Diagnosing VRM cooling issues/When water cooling isn't the best option

So I recently upgraded from my AMD FX-4300 to the FX-8350 for cheap; I put it into my system and since I didn't have a 3rd party cooler, I slapped on the stock cooler. Worked fine apart from from sounding like a jet engine about to take off. I could finally play DOOM at 60+ FPS on ultra and CS:GO at an acceptable framerate!

Eventually, I got tired of being deafened every time I turned on my PC so I decided to upgrade to an actually good cooler. After browsing around my local PC part retailer for a bit, I decided to get the Hyper 212X, but it turned out the site hadn't been updated and they were actually out of stock, so I got the Coolermaster Seidon 120V, an entry level water cooler (Also, it's compatible with AM4, Cool!).

There were some issues with installation, It would only fit in one way which meant I couldn't put the side panel on:



That's not really relevant, I just wanted to share my pain...

Anyway, I quickly started up CS:GO and was able to play without headphones for a good few minutes until IT happened. IT being the framerate dropping down to 30FPS then returning to normal after a few seconds. I brushed it off at first, these things just happen sometimes. Then it happened again, and again, and again at which point I'd just had enough and ragequit.

At first I thought it was the CPU overheating, I've heard horror stories of people buying AIOs and having super high temperatures from not putting it on right, so I opened up a terminal window and bought up 'sensors'
watch -n 1 sensors
got back into CS:GO and started playing (With bots). I watched the cpu temps rise up to 70 degrees then violently dropping down to 64-ish corresponding with the severe FPS drops in-game. 70 degrees is a bit warm but not critical enough to warrant throttling. so I searched around looking for others with this kind of problem and found this post on Toms Hardware. OP also had the same CPU and motherboard as me.

I scrolled past the obvious "GPU bottleneck" post and saw the the phrase "VRM throttling". Suddenly, it all made sense! The reason it was working fine with the stock cooler and not the water cooler was because the stock cooler had a fan pointing straight down at the CPU and VRMs while the water cooler didn't, providing absolutely no cooling to anything apart from the radiator! Luckily, I had a spare case fan and molex to fan connector adapter since the board only has 1 case fan connector and this happened:


It may look ugly, but, none of my games had framerate drops anymore and from where I sit, I can't even see it.

I did a few tests after the fan "installation" and in 'sensors', when I removed the fan the 'Vcore voltage' section showed dips from the normal 1.44-ish Volts down to 0.8-ish and after putting the fan back, the voltage stayed normal even after a few rounds of CS:GO. The CPU also rose up to 74 degrees, higher than before I put in the extra fan. That's fine though. Better than my i7-920 which got to 90 without an overclock.

first things first:

Power requirements (this cpu is from 2012, and is power starved; especially when you oc)
Thats why your VRM got sot so hot around your cpu.

Nice modding.

Next time get a good motherboard. Take it as a base of everything. If its crap/cheap, everything else will also be at that level.

Also watch this. If you're gonna run tests, they need to run for at least a minimum of 1 hour.

and unfortunately,

You're wrong. AMD stuff of that age wasn't meant to work well beyond 70 C. Since after awhile your temperature gradient max out and get near your max temp, you will either need to turn off your overclock, or undervolt your CPU, or get a better cooler as the current one isn't cutting it. You will need to upgrade to one of the huge Noctuas or get a 240 radiator AIO.

Yep 70°C is indeed pretty high for a FX cpu.
FX cpu´s dont have individual temp sensors per core, only 1 temp sensor for the whole package.
Socket temp should stay below 65°C

A friend of mine is fine with his single 120mm radiator on an overclocked FX-8350.
The only difference I could spot is that your mainboard has no VRM cooling to speak off. So either you mod some cooling on the chips next to the grey blocks, or invest in a new mainboard. And with Ryzen arround the corner, there are really nice boards for fairly cheap now.

No vrm cooling is a scary thing for FX 8350. Ideal is about 55 under load. Your cooling is not good enough. Expect a short life for the mb. Amd temps sensors are mostly terrible and unreliable and second 70 c is gonna cook stuff. At 70c you are not factoring in load and temps spikes. Not an ideal. Might work for awhile but the piper is coming.

Wenn it comes to the vrm implementations and temperatures.
This highly depends on which types of mosfets they have used on the board.
Its allready an older board in this case, so i think that wenn overclocking they will probablly trottle at a certain point cq temperature.

2 Likes

Alright then, how about this plan:
1. disable some cores/underclock/put FX-4300 back in
2. Play DOOM on Medium
3. Get a cheap/used actually good AM3+ motherboard as soon as I can afford one.

OR

  1. buy some tiny heat sinks from aliexpress/ebay/whatever and put them on the VRMs
  2. Keep playing DOOM on ultra
  3. Upgrade to Ryzen after VRMs cook themselves

Also, my motherboard is apparently rated to use 140W CPUs. OK then...

You could try buying some GPU VRMs then. They usually come in packs of like 10 and just stick on with thermal adhesive.

Pack of 10? That's cute.

There were some motherboards that had issues with 125 W TDP AM3+ CPUs and specified them to only be used with downdraft coolers. Those were 970 chipset motherboards.

I run a Noctua NH-U14S on my 8320E (95W TDP) and I have neither with it nor with my FX4170 (125W TDP) seen anywhere near the temperatures you are getting (tested with H.264 encoding under Linux, also reading temp via sensors.

You could get a PCIe slot fan bracket

That will bring the temps down as well.

I was talking about the copper ones.

As already mentioned, FX CPUs do not suffer heat well and your H2O cooler has inadequate capacity for the application at hand. For perspective, even after extended gaming sessions, my CPU and GPU rarely go above 39C. I'm using a pair of 280mm radiators to cool a X99 CPU and a GTX780. IMHO, if your water loop can't keep things under 50C, you need to fall back and regroup.

As you've found, VRM cooling is also important, especially when overclocking, which is why VRM water blocks are available for the more popular gaming motherboards. A well-aimed auxiliary fan, however, should be sufficient for your use.

You may also wish to revisit the overall fan/cooling/air flow situation in your case. Do you have unobstructed flow (hard disk blocking intake?), does the flow move in more, or less in a single, uniform direction, are you using a positive case pressure model, etc?