PC Processors

[s_vayner]

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.

[conraire]

I'm going to go with a lot of people haven't got their bios settings correct.  As my 3800x regularly boosts past 4.5Ghz, its advertised boost speed, in single core loads..  Which is how it's suppose to work with a good cooler, with a motherboard with a decent VRM layout. 

I may be currently running on an X370 board, but it was a $250 flagship board when I bought it new.  I'll be upgrading to either the Gigabyte Aorus Master or MSI MEG Ace here in about a week.

To those having stability issues, try adjusting cpu vcore load line calibration tightness until the system becomes stable, and doesn't randomly crash.  I had to with my 3800X.  Prior to adjusting that it'd be stable in bios, but windows would randomly blue screen after about 10 mins of use. 

Don't cheap out on a low end motherboard, and expect it to run a higher end processor full bore, some of the lower end boards have real crappy VRM's, and won't be able to handle the higher end Ryzen 3000 CPU's properly. Some people have a really bad habit of buying the cheapest thing possible.

Your ASUS Prime X570 Pro appears to be a decent mid range board.  SO it may have to do with bios settings.  What is the rest of your system configuration?  What other monitoring or OC software do you have installed other than Ryzen Master?

If you pay attention to any of the major reviewers on Youtube, etc.  You'd know that 4.1-4.2Ghz is the all core boost range for Ryzen 3000 cpu's.  Generally to get 4.2Ghz you need good cooling.  Some can hit 4.3Ghz all core boost, but it takes really good or exotic cooling to do so.  The advertised max boost is for single core under stock cooling, or lightly threaded under good cooling.  And if you have good cooling and a good VRM it will go past the advertised boost max. 

[s_vayner]

Hi, my board is Asus X570 Prime Pro which is not top of the line but its in higher end category. I'm not looking for all core boost to 4.6Ghz I know that's not possible. It's the single core performance that falling short. I've never seen single core to boost above 4.2Ghz, regardless of BIOS settings. That includes running PBO on/off/auto. Undervolting CPU, running lose mem timings, among other things. This is not isolated case, there is a large group of people that complaining about it on Reddit. Advertised boost speed for single core was 4.6Ghz i'm falling ~10% short of it. Also i shouldn't be required to buy $1000 motherboard or exotic cooling to get advertised speeds, but that's not even the case as people with top of the line boards and water cooling have same issues. Being Intel user for over 20yrs I'm not used to having jump thru hoops to get your CPU to operate at rated speeds, is that an accepted practice for AMD? And where is AMD response to this?

[conraire]

Well, thats why I asked what your system Config is.  And what other software you have installed. 

Shouldn't need PBO turned on to hit max boost, mine is disabled.  As is XFR on my board.  Under Volting may be the issue if you're still doing that.  The more you under volt, the lower the boost clock will be.  I noticed that when I was trouble shooting my 3800x, when I thought the 1.5v peaks were abnormal, apparently its normal behavior according to AMD.  The same goes for using a negative offset to under volt. 

[hardcoregames_]

The voltage varies a lot on my CPU as does others with Ryzen processors as this is one of the thermal management tools used.

If there is adequate cooling the CPU load can be higher than it could with inadequate cooling.

[s_vayner]

config is pretty basic Asus Pro x570 motherboard, 1TB gen 4 pcie nvme, 32gb Ripjaw 3600 memory and Ryzen 3900x, nvidia GTX1070. Its a brand new build with fresh win 10 pro install. i use couple monitoring tools cpuz, amd master and hwinfo. I run CB20 for single core thread to check performance but i also leave hwinfo running to record max speeds during different use scenarios. Like i said i tried every possible setting in bios and CPU never boosted past 4.2GHz in single core testing or normal usage. Multi core is about 4GHz boost.

[s_vayner]

also my temps never break 80C for cpu running stock cooler, in single core temps are even lower, so its not whats causing boost to fail.

[flank3rcz]

Im lazzy to read all comments, but propably issue is on your side. First, check your BIOS version on the board. I have best experience with BIOS with agesa 1003 (but not 1003abb). Be sure, if you monitoring your PC system, u have hardly closed proces AI Suite. exe. This software takes load of CPU.

Do install chipset driver for your motherboard (links bellow)

For single thread test you can use Superpi benchmark or Cinebench R15 single thread. For casual monitoring is best to use software HWINFO, after you run it, run "sensoring only". After check maximal values in HWINFO. Im sure, u will get something between 4540 to 4580 MHz. Software has higher refresh latency than real CPU changes on cores. But stil, it is very informative.

Do not use WIndows taskmanager and others sh*ts! It is not tweaked software for monitoring.

HWINFO software to download

AMD chipset driver (new X570)

BIOS 0808 for PRIME X570-PRO

If u need more, please write me email: simonek.michal@seznam.cz

[hardcoregames_]

My R5 2400G boost works fine. But my box is well cooled too.

[s_vayner]

this issue is with new Ryzen 9 processors i don't know about previous generation never used AMD till now.

[hardcoregames_]

s_vayner wrote:

this issue is with new Ryzen 9 processors i don't know about previous generation never used AMD till now.

Maybe you need to get a water cooling unit?

[shinkojiro]

PB2 just inst working very well with 3900x unfortunantly. It will take AGESA fixes to get it right. 4.6 on a whole CCX is certainly possible as I run 4.6 on CCX0 and 4.4 on all others on the norm using Ryzen Master, which allows those low-threaded tasks to function on higher clocks. Its also possible its windows scheduler once again doing its thing all wrong...

[trcoley]

shinkojiro‌ would you mind posting a screenshot of your OC parameters?  What voltage are you using?

[brucer]

Many are having this issue..  Its on amd and board manufacturers.. Amd still doesnt have the agesa code right...  Just another ongoing issue with amd and their early releases and poor support.  Welcome to amd..

I have an 3800x/Asrock x370 and have to run pbo to get it to hit the advertised boost and I'm running an air cooler and I can do an all core overclock of 4.4ghz and it doesnt get over 52c at full load, so it is not a temperature issue, my air cooler runs as well as a 240/280 radiator..

It is a bios/architecture and I feel voltage issue..  I'm sure amd knows about it, they are just so slow to resolve the issue and relay it to board manufacturers..

 I literally have the same board as conraire above does and I had to get a beta bios that wasnt listed on asrocks site and install the amd chipset drivers from amd's site and I still had to fight with it to get to the advertised boost and it still rarely hits it..

 I've even broke down and ordered a $300 x570 Taichi to try to solve the issue..

 A bigger issue is why are these Ryzen 3000 series cpu's pulling so much voltage at idle.. Its ridiculous and I'm almost certain it will greatly shorten the lifespan of the cpu. I also bet these Ryzens are not running at their rated tdp and no where near it..

[hardcoregames_]

MSI has had issues with the BIOS chip not having enough room for the latest AGESA updates so the A320, B450 and X470 models are now being revised to have 32MB BIOS chips which will have enough room. 

Part of the problem is that vendors have used too much space for value added features

MSI is calling the updated motherboards MAX so that consumers will be able to identify them easily. They should be available in a few weeks.

[s_vayner]

I've tried downvolting it and it actually helped with the boost a little i was able to get another 100Mhz out of it and a better benchmark score. and brought my voltage to acceptable 1.35v to 1.4v range and lowered temps even more. Before it would hit 1.5v almost constantly and 78deg at full load. I downvolted by -.075 with LLC set at 3 just to counteract higher load demands and i get 4.3ghz boost now and temps in low 70s under full load. I've tried offset of -0.1 but it BSODs almost instantly. I'm very hesitant to set any positive offset given this cpu already runs at 1.5v stock with max being 1.6v. I might try it just to see if it makes any difference in boost.

[ajlueke]

The OP is correct, that the advertised base/boost clocks should work out of the box and stock settings with the stock cooler. 

Now, AMD has said that not all of the cores on a particular CPU can hit the advertised boost clock, but at least one of them can.  You should check Ryzen Master and look at the boost clocks listed for each core.  One of them, on a 3900X should be capable of 4.6 GHz.  When you run a single threaded task, Windows should pin the tasks to that core preferentially.  This only works correctly with the latest Windows 10 builds, so if you haven't gone to the latest Windows 10 update, thread scheduling may not work correctly. 

If you have, and Windows is putting the workload on your fastest core, and it still doesn't hit 4.6 GHz, then I think you have a case to RMA that CPU and get a replacement.

[s_vayner]

yep got latest build of win 10, latest AMD Chipset drivers, 3AB bios, using amd power profile... I used AMD Master to check which single core gets prioritized and its the fastest core in the CC (the gold star one) so yeah basically this CPU doesn't perform to advertised specs and is about 10% lower. 

[sqwerty]

I've seen quite a few reports of people with lower boost clocks after updating their BIOS that their X570 shipped with (AGESA 1.0.0.1 or 1.0.0.2) to one with AGESA 1.0.0.3AB BIOS (which is funny considering this release was supposed to "improve system performance".

The fact conraire had to add LLC to get his CPU stable at stock shows AMD have pushed the silicon to the edge, if even his board (which I'm guessing from the price is the likes of CH6 or Taichi) has enough vdroop to BSOD then many lesser boards are going to suffer the same or worse (though I doubt many low end board owners bought 3800X...).

If you are running latest BIOS OP try an earlier one, if you don't want to do that then see what you can get with all core manual OC, or what seems to be popular with 3900X; per ccx OC.

hardcoregames™ : my 2600 boost worked fine and as expected, my 3700X not so much.

They've been pretty scummy with these imho, binning chiplets whereby if even a single core can hit the boost clock @1.5 VID for 2ms (and then swiftly throttle so as not to instantly self destruct) it's a pass.

Running auto settings and hw logging might reveal my 3700X hit 4390MHz on two of it's cores, at some time or another, but it don't mean jack when a 4300MHz fixed clock speed performs better because it doesn't throttle.

[s_vayner]

Exactly how i feel about it, AMD really put up some real false advertisement here. All intel chips run sustained advertised boost clocks as long as you stay in single core or low load territory. Also does this open possibility for RMA given that i have never seen my 3900x hit 4.4Ghz even for nanosecond never mind 4.6. possibly even class action lawsuit as well if they dont fix it. Regardless of outcome, long term they will lose more customers, I will not be coming back for any AMD products ever again after this. 

[sqwerty]

Well 4.2 max for a 3900X is clearly not right, did you check core performance boost is enabled in the bios? is the power plan selected allowing 100% CPU Maximum Processor state under the Processor Power Management tab? Did you check the voltages with hwinfo both requested voltage (VID) and actual (CPU Core Voltage (SVI2 TFN))

If everything looks good and you don't want to do any bios flashing or manual/ccx OC testing then yes just ask for a return.

[brucer]

Thats not the issue..

 The issue is the cpu is not hitting advertised boost clocks out of the box and its a well known issue..  A person shouldnt have to jump through hoops to install a cpu and have it hit its advertised speeds.. It should be removed from retail package, install it, load windows and drivers and it be there. This is just typical amd releasing products too soon, rushed to market with piss poor support, basically a BETA Product.. Yes this is on amd and their agesa code.. It'll take months for them to get this resolved as they'll have to update the agesa code, then relay that to board manufacturers then the board manufacturers to update the bios' on any various motherboard x370, x470 and x570 platforms..

In my opinion Amd is a slouch of a company, releasing chipset driver updates in a community news post.. really?

[sqwerty]

What's not the issue..? All of those things I said to check can reduce or completely remove boost operation..

I know full well it should "just work" but I was trying to offer suggestions of things for OP to try before resorting to filing for an RMA.

My 3700X also doesn't boost too well, but thankfully we can still tinker manually and they even offer a nice tool for it in windows (Ryzen Master). With it I can set one full CCX (the one with the gold star core in it) to 4.4GHz and the other CCX to 4.3GHz and (voltage is 1.35v no LLC), that's better than what they said it could do (and better than it can do in it's default "out of the box" state).

It's a similar story for my 1080Ti, manually setting a voltage/clock and fan curve results in better overall performance due to being able to run a solid 2GHz at <1v, where as stock it might boost a bit higher but with a load more voltage; it just ends up throttling until it's under 2GHz eventually anyway.

[s_vayner]

yeah unfortunately none of it fixed my boost performance, i was able to get extra 100mhz by setting offset to voltage -.025. but those spikes are very short lived. i know AMD released new chipset driver today will try that see if it makes any difference.

[ajlueke]

Unfortunately, I don't have a 3000 series to test with.  It does look like there are reports where everything is working fine, ie Techpowerup's review.

But there are a number of posts describing the issues s_vayner‌ described.  I don't really see a common thread between those that work and those that don't (motherboard manufacturer, Windows 10 version, driver versions).  So it is really hard to say what is causing the issue for those that have it, without going out and grabbing a 3000 series myself.

I will say, that you should be using the standard windows "balanced power plan".  If you are using the "Ryzen balanced" make sure to set the CPU minimum state to 5%.  It is set high by default and that can mess up the entire core boost algorithm.

[brucer]

and what you suggested would void any and all warranty, as soon as you enable pbo or set manual overclocking.. I dotn particularly care about the warranty, as I've manually overcloked mine to 4.4ghz all core.. but that still doesnt resolve the issue of it not hitting its advertised clock frequencies at default settings.. some people do not mess with the bios, they want to plug a cpu in and run it and not mess with it and get what they pay for!

 you know whats funny, the ccx with the gold star is one of my ccx's that never hit the boost frequency.. My best ccx is supposedly #6, yet its two and 5 that always boost the highest and its usually about 4.475ghz not 4.5ghz..

My 3800x is also suppose to run the low clockspeed of 3.9ghz, but it only does 3.6ghz and sometimes its even lower than that..

[david5343]

That is amazing. Did you passed P95 stressed test with 1.35V on 4.4&4.3GHz? 

Mine wont stable even on 4.3GHz with 1.43V. Only managed 4.25GHz on 1.41V all core to be able to passed Cinebench and 30minutes P95.

Default all core boost not even hit 4.2GHz... It's a joke.

[hardcoregames_]

AMD Ryzen Master shows the actual clock speed properly and I see my processor bouncing around as expected.

https://www.amd.com/en/technologies/ryzen-master

[sqwerty]

Edit: Deleted

[hardcoregames_]

I figure the overkill dual fan top count water cooler has to be able to get the most out of a CPU, if that cannot do it then maybe your thermal material sucks?

[conraire]

I should have my X570 MEG Ace some time this afternoon.  After which I'll be rebuilding my system, and doing a clean install of windows.  Then I should be able to tinker more with my 3800x, and see if I can get max boost under Ryzen master to report correctly. 

Some issues I've seen, are which software do you believe for boost frequency?  Ryzen master reports lower freqs than hwinfo64.

The Voltages and slightly high single thread/lightly threaded temps are apparently normal, due to the way the boost algorithm works, basically how aggressive it is.  This also has to do with power density on the Core complex dies.  If you're running 1 core at full boost, it'll run at 1.4-1.5v though with low current draw.  On the other hand a full thread load will run at a much lower voltage. 

Buildzoids recent video explains a ton about whats going on.

Checking out the boost behaviour of a Ryzen 3700X with an oscilloscope - YouTube 

[brucer]

Conraire.. I'm already watching that video myself..

 I was using the cpuid monitoring tool and also ryzen master, both were showing a little low..

 I'm thinking the low boost is happening on x370 and x470 boards mainly, and due to amd and the manufacturers being slow to update or not getting the bios updated correctly in a timely manner.. It could also be in poor voltage regulators on various boards not being up to the task to control ryzen3000...

[Ray_AMD]

We'd like to work with you to troubleshoot your issue.

Please open a service request so that we can work on on one on it.

[conraire]

I'm wondering if it's not a problem with Ryzen Master.  I'm seeing proper voltages in CPU-z, and boost speeds up to 4.55Ghz.  But in Ryzen master, it's not boosting above 4.4Ghz, 100mhz short of the advertised boost, even on the 4 fastest cores..  On the plus side, my CPU's cores all seem to be able to boost to 4.4Ghz while running cinebench R15/R20.  Temps are also much better under the new drivers, Idle in the mid 30s.  Mid 50c under single core load, and mid  70c under all core at just past 4.2Ghz. 

[s_vayner]

it seems you have won the silicon lottery my friend, unfortunately for every person that does there are 100's that don't.

[xlox]

i think you just don't understand the way this cpu works and delivers it's power , and you try to compare to intel , but it's 2 different ways of analysing the parameters of the cpu in my view... I had a 4930k that I OCed to 4.6Ghz (MANUALLY)

the difference now is my 3700x OC itself to 4.3Ghz (ALL CORES ! , 4.4 for 3700x for example IS ON ONE CORE , you should read reviews before buying (and it seem others here should too insteand of telling wrong things to mislead people) maybe you are a little bit faulty in that case, every reviews tells it , there is no lie . And as it is about OC , everything depends on a good cooler too and that's ok, it always have been like that ... i'm thinking that reviewer testing a 12cores without an average watercooling AIO (that's only what it needs) are just trying to lower the potential of that cpu, that's my point of view) ...

even the way you apply the thermal paste can affect the potential of an oc

instead of beeing concentrated on Ghz you should concentrate on performances these beasts delivers ... my 3700x out of the box is way faster than my 4.6ghz OCed 4930k even just in games ..

in fact you are complaining about something you don't really understand ...

the only thing i see that is weird is the fact you tell if you deactivate PBO , if i deactivate PBO my cpu doesn't OC over 3.8Ghz if I remember that's the difference, but that's a bios problem related to your motherboard

if you don't want your 3900x I would had loved to buy it from you if i didn't bought my 3700x before reading your comment, as 3900x are unavailable from the start in my usual store , and i took a 3700x because of that  lol

[s_vayner]

So let me get this straight: if I go to a bakery and order 10 bagels, pay for it and only get 9 then according to you that's ok. Because i should have read reviews of that bakery and educated myself that their bagels are so much better than anyone else's and also to get the 10th bagel i should have bought a premium cream cheese to go with those delicious works of art. But the funny fact is that most of you AMD fanboys would have an involuntary bowel movement if that hypothetical scenario happened to you. But when you get shafted by a corporate conglomerate and someone calls them out on it, somehow that person becomes the enemy of all good in this world? So thank you for that elaborate exercise in mental gymnastics, you'd get a gold if that was a sport btw, but still i don't agree with your premise. Advertised speed is 4.6GHz and that's that. CPU should hit it with no fail, and sustain it.

[xlox]

first you compare intel and amd , now you compare amd with a bagel ..still not the right way to analyse ... this 4.6Ghz is "up to" so there is a context that you refuse to understand

you only see the shape and not the substance, you should use the term of tolerance maybe in that case ... your bagel comparison is wrong

you complaining about 97% of the bagel , but the fact is that with a good cooler you should be able to achieve what you want , or maybe you think it's magical and there is no constraint ?

[s_vayner]

I am merely illustrating a concept of getting what you were promised when you paid for it. But if you saying my comparison is wrong because it's one cpu vs many bagels I guess you'd be ok with one bagel that someone took a bite off already? Also it's not 97% its 90%. i get 4.3 momentary boost not 4.6 and that only happens after incessant tweaking, carefully tailored benchmark and proper planetary alignment every second Thursday of the month . Real world usage: I've never seen boost past 4.2 on any single core. My benchmarks, even though very impressive, are still not on par with reviews and advertised speeds. So yes AMD owes me premium for that extra 10% that i didn't get. And most of the time that extra 10% translates to 30% price increase.

[xlox]

i don't know why you don't understand : 4,6 is on 1 core 1 thread ..

i have 3700x and never got over 4300mhz ... and that's ok because windows always use more than 1 core so it's a context

maybe if you can deactivate all core except 1 you will get you're 4.6ghz ... the fact is that i bought this cpu for the ipc gain and not for more or less 200 mhz and i get my ipc gain, maybe you should know why you buy something ... i think you should rma you re cpu because you will always be frustrated because you don't want to understand, ... one more time, you ve got the intel point of view

in fact it's not because you didn't see it going to 4.6 that it is not capable of it