Overclock.net banner
861 - 880 of 1025 Posts

·
Not a linux lobbyist
Joined
·
2,813 Posts
Thanks everyone.
Those idle clocks were really bothering me for some reason. Coming from a 8700k and not much experiance on what normal is for this new cpu...well thank you.

View attachment 2560249





View attachment 2560250
I know this is a bit late, but maybe some of these extra Windows power plan settings I put in another post might tune your cpu's behavior in a way you want. I really don't know as my last AMD cpu was a Phenom :p Not every setting seems to do something for me, and different stuff might work for you, but having a custom power plan that you can click active that does something you like -sometimes- is convenient.
Overclocking ADL - 12900k etc results, bins and discussion | Page 281 | Overclock.net
 

·
Registered
Joined
·
2,978 Posts
I run HWinfo64 all the time and haven't noticed any yet at 1866 fclk.
If you use it solely to look for WHEA errors you could just open Event Viewer anytime you want without using an app that polls the system. I assume you use it for other things though, right?
 

·
Iconoclast
Joined
·
32,312 Posts
Got my second ASRock B550 Phantom Gaming ITX/ax and dropped my 5800X into it, then saved a pair of overclock profiles, one with an auto offset, the other with a -5mV core voltage offset:


So, for anyone with a similar ASRock board, who cares, who hasn't figured this out for themselves yet, and wants to set a voltage offset on their 5800X3D, all you need to do is save your current profile to disk, open it in a hex editor, browse to 20E and set hex "1" for offset, then got to 211 and set whatever negative voltage offset you like, counting down from FF FF 1 00 00. Be mindful that it's all little endian (so -5mV is FB FF not FFFB).

Testing my sample now to see how much of a negative offset I can apply at -30 all-core CO before hitting instability or observing clock stretching.
 

·
Premium Member
Joined
·
20,831 Posts
If the price of this cpu is $400, there is no doubt it will be the No. 1 selling cpu in AMZ not the Ryzen5 5600X.
 

·
Registered
Joined
·
11 Posts
i don't like asus strix-e x570's 1.2.0.7 beta bios cos 3d is 3.4/4.5 but mb runs at 3.6/4450 and boost never stop at desktop and cause heat.it's like bug that 3000 series causes 60 degree idle.need another ABBA fix to this.there is no pbo2 in 1.2.0.7 beta bios or ryzen master in windows.
 

·
Registered
Joined
·
2,978 Posts
Does the 1.2.0.7 get rid of the fTMP and PSP causing system stutter?
 

·
Iconoclast
Joined
·
32,312 Posts
Negative voltage offsets work as expected on my ASRock board when applied via an edited profile. So far, no clock stretching or instability with -10mV set, but boost clock and performance under heavy load is reduced. Will probably need to switch to per-core COs to make any negative offset worth while.
 

·
Registered
Joined
·
2,978 Posts
How do you identify clock stretching?

Also, I've been testing CPPC enabled in bios and observing ST behavior with CBR20 + Ryzen Master. My 'Preferred Cores' are C 01 - Fastest core, C 02 - Second Fastest Core (using Natural Numbers). I have not changed any power settings like Heterogeneous Thread Scheduling Policy or anything at all. If my observation method is wrong let me know but I am not seeing the single thread sticking to the Preferred Cores. Is this normal thread scheduling behavior for 3D? I've also added CPPC Preferred cores to CPPC enabled and i get the same behavior.
 

·
Iconoclast
Joined
·
32,312 Posts
How do you identify clock stretching?
Typically I look at actual vs. effective clocks in HWiNFO; they should be nearly identical under load. I also verify with benchmarks...a loss of performance without a loss of actual clock speed is likely clock stretching.
 

·
Registered
Joined
·
11 Posts
Still haven't seen 4550 MHz on 1 or 2 cores using HWiNFO with 50 ms polling rate. 4450 MHz reported max. Did see around 4470 ish MHz reported inside Ryzen Master a few times. On AGESA 1.2.0.6c, latest version for my board.

With -20 to -30 mV offset all cores, this thing runs at such a low temperature and voltage though.
 

·
Registered
Joined
·
13 Posts
I tried increasing VCore SOC up to 1.25V and VDDP18 up to 2.0V

I believe the WHEA errors are caused by the AMD Boost/Voltage algorithm as the Prime95 threads themselves aren't failing with the WHEA errors but I also didn't let the software run for an hour+ because it's not stable; Eventually they may error but letting it run to find out wasn't a priority. Since the machine is not Prime95 stable at 1933 I didn't bother to much with 1967 or 2000 FLCK.

@1967 1:1:1 Memory WAS verified at boot with XMP toggled on "just because" but if it throws WHEA errors in windows then that information doesn't mean much.

Will mess with it more when a new BIOS is released.
Gigabyte just release AGESA 1.2.0.7 with BIOS 16d

2100 1:1:1 now Boots... it's not stable but it leads me to believe that the issue with stability at 2000 is BIOS related.

1900 is still stable. 1933/1967/2000 still give WHEA errors with Prime95 :/

Edit: Also the new BIOS seems to decrease performance slightly, but IDK how noticable 598 average CPU-Z Single core score is to 602. Would take that all day if it meant greater system stability.

Rectangle Font Software Screenshot Multimedia
 

·
Registered
Joined
·
23 Posts
I don't know why, but my latency seems a bit high. That's with all apps closed. Not clean OS install, also after swapping 3900X to 5800X3D did nothing (except new chipset driveris). Can it be the reason? My best result:
Font Screenshot Electronic device Technology Software
Rectangle Font Screenshot Electronic device Parallel



=== EDIT ===

Just noticed, Aida was done with tRAS 28, unstable. With tRAS 32 best latency was 63.4 ns.
 
861 - 880 of 1025 Posts
Top