Question Ryzen 7 3700X - CPU throttled?

[ch33zw1z]

Hi all,

I've just got done upgrading my CPU / MB, and have a question about what I should be expecting from the processor / if it's throttling. Specs:

CPU: Ryzen 7 3700X w/ stock wraith cooler and thermal grease
MB: Asus TUF GAMING B550-PLUS (at latest BIOS)
DIMMs: G.Skill Ripjaws F4-3000C15D-16GRK - two sets, set to run at rated speeds.
GPU: Nvidia Geforce GTX 1660 Super

The CPU is said to run up to 4.4Ghz, but it's topping out around 4 at the most. Scenario:

idle temps start around 41-45C

1. Kick off a Handbrake job que.
2. Hardware Monitor shows CPU temp increase to 74-76C, CPU fan speed increases
3. CPU-Z shows Core speed around 3.8-3.9 (sometimes 4.0) Ghz
4. Windows Task Manager is the same

My question is based around the expectation that the CPU can get to 4.4. Because it stopped at 74-76C and is consistently hovering there, and the CPU speed is consistently at 3.8-3.9, I'm thinking that the system is throttling this CPU a little to "save" itself.

Has anyone else had this experience?

Pictures attached

EDIT: Reading this, it appears it's probably just deciding to throttle based on sensor readouts, might need to apply aftermarket thermal paste and/or buy an aftermarket cooler to get the max boost clock.

Also, maybe the latest driver, currently running the v2.11 from Asus's website

https://www.amd.com/en/support/kb/release-notes/rn-ryzen-chipset-2-13-27-501

 

[DrMrLordX]

Word. Yeah those fans aren't too bad if you keep the speeds down. Try cranking them up just once to see how bad it can get. It's fun!

(good temps btw; 3700x is actually kind of hard to keep cool depending on how hard you push it).

 

[ch33zw1z]

Word. Yeah those fans aren't too bad if you keep the speeds down. Try cranking them up just once to see how bad it can get. It's fun!

(good temps btw; 3700x is actually kind of hard to keep cool depending on how hard you push it).

I ran the fans with the turbo profile for a day, was pretty loud. It didn't get to the hearing damage point, but you could definitely hear them! The standard profile seems just right.

Glad I didn't have to go liquid.....yet lol

 

[DrMrLordX]

Liquid doesn't help as much as you think. I slapped a massive MO-RA3 on my 3900x and I have to say that the improvement over say a well-equippred NH-D15 isn't as much as I thought it would be. Them hotspots make bulk cooling capacity less valuable than on something like a 14nm Intel chip.

 

[ch33zw1z]

So I have another question here regarding CPU verification / stress testing

This setup has been running like a champ since last year. However, I found it this morning with an unexpected power event / reboot.

I remember reading other threads back in 2021 about AMD CPU's failing stress testing and potential RMA's.

I would like to check the usual suspects for unplanned power events: CPU, Memory for starters.

What is the recommended tools / testing that others use?

 

[DrMrLordX]

Unplanned reboot? I'd look at your PSU first.

 

[ch33zw1z]

Unplanned reboot? I'd look at your PSU first.

Could be, recommended methods?

 

[Markfw]

Could be, recommended methods?

I was going to recommend this:

Are you a human?

www.newegg.com

I got one, but then when I just looked it up, its $40, and you may not want to spend that. But it works great. Otherwise swapping in a known good PSU with at least the same power rating is the only other method.

 

[ch33zw1z]

I was going to recommend this:

Are you a human?

www.newegg.com

I got one, but then when I just looked it up, its $40, and you may not want to spend that. But it works great. Otherwise swapping in a known good PSU with at least the same power rating is the only other method.

Nice little tool, out of stock though 😕

Not much has changed then, I guess I'll start shopping around for both the tool and another psu.

Current PSU is an EVGA 220-G2-0850-XR purchased in 2018. Would be great if I can find one that cables are compatible

 

[DrMrLordX]

Could be, recommended methods?

tbh I haven't tested a PSU in a long time. Been lucky with mine "just working" except for a Seasonic unit I had to RMA. But.

Easiest way (short of buying extra hardware as suggested by Mark) is process of elimination. Run some stress tests on other components, see if the system holds up. Stress test CPU, GPU, RAM. If you're still seeing random reboots during idle then you're looking at edge-case behavior which could take a long time to diagnose properly.

 

[ch33zw1z]

Thanks guys, the more things change, the more they stay the same 😉

I should've noted that it was idle, handbrake jobs had been running for days, it just finished up a job around that time.

Since I don't really want to worry about it too much, I'll likely just buy another PSU and give it a shot. I'll run stress tests while I'm looking around. I really like the EVGA modular, so eyes on this already.

https://www.evga.com/products/product.aspx?pn=220-P6-1000-X1

 

[DrMrLordX]

I should've noted that it was idle, handbrake jobs had been running for days, it just finished up a job around that time.

Interesting. Might have been a crash due to an erratic voltage swing when changing load state.

 

[Markfw]

Nice little tool, out of stock though 😕

Not much has changed then, I guess I'll start shopping around for both the tool and another psu.

Current PSU is an EVGA 220-G2-0850-XR purchased in 2018. Would be great if I can find one that cables are compatible

Its in stock at newegg, its in my cart ! I will remove it to be sure, but I added it to my cart after I posted. It uses standard cables, and that PSU has a at least 5 year warranty I bet.

 

[Markfw]

Thanks guys, the more things change, the more they stay the same 😉

I should've noted that it was idle, handbrake jobs had been running for days, it just finished up a job around that time.

Since I don't really want to worry about it too much, I'll likely just buy another PSU and give it a shot. I'll run stress tests while I'm looking around. I really like the EVGA modular, so eyes on this already.

https://www.evga.com/products/product.aspx?pn=220-P6-1000-X1

This one is $100 in FS/FT

https://www.evga.com/Products/Product.aspx?pn=210-GQ-1000-V1

here

https://forums.anandtech.com/threads/new-evga-1000-gq-1000w-gold-psu-power-supply-alphacool-monsta-240mm-radiator.2600948/

 

[eek2121]

Renoir has some improvements over Matisse. Though I must admit that I am surprised by the high sustained clocks during a fairly-intense workload like handbrake. Typical 3700x samples struggle to exceed 4250 MHz all-core even when static overclocked.

Renoir is also SLOWER than Matisse. Would be interesting to see the 4700g vs a 3700x at stock settings and same RAM timings. I don't know how much Handbrake likes L3 cache, but the difference could be quite high.

 

[ch33zw1z]

I ended up going with a NIB 220-P6-0850-X1

Side note: OCCT costs $8 a month for a user license? Seems a bit steep. I'm at a point now where I'd probably buy it once for $20.

Using P95 v30.7, build9 - just kick off basic tests, first time I've seen my cpu get to 71C since I installed the Noctua setup! Mostly hovering is mid 60's. OCCT basic stuff was barely touching it, temps in 50's.

 

[ch33zw1z]

Swapped out to a 5900X (and replaced PSU at the same time) and seeing roughly the same temps. Funny enough, pulling music off of cd's to MP3 / FLAC bumps it the highest so far to around 75C, with very minimal CPU usage (like 5%) *shrug*

5900X says Max temp of 90C (95 for the 3700X IIRC) in Ryzen Master, so I'll keep my eye on it, but I'm not too worried at this point.

Side note, it seems that HB and 1080p stuff hits 20 cores max, 4K will use it all. SD video was / is 8