Tuesday, September 3rd 2019

AMD Issues Statement on Low Ryzen 3000 Boost Clocks, BIOS Update Soon

After AMD's Ryzen 3rd generation launch many users have reported that they are not seeing the advertised boost clocks that AMD promises in their specifications. This has been an ongoing issue, with various tweaks tried, with limited success. This lead to serious allegations about "false advertising", and all AMD had to say up to this point was that these clocks are "up to".

AMD has now issued a statement regarding these lower than expected clock frequencies on Zen 2 processors, and it looks like there is indeed an underlying BIOS issue that's responsible. Let's hope that this new firmware gets released quickly and is able to restore faith in AMD's otherwise excellent track-record.
AMD is pleased with the strong momentum of 3rd Gen AMD Ryzen processors in the PC enthusiast and gaming communities. We closely monitor community feedback on our products and understand that some 3rd Gen AMD Ryzen users are reporting boost clock speeds below the expected processor boost frequency. While processor boost frequency is dependent on many variables including workload, system design, and cooling solution, we have closely reviewed the feedback from our customers and have identified an issue in our firmware that reduces boost frequency in some situations. We are in the process of preparing a BIOS update for our motherboard partners that addresses that issue and includes additional boost performance optimizations. We will provide an update on September 10 to the community regarding the availability of the BIOS.
Our resident Ryzen memory overclocking guru 1usmus is reporting the same on his Twitter account:
SMU FW 46.44.00 will improve the boost of processors of the Zen 2 generation, very soon in all BIOSes. ETA: 10 September
Add your own comment

86 Comments on AMD Issues Statement on Low Ryzen 3000 Boost Clocks, BIOS Update Soon

#51
trparky
Darmok N Jalad
Anandtech has an article running the 9900K with a 95W cooler. You still get 5.0GHz on a single core. Where you see a big performance loss is under multi-core load, and then the chip runs at the rated base clock. Give it more thermal headroom, and it runs at much higher clocks.
5 GHz on a single core, big freakin' deal!!! I expect more, a hell of a lot more! I expect to be able to run at (or at least, close to) 5 GHz on all cores; that's where your TDP value goes through the roof.

Hell, my 8700K gets pretty hot as I ramp up the workload on it thanks to Intel's toothpaste.
Posted on Reply
#52
HugsNotDrugs
EarthDog
It doesnt though... that is the point..and why amd has reacted now 3x times over clocks in these CPUs.;)
It does technically reach full boost under certain workloads: Amd/comments/cfli2n
Posted on Reply
#53
ssdpro
Tough wording in this article. I am not much for AMD but it seems pretty minor. Some chips or boards aren't running a specific core(s) at a specific frequency under some scenarios. AMD acknowledges the defect and responds with a promised fix. I don't really see any false advertising, just a bunch of really nice chips running close to max specification. Problem discovered, problem hopefully soon to be fixed.
Posted on Reply
#55
Nordic
HugsNotDrugs
It does technically reach full boost under certain workloads: Amd/comments/cfli2n
How can you explain that 6% of 3900x owners can hit the max advertised boost commonly, while 94% cannot under nominal conditions? I for example have not been able to achieve more than 4525.3mhz under any workload I have tried.
Posted on Reply
#56
EarthDog
HugsNotDrugs
It does technically reach full boost under certain workloads: Amd/comments/cfli2n
for that guy... but not for everyone and not for a majority. Please let's not rehash this crap again...amd already admitted there is something up and will be fixing it.

Remember, you said "quite readily" which is obviously not true and mainly what I responded to. If the chip 'quite readily'did that for a majority, we wouldnt be here right now, boss. :)



The post mortem denial from people is offputting.
Posted on Reply
#57
R-T-B
trparky
5GHz on a single core, big freakin' deal!!!
Industry norm for boost man...
Posted on Reply
#58
Darmok N Jalad
trparky
5 GHz on a single core, big freakin' deal!!! I expect more, a hell of a lot more! I expect to be able to run at (or at least, close to) 5 GHz on all cores; that's where your TDP value goes through the roof.

Hell, my 8700K gets pretty hot as I ramp up the workload on it thanks to Intel's toothpaste.
But that's not what it was designed to do. These multi-page forum threads may rage on about boost clocks, but simply put, multi-core CPUs take an opportunistic approach to boosting. Under light or low-threaded loads, you get higher clocks. This is a fantastic design idea, because the alternative is to not allow any core to exceed the safe all-core max load. So instead of a single core hitting its max capable clock, you get a much slower top clock regardless of load. Also, we're talking about 4400MHz+ clock rates. A modern CPU could have boosted to a max clock and put away instructions multiple times before a frequency polling tool even had a chance to take a measurement. Boost is an engineering and consumer win. It's just marketed and understood poorly.

I'm not here to excuse AMD, but they need a chance to make it right. We're talking about a 12C/24T CPU for $500 that can drop into existing boards. Two months ago, it cost $1200 for a 12C/24T CPU alone. What did a $500 CPU get you last year? To me, this is the same early adopter headache that shows up with every new product--in 6 months, everyone will be fuming about something else.
Posted on Reply
#59
Bitgod
Sweet, now we only have to wait 2 or 3 months for the firmware to show up...
Posted on Reply
#60
Crackong
September 10.

Together with 3950x ?
Posted on Reply
#61
Liquid Cool
Darmok N Jalad
To me, this is the same early adopter headache that shows up with every new product--in 6 months, everyone will be fuming about something else.
Since were talking about AMD...I'm in complete agreement.

When has AMD been "good to go" right out of the box? I can't recall when this was the case(and I was a member of AMDMB, so I've been around awhile!) and frankly, I know better than to expect it. I waited a solid year to purchase my Ryzen 5 2600 and I couldn't be more pleased with the processor. Simply put...144.99 in the past has never netted me something even close. Coupled with my 88.00 RX 580 and I'm good for....years. I absolutely LOVE my computer! Haven't had a single hiccup since I tossed it together.

Having just come back to AMD with the release of the ryzen 3 2200g, one thing I did notice....drivers are much much better. Even the catalyst software isn't half bad to look at!

Best,

Liquid Cool
Posted on Reply
#62
jmcosta
trparky
Really. Security issues and lying about their TDP values and what they mean. Yeah...
I don't remember Intel lying to the public about the security flaws but they did knew about them, still was best to keep a secret from the public at that time or that would have generated more problems to the consumers.
Their TDP values are correct, ~95w for base frequency, they never advertise the chip TDP under boost on the box.
Posted on Reply
#63
trparky
jmcosta
Their TDP values are correct, ~95w for base frequency, they never advertise the chip TDP under boost on the box.
Well, when a good majority of their users run their chips at boost speeds one would think that they should advertise their boost TDP. That way, we as enthusiasts, can make better-informed decisions about our builds. It would certainly make choosing high-end coolers much easier since we would be able to know what we need for the rated speed that we want.

For instance, if the chip has a TDP of 200W when running at say... 4.7 GHz, we would know right away that we would need either a high-end 240mm radiator or a BeQuiet Dark Rock Pro 4. Or, if we don't necessarily need those speeds we could then choose a lower-cost solution and save some money and put it towards other components in our system such as perhaps a bigger or better SSD or faster GPU.
Posted on Reply
#64
Khonjel
EarthDog
You just posted in a thread where AMD "made a statement" admitted to there being an issue they will correct with firmware... o_O

It was NEVER what you inferred (intentionally nerfing boost), and people that said hardware issues, were mostly just wrong as plenty of people had the right hardware (board, cooling / nominal conditions) and still can't hit it (raises hand).
They admitted and said fix is incoming is the best case scenario. I consider the worst case is if they admitted and said it was "designed to run slower" for some weird-ass reason they come up with.

Now if the fix actually works is to be seen.
Posted on Reply
#65
R-T-B
HugsNotDrugs
Like full boost under certain workloads? It does.

Amd/comments/cfli2n
So... noop instructions. As in noninstructions. It can boost while literally doing nothing.

Wow. Yeah... really doesn't count.
Posted on Reply
#66
R0H1T
Nordic
How can you explain that 6% of 3900x owners can hit the max advertised boost commonly, while 94% cannot under nominal conditions? I for example have not been able to achieve more than 4525.3mhz under any workload I have tried.
Your source, I assume you have personally verified this with little margin of error?
Posted on Reply
#67
Khonjel
BTW der8eur's survey is specifically with pbo turned off. He said that pbo is hit and miss. I'm paraphrasing here but he said it downclocks for some while upclocks for others. While pbo technically voids warranty, I doubt many people would have it disabled. Hell, in my Tomahawk it was on by default for some reason.

He also ditched entries that the posters specified achieved with pbo. Now this leads me to believe that as per human trait of forgetting minute details some people could've forgot to or didn't state that their results were achieved with pbo and then the percentage of boost reaching chips (per his criteria) could be even lower.

Ultimately his survey isn't scientific and his numbers shouldn't be taken as gospel. However that does not make his data is invalid, not at all.
Posted on Reply
#68
Mussels
Moderprator
Bravo to AMD.
From what i read prior to this post, almost everyone got within 25Mhz of the targeted clock - it was hardly an epic massive failure.

AMD are still fixing it, with the magical power of BIOS updates - and they fixed it FAST too.
Posted on Reply
#69
Chomiq
Sometimes I wonder who's more butthurt about it, people that got the CPUs and expected more or people that weren't even considering getting them in the first place?

I'm happy to see that AMD finally admitted that something was wrong on their part. Partially, but still.
Posted on Reply
#70
Calmmo
We'll see if it gets actually "fixed".

It is surprising reading this thread however, I always underestimate the number of id..
I mean fanboys on either side of the fence.
Posted on Reply
#71
Nordic
Mussels
Bravo to AMD.
From what i read prior to this post, almost everyone got within 25Mhz of the targeted clock - it was hardly an epic massive failure.

AMD are still fixing it, with the magical power of BIOS updates - and they fixed it FAST too.
By what I have read and experience it is typically 100-200mhz less. In my case it is 75mhz less. 75mhz won't make or break the system performance but that isn't the point.
R0H1T
Your source, I assume you have personally verified this with little margin of error?
I am quoting debaurs survey data which was on the tpu front page. There is another thread talking about just that. You can argue against his methodology and data, but I am quoting a legitimate source.
Posted on Reply
#72
Xaled
R-T-B
Not at all, no.

I just assume a product will do what it says.
So no reaction when Intel lies about boost clock
partial lie is a lie too and intentionally misleading is even worse

Posted on Reply
#73
I No
Xaled
So no reaction when Intel lies about boost clock
partial lie is a lie too and intentionally misleading is even worse


What is so misleading about those 2 SKUs? They all hit their boost clocks. It doesn't matter if you cool them down with a garden hose or with rocks, if put to work those 2 will reach the specified boost clocks 100% of the time, they might throttle after 1 second but hey the boost will be reached. The only difference is the silicon lottery you get to play for the max OC. I don't mind AMD's line-up even though they are running different clock speeds depending on the motherboard and whatnot and i'm willing to bet nobody will make a big fuss out of this, but if somebody decides to take it to court they would have a valid lawsuit which AMD is trying to avoid like the plague after the Bulldozer one.
Posted on Reply
#74
Xaled
It
I No
What is so misleading about those 2 SKUs? They all hit their boost clocks. It doesn't matter if you cool them down with a garden hose or with rocks, if put to work those 2 will reach the specified boost clocks 100% of the time, they might throttle after 1 second but hey the boost will be reached. The only difference is the silicon lottery you get to play for the max OC. I don't mind AMD's line-up even though they are running different clock speeds depending on the motherboard and whatnot and i'm willing to bet nobody will make a big fuss out of this, but if somebody decides to take it to court they would have a valid lawsuit which AMD is trying to avoid like the plague after the Bulldozer one.
It is definitely misleading because 80% of people think all cores are going 5.0 GHz

The Bulldozer is no different that Intel's Hyperthreading hypocracy. 90% of people think Threads are actual cores and that's why Intel made them at first and lied misleaded 100 of millions of people
Posted on Reply
#75
R0H1T
Nordic
I am quoting debaurs survey data which was on the tpu front page. There is another thread talking about just that. You can argue against his methodology and data, but I am quoting a legitimate source.
And that's about as relevant as the grounds on which AMD settled their latest lawsuit. Which is to say that if you confine the argument to a set of variables which probably don't even represent a quarter of the (whole) set (or less than 1% of all users) then sure, we can all claim whatever we want to.
Posted on Reply
Add your own comment