AM4 R 1700 VRM MOS Temps too high?

Toggle sidebar Toggle sidebar
N

nobspls

Reputable
Mar 14, 2018
902
12
5,415
I recently put together a AIO water cooled build using the following:

R7 1700
Gigabyte AB350M-DS3H
Cooler Master Masterliquid Lite 120 (AIO)

I am overclocking the R7 1700 to 3.8 Ghz. The Gigabyte board despite its misleading name is actually a X370 board, but only lets set +/- voltage to DVID core (so dynamic voltages).

In any case I got stable setup with +0.096v at 3.8 Ghz. The issue is with the VRM MOS temps. I got a 80 mm blowing on the VRM.

I am running Prime 95. If I do the standard blend test, the VRM MOS reported by HW monitor ver 1.35 (https://www.cpuid.com/softwares/hwmonitor.html)

as well as Gigabyte SIV (hardware monitoring info) both reports about 79C in this scenario. The CPU temp itself gets to about 65C. Which seems just fine.

However if I have Prime95 do the small FFTs test (stress out that AVX crap), then VRM spikes all the way up to 107C but the CPU is only reporting 75C.

Most will say 107C is too high correct? But does the small FFTs really count? I do not dare overclock any higher because it seems this might be a limit.

Any one have any comments or suggestions? Advice?
 
Solution
n0ns3ns3
small FFT is not an anomaly, it's a power virus.
there are not many real life usages that can pull that much power. It is mostly used for stress testing to ensure that system is stable even under this extreme load.
If all you do is gaming, you have nothing to worry about.

Regardless of the chipset, this board is on the "low end" side in terms of power capabilities.

as for the VRM phases count it's a bit tricky. you start with counting the inductors - those are the black cubes around the CPU. on your MB it's 4 to the left and 2 above CPU socket. that means that it's either 4+2 or 3+3. the first number is for powering the CPU cores and the second for the rest like memory controller and iGPU. I guess that it's 4+2 as it was probably...
Sort by date Sort by votes
n0ns3ns3

n0ns3ns3

Illustrious
May 25, 2016
8,065
4
42,715
This board is just a typical B350 with 4 phases (core) VRM.
This is not enough to power overclocked 8 cores in continuous stress like small FFT. probably for some encoding tasks as well.
the 107C is way too high. the VRM max temp is somewhere in 105-125C depending on the model. The problem is that most probably the temp you see is not the actual, but one reported by the thermal sensor in the VRM area. it can be significantly different from the actual.
also, is the VRM temp tends to climb for very long time. for example my system will shutdown (due to VRM overheat) after about 60-120 minutes of "abuse" with max power draw on CPU.
you can somewhat improve the temp by providing good airflow over the VRM area.
 
Upvote 0 Downvote
N

nobspls

Reputable
Mar 14, 2018
902
12
5,415
BTW the CPU Vcore reported by CPU-Z is only about 1.296V for the max value I am seeing.

Also the board is an X370 not a B350. See:
https://www.gigabyte.com/Motherboard/GA-AB350M-DS3H-rev-1x#sp
O7DNUtD.jpg


And how do you determine how many phases the mobo is doing?

And for normal gaming loads, the CPU does not even crach 50C the VRM report around 55C. It seems those small FFTs are really an anomaly.
 
Upvote 0 Downvote
Z

zyh1987

Estimable
Mar 16, 2018
2,146
9
3,065
You will need a mid range x370 board to support a oc 8 core. Which means you will at least need either proper 6 phase vcore vrm or a mimic 8 phase ( 4 phase with double amount of hi /low mosfet, choke, cap) anything lower than that you either not getting enough power output, or shoots voltage regulation or vrm over heating. For benchmark, you can temporarily boost the clock, but for prolong use, I would suggest get a better board, if you are budget tight and still want to maintain the oc, you can zip tie a fan and let it blow air directly at the vcore vrm which is between your up plate and your cpu.
 
Upvote 0 Downvote
Z

zyh1987

Estimable
Mar 16, 2018
2,146
9
3,065
The best way to measure voltage is to use a volt meter. CPUz shows you the voltage request by cpu, whether your mobo can deliver, it’s a different story.

To count the phase, it’s tricky, first you would need to remove the vrm heat sink, count the number of phases as a whole, then you will need to figure out the vrm is for vcore or for soc, you can do that by measure the resistance between the choke and the cpu vcore caps if it’s zero then it’s vcore vrm if not it’s soc. then you can count the number of driver chip To determine the true phase. Occasionally, you will find no driver chip as the vrm controller integrates them, then you would need a spec sheet to make an educated guess. Keep in mind that most vrm controller produce maximum of 8 phases.
 
Upvote 0 Downvote
n0ns3ns3

n0ns3ns3

Illustrious
May 25, 2016
8,065
4
42,715
small FFT is not an anomaly, it's a power virus.
there are not many real life usages that can pull that much power. It is mostly used for stress testing to ensure that system is stable even under this extreme load.
If all you do is gaming, you have nothing to worry about.

Regardless of the chipset, this board is on the "low end" side in terms of power capabilities.

as for the VRM phases count it's a bit tricky. you start with counting the inductors - those are the black cubes around the CPU. on your MB it's 4 to the left and 2 above CPU socket. that means that it's either 4+2 or 3+3. the first number is for powering the CPU cores and the second for the rest like memory controller and iGPU. I guess that it's 4+2 as it was probably build with CPUs without iGPU in mind.
For mother boards with higher number of inductors there could be few options:
1. true phases - up to 8 (i'm not aware of voltage controller capable of more than 8 phases).
2. almost true phases - any usable number. same as above, but using doublers to split the pwm signal from voltage controller to the phases.
3. fake phases - double the amount of components without actual split. so 2 sets of inductors + mosfets are connected to the same line acting like a single phase. a bit better efficiency and temperature.

someone created a thread here on Tom's with a table listing the MB, chipset and the VRM arrangement. you can try to search for it.
if you wish to see numbers, check out this trash talk about B350 motherboards https://www.youtube.com/watch?v=ZGrxhf_xZWI
This guy is exaggerating a bit since he is looking at it from a pro overclocker point of view (max OC that should pass stress test), but you can have an idea about what is not that good.

There is also great importance to the actual components used to create VRM as they differ in temperature limit, current capabilities, efficiency and other features like thermal protection, switching frequency etc.
 
Upvote 0 Downvote
Solution
N

nobspls

Reputable
Mar 14, 2018
902
12
5,415
Well I am trying to do this stuff on the cheap and see just how much OC I can get away with.

I got the R7 1700 for basically $170 with another $30 rebate from AMD, so essentially $140 for the CPU
The Gigabyte board was only $70, an the AIO was $40 with $10 rebate so $30.

The whole package is basically $240. I do not plan on running heavy loads 24/7 on this thing, but the last thing I need is the thing to randomly keel over in the middle of gaming. I might occassionally run over night video compression transcoding on it, but probably nothing more than 4 hour stretch at a time.

If the board is worn out after 3 years of use like this, I'll call it good.
 
Upvote 0 Downvote
n0ns3ns3

n0ns3ns3

Illustrious
May 25, 2016
8,065
4
42,715
you had an excellent deal.
you can test the VRM temps during the "video" load (whatever you do with it). If it's below 80C you have nothing to worry about.
as for longevity, the usual suspects to die are capacitors. they are typically rated for 5-10K hours at ~100C. every 10C drop practically more than doubles the lifespan. So 3 years of non "production" workload is non a sci-fi XD
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom