[SOLVED] Are my temperature readings okay for Ryzen 3700x CPU?

[boagz]

So I just built a new system with the Asus strix x570 mobo and the Ryzen 7 3700x. I'm using the scythe fuma 2 cooler. After install the temperature readings on this cpu vary pretty wildly at idle. I would say, after having the computer on for a bit, they spend most of the time between 40-49 C but often spike to approx 57C (though the highest spike I've seen was 67C but this was rare and it dropped back down shortly after). A typical temp reading on something like core temp might look like:

40C
49C
48C
58C
55C
53C
51C
49C
45C
56C
........

When running prime95 for about 20minutes or so my CPU temps do the same thing but jumps back and forth between 69C and 88C (88C was the highest spike), with most temps hanging out within the 72-79C range. My questions are as follows:

1.) Is it normal for a CPU to temperature jump like this? Especially at idle? I already took off my CPU cooler and reapplied thermal paste to no avail.

2.) After some searching I found these threads (thread 1, thread 2) which eventually talk about how AMD changed the way temperature are measured in Ryzen CPU's so they actually read around 20C higher then they actually are. Is this true?

 

[CountMike]

So I just built a new system with the Asus strix x570 mobo and the Ryzen 7 3700x. I'm using the scythe fuma 2 cooler. After install the temperature readings on this cpu vary pretty wildly at idle. I would say, after having the computer on for a bit, they spend most of the time between 40-49 C but often spike to approx 57C (though the highest spike I've seen was 67C but this was rare and it dropped back down shortly after). A typical temp reading on something like core temp might look like:

40C
49C
48C
58C
55C
53C
51C
49C
45C
56C
........

When running prime95 for about 20minutes or so my CPU temps do the same thing but jumps back and forth between 69C and 88C (88C was the highest spike), with most temps hanging out within the 72-79C range. My questions are as follows:

1.) Is it normal for a CPU to temperature jump like this? I already took off my CPU cooler and reapplied thermal paste to no avail.

2.) After some searching I found these threads (thread 1, thread 2) which eventually talk about how AMD changed the way temperature are measured in Ryzen CPU's so they actually read around 20C higher then they actually are. Is this true?

Temps are corresponding with load , nothing unusual about that. The reason they are changing with continuous load like in prime95 is it's throttling due to high temperatures, Ryzen works best at up to 62-65c, over 70 it already looses some of boost. Over 80c loss is substantial.

 

[drea.drechsler]

... so they actually read around 20C higher then they actually are. Is this true?

Yes, that MIGHT be true. That offset depends on CPU and BIOS version in your motherboard. But don't worry about that too much as temperatures are never accurate anyway, although they do follow processing load.

Your temps look perfectly normal to me (I also have a 3700x). The temp sensors are very close to the working bits in the CPU die, there are lots of them and it reports the highest one at any time. So when one core boosts you see it briefly spike in temp, then fall away as the boost goes down. The next core boosts and it goes back up and repeats. Only under continuous processing load will the spikes start averaging up into a steady reading.

The best temp readout is one that averages for you; RyzenMaster does that and HWInfo also shows an average reading. That average is a better indication of the thermal state of the CPU as a whole.

Your temps in Prime95 looks really good, averaging in the 70's as it is. Even the big jumps into the 80's is normal. Yes, the CPU will reduce boosting in that temp range but Prime95 is hardly a reasonable processing load.

 

[Avro Arrow]

As the previous answers said, those temps are more than fine. I commonly run my R7-1700 with a 20% overclock (3600MHz) when gaming and I've had it since March of 2017 with no issues whatsoever.

I'm receiving my new R5-3600X today so I'll probably wait until motherboard prices come down and start playing with liquid cooling to see just how far I can push it at 1.55v. Up to now, it has only ever used the enclosed Wraith Spire cooler.

Remember that CPUs are INCREDIBLY HARD TO KILL because all AMD64-based (x64) CPUs have heat sensors that are monitored by the motherboard BIOS. The BIOS knows what temperatures are dangerous for all the CPUs that the motherboard supports and won't let the CPUs get in that range because the BIOS will use the motherboard's power-control system to throttle the CPU and even shut the computer in extreme situations.

In the (so rare as to be almost unique) event that the BIOS control fails, modern CPUs have a built-in fail-safe called #THERMTRIP that performs an immediate emergency shut-down if the CPU reaches a dangerous temperature. It works like a circuit breaker and cuts all power to the core in a matter of microseconds.

 

[drea.drechsler]

...
I'm receiving my new R5-3600X today so I'll probably wait until motherboard prices come down and start playing with liquid cooling to see just how far I can push it at 1.55v.
...

I have to think 1.55V in a fixed all-core could be very quickly fatal for any Zen 2 CPU on the current 7nm node. Overclockers are seeing early degradation even with 1.325V in a fixed all-core. The current advice is not more than 1.275, but testing to determine your processor's FIT voltage for maximum loads is the best approach.

Not really on-topic I know, I just had to toss that in so nobody else gets notions that that's a feasible goal.

 

[Avro Arrow]

I have to think 1.55V in a fixed all-core could be very quickly fatal for any Zen 2 CPU on the current 7nm node. Overclockers are seeing early degradation even with 1.325V in a fixed all-core. The current advice is not more than 1.275, but testing to determine your processor's FIT voltage for maximum loads is the best approach.

Not really on-topic I know, I just had to toss that in so nobody else gets notions that that's a feasible goal.

You make a fair point and I agree. This is the "official" word from AMD but I wouldn't try it for long and I wouldn't try it if I was worried about losing the CPU. Maybe I'll just take it up to 1.5V, I know that it can handle that because I tried it once before with CineBench. Right after it was over, it was back to 1.325V. It was a proof of possibility more than anything. It did get pretty damn hot but oddly enough, it didn't get hot enough to throttle. Of course, if I kept it there for longer than the couple of minutes that it was, who knows what could have happened.

Funny thing, 22 years ago, CPUs ran at 2.2V:

NOW THAT'S JUST PLAIN CRAZY!!!!

 

[drea.drechsler]

You make a fair point and I agree. This is the "official" word from AMD but I wouldn't try it for long and I wouldn't try it if I was worried about losing the CPU. Maybe I'll just take it up to 1.5V,
.....

I'd like to know what you read that suggests 1.55V is OK. Even the 1.5V guidance I've seen is that it's algorithm will push volts that high but for single core boosting to max clocks, which ONLY occurs in light 'bursty' loads (as they put it.) To put that in perspective, it will only do it for a few milliseconds at a time, when temperature allows and when one SINGLE core is seeing a processing transient that merits it. Those conditions are hopelessly impossible for us to replicate in an all-core overclock.

AMD's position on "what's a safe voltage limit" for all-core overclocking of Zen 2 has always been "leave it in Auto".

That K6-2+ processor used AMD's .18micron processing node... or .180micrometer feature size. I don't know if it's fair, but compared to 7 nm that would probably be something akin to comparing a 000 bus bar against a 20ga hookup wire. Major current carrying difference, and far, far, far more tolerance for electron migration to work it's evils without impact to conducting path integrity.

 

[CountMike]

Before that they run at 3.3v, that's why PSU still has that voltage among others but I digress, higher voltage is dangerous only under high loads. I remember my first Ryzen, 1600x on Asus prime x370 with initial BIOS, first time I turned it on it started at 1.55v. That got fixed with next AGESA and with some negative voltage offset it dropped to 1.4v. Didn't do any damage though because under full boost it would drop under 1.22v.

 

[boagz]

Awesome, thanks everyone for you input. A big help for sure

 

[Avro Arrow]

I'd like to know what you read that suggests 1.55V is OK. Even the 1.5V guidance I've seen is that it's algorithm will push volts that high but for single core boosting to max clocks, which ONLY occurs in light 'bursty' loads (as they put it.) To put that in perspective, it will only do it for a few milliseconds at a time, when temperature allows and when one SINGLE core is seeing a processing transient that merits it. Those conditions are hopelessly impossible for us to replicate in an all-core overclock.

AMD's position on "what's a safe voltage limit" for all-core overclocking of Zen 2 has always been "leave it in Auto".

That K6-2+ processor used AMD's .18micron processing node... or .180micrometer feature size. I don't know if it's fair, but compared to 7 nm that would probably be something akin to comparing a 000 bus bar against a 20ga hookup wire. Major current carrying difference, and far, far, far more tolerance for electron migration to work it's evils without impact to conducting path integrity.

Ok, it has become apparent to me now that you think I'm talking about running my R5-3600X at 1.55V. I think that you only half-read my post and then went off in some crazy direction with it. I'll highlight the parts of my original post that made it clear I was talking about my R7-1700 and not my R5-3600X.

"As the previous answers said, those temps are more than fine. I commonly run my R7-1700 with a 20% overclock (3600MHz) when gaming and I've had it since March of 2017 with no issues whatsoever.

I'm receiving my new R5-3600X today so I'll probably wait until motherboard prices come down and start playing with liquid cooling to see just how far I can push it at 1.55v. Up to now, it has only ever used the enclosed Wraith Spire cooler."

#1 - "I commonly run my R7-1700 with a 20% overclock (3600MHz) when gaming and I've had it since March of 2017 with no issues whatsoever."
- I was already talking about how I overclock my R7-1700.
#2 - "I'm receiving my new R5-3600X today"
- This means that I hadn't yet received my R5-3600X at the time of the post.
#3 - "Up to now, it has only ever used the enclosed Wraith Spire cooler."
- Since I hadn't yet received my 3600X, I'was clearly talking about my 1700.

So, as for where I got the idea that original Zen could handle 1.55V was here:

Since its release, the Ryzen 1700 have been getting overclocked out of the box. But there is one thing that has been unclear even with some research, the max temp. Most people say stay under about 80-85 degrees, but no exact answer. So i sent AMD and email about the R7 1700 (which is what i have) and overclocking. Here is the email they sent back

The maximum warrantied voltages are programmed into the CPU when it is made. The maximum temperature for Ryzen 7 1700 is 75C.

For overclocking, we cannot guarantee safe operation as this action voids AMD warranty. A voltage of up to 1.55v can be used but requires very robust cooling solution (liquid cooling is suggested).

Thank you for contacting AMD.

This is why I was talking about using liquid cooling at 1.55V, because it's what AMD themselves said to do.
Here's the link to the post:
https://forums.tomshardware.com/threads/psa-ryzen-r7-1700-official-max-temp-and-voltage.2981631/
As for the AMD K6, that was meant to make you laugh because it shows just how far we've come. That's why I had put this at the end of it:
NOW THAT'S JUST PLAIN CRAZY!!!!
Now for some reason, the picture that I thought I'd posted didn't come up (doesn't matter, it was just a picture of an AMD K2) but it was for humourous effect, not because I was trying to justify pushing a 7nm CPU to 1.55V.

I flashed my ASRock X370 Killer SLI to accept the 3600X and I'll buy a cheap B350 board to use as the platform for my R7-1700 if and when I decide to see what it can actually do maxed out. Right now however, motherboards are too bloody expensive so I'll wait for their prices to come back down to Earth. As it turns out, the courier screwed up so I STILL don't have my R5-3600X. Grrr

 

[gingerrankin]

Hello,
This is a common problem or characteristic of the Ryzen 3000, due to the architecture of the same. I tell you about my experiences with my ryzen 7 3800x.
First of all, you need to rule out problems with cooling. For this, you must pass a stress process (prime95, cinebench ...). If it does not exceed 85ºC there is no problem with your cooler.
There are various procedures that serve to partially mitigate this behavior, but none fully resolve it. If you really want a quiet CPU you should look to Intel.
Not all cores in a Ryzen 3000 have the same quality. There is only one that achieves the maximum frequency promised by AMD. In light tasks, windows try to squeeze the preferred core to the maximum, even if this means raising the temperature momentarily. This does not endanger your CPU.
You should avoid the “observer effect”. Many monitoring tools stress the CPU themselves. You must use to monitor Ryzen Master or a tool adapted to Ryzen 3000 operation (such as HWinfo64 with 2000 millisecond refresh option). In no case should you use more than one tool at a time. Attention to the bloatware on your motherboard!
Adapt the CPU fan curve so that it is flat up to say around 65ºC with an acceptable value for you (for example 40%). The exact value depends on the quality of the CPU fan, so a better cooler will help. From 65ºC up to the maximum.
Update the latest version of the BIOS for your motherboard, AMD chipset and Windows. With the latest versions of the AMD chipset, special Ryzen power plans will appear to improve this aspect.
Additionally, a user has created a power plan "sz Ryzen Balanced V4" that promises lowers the temperature without affecting performance. You can search for information on the internet.
Using undervolting (negative offset) in stock does not usually help. Some users indicate that with manual OC + undervolting can help, but I have not verified it.
I hope I've been helpful.
Regards!

 

[John Chesterfield]

If you really want a quiet CPU you should look to Intel.

I'd agree with almost all of what you've said apart from this. If you're using the stock cooler with your 3800x, then it'd be no surprise that you'd think the 3800x is noisy. AMD bundle a cooler with their chip and Intel don't, so with Intel you need to buy a cooler anyway and it'd be quieter than any stock offering. If you're using an aftermarket cooler, no reason not to be able to tune that up to be silent.

The stock cooler is functional but not quiet. I've owned Intel rigs for the last 15 years before this year and yes, they were quiet. But that was because I bought a new cooler for them every time, just like I did with this Ryzen 3700x. I didn't even bother with the stock cooler and just whacked the Dark Rock 4 Pro straight onto it, because stock is merely functional.

As a result, the Ryzen is even quieter than all of those Intels I had. And they all had equivalent coolers, the i7 had a very good Phanteks cooler and my friend will be very happy with how silent that is when he gets it.

If you really want a quiet CPU, you need to get decent cooling for it should have been what you said, because that's the truth of the matter. It isn't about Intel or AMD.