Overclock.net banner

1 - 12 of 12 Posts

·
Registered
Joined
·
2,052 Posts
Discussion Starter #1 (Edited)
I built my rig a couple of weeks ago and noticed I've been having this weird hanging issue on startup. The motherboard (B550 Steel Legend mATX) will hang on the checking hardware portion of the post, usually after a restart or a bios reconfiguration. That will force me to cut power via the PSU and restart after 30 secs and then it will work.

I decided to update my bios because I was still on VER 1.00 and there was already three updates available. I updated to VER 1.10 which was the latest stable release. After the update my G.Skill 3600 RAM refuses to run at its rated XMP speed. Boot loops if you try entering the bios @ 3600 and Aida crashes within 30sec of the test starting. Only setting it back to 2133 gets it working stably.

So I went back to the bios page and downloaded the beta bios. I checked some reviews on it and while there was only one review it was fairly thorough with benchmarks showing an improvement in speed. And the reviewer noted no stability issues. Okay so I updated the bios and try the RAM again. Now it seems stable but the chip is running hotter then ever, got all the way to 105C before the damn thing cut power to prevent damage.

My temps seemed kinda weird from the start, Bios states the 3900x is running at around 60c just in the bios. HWmonitor seems to put it at 45-50c at idle and around 87c at load with jumps into the high 90's. Temps seem to vary wildly. With some jumps going as much as 15c in a single increment. I'm running the Deepcool Castle 240EX AIO, which should be able to handle this chip.

I plan on reseating the AIO once I get home but this is my first AMD chip and was wondering if you guys had any advice for something I may have missed. I'm coming to realize these chips require more fine tuning then Intel and I enjoy that, I just want to make sure I'm not missing anything obvious. Thanks.

PS: Also I'm looking into undervolting this chip as I've read many posts talking about the default mobo setting pumping way too much voltage to try and get more boost. As this is my server/multi use rig, I'd prefer stability over a few extra mhz, so any advise on what negative offset voltage to use would be awesome.
 

·
Registered
Joined
·
36 Posts
Your CPU temps seem way too high - I suspect you may need to repaste. My 3900x idles at about 40, with the stock air cooling (and and extra chassis fan). I do not know your bios (my mboard is an ASUS TUF-Gaming-570 plus wireless). to change to 3600 for ram on mine, you have to use the DOCx setting, which also adjusts the ram voltage. Just switching to 3600 will probably not work, because the ram needs more voltage (1.35 for mine). If you haven't already done so, it might be wise to set everything to "optimized", and then tweak from there.

Good luck - we'll try to help.

Joe
 

·
Registered
Joined
·
2,052 Posts
Discussion Starter #3
Your CPU temps seem way too high - I suspect you may need to repaste. My 3900x idles at about 40, with the stock air cooling (and and extra chassis fan). I do not know your bios (my mboard is an ASUS TUF-Gaming-570 plus wireless). to change to 3600 for ram on mine, you have to use the DOCx setting, which also adjusts the ram voltage. Just switching to 3600 will probably not work, because the ram needs more voltage (1.35 for mine). If you haven't already done so, it might be wise to set everything to "optimized", and then tweak from there.

Good luck - we'll try to help.

Joe
Thanks for the quick reply. And yes I agree they are way to high, I should have stressed tested it before performing the bios update but alas, I did not. I should mention that I have everything set to auto in the bios except the RAM which I selected XMP for, as in I set it to XMP not manually selecting the XMP values. Before the bios update I set the RAM to XMP and was playing with manually choosing the Fclock to 1800 but never got that far as that's when I noticed my temps going nuts.

I'll do a repaste/reseat and post back soon. Thanks again.
 

·
Registered
Joined
·
2,052 Posts
Discussion Starter #4
Update:

Started to take apart my rig when I thought I would check my mounting pressure...quarter turn all around, I'm an idiot. Plugged it back in and saw a more consistent temp but only a drop of about 5c, it's high but at least it wont thermal-shutdown on me now. Started to tweak some bios settings. Ended up with this.
XMP: Enabled
Vcore offset across base and boost: -100mV
Fclock: Manually set to 1800

Aida64 now states 76c on full load with some spikes to 85c. Which is still a bit high but I assume this is normal for this chip based on some of the posts I looked up. Any other advice you guys might have would be appreciated. I'm going to keep researching this, seems there's a lot to learn about Zen 2.
 

·
Registered
Joined
·
36 Posts
Those temps seem more normal. The memory setting I found that switched everything to 1800/3600/1.35 was D.O.C.P. Seems like you are there already. I found settings in my bios to kick fan speeds up a bit - that helped. My pch fan was running full seed all the time. Turns out it is near my graphics card (2070 Super). A kind person pointed out that, if I was willing to tolerate a little more noise, kicking the graphics cards fans up a little (not much) would bring the pch temp down. You may be able to find a thread with owners of your motherboard - they may have specific settings that can help. Lots of interesting things to try here.

Welcome,

Joe
 

·
Registered
Joined
·
590 Posts
There's one thing nobody really mentions when talking about the idle temperature which is what power plan they are using when takin note of it. If you are using high performance one it will idle around 55-60 depending on what's poking the CPU in the background. Using balanced is around 50-55 and power saver always sub 45C. On load high performance and balanced are the same around 77c, in a cinebench20 run for example. So don't stress out about that too much. The 3900x runs "hot". I cool it with three 360 rads and in these hot days I'm getting around 60-70c while playing warzone and 70-77c while encoding. If is a long video it bounces around 75-82C.

I'd disable pbo as it never worked as intended. It's only going to inject more voltage unnecessarily.

Memory wise try fclk to match mclk.

Try using dram calculator to tweak your ram, sometimes you will need more voltage than dram calculator will suggest, if your system is unstable remember this, add some more voltage to you ram. I mean if xmp does not work primarily.

Depending on the undervolt (I'm talking about the negative offset) you might lose the high single core boost. The less vcore you use the smaller will be the boosted clock. Stock parameters for the CPU in the bios should be fine. I've used for a while a negative offset of -0.1 and well, temps were better but the max boost was around 4450, not the usual 4600-4625 when leaving the CPU stock with pbo off.


Another option is setting a fixed vcore, most people can do 4.2ghz all core with 1.2v (or even less), however you will leave some single performance on the table if you use fixed clock, but the thermals will be way better. Try and get to a consensus about it.
 
  • Rep+
Reactions: Redwoodz

·
Registered
Joined
·
393 Posts
I will link to a custom power plan that helps tremendously tomorrow if I remember. The chips throttle down to 2,200 MHz and and less opportunistic about spinning up. Performance isn’t hurt at all. I was playing factorio tonight and the the CPU sipped power and the temps averaged 40C.
 

·
OG AMD
Joined
·
8,803 Posts
There's one thing nobody really mentions when talking about the idle temperature which is what power plan they are using when takin note of it. If you are using high performance one it will idle around 55-60 depending on what's poking the CPU in the background. Using balanced is around 50-55 and power saver always sub 45C. On load high performance and balanced are the same around 77c, in a cinebench20 run for example. So don't stress out about that too much. The 3900x runs "hot". I cool it with three 360 rads and in these hot days I'm getting around 60-70c while playing warzone and 70-77c while encoding. If is a long video it bounces around 75-82C.

I'd disable pbo as it never worked as intended. It's only going to inject more voltage unnecessarily.

Memory wise try fclk to match mclk.

Try using dram calculator to tweak your ram, sometimes you will need more voltage than dram calculator will suggest, if your system is unstable remember this, add some more voltage to you ram. I mean, if xmp working, primarily.

Depending on the undervolt you might lose the high single core boost. The less vcore you use the smaller will be the boosted clock. Stock parameters for the CPU in the bios should be fine. I've used for a while a negative offset of -0.1 and well, temps were better but the max boost was around 4450, not the usual 4600-4625 when leaving the CPU stock with pbo off.


Another option is setting a fixed vcore, most people can do 4.2ghz all core with 1.2v (or even less), however you will leave some single performance on the table if you use fixed clock, but the thermals will be way better. Try and get to a consensus about it.
I will link to a custom power plan that helps tremendously tomorrow if I remember. The chips throttle down to 2,200 MHz and and less opportunistic about spinning up. Performance isn’t hurt at all. I was playing factorio tonight and the the CPU sipped power and the temps averaged 40C.
Finally someone get's it. Tired of reading all these post about high idle temps when they've disabled all power saving features. Well said.
 

·
Registered
Joined
·
393 Posts
Waiting on a DNS update for my web server. The download link will be at http://hwgeeks.com/downloads/ryzen_power_saver when it is done. I will post an update later when it's live. I want to create a thread here anyway for discussion.

EDIT: The download link will go live shortly after the DNS issue is resolved. I want to include the overclock.net discussion link in the readme, but I can't create the discussion until the page is live...chicken or the egg... :)

EDIT: The site should be up and running and the download link is active.
 

·
Registered
Joined
·
2,052 Posts
Discussion Starter #10 (Edited)
The saga continues.

I switched back my IF clock to its AUTO setting default for troubleshooting because something else is happening now. So originally I thought the powercut/shutdown was due to hitting the thermal limit, but now I'm sitting pretty at 72c under normal full loads I still randomly lost power during a cinebench R20 run. But this only happened when I run at vcore on auto. With vcore on a -100mV offset, it doesn't seem to happen, at least not yet. Also I noticed absolutely zero performance lose from the undervolting by 100mV, which is weird right?

On cinebench R20, I get 6990 with vcore on auto and I get 7002 (give or take a few) with vcore set to -100mV offset. Which shouldn't be happening right? I should be getting a small decrease in performance. Something else I noticed that was really interesting. Theres two chiplets on a 3900x, with vcore on auto, I'm getting one chiplet using between 13-15w per core on full load. while the second chiplet uses 8w-ish per core. Now if I set the vcore to the -100mV offset I get the same results on the first chiplet but the second chiplet is now getting around 10-11w per core. So bringing vcore down brings wattage up? what?

So something is going on but I just don't know what yet. The random shutdown is plaguing me because I'm having so much trouble isolating it. Event viewer reports nothing other then immediate power loss. I'm leaning towards mobo but the reviews for the B550 Steel Legend mATX are all 5 stars. I really hope it's not a mobo issue.

Full specs:
Motherboard: Asrock B550 mATX Steel Legend
CPU: 3900x
RAM: G.Skill Trident Neo DDR4 3600 16-16-16-36 2x8GB (soon will be 4x8GB)
GPU: GTX 960 SSC (for encoding)
PSU: EVGA G3 750w
Cooler: DeepCool 240EX castle AIO
Running 2 4TB WD Blacks with an SSD.
 

·
Premium Member
Joined
·
20,111 Posts
The saga continues.

I switched back my IF clock to its AUTO setting default for troubleshooting because something else is happening now. So originally I thought the powercut/shutdown was due to hitting the thermal limit, but now I'm sitting pretty at 72c under normal full loads I still randomly lost power during a cinebench R20 run. But this only happened when I run at vcore on auto. With vcore on a -100mV offset, it doesn't seem to happen, at least not yet. Also I noticed absolutely zero performance lose from the undervolting by 100mV, which is weird right?

On cinebench R20, I get 6990 with vcore on auto and I get 7002 (give or take a few) with vcore set to -100mV offset. Which shouldn't be happening right? I should be getting a small decrease in performance. Something else I noticed that was really interesting. Theres two chiplets on a 3900x, with vcore on auto, I'm getting one chiplet using between 13-15w per core on full load. while the second chiplet uses 8w-ish per core. Now if I set the vcore to the -100mV offset I get the same results on the first chiplet but the second chiplet is now getting around 10-11w per core. So bringing vcore down brings wattage up? what?

So something is going on but I just don't know what yet. The random shutdown is plaguing me because I'm having so much trouble isolating it. Event viewer reports nothing other then immediate power loss. I'm leaning towards mobo but the reviews for the B550 Steel Legend mATX are all 5 stars. I really hope it's not a mobo issue.

Full specs:
Motherboard: Asrock B550 mATX Steel Legend
CPU: 3900x
RAM: G.Skill Trident Neo DDR4 3600 16-16-16-36 2x8GB (soon will be 4x8GB)
GPU: GTX 960 SSC (for encoding)
PSU: EVGA G3 750w
Cooler: DeepCool 240EX castle AIO
Running 2 4TB WD Blacks with an SSD.
Weird indeed. If nothing else help, then i suggest replacing the motherboard. May i suggest picking the Gigabyte B550M AUROS Pro. Got one for a few days now and absolutely stable. Though, i only have an R5 3600 on it. Have not updated the BIOS and my RAM - GSKill 3600 CL16 is running in DOCP.
 

·
Registered
Joined
·
36 Posts
intresting info:
http://blog.livedoor.jp/wisteriear/archives/1077751705.html (google translate)
We have confirmed that the Ryzen 9 series such as Ryzen 9 3950X and Ryzen 9 3900X are combined with AMD B550 motherboard, and especially the power failure occurs.

Regarding the environment where symptoms occur, it is "Ryzen 9 and B550 motherboard combination regardless of manufacturer". It seems that B550 motherboards occur regardless of manufacturer such as ASRock, ASUS, GIGABYTE, MSI, and we have confirmed similar symptoms with products from 4 companies.
I have not confirmed the occurrence of the same symptom with a combination of lower CPU such as Ryzen 7 and B550, or with Ryzen 9 in combination with X570 or B450.

Symptoms can occur regardless of manufacturer, but the frequency and scene (whether startup or idle) tends to change depending on the motherboard product. We have also confirmed that some products do not exhibit the symptom in the initial BIOS, and the symptom occurs in the latest BIOS.
As an example, ASUS ROG STRIX B550-F GAMING (WI-FI) does not cause symptom in BIOS: 0243, but it does occur in 0603 and 0805. With ASRock B550 PG Velocita, the symptom does not occur in BIOS:1.00, but it occurs in 1.10 and 1.10A.

https://www.techporn.ph/review-asrock-b550-pg-velocita/
Now I had one concern with this board which is stability. Out of all the samples I’ve tested, the B550 PG Velocita with the Ryzen 7 3700X is the first to randomly display BSOD (not frequent, maybe one in 24 hours). I’ve tried everything, even approaching ASRock for help. Fix I did here is to pump a consistent amount of vCore – undervolting it even and it works.

on my b550 taichi i have random bsods at idle (bios 1.15A), after disable global c-state control i fixed issue
idle voltages (hwinfo)
vcore
all auto (windows power plan high perfomance, cpu freq 3600) - min 0.350v - bsod
all auto (windows power plan high perfomance, cpu freq 3600), Global C-state Control Disabled - min 1.024-1.040v - normal
all auto (windows power plan high perfomance, cpu freq 3600), Power Supply Idle Control Low - min 0.350v - bsod
all auto (windows power plan high perfomance, cpu freq 3600), Power Supply Idle Control Typical - min 0.976v - bsod
 
1 - 12 of 12 Posts
Top