Processors

[entropie]

Hello,

first of all I want to congratulate you on this new awesome Ryzen 9 processor. I ordered one 3 days after launch and got it 5 days later in Europe, so thumbs up for that. It boosted my headless Linux gaming rig to the best possible level and the CPU doesnt even sweat yet. Star Citizen + wine + dxvk + h.264 encoding + stream recording running easily all at once and the wraith prism is barely spinning.

So my question is, I was wondering what I am doing wrong, cause it wont hit the 4,6 Ghz single core boost. I got it pretty close but I am still not there. I am running on a cheap and plain MSI B450 Tomahawk and had a R7 2700X before. Updated to Agesa 1.0.0.3ab and besides that I havent changed a BIOS setting, except putting the Ram @3200 Mhz what it can handle. Are there maybe any BIOS settings I forgot to switch or is it yet due to the Agesa version, which doesnt manage the 4,6 Ghz yet? I have the version 19 of the actual BIOS for my board atm, there was just a new beta BIOS released, I havent gotten yet.

I have made some single core benchmarks, where I let a Fast Fourier Transformation run contained to every of the 24 logical cores and the result is:

Highest I get is 4,44 Ghz on Core 13, I tested all of them:

https://openbenchmarking.org/result/1907307-HV-SINGLECOR71

On Core 0 the average fps are highest with 4,23 Ghz:

https://openbenchmarking.org/result/1907306-HV-DSFDS326725

My Geekbench results are pretty good, in single core it is on par with the 9900K and on multi core it beats them all.

https://browser.geekbench.com/v4/cpu/14110048

Thank you for your hints and keep up the good work !

Greetings

~ent

[s_vayner]

there's a couple of threads about it going already. i started one before, there are few replies there. but short answer is nothing you can do right now. AMD dropped the ball and released unfinished product with shaky support. you can try play with CPU voltages but it wont make big difference.

[entropie]

hmhm, I see. Thank you both for answering.

Well then lets just hope that will get better with upcoming Agesa updates. I really love that cpu, the application powergain put me in the easy 100 - 180 fps range even when gaming headless .

[brucer]

You're not getting advertised boost frequencies because amd launched a beta product..

 All amd has done is update the chipset drivers but that has further degraded performance.. The new chipset driver has just downclocked the cpu even further..  Now mine wont hit the lowend sustained frequency or the boost frequency now..  It downclocks to 3.875ghz and it suppose to be at 3.9ghz

I'm so sick of amd releasing beta products to consumers..

Amd should be sued over this..

[entropie]

I think I have debunked the Ryzen mystery a bit more. The idea is under *Update 2* but I copied everything, so you get the history:

-----------------------------------------

https://www.reddit.com/r/Amd/comments/cpp38k/r9_45_ghz_boost_tried_to_create_a_nominal/ 

So after reading about AMDs nominal environment claim, I tried to produce one. On Linux I shut down the display manager and had only the kernel and the console running. Every program besides that was shut down. Then I ran a single core benchmark, which I doubly so confined to a single core, to get it as basic as possible. The results were, I never exceeded 4,5 Ghz. Core 12, ranging from 0-23 seems to be my fastest. The test was performed 3 times in a row, so thats why there are the downspikes in the graph. The cpu held the upper single core boost pretty steady.

Core 12 PBO auto: https://openbenchmarking.org/result/1908128-HV-RYZEN390037

Core 12 PBO set to enabled: https://openbenchmarking.org/result/1908127-HV-RYZENSING99

Core 12 enhancement mode 3: https://openbenchmarking.org/result/1908124-HV-RYZEN390088

Core 12 enhancement mode 4: https://openbenchmarking.org/result/1908127-HV-RYZEN390043

I am only running on a B450 Tomahawk on stock cooler. I doubt tho it can be temperature related, cause the tests are pretty short and the fan doesnt even spin up (its not impossible tho). But ofc it can be board related. I have the BIOS options to manually adjust the voltage and power requirements for PBO but I dont want to temper with that, being uninformed. Agesa is 1.0.0.3ab.

*UPDATE*

Thanks to redditor chapstickbomber, who suggested changing the testing method to an integer arithmetic based, I was able to up the maximum a few Hertz. I got it to 4,524 Ghz now. The test is basically asking for short bursts of cpu power and you can see that, while testing, the clock runs pretty high at around 4,5 Ghz constantly, when before it only hit that value on occasion at all. So it may be all along that those CPUs can hit 4,6 Ghz, its just the tests used that reflect other variables.

Here is the bench (Core 0, Thread 12 is still the champ): https://openbenchmarking.org/result/1908137-HV-RYZEN390008

*UPDATE 2*

I went down with the thesis that the R9 3900X will only reach its max advertised single core clock of 3.6 Ghz running no operation. So I wrote a loop in c and put 2000 nop commands iterated 10 million times in the assembler file. It still is only going up till 4,524 Ghz max but the intresting thing is, I ran the test on Core 0 again with thread 0 and 12 and if I was taxing thread 0, thread 12 showed the highest frequencies and vice versa. So that looks like the nop - thesis is a pretty good one. Obviously I am not able to beat the 4,524 Ghz cause the lacking 0,476 Ghz are prolly used for the OS.

I made a shady youtube video of the test since screen recording was dragging down the max frequency. Sorry for that, but if you have good eyes, you may see a thing (set it to HD):

https://www.youtube.com/watch?v=IDhriFlZjwo

[s_vayner]

so is it safe to say that AMD at a minimum misled, but most likely just plain simple lied to its customers about capabilities of its processors?

[s_vayner]

Also I find it amusing that someone would even consider that reaching 4.6Ghz boost under no load scenario would justify AMD's false claims. This is like having a car that can reach 200mph but only without a driver inside.

[entropie]

I am a bit pushed off by the heavy complaints going on around the boost clock. Somehow the fact that AMD delivered a monster CPU which is performing extraordinarily well for me gets brushed away under the carpet. The CPU runs for me now stable around 4,3 and 4,4 Ghz, when I am for instance compiling a larger programm, I can see how this frequency holds. Those boosts to 4,524 Ghz are really short and even if it was going to 4,6 Ghz, those would be only short bursts as well. But that is what AMD sold us from the start, the base frequency is 3,8 Ghz. The performance achieved that way, puts the cpu for me on place 1 in geekbench scores and it is what I see when playing games or compiling software. The CPU runs like a monster and barely sweats. 

The thing is, tho you may be able to push the intel easily over 5 Ghz, the frequency is not everything, cause the 3900X still beats it. 

So I am not really pushed off by the 3900X "only" reaching 4,524 Ghz for me. I will wait for Agesa Updates and I am pretty sure, we will see the 4,6 Ghz. It would have been more clever, if they sold the 3900X with 4,5 Ghz max boost clock but I can understand their position, they have to beat intel and cpufreq is most often the only thing people look at, when buying a cpu. 

[brucer]

The truth comes out.. amd ninja nerfed ryzen3000 cpus after launch.. Radeon 7 “100%” End of Life, Says Puget Systems | AMD Reduced Ryzen 3000 Stock Boost Clocks - YouTub...