Temp issues on stress test

Associate
Joined
21 Dec 2019
Posts
19
Hello all, I am new to the forum and still relatively new to PC gaming and overclocking. My setup consists of:
MSI Meg Ace x570
Ryzen 7 3800x
Aorus RTX 2080ti waterforce Xtreme (AIO Version)
4x8gb 3600 Corsair Vengeance RGB Pro
Corsair H110i pro
Phanteks Enthoo Evolv X.
Front fans all Corsair LL 120's.
Top fans are ML 120's
I have left the rear as the Phanteks 140mm that came with the case.
EVGA Supernova 750w PSU

I recently bought the Evolv case as my last case (Corsair Spec Omega) was a bit cluttered after getting the 2080ti. This time around I decided to try manual overclocking rather than using the onboard button. Following an MSI video guide, I first set my Precision Boost to enhanced mode 2, testing each one first. I then started to manually overclock the CPU and voltage. I feel this went well, increasing the clock speed slowly and then testing at each point using Cinebench R20, Heaven and even Memtest on the RAM as I was also slowly increasing the FCLK frequency (only after CPU clock speed was proven to be stable so that I knew which parameter was stable).

All was good, system stable, no bugs or crashes and temps were never over 70°C during these tests. I got up as high as 4.4Ghz @ 1.37v before I remembered that I should also stress test the system. Using Prime 95 on a blended test, it was only a matter of minutes before my temps on the CPU were reaching over 90°C, at one point spiking to 100°C. Luckily I was monitoring and was quick to stop the testing..I started reducing the clock, then voltages to check for any improvement.

Currently sitting at 4.125Ghz @1.375v - temps during Cinebench, heaven etc never exceed 60°C. However 7mins in Blended Prime 95 stress tests, temps are rocketing up into the 90's. I think I will next try returning all to default and get a base line on Prime95 as I hadn't done this first time around. But I am wondering if any of you have any ideas as to why I am having such issues with temps?

The system was cleaned only 2 days ago as this was when I rebuilt in the new case. points of note: 2x fans mounted on the Corsair AIO are controlled by the H115i through icue. I have monitored these and are increasing speed with temp. 2 out of the 3 front mounted fans are controlled by the GPU, as these are mounted on the Aorus AIO. The additional front fan and the rear fan are set to PWM and smart fan mode, controlled by the CPU temp. These also increase speed with temp. I also applied the thermopaste this time around using the X method rather than the pea size dot.
I am looking into a proper liquid cooling setup, however I do feel my system should be dealing with temps better than it is now as the system stands.
 
OK so with a little investigation, looking into posts on this forum regarding overclocking CPU voltages, I noticed that most of you are between the 1.27-1.32v range. The video I watched from MSI, the guy defo recommends 1.37v mark. After lowering initially to 1.31v, keeping the CPU clock at 4.125Ghz I ran a quick stress test, after 15mins, my max recorded temp was 87°C. A huge improvement. I think a little more tinkering and learning is required. Still thinking about getting a liquid cooling setup. Got some stuff to sell first.
 
Not sure on AMD but the newer Intel CPU's with AVX will run very hot/throttle with new builds of Prime95 (unless they have a different version for this now), I just stuck to the older build before this change.

To run on the newer builds Intel users had to set an offset for AVX.

Thanks bud. Looked into AVX since you mentioned it and seems to be a common issue. I will do a few runs tomorrow with it disabled and see what the difference is. This raises a new dilemma for me. If the small FFTs part of the test is designed for max heat max load, and my CPU has a common issue with this test, but has been proven stable on all other benchmarks and parts of the test, then do I take the CPU clock and voltage as stable or not. The heat and loading applied during the small FFT part of the test will never be seen during normal use
 
Not sure on AMD but the newer Intel CPU's with AVX will run very hot/throttle with new builds of Prime95 (unless they have a different version for this now), I just stuck to the older build before this change.

To run on the newer builds Intel users had to set an offset for AVX.

What version excludes the AVX?
 
what does the temp get to with your normal use as that's the only important temp you have to worry about.

Your voltage is pretty high and with only a H100i pro on the cpu I'm not surprised you are getting such high temps with a prime stress test.

My temps remain below 60°C mostly, rarely going above. This is at 1440p on Ultra settings. I have also carried out testing using the following, Prime95 with AVX off, Cinebench R20, Memtest, OCCT (only a 30min test) and heaven. Temps are usually in the 65-75°C in the stress tests but below 60°C in benchmarks. No crashes or bugs. No errors found. That is at 4.2Ghz and 1.31v. I will run a long stress tomorrow to make sure the system is OK.

Definitely going to purchase a custom loop though, but it's all time and money.
 
OK mate, but you also said that with just the H110i you weren't surprised that I had high temps. So with a custom loop (which will have EK SE 240 and 420 rad's) you're saying I won't see much difference? Is there any point in going with full custom loops over AIO's except for aesthetics?
 
won't be much difference. got a custom loop on my 3900x and 5700XT. had AIO on it before. Very little difference due to the chiplet design and 7nm.

personally wouldn't go with EK though.

Any reason for that mate? I was going to get Hardware labs rads etc. But thought it would be easier just to get everything from the one supplier.
I suppose the custom loop with bigger rads will be able to handle/dissipate more heat over the AIO's.
 
expensive. looked at it before I got my system. pump isn't as powerful as other D5's. Paying for a name there. with the pump make sure you have the inlet/outlet you want.

I've decided against a custom loop. For the amount I get time to play on the pc (demanding job and demanding family ) it will just be an expense for the sake of it. So I have just purchased a better AIO - Thermaltake Water 3.0 ARGB 350mm. Should see a big difference over the H100i.
 
Last edited:
Well that is personal preference. I am over the moon with the Thermaltake. Huge improvement in temps. Even running AVX on in Prime95, temps are below 80°C. In gaming they remain around 48°C and idle is about 23°C. Done a 2 hour test using Aidia64 and the highest recorded temp was 72°C, running at 4.3Ghz and 1.3v
 
Thanks very much for the info and recommendations, I will have a go with the settings you recommended. I have used Intel Burn Test as well, just forgot to write about it. Tried on standard, high and very high, all passed without issue.

Can I just ask, what basis are you guys saying the Thermaltake AIO isn't that good? I looked into quite a few reviews, all saying it was great. Certainly better the h100i pro, and beating the likes of the Kraken and Be Quiet AIO's in some circumstances. I knew it wouldn't be the best on the market, but thought it would be one of the better ones at least? I know every system is different and there are variances in comparisons
 
Sorry I meant the Corsair H110i wasnt much good, pretty much an entry level AIO, id imagine the Thermaltake one was much better, as for running IBT, anything other than standard, ie high, very high, extreme, and custom, just tests more RAM, hence why its slower, its still the same amount of abuse as standard settings on the CPU.

OK bud that makes sense. Anyway I put in the settings you recommended and my system was stable throughout testing. However I did notice that my voltage was sitting at 1.33v during the IBT on LCC mode 5. I lowered the CPU voltage slightly to 1.37v and then LLC mode 6, and I am getting around 1.25v during the IBT. Is this OK or what would you do differently.

Can you explain to me, what the benefits of overclocking via the LLC are over just setting a lower CPU Voltage?
 
Last edited:
Back
Top Bottom