Question 3900X not boosting properly even after ABBA bios update? (ASUS X570 Strix F board)

Status
Not open for further replies.

Kocicak

Senior member
Jan 17, 2019
982
973
136
I finally got the 3900X. I run it on Asus X570 Strix F board with the latest 1201 bios. It is under the AIO water cooler, temperature in R20 does not go over 70°C. Everything is on default.

If I run 1 thread in R15, I never get not only 4650 MHz as reported by some, but not even 4600 MHz. HWinfo reports three cores hitting maxima between 4500 and 4600 MHz.

Are there any settings one needs to make so that it works properly? Is the bios on this board still broken? Or did I simply get a lemon CPU?

If I run 4 threads in R15, I get consistent maxima on many cores of 4366 MHz. Is this typical behavior of the CPU?
 

DrMrLordX

Lifer
Apr 27, 2000
21,617
10,825
136
CBR20 is AVX/AVX2 capable, so it will lower your clockspeeds somewhat. Try something lightweight like SuperPi 1.5XS and see if you can get it to peg one core @ 4.6 GHz non-stop. Or just try CBR15. You're going to see boost variance based on workload, which gets a little weird.

And yeah you are not going to see more than one core stay at 4.6 GHz constantly. Not without static OC and some insane cooling. If you really want to try and optimize your boost clocks with default behavior you can try messing with LLC settings and voltage offsets.
 

Justinbaileyman

Golden Member
Aug 17, 2013
1,980
249
106
You could try downloading Handbrake and get the 4k file of Big Buck Bunny which is a free open source computer generated/animated cartoon free of charge and give it a quick format conversion to 1080p which will max out all your cores at there max speed which should show you at which speed they will staying running at while maxed out ect..
Its also a great test to see if your cooling is up to snuff for when you maxing out all your cores under full load..
 

DrMrLordX

Lifer
Apr 27, 2000
21,617
10,825
136
I was running one thread in R15. I used R20 just to test how cooling works.

Oh okay.

With LLC set to Off ("Standard") I can get stuff like SuperPi 1.5 to run a single core nonstop at 4600 MHz. It runs 4575 MHz if I use LLC Off + -.1v offset. So I lose maybe 25 MHz single core, but I gain a bunch of performance in MT workloads, so to me it's worth the tradeoff. You are probably not going to see 4600 MHz in CBR20 ST. 4650 MHz is not happening without PBO + AutoOC, and I don't even bother with that crap to be perfectly honest. PBO has been nothing but a headache for me.
 

Kenmitch

Diamond Member
Oct 10, 1999
8,505
2,249
136
Are there any settings one needs to make so that it works properly? Is the bios on this board still broken? Or did I simply get a lemon CPU?

It's hard to say how Asus configured the uEFI with the default settings without actually looking around in there and experimenting.

Have you tried just loading the optimized defaults and enabling PBO in Ryzen Master to see what happens?

PBO doesn't do much for my 3700x other than give me 100-150 MHz higher sustained all core boost speeds. My 3700x doesn't look to have single core boost speed issues as 6 of 8 will do 4400 MHz and the other 2 tap out at 4350 MHz on my current ABBA uEFI according to Hwinfo.

Ignoring those max boost clocks how's the rig perform otherwise?
 

Guru

Senior member
May 5, 2017
830
361
106
I finally got the 3900X. I run it on Asus X570 Strix F board with the latest 1201 bios. It is under the AIO water cooler, temperature in R20 does not go over 70°C. Everything is on default.

If I run 1 thread in R15, I never get not only 4650 MHz as reported by some, but not even 4600 MHz. HWinfo reports three cores hitting maxima between 4500 and 4600 MHz.

Are there any settings one needs to make so that it works properly? Is the bios on this board still broken? Or did I simply get a lemon CPU?

If I run 4 threads in R15, I get consistent maxima on many cores of 4366 MHz. Is this typical behavior of the CPU?
Cinebench does not test your cores max clock speeds, in fact it never uses max boost on any cpu. A better test would be to try it out in games and monitor your cpu with hwinfo and see what the max clocks are.

Damn I hate when newbies who have no idea what they are doing are endlessly complaining over nothing. Believe me your cpu is 99.9999% percent boosting properly if you've set everything up properly, meaning ryzen power plan in windows and default bios settings. Either way 100MHz more on 1 or two cores won't change anything believe me. This has been tested and at best you get 1-2fps more, usually its like 1fps or less in terms of "gains".
 

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,654
136
Did some tests with 20. Cinebench didn't try to find the best core, ignored affinity, and didn't even get to boost and this is with PBO on. With everyone else on this. Overall the 3900x and prob the 3950x are clocked within inches of the silicon and socket. Probably more than any other chip the cores aren't all equal and it takes some smart processing to make sure the golden cores are prioritized (I thought that an update happened for this but I am not seeing it, might try to update the chipset when I get home). Just keep in mind the 3900x numbers we saw owning everything wasn't running nearly as well as it does now with abba.
 

Kocicak

Senior member
Jan 17, 2019
982
973
136
I tried SuperPi and the maximal sustained speed is 4541 and 4566, recorded maximum is 4591.
Now I see max. 4616, but that was while doing almost nothing.

Overall performace is probably normal - R20 score is 7179/524.

I just had the idea that after the update processors with light load really run at 4600 MHz, mine seems to run approx. 50 MHz lower.

It also does not overclock well, for all core 4300 MHz I need more than 1.4V (it worked at 1.425) (R20 = 7658)
4200 MHz all core worked at 1.35V. (R20 = 7509) I did not try optimise things or run extensive stability tests, but overclocking wise it does not seem much more better than my 3600.

4600 MHz at 1.5V did not even boot... :D
 

DrMrLordX

Lifer
Apr 27, 2000
21,617
10,825
136
@Kocicak

You might have lost the silicon lottery (tm), who knows?

What voltage levels are you seeing in CPU-z attempting those all-core OCs, btw? There is a lot of vdroop on a 3900x. Just because your UEFI is saying 1.4v (or whatever) does not mean that's what is getting to the cores, per se. Also that MT score @ 4.2 GHz is pretty good, so I wouldn't be too upset about it.

If you want more MT performance out of your chip in default boost mode, I highly recommend choosing the lowest LLC setting possible (LLC Off, LLC0, LLC1, or whatever is the least you can pick) and then start messing with negative voltage offsets. For me it is an offset of -.1v but for others it differs. Check out your CBR20 MT scores. Mine will do nearly 7400 with those settings and an aggressive mem OC. Temps get lower too.
 

Guru

Senior member
May 5, 2017
830
361
106
I tried SuperPi and the maximal sustained speed is 4541 and 4566, recorded maximum is 4591.
Now I see max. 4616, but that was while doing almost nothing.

Overall performace is probably normal - R20 score is 7179/524.

I just had the idea that after the update processors with light load really run at 4600 MHz, mine seems to run approx. 50 MHz lower.

It also does not overclock well, for all core 4300 MHz I need more than 1.4V (it worked at 1.425) (R20 = 7658)
4200 MHz all core worked at 1.35V. (R20 = 7509) I did not try optimise things or run extensive stability tests, but overclocking wise it does not seem much more better than my 3600.

4600 MHz at 1.5V did not even boot... :D
Man read just ONE review before you start a thread or write more comments. NO Ryzen 3000 overclocks more than 4300mhz on all cores, especially not the 12 core ones. I've seen some people post 4400MHz all core on 3800x, but I think that is probably with some very advanced cooling and probably enough to just run a benchmark before it becomes unstable.

These processors run faster when left to run out of the box as the boost frequencies on 1 or two cores are much faster and overall the processor boost much better depending on workload and you get better results.

Basically the only benefit from all core of 4.3ghz would be encoding, most other apps would run better just the cpu normally boosting.
 

Kocicak

Senior member
Jan 17, 2019
982
973
136
To sum it up, I got 10% more light load performance and twice as much multithread performance for three times as much money I payed for Ryzen 3600.

And my CPU does seem average at best. I do not feel I got any real quality for the price premium above the performance improvements, I will probably return it.
 
Last edited:

lightmanek

Senior member
Feb 19, 2017
387
754
136
To sum it up, I got 10% more light load performance and twice as much multithread performance for three times as much money I payed for Ryzen 3600.

And my CPU does seem average at best. I do not feel I got any real quality for the price premium above the performance improvements, I will probably return it.

In a perfect world you would get three times performance for three times the money, but we live in a different reality, where you pay premium for the better stuff. There are many car, house, electronic or nature analogies I could give, but I think you know them all.
 

TheGiant

Senior member
Jun 12, 2017
748
353
106
To sum it up, I got 10% more light load performance and twice as much multithread performance for three times as much money I payed for Ryzen 3600.

And my CPU does seem average at best. I do not feel I got any real quality for the price premium above the performance improvements, I will probably return it.
yep
that is why we need moar coarz...to find out that we don't
 
  • Like
Reactions: Zucker2k

Kocicak

Senior member
Jan 17, 2019
982
973
136
... you pay premium for the better stuff.
I know that.

There are many car, ... , electronic ... analogies I could give
Well, if you get some premium car or electronic equipment, is usually comes with added quality or higher standards even beside the main spec characteristic.

When I got 3900X, I put it in a new motherboard with updated bios which supposedly fixed the boost behavior, paying quite high premium over the performance difference, I expected to get some confident sustained 4600MHz action happening on various cores. Instead I got this old crapola with spotty 4550 MHz on two same cores. What is going to happen if these two cores wear out???

I begin to think that to get the highest quality silicon in consumer market, you need to get 3800X or the upcoming 3950X.
 

Kenmitch

Diamond Member
Oct 10, 1999
8,505
2,249
136
To sum it up, I got 10% more light load performance and twice as much multithread performance for three times as much money I payed for Ryzen 3600.

And my CPU does seem average at best. I do not feel I got any real quality for the price premium above the performance improvements, I will probably return it.

Sounds like your chip performs pretty much just like all the reviews said it would. Why did you buy the 3900x if you don't have any need for it's superior multi core performance?
 

Elfear

Diamond Member
May 30, 2004
7,097
644
126
I know that.

Well, if you get some premium car or electronic equipment, is usually comes with added quality or higher standards even beside the main spec characteristic.

When I got 3900X, I put it in a new motherboard with updated bios which supposedly fixed the boost behavior, paying quite high premium over the performance difference, I expected to get some confident sustained 4600MHz action happening on various cores. Instead I got this old crapola with spotty 4550 MHz on two same cores. What is going to happen if these two cores wear out???

I begin to think that to get the highest quality silicon in consumer market, you need to get 3800X or the upcoming 3950X.

Would you really be completely satisfied with your 3900X if it boosted another 50Mhz on a couple cores? That's like a 1% performance boost at the very most (if performance scales linearly with clock speed). If you're mad that you spent 3x as much money for 2x the MT performance of a 3600, an extra 1% isn't going to do anything for you.
 
Status
Not open for further replies.