Need Help Demystifying Ryzen 3000 - Overclock.net - An Overclocking Community

Forum Jump: 

Need Help Demystifying Ryzen 3000

Reply
 
Thread Tools
post #1 of 16 (permalink) Old 10-21-2019, 01:57 PM - Thread Starter
New to Overclock.net
 
MNiceGuy's Avatar
 
Join Date: Sep 2007
Location: Belle Plaine, MN
Posts: 4,629
Rep: 342 (Unique: 278)
Need Help Demystifying Ryzen 3000

I've exclusively built Intel machines from the introduction of LGA775 onward but I decided it was time to try Ryzen since everyone seems to be raving about the performance and value.

I went with an X570 Aorus Ultra and a 3900X for my new setup. I updated the BIOS to the latest version (F6b) and away I went. Windows 10 installed without an issue and there were no signs of instability.

Once I got going I noticed that my case fans were constantly ramping up and down. I installed Ryzen Master and found the temperature fluctuating between 45*C and 70*C. The instant I attempted a video encode the temps shot to the upper 70's before I canceled it; easily besting the thermals of the notoriously hot 9900K I have in another setup.

Cooling:
Corsair H150i Pro with x3 120mm (mounted top as intake)
Lian Li Dynamic O11
x3 120mm case fans (mounted side as exhaust)

I went to the internet and found what appears to be some general confusion about how Ryzen 3000 works and what to expect while using it. Most of this seems concentrated on voltage and boost behavior.

I chose to set CPU voltage to "Normal" as opposed to "Auto" which I found posted in numerous places. For my motherboard, this meant 1.225v. Once I got into Ryzen Master again, supposedly one of two tools (the other being CPU-Z) that will accurately represent Ryzen 3000 telemetry without introducing the "observer effect", I found that the voltage was right around that 1.225v and I was idling in the high 30's. Great! It's worth mentioning I was using the "Ryzen Balanced" Power Plan.

I then loaded up another MP4 encode and attempted to render it. Temps shot up into the 70's and I saw 1.45v+ in Ryzen Master. When I stopped the encode, voltage went back down to 1.225v and the system cooled off. I could reproduce this at will. I noticed that when setting Ryzen Master to Manual Mode, the "Peak Core(s) Voltage" was set to ~1.45v. I lowered this down to 1.25v and went back to another encode.

This time, temperatures were sub-70 but I noticed that Handbrake was only using ~85% of the CPU potential. If I reset the Peak Core Voltage back to default I got 100% utilization again but I also got high temps again as well.

Reminding everyone of my AMD-ignorance mentioned above, this 3900X seems like a basket case. My PC is used for gaming about half the time and video editing and rendering the other half. I love the performance boost in Handbrake and Premiere but at the same time I feel like I'm trying to cool a fusion reactor. Yes, I did check to make sure I was getting good contact between CPU/TIM/AIO.

For reference, my 9900K with the same case/cooler peaks around 70*C after a long encode using Handbrake (which uses AVX). In fairness I do have a -100Mhz AVX offset (so 4.6GHz per core when fully loaded with AVX work)

I'm also concerned about that 1.45v+ I'm seeing on sustained heavy load with a 7nm CPU. Assuming I could somehow cool this thing I can't imagine that's a good voltage for sustained load. I've confirmed this voltage behavior with CPU-Z and HWInfo as well. All three agree.

What am I doing wrong?


Last edited by MNiceGuy; 10-21-2019 at 02:01 PM.
MNiceGuy is offline  
Sponsored Links
Advertisement
 
post #2 of 16 (permalink) Old 10-21-2019, 03:03 PM
New to Overclock.net
 
The Stilt's Avatar
 
Join Date: May 2013
Posts: 2,555
Rep: 859 (Unique: 327)
In multicore workloads (e.g. Cinebench R20 NT) the 3900X should typically run at =< 1.325V, given that it operates at stock settings.

Some ODMs unfortunately intentionally misconfigure the power management of the CPUs, practically breaking it. That it done to make their boards to perform / appear better in the reviews.

If your CPU really runs at 1.450V in multithreaded workloads, it definitely isn't running at the stock settings.

For starters you could try enabling PBO in manual mode and using the default settings: 142W PPT, 95A TDC, 140A EDC and PBO scalar 1x.

If the CPU still runs at abnormally high voltages, then the power management most likely has been intentionally misconfigured.

That being said, the 3000-series Ryzens do run extremely hot.
The high temperature and the general issues in cooling them is entirely normal, since the 3000-series CPUs have the highest power density of any CPU ever released.
A stock 3800X CPU can easily hit > 1.48W/mm² intensity (~ 110W for the CCD only), whereas a 9900K for example has theoretical intensity of ~ 1.15W/mm² when its drawing 200W of power.

Anything under 80°C can be considered as good.
The Stilt is online now  
post #3 of 16 (permalink) Old 10-21-2019, 03:20 PM - Thread Starter
New to Overclock.net
 
MNiceGuy's Avatar
 
Join Date: Sep 2007
Location: Belle Plaine, MN
Posts: 4,629
Rep: 342 (Unique: 278)
Quote: Originally Posted by The Stilt View Post
In multicore workloads (e.g. Cinebench R20 NT) the 3900X should typically run at =< 1.325V, given that it operates at stock settings.

Some ODMs unfortunately intentionally misconfigure the power management of the CPUs, practically breaking it. That it done to make their boards to perform / appear better in the reviews.

If your CPU really runs at 1.450V in multithreaded workloads, it definitely isn't running at the stock settings.

For starters you could try enabling PBO in manual mode and using the default settings: 142W PPT, 95A TDC, 140A EDC and PBO scalar 1x.

If the CPU still runs at abnormally high voltages, then the power management most likely has been intentionally misconfigured.

That being said, the 3000-series Ryzens do run extremely hot.
The high temperature and the general issues in cooling them is entirely normal, since the 3000-series CPUs have the highest power density of any CPU ever released.
A stock 3800X CPU can easily hit > 1.48W/mm² intensity (~ 110W for the CCD only), whereas a 9900K for example has theoretical intensity of ~ 1.15W/mm² when its drawing 200W of power.

Anything under 80°C can be considered as good.
Thanks. That's really good info.

Just to make sure it isn't observer error, I have seen the occasional 'flicks' to ~1.45v when idling or under light load that various forums and sites have experienced. Using Ryzen Master and CPU-Z as my guide, I am seeing a constant, unwavering voltage in the 1.45v area when the CPU is 100% loaded for an extended period of time (i.e. video encoding).

As I mentioned, if I use Ryzen Master to limit the peak voltage to 1.225v then things cool down but CPU utilization (Windows Task Manager) goes down as well without any change in workload. Admittedly I didn't check to see if there was an actual, measurable performance hit in this state to rule out the fact that Task Manager wasn't sure how to correctly represent usage.

My board does have LLC which in my experience has been a way to combat vDroop under heavy demand. I thought maybe this was the world's most overly-ambitious LLC implementation but none of the the various settings impact my problem.

I'll have a look at what you suggest though to see where that gets me.

MNiceGuy is offline  
Sponsored Links
Advertisement
 
post #4 of 16 (permalink) Old 10-21-2019, 04:30 PM
New to Overclock.net
 
The Stilt's Avatar
 
Join Date: May 2013
Posts: 2,555
Rep: 859 (Unique: 327)
All Ryzen 3000-series CPUs can use up to 1.5000V during the peak boost, thats entirely normal. Peak boost obviously isn't achieved when multiple cores are being stressed, so the voltages in multicore workloads will be significantly lower.

Load-line should not be adjusted, if the CPU is operating at stock conditions (i.e. not in manual OC mode).
The CPU will modify its voltage requests based on the required and the received voltages, so using load-line at stock is completely unnecessary and therefore not recommended.
The CPU expects a certain amount of droop and it is able to modify the VRM load-line slope on its own too (through SVI2 interface). Because of that, the external overrides are unnecessary and also prevent the CPU from controlling things as intended.
The Stilt is online now  
post #5 of 16 (permalink) Old 10-22-2019, 04:11 PM
New to Overclock.net
 
MikeS3000's Avatar
 
Join Date: Mar 2017
Posts: 53
Rep: 2 (Unique: 2)
I'm running a similar motherboard as you (Aorus x570 Pro Wifi). I have owned the 3900x since early August and was frustrated with the temperatures under load with my h100i v2. I would peak at up to 81c with multi-core benchmarks like CB20. This AIO performed great on my i7-7700k overclocked but wasn't doing well on the 3900x. Out of curiosity I purchased a Noctua NH-D15S and added a 2nd 120mm fan. For whatever reason this dropped my load temperatures to the low or upper 60's during heavy multi-core workloads at stock settings. The CPU was boosting higher and benchmarks were improved. There are some reports that some AIOs do not have great coverage of the CPU or just can't handle the 3900x. Going back to air cooling has been my answer and I'm selling my H100i v2 on ebay.
MikeS3000 is offline  
post #6 of 16 (permalink) Old 10-22-2019, 05:44 PM
Hardware Rep
 
GBT-MatthewH's Avatar
 
Join Date: May 2016
Posts: 520
Rep: 49 (Unique: 37)
You should be in the 1.3 range, not 1.45. First thing to establish a baseline is not to adjust voltages. Load defaults and test first.

You may also want to try the latest BIOS here on the forums, F6E: https://drive.google.com/open?id=1rV...tV7-bDcGlBVGry

You may also find alot of answers in the X570 AORUS thread here: https://www.overclock.net/forum/11-a...rs-thread.html
GBT-MatthewH is offline  
post #7 of 16 (permalink) Old 10-22-2019, 07:41 PM
New to Overclock.net
 
VPII's Avatar
 
Join Date: Dec 2014
Location: Hout Bay, Cape Town, South Africa
Posts: 669
Rep: 9 (Unique: 8)
Quote: Originally Posted by MNiceGuy View Post
I've exclusively built Intel machines from the introduction of LGA775 onward but I decided it was time to try Ryzen since everyone seems to be raving about the performance and value.

I went with an X570 Aorus Ultra and a 3900X for my new setup. I updated the BIOS to the latest version (F6b) and away I went. Windows 10 installed without an issue and there were no signs of instability.

Once I got going I noticed that my case fans were constantly ramping up and down. I installed Ryzen Master and found the temperature fluctuating between 45*C and 70*C. The instant I attempted a video encode the temps shot to the upper 70's before I canceled it; easily besting the thermals of the notoriously hot 9900K I have in another setup.

Cooling:
Corsair H150i Pro with x3 120mm (mounted top as intake)
Lian Li Dynamic O11
x3 120mm case fans (mounted side as exhaust)

I went to the internet and found what appears to be some general confusion about how Ryzen 3000 works and what to expect while using it. Most of this seems concentrated on voltage and boost behavior.

I chose to set CPU voltage to "Normal" as opposed to "Auto" which I found posted in numerous places. For my motherboard, this meant 1.225v. Once I got into Ryzen Master again, supposedly one of two tools (the other being CPU-Z) that will accurately represent Ryzen 3000 telemetry without introducing the "observer effect", I found that the voltage was right around that 1.225v and I was idling in the high 30's. Great! It's worth mentioning I was using the "Ryzen Balanced" Power Plan.

I then loaded up another MP4 encode and attempted to render it. Temps shot up into the 70's and I saw 1.45v+ in Ryzen Master. When I stopped the encode, voltage went back down to 1.225v and the system cooled off. I could reproduce this at will. I noticed that when setting Ryzen Master to Manual Mode, the "Peak Core(s) Voltage" was set to ~1.45v. I lowered this down to 1.25v and went back to another encode.

This time, temperatures were sub-70 but I noticed that Handbrake was only using ~85% of the CPU potential. If I reset the Peak Core Voltage back to default I got 100% utilization again but I also got high temps again as well.

Reminding everyone of my AMD-ignorance mentioned above, this 3900X seems like a basket case. My PC is used for gaming about half the time and video editing and rendering the other half. I love the performance boost in Handbrake and Premiere but at the same time I feel like I'm trying to cool a fusion reactor. Yes, I did check to make sure I was getting good contact between CPU/TIM/AIO.

For reference, my 9900K with the same case/cooler peaks around 70*C after a long encode using Handbrake (which uses AVX). In fairness I do have a -100Mhz AVX offset (so 4.6GHz per core when fully loaded with AVX work)

I'm also concerned about that 1.45v+ I'm seeing on sustained heavy load with a 7nm CPU. Assuming I could somehow cool this thing I can't imagine that's a good voltage for sustained load. I've confirmed this voltage behavior with CPU-Z and HWInfo as well. All three agree.

What am I doing wrong?
Because I got somewhat irritated with the cpu vcore floating around during normal operation and the fact that I very rarely run a cpu at stock I decided to do a manual overclock on the my 3900X. Currently it is running happily at 4.267ghz using a manual voltage set in the bios of 1.2375v. This significantly reduced load temps and even though I've lost some single core performance I've gained about 267mhz on multicore performance.

Now I'm not saying you should do this, it is just to show you what can be done if you able to do a manual overclock.
Attached Thumbnails
Click image for larger version

Name:	CB20 4.267Ghz.jpg
Views:	19
Size:	787.2 KB
ID:	301962  

Click image for larger version

Name:	IBT 4.267Ghz Check Temps.jpg
Views:	13
Size:	735.3 KB
ID:	301964  


AMD
(9 items)
CPU
AMD Ryzen 2700X
Motherboard
Asus Crosshair VII Hero
GPU
GALAX RTX 2080 TI OC
RAM
G.Skill f4-3200c14d-16gtzr tridentz rgb
Hard Drive
Crucial MX300 M.2 SSD
Power Supply
Corsair HX1200i
Cooling
CoolerMaster MLW-D24M-A20PC-R1 MasterLiquid ML240L 240mm RGB Liquid CPU Cooler
Operating System
Windows 10 Pro
Monitor
Dell U2717D 27" UltraSharp InfinityEdge LED Monitor
▲ hide details ▲
VPII is offline  
post #8 of 16 (permalink) Old 10-23-2019, 11:50 AM - Thread Starter
New to Overclock.net
 
MNiceGuy's Avatar
 
Join Date: Sep 2007
Location: Belle Plaine, MN
Posts: 4,629
Rep: 342 (Unique: 278)
Here are my results after doing a BIOS reset and a fresh Windows install. I did update to the current chipset drivers available for X570 on AMD's page. The only software added to this PC at the moment is Handbrake and Ryzen Master.

In the BIOS, I have only enabled XMP and changed CPU Voltage to "Normal" per various recommendations I've seen. According to the BIOS, that should result in 1.225v. This is what I consider the minimum work a consumer would be expected to do after purchasing this hardware.

Screenshot below is encoding a 1080/60 MP4 to 1080/30 MP4. Within a minute or two I was over 80*C and sustained voltage was in the 1.35v area. This is using the included AMD cooler on an open-air test bench. Ambient is 20.5*C.

Now, for what it's worth, this setup is noticeably faster than my 9900K setup for this kind of work. Ryzen has an average FPS of 248 where the Intel managed 191 under the same conditions. To be fair, I am running the 9900K bone stock with a -100Mhz AVX offset to keep temperatures in the high 60's to low 70's during sustained encoding jobs. That limits me to 4.6GHz across all cores when fully loaded.

I want to like this setup but that voltage and temperature still concern me. Perhaps Ryzen just does things differently than I'm used to but with the Intel setups I've been building for years, these metrics would be concerning to say the least.
Attached Thumbnails
Click image for larger version

Name:	Ryzen Master 2.JPG
Views:	19
Size:	251.8 KB
ID:	302054  


MNiceGuy is offline  
post #9 of 16 (permalink) Old 10-23-2019, 11:55 AM
Hardware Rep
 
GBT-MatthewH's Avatar
 
Join Date: May 2016
Posts: 520
Rep: 49 (Unique: 37)
Quote: Originally Posted by MNiceGuy View Post

Screenshot below is encoding a 1080/60 MP4 to 1080/30 MP4. Within a minute or two I was over 80*C and sustained voltage was in the 1.35v area. This is using the included AMD cooler on an open-air test bench. Ambient is 20.5*C.
This sounds 100% in line with what to expect from defaults and a stock cooler.
GBT-MatthewH is offline  
post #10 of 16 (permalink) Old 10-23-2019, 12:29 PM - Thread Starter
New to Overclock.net
 
MNiceGuy's Avatar
 
Join Date: Sep 2007
Location: Belle Plaine, MN
Posts: 4,629
Rep: 342 (Unique: 278)
Same test conditions but with PBO (called "Core Performance Boost" to Gigabyte owners) disabled.

Lower voltage and lower temperatures. Looks like the cores are holding around the 3.8GHz mark.

Interestingly, there wasn't a noticeable performance hit in Handbrake. In the screenshot you can see it's running an average FPS of 249 which is within the range of the PBO-enabled test.

I'm trying to wrap my head around this but to me, PBO works like this:

There are limits set to temperature, power, and current. According to Ryzen Master, this is 95*C, 142W, and 95A/140A respectively. PBO will run the cores as hard as they will go without violating any of these predefined values. Kind of neat really. The sustained voltage of 1.35v still seems like a lot for a 7nm chip but on the other hand I didn't design the thing. Barring some sort of defect with my motherboard/BIOS, I can only assume this is how AMD intended the chip to function.
Attached Thumbnails
Click image for larger version

Name:	Ryzen Master 3.JPG
Views:	12
Size:	269.7 KB
ID:	302060  



Last edited by MNiceGuy; 10-23-2019 at 12:34 PM.
MNiceGuy is offline  
Reply

Quick Reply
Message:
Options

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in



Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off