Showing results for 
Search instead for 
Did you mean: 


Adept III

Ryzen 9 3900X Boost Frequency

To AMD why your new processors never hit advertised boost speeds in single or multi core performance. I have Ryzen 3900x with Asus X570 Prime Pro board running latest bios and AMD chipset drivers. My single core boost is maxed at 4.2Ghz and multi core at 4Ghz. Doesn't matter if PBO enabled or not. There are 1000's of people complaining about this and you have not even acknowledged this as a problem. Your false advertisement about boost speeds is completely unacceptable. I was always Intel customer but this time I took a chance on your company and trust me this is first and last time i will buy anything from you. You hand selected CPUs for reviewers to get the hype going and stuck your paying customers with garbage. I will be filing a RMA claim and will be exchanging this processor until i get one that performs to advertised specifications.

195 Replies

Whether or not you hit the max on your CPU replies on many factors... From temperature to RAM, the motherboardand your BIOS settings. You might might want to post more data. 


Eddie, AMD release this product saying that the Max Boost is 4.6. Then retracted and said that at least one core would hit 4.6.  Not, one core will hit Max Boost is 4.6 if you have the right cooler with the right RAM at the right speeds under the full moon with Venus in it's 8th House. 

My point is that none of my cores are hitting 4.6. If you read my post, I listed what you ask but RAM. Which is G.SKill 3200C14 Samsung B-die. But, it doesn't matter. Because, if you have a compatible MB and compatible RAM with a stock cooler, one of your cores should hit 4.6 as advertised.


It's not that simple.

The AMD changes can be seen as a clarification, not as a retraction of promised performance.

Why isn't it that simple? Because having b-Die RAM is only a part of the equation. There's plenty of ways to not even get 3200 from your RAM. Only needs an unstable timing setting, a power source that is unstable or VRMs that generate too much heat. Nothing you can blame the CPU for.

But keep dreaming of the simplicity of things. 


Max boost has always been 1 core... What they clarified is that the max boost is under ideal conditions.

I will reiterate that whatever they are considering to be ideal conditions is dumb, because if you just put CCD0 at 4.6 manually you'll get much higher single core test scores than what stock/PBO/Auto OC will give you. A post by Shamoo from ASUS says they since reduced the temp limit of "Something" to 75c which resulted in slightly lower boost speeds, but I am getting ready to ask Shammo what that "Something" is because my single core tests have never exceeded 47c according to Ryzen Master. Right here you can see that theres not a problem with temp, at all. Not any temp that we can see.


A while ago I made a program that can saturate all cores equally

I use it to test thermal stability of a repair of a CPU cooler

My R5 2400G is 3.6 GHz design but wth the Goldback tool it is 3.7 and stable

The core boost works, but keep in mind the actual max speed will vary with the threads in use


I'd say if Cinebench can't properly give the same '4.6ghz' score in its single-core runs with PBO than it can with a manual 4.6ghz something is really wrong with the boosting. This isnt just limited to cinebench either. Gaming benchmarks I dont get near the FPS scores with PBO as I do with my manual.

It didnt even do this with my 1800x or 1950x. Single core score whether manually set to 4.1ghz or left on auto were the same then, so whats changed is the 'new' Precision boost 2 algorithm. Its even been jacking with Zen and Zen + cpu's where people updated their boards bios to the latest AGESA. You no one should be OK with their processor suddenly not performing as advertised because of bad bios and agesa revisions, and it is down to bad bios and agesa revisions. These processors can do the advertised boost on a single core, easily. All-core is going to be up to cooling solution, but a single core load does not, if you attached your cooler correctly, get to a temperature that should be concerning. It sure as hell doesnt for me.


All I know is that my CPU is not the problem that my GTX 1060 is for gaming at 4K UHD


Yeah I can imagine :-)


My issue is that no one can seem to get these "ideal conditions". And if no one can get them then it's probably not true.

I've made another post for testing and results. Please post a pic of your results following the guidelines posted there. I really want to know if anyone is getting 4.6 under any conditions. Thanks! 


There are responses of people that do hit the promised speed. 

But you won't see them here. If a Product works fine, why go on to a forum and shout it from the rooftops... It's always the complainers that publish it. 


I'm pretty sure they arent really hitting the promised speed. What they need to do is actually use the histogram of Ryzen Master as tarvaln‌ suggests. HWiNFO is showing higher but all of my PBO single core scores are in line with the peak of 4400mhz that Ryzen master is showing. Right now Ryzen master is the only thing that can show what the processors are really doing. HWiNFO max speeds are simply anomalies. 

Adept III

I know is not the same as a Ryzen 3900X, but If Im not mistaked I did post in this forum my CPU Ryzen 5 3600 hitting the 4.2 (4192MHz) on Cinebench Single Core test. I showed the HW info data, but its the same on Ryzen Master, any of the two shows 4192MHz, I just like HWinfo cause its have soo much more info in a small Windows.

The frecuency was achived with only PB enable on BIOS, I didn't need PBO or Autooverclocking, my RAM is Adata XPG 8x2GB DDR4 3200Mhz (no samsung-b dies, I believe they are hynix) This was with stock cooler and BIOS F41, AGESA 1003 ABB, Mobo: Gigabyte B450 Gaming X.

Im at work right now so I can't re-post the screenshoot. But is out there.


And, thats not really what happened.


The heck it isnt, its exactly what has happened.. I've seen it with my own eyes and on my own system.


I see the same thing, but there is no confirmed intent to lower boost speeds.


Do you actually think amd would incriminate themselves?


Yea, saw that. "Shamino said this" and "The Stilt said that" Neither of which are AMD employees.

Then you have stuff like this:2019.08.29-20.46.png

Oh look, 4600 on 3 cores simultaneously, even the the one that is not one of the 'fastest' in the CCX.

AMD tried something and it didnt work to better stock operation, thats it. You're unwilling to do simple things that would disprove any claim that these processors cannot do their advertised speeds. Sure, there are hickups in bios that are preventing it from doing it out of the box as they should, but they are just hickups. There is no scheme or false advertisement going on.

Adept I

Not sure if anyone else noted this, but I found an odd behavior between my 3900x and Trident Z Neo 3600mhz, 14CL ram. I saw my 3900x hitting the advertised boost clocks, hell, I have at least 4 cores that can and 2 of them actually go slightly over the 4.6ghz.

However, when I bought G Skill's new Trident Neo ram and overclocked it to 3600mhz at 14 CL (Which is what the ram is intended for) I noticed the 3900x would no longer boost over 4.25ghz. After much tinkering, I found the ram frequency is what is holding the CPU back. My ram is now running 3200mhz and the cpu boosts as it should.

Has anyone else found this odd coincidence as well? Perhaps those who are having issues with their CPU's boosting should try resetting the bios and see how it boosts, then tinker from there.

EDIT: I was running Bios 2606 for the above findings. I upgraded to the newer version 2703. This did not completely fix it though. Ram at 3600mhz causes the cpu to boost no higher than 4.475ghz. An improvement, but still not where it should be. At 3200mhz it boosts no higher than 4.575ghz which is less than the last Bios version.

My PC: 3900x, Asus X470 Crosshair vii Hero wi-fi, 32gb Trident Neo 3600mhz ram, RoG Strix RTX2080, Samsung 970 Evo, EVGA 750w PSU, CoolerMaster H500M, Windows 10 v1903


I noticed that if I set RAM to the XMP profile, boosting breaks completely.  I noticed that RAM speed is set by the profile on the main UEFI page.  Infinity Fabric and Northridge speed to not scale correctly with RAM, when set there.  Instead, I set the RAM clocks in AMD CBS and Overclocking menus.  Setting it here caused the Northridge and Infinity fabric to scale correctly, and also restored the normal boost behavior.

Good call, this helped but didn't get it to 4.6ghz for me. I'm boosting to 4.5 now with 3600mhz ram. Hopefully motherboard makers (Asus for us) get on this quickly.


That's where I got too as well.  If you set the RAM clock speed in all three places, you will see Northridge speed and Infinity Fabric scale correctly on auto with RAM.  Boosting behavior then functions normally, but mine is stuck around 4.5 GHz as well.

I think the fact that some of these settings are in three places is causing issues.  

Journeyman III

I'm trying to duplicate allstate findings. I'm having an issues setting RAM with Ryzen Master. I can't get it to set over 1.325 with MEM VDDIO. My RAM is G.Skill F4-3200C14D-16GVK. XMP setting is 14-14-14-34-1.35. My PC won't POST with a DRAM voltage setting that low and it defaults the RAM back to stock settings. Is anyone else having this issue? Also, is this the right setting for DRAM voltage. I'm going off Not sure what I am doing wrong.

Also, Ryzen Master is says my Coupled Mode is on. Could that be a false positive? In BIOS I set the Fabric Clock to equal Mem Clock manually. I don't leave it on "Auto" I read that Auto is for Fabric Speed is broken.



I don't typically use Ryzen Master and do everything through the bios. I know all these programs have come along way but I just can't do it lol. I wish I could help but I saw the same issue the one time I considered using it to OC my ram. Never tried again after that.

As for coupled mode, it's my understanding that any fabric OC 1800 and lower will read coupled which is a good thing. That's your 1:1 IF to Mem Freq ratio. Anything higher is uncoupled and will usually force a 2:1 ratio.

Journeyman III

AMD has stated it will release a new BIOS to address the issues we are seeing. AMD Announces BIOS Fix for Ryzen 3000 Boost Clocks, Update Comes September 10th 

Some speculation on my part. AMD doesn't release MB BIOS'. They release a package to MB vendors and then those vendors have to incorporate it into their BIOS for their boards. I don't know if the vendors will have new BIOS' on the 10th.
I'm beginning to believe this may be a hardware issue. A BIOS update can't fix that.
For the 3900x AMD gave each CPU one "good" chiplet and one "ok" chiplet. That's why they claim at least 1 Core can hit advertised Boost. AMD has also stated they thoroughly binned their chiplets. So, I wonder if they yields weren't as good as they hoped and they release CPU's with two "ok" chiplets. Saving the "good" chiplets for the 3950x. That would account for what we are seeing.

As to why AMD would do this. I don't think AMD was wringing theirs hands in their secret lair thinking how to lie to people. I do think something went wrong or did go as well as they hoped and they are trying to fix and/or cover themselves. I wish they would be for forthcoming about what is going on. They just settled their lawsuit about the bulldozer series. I hope they use the lessons from that and be more transparent, so there won't be a "Zen 2" lawsuit.

If you haven't, I recommend watching these two videos.

Ryzen 3000 Update! Boost Clocks, Voltages, and More! - YouTube 

It's worse than I thought - RYZEN 3000 Boost Survey - YouTube 


"I'm beginning to believe this may be a hardware issue. A BIOS update can't fix that. "

As I have demonstrated ad nauseam, the silicon CAN hit the boost clocks even at the voltage precision boost already uses.  So this isn't a hardware problem.  Those two videos suggesting there may be flaws in the actual silicon isn't supported by any data currently available, what's more, it is actually fairly easy to debunk, as I have already done.

Suggesting causes that aren't supported by any data is pure speculation and needs to be labeled as such.  Failure to do so is not only negligent journalism, it is poor science and damages the reputation of those who engage in it.  

Tarvaln wrote:

Some speculation on my part. AMD doesn't release MB BIOS'. They release a package to MB vendors and then those vendors have to incorporate it into their BIOS for their boards. I don't know if the vendors will have new BIOS' on the 10th.
I'm beginning to believe this may be a hardware issue. A BIOS update can't fix that. 

I said it was speculation. Please don't cherry pick my statements.
I'm glad you were able to hit 4.6 on one of your cores. I could not duplicate your result. 


Hi Tarvain. I mentioned in a comment above that on one version of Bios, I actually hit 4.625ghz on 2 cores and 4.6 on two others. I'm sure that wasn't simultaneous, but still better than what I hear others are getting. That was also at 3200mhz on the ram. At 3600mhz and that Bios version, I was getting 4.250ghz only. When I upgraded, I found that I was boosting to about 4.5gz at 3600mhz and getting 4.575ghz at 3200 & 3466mhz ram. This tells me it is a Bios issue in conjunction with the ram. Something to think about when testing yours out.

For everyone, feel free to see a thread below. There is a Google spreadsheet that someone started with other findings that I have added to as well. By all means, add yours to the list. The more there is added, the more we can all find trends as to the cause of this.


I have noticed the identical behavior allstate.  When setting RAM, you have to make sure to set the clockspeed in both the main UEFI page and AMD overclocking, otherwise boosting is limited to 4.25GHz for some reason.

I too have found that I am limited to around 4.5 when running my 4 DIMMs at 3600CL16 1T, with 1:1 northbridge and infinity fabric scaling.  My SOC voltage is pushed up to around 1.1V in that setup, so I'm not sure if that somehow impacts the vcore voltage from the algorithm?

In any event, I'll wait until AMD releases their statement on the 10th about the microcode update, and hopefully we won't have to wait long for motherboard vendors to get the new UEFI releases out.  No guarantee it will fix our issues, but I guess we'll wait and see. 


Yea, I saw your statement above and did some test between 3200c14 and 3600c16. I didn't hit 4.6 on Cinebench, CPU-Z, or Geekbench with either. I am unable to get Ryzen Master to set Mem voltage correctly so I did all my settings in BIOS. The scores were about the same. I'm unable to get over 4.4Mhz on anything I've tried and tested so far. I've only tested on the last firmware update on my BIOS which is 1005 released 8/12.

One thing I haven't done is manually overclock a single core, cores, or shutdown all but one core on any test. I haven't done this because it's not a standard operating parameter for this CPU. Also, if you manually overclock your CPU then you're no longer testing a boost you're testing an overclock. 


Did you? What were your settings?


Not sure if you are asking if I tried or if I achieved. I tried. I did not achieve.

As for settings. I'm keeping my RAM at 3200c14 with timing from "Safe" on DRAM Calculator. CPU, I've tried different things between tests. Setting should be listed in the Test/Results topic. 

As, you, ajlueke, and myself have noted. Even getting the settings to work properly can difficult to not possible with Ryzen Master. I've tried a few OC's in Asus BIOS and have had problems. There are discrepancies between HWiNFO and Ryzen Master. You may have to set your BIOS, Ryzen Master and a motherboard tool in order to get correct settings working. Ryzen Master may be reporting things incorrectly. Again, I want to thank you and ajhlueke for testing and pointing those out.


processor settings, to test boost clock speed on manual


Forgive me, but the statement:


"I'm beginning to believe this may be a hardware issue. A BIOS update can't fix that." 

indicates that you have come across new information "presumably the videos you linked" that have caused you consider that a hardware fault is to blame.  Furthermore, I find it puzzling how an individual could speculate on whether or not they believe something as they should probably just know, which lead me to conclude that this was a new statement and no longer part of the speculation that preceded it.  And, beyond that, if we really want to analyze it, the statement doesn't seem to contain any speculation at all from what I can tell.

"Some speculation on my part. AMD doesn't release MB BIOS'. They release a package to MB vendors and then those vendors have to incorporate it into their BIOS for their boards."

Well this isn't speculation, it is true, AMD releases AGESA microcode that motherboard vendors incorporate into their UEFI releases, great.

"I don't know if the vendors will have new BIOS' on the 10th."  Also not speculation, simply a statement.

"I'm beginning to believe this may be a hardware issue. A BIOS update can't fix that."  Again, not speculation, just a statement of what the author believes. 

"I said it was speculation. Please don't cherry pick my statements."  I would offer, that I did not "cherry pick" anything, but that the sentence structure and verbiage used led me to the most logical conclusion.  Since that is not what was intended, I would suggest the following correction. 

AMD doesn't release MB BIOS'. They release a package to MB vendors and then those vendors have to incorporate it into their BIOS for their boards. I don't know if the vendors will have new BIOS' on the 10th.
Some speculation on my part, this may be a hardware issue which is something a UEFI (these are not technically BIOS anymore) update can't fix that. 

"I'm glad you were able to hit 4.6 on one of your cores. I could not duplicate your result."

I seem to be failing to get the point across. The 4.6 GHz is only possible if I disable precision boost completely, manually set the voltage to the boost voltage, and then run the CBR20 single core benchmark.  This was merely an experiment to see if a single core could hit 4.6GHz at the voltage used by precision boost with the algorithm removed from the equation, and it can.  Once precision boost is turned back on, the core no longer boosts to 4.6 despite the identical voltage being applied in my manual test.  This indicates that the silicon itself is not to blame, and can hit the advertised clocks within the voltage threshold already used.  The problem must lie elsewhere.  This isn't a solution however, as running the system with a manual voltage of 1.48 would be decidedly unsafe for multicore boosts. 

So again, I stand by my statement that  the creators of those videos should make sure to only list conclusions supported by the data. And that "failure to do so is not only negligent journalism, it is poor science and damages the reputation of those who engage in it."  

I hope your detail analysis of my speculations bring you some happiness.




Looks like AMD will get motherboard vendors an update on 9/10.  Time will tell when the vendors will push those updates to all the different models.


Nice! Thanks for the heads up!


Lemme know when my R5 2400 works on my X570 board......


I see amd is now working on a 3rd update to the lack of hitting advertised frequencies..