomshardware.com

"I hit 4.425 GHz single core today with just a few adjustments to my PBO settings." (from one of your posts on the link you shared)

Running with PBO enabled doesn't count as hitting the advertised max boost speed.

 

[st379]

Lol people calling fanboys after those articles?
https://www.tomshardware.com/news/amd-ryzen-3000-boost-clock-controversy-intel-attack,40231.html
https://forums.tomshardware.com/thr...eads-to-false-ryzen-benchmark-scores.3520916/
https://www.tomshardware.com/news/a...ing-advertised-boost-speeds-survey,40291.html

It is very obvious who payed on this articles.

AMD 3rd Gen Ryzen Boost Clock Investigation from hardwareunboxed

I did not see tomshardware publishing this article from hardwareunboxed excellent review on the boost problem while swapping a lot of motherboards and reaching different boost clocks.

After all you love taking articles from youtube and publish it on site.

Not to mention just buy it and athlon 200ge excellent review when a 55$ cpu paired with 1000$ gpu and the igpu completly ignored in order not to make the core i7 igpu looks bad.

I want to know about weak cores and scheduler problems not to read Intel sponsered articles.

 

[DMAN999]

"I hit 4.425 GHz single core today with just a few adjustments to my PBO settings." (from one of your posts on the link you shared)

Running with PBO enabled doesn't count as hitting the advertised max boost speed.

It hit 4.375 GHz on 5 different cores with nothing changed in the BIOS except enabling the XMP (DOCP) profile.
Which is close enough for me.
my 3700x absolutely meets my performance expectations, even if at stock settings it is .25 GHz below the stated Max Boost.

https://imgur.com/mja5k5V

View: https://imgur.com/mja5k5V


And by just enabling PBO it hit and even exceeded 4 Ghz.

And if you look on the AMD website it does state that 4.4 GHZMax boost may or may not be reached depending on several factors:

https://imgur.com/K2nwE2y

View: https://imgur.com/K2nwE2y


PS
Why is it only people with low forum post counts who are talking trash in this thread ???
Do you guys just wait around for something to complain about before posting ?
How about actually helping people with PC issues, instead of complaining about something that apparently doesn't even affect you since you most likely have an Intel based PC.

 

[redgarl]

Lol people calling fanboys after those articles?
https://www.tomshardware.com/news/amd-ryzen-3000-boost-clock-controversy-intel-attack,40231.html
https://forums.tomshardware.com/thr...eads-to-false-ryzen-benchmark-scores.3520916/
https://www.tomshardware.com/news/a...ing-advertised-boost-speeds-survey,40291.html

It is very obvious who payed on this articles.

AMD 3rd Gen Ryzen Boost Clock Investigation from hardwareunboxed

I did not see tomshardware publishing this article from hardwareunboxed excellent review on the boost problem while swapping a lot of motherboards and reaching different boost clocks.

After all you love taking articles from youtube and publish it on site.

Not to mention just buy it and athlon 200ge excellent review when a 55$ cpu paired with 1000$ gpu and the igpu completly ignored in order not to make the core i7 igpu looks bad.

I want to know about weak cores and scheduler problems not to read Intel sponsered articles.

I posted hardwareunboxed investigation on the day it was published... and I was called a fanboy when it was obviously a BIOS issue and pointed it out.

By the way I stated exactly the same thing as you did.

Toms, I am not the only one, hardwareunboxed, Good Old Gamers, GamerNexus, AdoredTV... called you out at many times over issues about questionable content decisions.

If I want an honest analysis, I am going to anandtech, at least Ian and Ryan are doing it professionally...

 

[redgarl]

Why do you continually slant your headlines to be as anti-AMD as possible? Besides being click-bait? We just got through with this:

https://forums.tomshardware.com/thr...eads-to-false-ryzen-benchmark-scores.3520916/

The present article should have been entitled "AMD's Ryzen Boost Fix Works (but Win Scheduler still weak)". Instead, you give the impression that the AMD fix is actually worthless.

It's a shame that you do all this careful technical work only to have some biased or ignorant editor slap on misleading headlines. You come off looking like Intel tools.

I am not agreeing with some broken analysis from toms who did not even switch the motherboard on a simple fault finding analysis, so of course I am a fanboy for pointing out the issue... (sarcasm)

I am having problems when I know better than the so called tech press posting garbage. If they want to be considered experts, they should act like experts. This was fault finding 101 right there... and they went locked and loaded without even taking that for account. They had a goal and they went for it without even doing their due diligence. That`s, my friend, is the work of an amateur.

Call me a fanboy pro-AMD like you want, but in reality I am a critic of biased conclusions when they are manipulating the public views. This is totally unacceptable from a source that claim to have the interest of their readers at heart.

The funny part is I am defending AMD in the process, but my goal is to have content I can trust. I cannot stand propaganda and unfortunately, this is propaganda.

If Toms was so concerned about their readers, we would have an analysis of the vulnerabilities over Intel CPUs not missing 25MHz of boost clock because of a new CPU and chipset release. It took 1 years for my Phenom II X3 720 to be unlocked, stable and matching a 950 in OC... so this is normal in my book since the whole lineup is not even released (3950x).

By the way, I am an Electrical Engineer, P.Eng and a PMP... and looking to do my master in AI in the next couple of years. You can downsize my comments or who I am all you want, it is still an issue that other people are also perceiving.

 

[redgarl]

It hit 4.375 GHz on 5 different cores with nothing changed in the BIOS except enabling the XMP (DOCP) profile.
Which is close enough for me.
my 3700x absolutely meets my performance expectations, even if at stock settings it is .25 GHz below the stated Max Boost.

https://imgur.com/mja5k5V

View: https://imgur.com/mja5k5V


And by just enabling PBO it hit and even exceeded 4 Ghz.

And if you look on the AMD website it does state that 4.4 GHZMax boost may or may not be reached depending on several factors:

https://imgur.com/K2nwE2y

View: https://imgur.com/K2nwE2y


PS
Why is it only people with low forum post counts who are talking trash in this thread ???
Do you guys just wait around for something to complain about before posting ?
How about actually helping people with PC issues, instead of complaining about something that apparently doesn't even affect you since you most likely have an Intel based PC.

You are pointing out what I am screaming at for the last month or so. It was only a matter for Intel fanboys not even owning an AMD platform trying to fight Intel argument war. Intel even used der8aur (or whatever is name) as an argument that AMD purposely lowered their boost clock for longevity issues. Of course that FUD propagated and toms helped in the process.

AMD users didn`t really cared because performances were there even if they were missing 25MHz, but fanboys made an argument that this 25MHz was night and day... and this idea was sold like it was here from toms to.

Anyway, I am done on this subject. It was as worst as the exclusive Lisa Su article about leaving AMD at WCCF. I slamed them as hard as I did here. It was questionable at best.

 

[cryoburner]

While the new firmware delivers smaller boosts than expected...

Your 3900X originally boosted to within 99.5% of it's rated clocks, and now you found it to exceed those rated boost clocks by 1%, a roughly 1.5% improvement. Perhaps there isn't a particularly noticeable impact on actual performance, but how much of a performance improvement where you expecting? These are all more or less margin-of-error differences, and the entire situation was arguably a non-issue to begin with. Whether a chip slightly misses or slightly exceeds its max boost clocks on a given system should have virtually no impact on the product's overall value, or one's recommendation of it.

Unfortunately AMD has been criticized due to some chips' inability to hit the rated boost frequencies...

By who, their competitor and some click-bait journalists? Has Tom's even tested a chip that has fallen short of its max boost clocks by more than a percent? Perhaps there are some chips out that miss their lightly-threaded boost clocks by more significant margins, but at least the vast majority seem to perform just fine. An online survey of anonymous internet denizens who may or may not even own the processors in question, let alone have their systems configured optimally, doesn't seem like data worth basing criticism on.

It was nice to see a detailed analysis of what the new BIOS was doing on this particular motherboard with this particular sample of this particular chip, but ultimately, a sample size of one doesn't really make for something to base any definitive conclusions on. There are other processors and other boards from different manufacturers that may be affected differently by the BIOS changes.

In all, there seemed to be an excessive amount of negativity for an article covering improvements to a BIOS for a series of processors that this site was already recommending. And the introduction and conclusion completely glossed over the improvements to power consumption, instead focusing on the lack of tangible performance improvements for a chip that was performing as expected to begin with.

 

[joeblowsmynose]

It hit 4.375 GHz on 5 different cores with nothing changed in the BIOS except enabling the XMP (DOCP) profile.
Which is close enough for me.
my 3700x absolutely meets my performance expectations, even if at stock settings it is .25 GHz below the ...

But 25mhz of 4400mhz is 0.57 % of the total frequency of that chip, and then if we divide that by eight cores to get total performance amount per chip (0.07%), then divide it some more to consider potentially only temporary boosting perhaps (like Intel does), and can't you see we have a VERY serious performance problem here? I mean, have you not been reading the headlines, the comment threads, and Intel presentations? This is some serious problem!

Just look at all the major concern on this thread even -- this is some serious stuff we're talking about here ! You should be very concerned! Very VERY concerned, like the rest of us! [/s]

 

[DMAN999]

But 25mhz of 4400mhz is 0.57 % of the total frequency of that chip, and then if we divide that by eight cores to get total performance amount per chip (0.07%), then divide it some more to consider potentially only temporary boosting perhaps (like Intel does), and can't you see we have a VERY serious performance problem here? I mean, have you not been reading the headlines, the comment threads, and Intel presentations? This is some serious problem!

Just look at all the major concern on this thread even -- this is some serious stuff we're talking about here ! You should be very concerned! Very VERY concerned, like the rest of us! [/s]

LMAO !!!!!
I am Very concerned that this CPU performs as well as it does.
I'm really worried that it might just explode at any given moment and take out my PC, me and anyone else that just happens to be standing in the vicinity at the time. LOL

 

[joeblowsmynose]

LMAO !!!!!
I am Very concerned that this CPU performs as well as it does.
I'm really worried that it might just explode at any given moment and take out my PC, me and anyone else that just happens to be standing in the vicinity at the time. LOL

Well if AMD's boost clocks aren't exactly up to expectation, maybe their chips inability to explode and harm people isn't up to our expectations either, creating a risk ... we can never know with all this AMD "shadiness" going around ... lol.

I'm all for making fun of silly and stupid marketing efforts, like "28 core 5ghz coming this fall!", and "Real world desktop applications! (all stats based on laptop and tablet data)" and "We're better at running Internet Explorer and Windows Media Player!" and "Why pay less for a 9900K and overclock when you can pay more for a 9900KS and not OC it!" (ok I don't know about that last one, but knowing Intel they'll charge more for the KS)

I'm good for making fun of AMD too, but I have no recent examples to work with. Well they did promise 5ghz zen2 ... oh wait, that was AdoredTV that promised that, not AMD. Never mind. This issue current isn't really good for poking fun anyway, especially now that its "fixed". (although I will re-iterate that I think the concern was valid, it just wasn't a big a deal as many have been trying to make it out to be). ... ok, I got one for AMD ... "Why buy the 3600 when you can pay $50 for an 'X'!" - Steve Burke beat me to that one though.

I am glad that AMD has applied a fix for this issue, but I really think this whole thing was blown WAY out of proportion.


I am a little concerned though that CPUs are now being manufactured to perform like GPUs - boost clocks are based on individual silicon quality and not necessarily a static number - I noticed that many zen2 samples actually boost above their rated boost clocks as well - indicating that individual silicon quality is indeed at play here- a sort of new take on the "silicon lottery" I suppose. I guess this is they way things are headed.

 

[RodroX]

Well, I'am a proud Ryzen 5 3600 owner, and after .... 6 years of having the same intel cpu on my PC, I can say that man, the change was worth it and huge (coming from a core i5 3570).

I know, I know, many would say "well coming form that CPU anything will do better", the fact is that the core i5 3570 was a beast for the money back in the day, and still is more than decent CPU for many games.

THE PROBLEM, well is fairly easy, today intel has nothing on the 199ish dolar range to offer better than the Ryzen 5 3600, zero, nothing at all.
Even thought not many people notice this, or want to see this, Intel hasn't have nothing worth buying on that price range since Ryzen 5 2600 was out to the wild either (good news for the ones that when the AMD way, bad news for the ones that got caught into the core i5 8400 and the likes) - (don't get me wrong Im not a fan boy, I know that i5 8400 still a beast for most games, but what about next year?, and the next one after that?) .
Of course if you have infinity budget you could go and get yourself a i7 8700/8700K/9900K (fan boys, please stay away of the 9700K, cause intel switched HT off on that one to be able to charge you more for the same CPU with it -> 9900K, oh Intel you're such a naughty boy).

On topic: " ... but the increased peak frequencies often don't equate to noticeable differences in performance "

Its been weeks, and I keep saying and writing this, either be AMD or Intel, honestly, please, 25 to 50 MHz clock diference, in almost every situation wont impact on the system performance that much, if it does at all. Yes it will depending on the software too.

This is not news, this is basic overclocking. Anyone that read about it , tried, succeeded or failed at overclocking knows that 25 MHz diference wont be a huge impact on daily use, even if you spend the whole day rendering 4K videos.
You're better off optimizing RAM, getting a better SSD or a better Motherboard.

Cheers

 

[Arbie]

THANK YOU for fixing the title!

Fair is fair.

Fanboy out.

 

[kenjitamura]

If this is a Windows scheduler issue I'm wondering if testing with linux and trying out the different scheduler configurations would yield the expected results.