Definitive guide to configuring the Ryzen 3900X/3950X and all other 3000 Series CPUs

[moproblems99]

Hi

I joined specifically to thank you for your terrific post.

I cannot believe all of the negative reaction to your post. It's disappointing and there is absolutely no need for it.

People should be offering additional information - not attacking a post that was fantastic and intended to help people.

Regards

Johnno

I agree it is a good post and offers a step by step of how to make it happen. HOWEVER, calling it The definitive guide was a poor choice of words. I can't even call it the best method. It worked for HIM and that is great. But he seems to have issues that no one else seems to have and begs the question if this could be done more permanently through the UEFI. Not to mention you lose about 5% of single core performance which according to everyone with an Intel CPU is all that matters. Single does matter, specifically with gaming. It will be less of a thing in the future but it is important still.

I agree with you that his post and process getting shit all over was unwarranted and should have been avoided but the blame is spread on that.

Edit: On a side note, you can get 98% of his multicore performance and keep all single thread performance simply by adjusting a few simple UEFI settings.

 

[mtcn77]

Edit: On a side note, you can get 98% of his multicore performance and keep all single thread performance simply by adjusting a few simple UEFI settings.

As an observational study which would they be?

 

[moproblems99]

As an observational study which would they be?

The biggest ones for me have been LLC and limiting EDC.

 

[mtcn77]

I didn't even know TDC worked.

 

[moproblems99]

I didn't even know TDC worked.

My mistake. EDC.

 

[mtcn77]

Ah, the rancorous debate about EDC. Time to abandon this thread...
Other best practices that I am aware of, but too intricate to fulfill;

1. IHS lapping,
2. Liquid metal paste,
3. Tighter mounting pressure,
4. Processor affinity,
5. Undervolt balance.

There is a way to register the same multicore taper target using edc alone, but that level of manual control - 4.1-4.3 ghz all core - is not yet achieved by anybody.
This person be the judge - setting 'off' manually the stock settings make the 3600 run at its base clock - so with that in mind, losing a couple points for a set voltage threshold that makes the biggest impact in operating temperature is acceptable in my book.

My findings:

Everything at stock: Cinebench score 3437, temps around 75
Ryzen Master manual, no changes:
Cinebench score 3230, temps around 75
Ryzen Master manual, 1.025 peak voltage: Cinebench score 3208, temps around 53

 

[Michael Nager]

I agree it is a good post and offers a step by step of how to make it happen. HOWEVER, calling it The definitive guide was a poor choice of words. I can't even call it the best method. It worked for HIM and that is great. But he seems to have issues that no one else seems to have and begs the question if this could be done more permanently through the UEFI. Not to mention you lose about 5% of single core performance which according to everyone with an Intel CPU is all that matters. Single does matter, specifically with gaming. It will be less of a thing in the future but it is important still.

I agree with you that his post and process getting shit all over was unwarranted and should have been avoided but the blame is spread on that.

Edit: On a side note, you can get 98% of his multicore performance and keep all single thread performance simply by adjusting a few simple UEFI settings.

In one sense of the word, "Definitive", as in "The absolute last word" I could agree with you, and that was not the definition which I had in mind when I wrote that.

By "Definitive" I was thinking of the universality of my solution which takes the vagaries of which brand of motherboard and the quality - or lack thereof - of the motherboard BIOS and of course if you are dealing with an X370, X470 or X570 or the eqivalent B chipset. What I wrote is valid and relevant to everybody.

Also the way I wrote it, I defined exactly which steps to take to get to the result and it is complete from start to finish.

There is nothing vague or wishy-washy about what I wrote.

The only thing that I have found is that with an ASUS X570 board you have to set the D.O.C.P if you want to change anything on the fly in Ryzen Master (like the Vcore voltage or clockspeed etc) in Ryzen Master without having to reboot.

 

[moproblems99]

In one sense of the word, "Definitive", as in "The absolute last word" I could agree with you, and that was not the definition which I had in mind when I wrote that.

By "Definitive" I was thinking of the universality of my solution which takes the vagaries of which brand of motherboard and the quality - or lack thereof - of the motherboard BIOS and of course if you are dealing with an X370, X470 or X570 or the eqivalent B chipset. What I wrote is valid and relevant to everybody.

Also the way I wrote it, I defined exactly which steps to take to get to the result and it is complete from start to finish.

There is nothing vague or wishy-washy about what I wrote.

The only thing that I have found is that with an ASUS X570 board you have to set the D.O.C.P if you want to change anything on the fly in Ryzen Master (like the Vcore voltage or clockspeed etc) in Ryzen Master without having to reboot.

I think most of the friction stemmed from the fact no one understood your definition definitive. Certainly not what I thought it was but then, again, I don't feel like coughing up the low-thread performance for just a little bit more all-core. But to their own.

 

[Michael Nager]

Ah, the rancorous debate about EDC. Time to abandon this thread...
Other best practices that I am aware of, but too intricate to fulfill;

1. IHS lapping,
2. Liquid metal paste,
3. Tighter mounting pressure,
4. Processor affinity,
5. Undervolt balance.

There is a way to register the same multicore taper target using edc alone, but that level of manual control - 4.1-4.3 ghz all core - is not yet achieved by anybody.
This person be the judge - setting 'off' manually the stock settings make the 3600 run at its base clock - so with that in mind, losing a couple points for a set voltage threshold that makes the biggest impact in operating temperature is acceptable in my book.

In my experimentation with the Ryzen R5 3600X I had the following result running Prime95 with small FFTs for over three hours using the Noctua NH-U12A air cooler:

Running CineBench at 4.225 GHz:

The CineBench result was:

I think most of the friction stemmed from the fact no one understood your definition definitive. Certainly not what I thought it was but then, again, I don't feel like coughing up the low-thread performance for just a little bit more all-core. But to their own.

The other thing is that English (although I am a native speaker of it) is my second language, German being my first.

In German we use the word "Definitiv" to mean "without ambiguity" or "straightforward".

Which is another reason why that word naturally occurred to me when I was creating a title for the thread.

 

[Michael Nager]

I think most of the friction stemmed from the fact no one understood your definition definitive. Certainly not what I thought it was but then, again, I don't feel like coughing up the low-thread performance for just a little bit more all-core. But to their own.

Not to put too fine a point on it, but I really am getting sick and tired of the "Single Core Performance" bullshit.

When does your system ever run on a single core outside of artificially constraining CineBench R20 or some other benchmark to one core?

As soon as you get more cores involved the performance of the Ryzen 3rd Gen CPUs tanks.

Ryzen 3000 CPUs will boost when there is fuck all to do, and you know, that's really when I would want my CPU to boost to the maximum, when it is doing ABSOLUTELY NOTHING!

Have you taken a look at how much voltage and wattage is punted into the CPU when it is doing a single core workload?

The first time I saw it, I thought to myself, "You have to be fucking kidding me".

Because of the Tech Media and Tech YouTubers - professional shills in other words - bringing up this bogus long outdated measurement metric that people like you parrot it as if it is supposed to have any meaning. It is an Intel talking point; however, that being said, when you run something like CineBench R20 all core with a Ryzen 3rd Gen CPU the clockspeed of the individual cores goes down, even though a disproportionate amount of voltage is being applied.

But let's just see what "Single Core advantage" we actually get when I run the system constrained to say two cores (with SMT Off) and run CineBench R20 with the single core benchmark - it should then be able to boost gangbusters don't you think?

WOW I am getting a clockspeed of 4.725 MHz on Core 0 (which you can see from the voltage it is drawing is doing fuck all) and Core 1 did manage a maximum of 4.7 GHz, but is at the time the snapshot was taken running at 3.4 GHz, but using 1.425 volts to get there.

At 4.7 GHz the CineBench single core Benchmark must have been phenomenal:

Yeah, not so much.

So what about two Cores with four Threads (i.e. two Cores with SMT On):

Still hitting that magical 4.725 GHz maximum clockspeed on on of the Cores and 4.675 on the other. This time you can see that both Cores are doing something and managing a dizzying 3.4 GHz at 1.44 Volts - what a staggering return on investment.

But surely this time, with CineBench R20 getting a clear shot at one Core on its single Core benchmark with all MUST be killing it right?

Yeah, still a bit underwhelming for 4.725 GHz dontcha think?

But surely if we run the system with all Cores and Threads enabled the results will make me and my guide look foolish:

Well the average voltage has certainly gone up and the Cores were still hitting 4.5 GHz so this score has to be stellar:

Are you impressed? I certainly am not.

All the above tests were conducted with the system at stock, with no input from Ryzen Master whatsoever.

So now we will see how much I lose on the single core performance compared to what I gain on the all core performance if I configure the system my way, but first the voltages:

The voltages are looking a hell of a lot healthier, but at only 4.3 GHz maximum the CineBench R20 and especially - according to you - the single Core score must be abysmal. Let's just see shall we?

WHAT A CATASTROPHE!! I LOST 2.53% ON SINGLE CORE PERFORMANCE, NOTHING CAN MAKE UP FOR THAT!!!

That can barely be ameliorated by the fact that the all Core result is 12.35% higher.

I am obviously a delusional person and need to find the nearest bottle of bleach to drink to rid humanity of my foolishness.

You will be want to say, "Surely you can't be satisfied with that?".

To which I reply, "You are correct, and stop calling me Shirley".

I don't really need the 32 Threads except when I am editing a video for instance, for games and most other things 32 Threads are either of no use, or can be detrimental. So why not run the CPU with SMT Off as a straight 16 Core CPU?

First let's look at the voltages and the clockspeeds

I upgraded to the newest version of HWInfo since doing the previous screenshots and it isn't doing a great job of reading the voltages not that it was doing a brilliant job before where you see a maximum of 1.306 as a maximum where the maximum was set for 1.3 (as it was this time as well).

But let's get to the meat and potatoes, the benchmark result:

The advantage, aside from the insignificance of the single core score, is that it runs a lot cooler in this mode.

 

[moproblems99]

Not to put too fine a point on it, but I really am getting sick and tired of the "Single Core Performance" bullshit.

So single thread is a little bit a misnomer. You'll see later on I relabeled it as low thread. 3900x and 3950x spend a lot of time in low thread. That is why low thread is important.

 

[Michael Nager]

So single thread is a little bit a misnomer. You'll see later on I relabeled it as low thread. 3900x and 3950x spend a lot of time in low thread. That is why low thread is important.

Which is also the reason why I run my 3950X mostly with SMT disabled as a straight 16 Core 16 Thread CPU.

The thing is that when I saw "single Core (Thread) performance" I was pretty sure that you could not be stupid enough to believe that and that you must be trying to wind me up.

The thing is that I have put a lot of thought and a lot of work into my new rig and how it is configured.

The CPU (and the rest of the system) is rewarding me for my efforts and I decided to share my experience with people; which the other "Subscribe", "Donate to my PayPal", "Join my Patreon", "Bitch about Ad revenue", motherloving lazy (or stupid) bastards who get their shit for free on top of anything else have not seen fit to do.

I am not counting the likes of Buildzoid, Der8auer or Gamers Nexus to that crowd. Their focus is more on extreme overclocking and my methodology would probably be unworkable for them, they also work hard at what they do, and don't indulge in "Access Journalism" or outright Shilling the way that the aforementioned above do.

For instance a site called "Hardware Unboxed" got a 3900X and managed to fry it when they put it in an unnamed motherboard (I am going to bet on ASUS) and yet this didn't prompt them to look more closely.

All they did make it so that this didn't happen in future - the immediate frying that is - but they have done nothing with regard to looking into the degradation of the 3rd Gen chips they have over time.

 

[moproblems99]

Which is also the reason why I run my 3950X mostly with SMT disabled as a straight 16 Core 16 Thread CPU.

The thing is that when I saw "single core (Thread) performance" I was pretty sure that you could not be stupid enough to believe that and that you must be trying to wind me up.

The thing is that I have put a lot of thought and a lot of work into my new rig and how it is configured.

The CPU (and the rest of the system) is rewarding me for my efforts and I decided to share my experience with people; which the other "Subscribe", "Donate to my PayPal", "Join my Patreon", "Bitch about Ad revenue", motherloving lazy (or stupid) bastards who get their shit for free on top of anything else have not seen fit to do.

I am not counting the likes of Buildzoid, Der8auer or Gamers Nexus to that crowd. Their focus is more on extreme overclocking and my methodology would probably be unworkable for them, they also work hard at what they do, and don't indulge in "Access Journalism" or outright Shilling the way that the aforementioned above do.

For instance a site called "Hardware Unboxed" got a 3900X and managed to fry it when they put it in an unnamed motherboard (I am going to bet on ASUS) and yet this didn't prompt them to look more closely.

All they did make it so that this didn't happen in future - the immediate frying that is - but they have done nothing with regard to looking into the degradation of the 3rd Gen chips they have over time.

To each their own. I'm not particularly worried about the high voltage. Only time will tell.

 

[Michael Nager]

To each their own. I'm not particularly worried about the high voltage. Only time will tell.

$750 is a lot of money to just piss away.

 

[moproblems99]

$750 is a lot of money to just piss away.

You are assuming there is actually something wrong.

 

[Michael Nager]

You are assuming there is actually something wrong.

I don't make assumptions.

I do however tend to err on the side of caution, if I err at all.

I did a lot of reading about the 7nm Node before AMD brought out the 3rd Gen, and I also know AMD's track record with voltage and clockspeed - GCN comes to mind for instance.

 

[moproblems99]

I don't make assumptions.

I do however tend to err on the side of caution, if I err at all.

I did a lot of reading about the 7nm Node before AMD brought out the 3rd Gen, and I also know AMD's track record with voltage and clockspeed - GCN comes to mind for instance.

I apologize if my previous comments in the beginning of the thread came off as condescending as they weren't meant to be. I am glad you found a way that works you. Different strokes for different folks. I am mostly fine letting Ryzen do its thing. When I get more time to really play around with it, I may. If I could get a constant 4.5+ all-core, I would be tempted to try.

 

[lorry]

I apologize if my previous comments in the beginning of the thread came off as condescending as they weren't meant to be. I am glad you found a way that works you. Different strokes for different folks. I am mostly fine letting Ryzen do its thing. When I get more time to really play around with it, I may. If I could get a constant 4.5+ all-core, I would be tempted to try.

Who wouldn't though

 

[Michael Nager]

I apologize if my previous comments in the beginning of the thread came off as condescending as they weren't meant to be. I am glad you found a way that works you. Different strokes for different folks. I am mostly fine letting Ryzen do its thing. When I get more time to really play around with it, I may. If I could get a constant 4.5+ all-core, I would be tempted to try.

I won't deny that the way I achieve my goal is a bit of a kludge - it could be worse though - and an improvement would be if Ryzen Master would implement a feature to automatically load the last used profile on the next boot.

Truth be told I was a bit surprised at the hostility that confronted me here, because from what I have seen there has been no suggestion or advice that I have been able to find that kills as many birds with one stone for me, and others, as the one I developed for myself.

Buildzoid's videos did set me off a bit. When watching them I was thinking, "That's all well and good, but you are just bunging a band-aid on a gaping wound".

What I find most unfair about the whole situation is that someone looking for advice on how to configure their systems are getting all kinds of shit told to them, and if they are inexperienced they will be in a world of pain if someone who sees advice for one motherboard manufacturer and tries to apply that advice to a motherboard from another manufacturer.

You have the situation, where BIOS options (for instance with ASRock) don't even work if you set them.

The other thing is that monitoring software just doesn't work properly when some motherboard options - particularly voltage options - are set in the BIOS.

So the guide I wrote is not a solution looking for a problem, the problem is real and it exists.

I help a lot of people who are building their own systems (because they don't have a lot of money) but don't have much of a clue even about that - and that's pretty much the easiest part to do.

You and I can talk about the LLC, but there are many more people out there who, when I say, "Bung the Vcore LLC and SOC LLC onto the third highest value", they ask, "Where?" and I will answer "Under the Tweaker Option in the BIOS" and they ask, "What's a BIOS? Is that something I have to plug into the computer? Because I didn't get one of them with the computer".

I won't deny there have been a number of times where I have felt myself losing the will to live in these kind of discussions.

I talked to one guy yesterday from Greece who contacted me on a Discord Server after someone told him about me, and he has built a 3900X system but it kept shutting down. I can't remember exactly what he was using it for (he told me, but I wasn't interested or listening), but it was something that made use of the threads.

The thing is that he bought the CPU and a high quality motherboard (ASUS X570 Crosshair VIII Hero WiFi) and wanted to make do with the stock cooler until he could afford something better. He had run the gauntlet of getting advice from other people and watching YouTube videos and was both somewhat frantic as well as being confused, disillusioned and feeling pretty bad.

Turns out that his CPU was overheating using the APP and shutting down - never mind throttling.

So I walked him through what I basically wrote in the original post, but luckily I had talked to someone a few days ago with an ASUS board and to be able to make changes on the fly to a profile without needing to reboot, you have to set the DOCP (which gives me problems when I set XMP on my GigaByte motherboard, so the exact opposite situation).

Anyway, after walking him through it, with the application running he is hitting 73-75 °C.

I couldn't tweak it to the max (I think we got it to 4.225 GHz all core), but with a profile set for turning SMT Off he can run games at 4.375 GHz (which is all his system would run stable at). The reason for this I suspect is that although he got the same RAM from G.Skill (4x8 GB in two different 2x8 GB packs) one of the sets of RAM had Hynix chips and the other Samsung. I find that a really shitty move by G.Skill.

It was a lot of work and took me about four or so hours on voice to walk him through it and of course test it.

 

[lorry]

thing is that there are so many 'experts' out there and so many new people with high spec setups, I'm one myself, although I do have old knowledge. The cost of building a quite high end setup has come down a fair bit in recent years and from my recent experience the building side of it is pretty much Lego skillset these days.

On a side note, you can be a Great, even The best player in the world at something, but a crappy teacher. it's not as easy the other way around though.

 

[moproblems99]

I won't deny that the way I achieve my goal is a bit of a kludge - it could be worse though

No denying that, there are likely still worse ways. I still think you have something else going on keeping things from working the way you expect. But I am with you on just getting things working and moving on at times.

Truth be told I was a bit surprised at the hostility that confronted me here, because from what I have seen there has been no suggestion or advice that I have been able to find that kills as many birds with one stone for me, and others, as the one I developed for myself.

Well, I think a lot of it was mostly a misunderstanding but there are a lot of "experts" here and don't like to except there are alternate ways of doing things.

Buildzoid's videos did set me off a bit. When watching them I was thinking, "That's all well and good, but you are just bunging a band-aid on a gaping wound".

I think that is how a lot of people feel about your method. That said, AMD made Ryzen master for some reason - either because board partners make shitty UEFIs or Ryzen is more finicky than it needs to be.

What I find most unfair about the whole situation is that someone looking for advice on how to configure their systems are getting all kinds of shit told to them, and if they are inexperienced they will be in a world of pain if someone who sees advice for one motherboard manufacturer and tries to apply that advice to a motherboard from another manufacturer.

Well, there isn't much that can be done about that. At some point, a user just has to understand what they are dealing with or they will just get what they get from the factory. In that sense, Ryzen is not a bad choice as you get pretty much what you are gonna get from the factory (Let's pretend these last five pages of discussion didn't occur).

You have the situation, where BIOS options (for instance with ASRock) don't even work if you set them.

Honestly, I can't say I have experienced that. With the exception for RAM settings that must be done through the tweaker section. Which sucks because the other areas are laid out better than the tweaker secion.

 

[RealNeil]

I am mostly fine letting Ryzen do its thing.

I'll probably do the same with my 3800X setup once I build it. If it works, don't fix it.

It's what I'm doing with my i9-9900K machine and it's doing fine.

 

[Michael Nager]

Honestly, I can't say I have experienced that. With the exception for RAM settings that must be done through the tweaker section. Which sucks because the other areas are laid out better than the tweaker secion.

To your points:

No denying that, there are likely still worse ways. I still think you have something else going on keeping things from working the way you expect. But I am with you on just getting things working and moving on at times.

I think that is how a lot of people feel about your method. That said, AMD made Ryzen master for some reason - either because board partners make shitty UEFIs or Ryzen is more finicky than it needs to be.

I think we can kill two birds with one stone here.

The conclusion that I came to - and that was my initial breakthough in cracking my problems - was to only use what we would classically consider to be the BIOS to configure the motherboard, and nothing else. To configure the CPU and RAM I used the real BIOS, namely the AGESA portion contained in "AMD CBS" and "AMD Overclocking" I switched to Ryzen Master, which is a lot more convenient for configuring the RAM because to do it manually in the "AMD Overclocking" and then changing the RAM timings under "DDR and Infinity Fabric Frequency/Timings" you have to enter the values in hexadecimal.

"Load the XMP profile", I hear you say, "Nothing could be simpler than that", I hear you say.

Yeah, about that, if I load the XMP profile, with a clean BIOS then the first thing I get are the three beeps (I am old-school and although motherboard manufacturers no longer see fit to include a little speaker, I have one installed, because fuck little LEDs or BIOS code displays, just hearing the series of beeps allows me to pinpoint the source of problems a shit-ton easier) that tell me there is a RAM problem.

This is followed by a number of sometimes hangs, sometimes a clean series of reboots, aka "RAM Training" until the system boots into the OS. So it seems to me that the SPD of the RAM is read by the BIOS (which I call "GigaByte Master") and then that attempts to update the real BIOS, namely the AGESA, and there is a negotiation back and forth.

Sometimes it just results in the system shitting itself and giving me the "Cleared CMOS" message and I have to go in and load the XMP profile again - rinse and repeat the procedure outlined in the previous paragraph.

What is certain, at least with the motherboards I have worked with, that setting the XMP profile IS NOT what configures the RAM, but rather it is the negotiation between setting the XMP profile in the BIOS and then what the RAM timing portion of AGESA agrees with is important.

For instance, if I set XMP and then go through the "RAM Training" I get a value for tRFC of 648 although the XMP profile of my RAM states 312. A number of other timings are worsened.

The thing is though, if I copy out the SPD values displayed by loading the XMP profile, then turn XMP off again, load Windows and put the SPD values into Ryzen Master, it accepts and applies the values with no problems whatsoever.

My way of doing it is admittedly kludgy, but it does have one thing going for it, IT WORKS!

For any benchmark you care to name, Buildzoid, in a non-exotic cooling scenario (LN2), will not be able to match any score I can achieve given his methodology for overclocking his 3950X (aside from a single core benchmark, which is essentially meaningless and only being clung to by Intel because they have nothing else) and that I can not only beat his scores easily, but beat them with less voltage and lower temperature.

There are a lot of steps I have taken along the way and you reminded me of the time I spent months ago trying to work out what the hell was happening with regard to the RAM problems I was having.

I am not saying this to pat myself on the back, and if anyone can show me a better way of achieving my goals, I will abandon my methodology in a heartbeat. I am not wedded to it. But looking around, there is nothing, a gaping vacuum.

Well, I think a lot of it was mostly a misunderstanding but there are a lot of "experts" here and don't like to except there are alternate ways of doing things.

Back in the mid 1990's I set up a network for one of the biggest porn producers in Germany. After getting it all set up, and writing a database solution for her, because the one she had was a total rip-off, I was invited to a party she was giving.

And no, it was a party just like any other, not what you are thinking.

At the party she introduced me to people as, "My computer expert", and this made me very uncomfortable, and I explained to those people that I was not then nor could I ever be an "Expert", because just when I start to think I know what the hell I am doing, things change and in many cases my current knowledge is more of a hindrance than benefit moving into the future and that "Permanent Scholar" was the highest title I could aspire to.

How much do I know about computers?

Well I'd like you to stretch out your arms as far out from your body as you can, now bring your thumb and your forefinger together so that they are as close to touching as you can. Now look at the distance between your thumb and your forefinger, that is how much I think I know, and the distance of the spread of your arms is how much I feel woefully ignorant about in terms of computer knowledge in comparison. That's not really a great answer from having been a techie for 38 years when someone asks me, "What do you know about computers?", and I have to answer, "Essentially fuck all".

Well, there isn't much that can be done about that. At some point, a user just has to understand what they are dealing with or they will just get what they get from the factory. In that sense, Ryzen is not a bad choice as you get pretty much what you are gonna get from the factory (Let's pretend these last five pages of discussion didn't occur).

That would be all well and good except for the fact that they are leaving a hell of a lot of performance on the table. What worries me more though is that you are left at the mercy of motherboard manufacturers that they have the voltages under control - which they don't in my opinion - and that something you paid good money for will not decline in performance over the short term and be fried over the medium term.

With my original post I went beyond whining and showed how to configure a 3rd Gen Ryzen and be within specs to get a lot more performance for your money than you otherwise would have.

TANSTAAFL applies though, and the kludginess of my solution is a bit of a downside - which AMD could go a long way to remedying if they decided to expand their configuration software with some options. But who am I kidding, it's AMD, THAT'S never going to happen.

I'll probably do the same with my 3800X setup once I build it. If it works, don't fix it.

It's what I'm doing with my i9-9900K machine and it's doing fine.

If you try to configure your 3rd Gen Ryzen as if it were an Intel CPU then you will be in a World of Pain, with a good chance of frying your CPU.

 

[RealNeil]

If you try to configure your 3rd Gen Ryzen as if it were an Intel CPU then you will be in a World of Pain, with a good chance of frying your CPU.

I'm not into pain anymore.

I just sold my i7-8700K system to my next-door neighbor.
That one was tweaked in many ways.
I de-lidded it and had very low temps with it. All cores were running at 4.972GHz.
I had SLI GTX-1080-FE GPUs in it and 32GB of DDR4-3200 MHz. RAM installed.
It seemed like I was always looking for a way to tweak it to get a faster benchmark score out of it.

Then, someone sold me an i9-9900k CPU and board.
I knew that it was a PITA to de-lid this one, so I got a good 280mm AIO for it.
I got 32GB of 4133MHz. RAM for it.
I got two 1TB Crucial Gumstick drives for it to run in a RAID configuration.
For DATA, there is a pair of 5TB Toshiba X-300 SATA drives in RAID configuration.

I put it all together in a high-airflow chassis with five 140mm Cougar Vortex fans. They move a shit-ton of air.
The SLI GTX-1080-FE GPUs moved over to this PC.
I went into the BIOS (ASUS ROG Strix Z-390-E Gaming) and set the RAID up. I set the Intel XMP (Extreme Memory Profile) to XMP-one.

I was ready,
Let the tweaking begin!
I started trying to OC the CPU.
No Joy.
no matter what I did to tweak the CPU, it failed to boot. After a few days of screwing around with it, I set the XMP-1 memory profile and left the CPU on stock settings.
It booted and ran fine. It also scored significantly higher in benches than the 8700K ever did.
I looked and found that it was boosting to over 5GHz. speeds. (not sure how many cores though)

Thinking about it, I decided to just leave it alone to do its own thing. It works, and I'm happy with it.

I'll probably set the Ryzen box up to do its own thing at first and see how it does.
I don't want to constantly massage my PS anymore. I want to just do some gaming with them.

 

[Waldorf]

@mtcn77
for anyone that put more than one machine together and can read a manual,
should not have any probs using LM.
even when i didnt care to do more than apply it (not removing it after couple of month and replacing it a few times to "saturate" block/HS),
twisting the block removed it without trouble from the cpu after being installed for +1y.
and using it for now +6y, showed me the "gap" between a quality (cooler) base and the cpu HS is a lot less
than what i would need from even the thinnest coat of ("normal") TP i could apply.

and it was easier to cover all of the HS evenly, when a (greasy/oily) TP was previously installed,
which always gives me a hard time applying new/different TP.


@Michael Nager
yet here i am, running aorus ultra with xmp loaded in bios, even managed to lower the timings from stock 18-22-22-42
(to 16-19-19-36/1, haven't tried more) at the same V, and on "crappy" micron E die.
no ram retraining happening.