Is this laptop overheating or just old tech?
Is this laptop overheating or just old tech?
Hi guys! First things first: here are the specs of my rig. My processor is an i7-9700K, and it runs from NZXT Kraken X72 to Asus ROG Strix Z390-F GAMING. The graphics card is a 2070 O8G from Asus ROG Strix RTX. I also have some RAM: Corsair Vengeance RGB with 16GB (two 8GB sticks) running at 3000MHz on DDR4, stored in Samsung memory using the EVO Plus drive. My storage has a mix of hard drives and an NVMe M.2 card from Toshiba P300. One stick is 500GB while another holds 2TB spinning fast with a 64MB controller. I also got some cooling sticks from Corsair RM750x and an NZXT H series tower in Window Black.
I started feeling unstable when I pushed the CPU to 5GHz overclocking mode for gaming. So, before trying to fix it all over again, I updated my BIOS to the newest version and that reset everything. Now I'm really struggling to get back to the same speed I used before. I thought maybe my CPU just got weak from age or use, so I wanted to know if something is wrong with the processor itself.
Last time things worked great: Multiplier was 50, XMP turned on, AVX and MCE stayed automatic, LLC was set at 5, manual voltage at 1.385v, VCORE @ load at 1.350v, VCCIO at 1.250v, VCCSA at 1.150v. Back then it passed 8 hours of Realbench, 2 hours of Prime95, and 64MB of Intel Burn Test & OCCT Linpack NonAVX without any problems.
But now? It's hard to pass 2 hours of Realbench or even do 15 minutes of P95 tests. My temps have always been fine on the GPU side, hovering between 50 and 65 degrees under load during gaming, so I don't think that caused any damage. So what happened over those last 10 months? Or am I doing something wrong here?
Thanks a bunch in advance! Best regards
New BIOS versions can mess with overclocking. You need to reset the system back to stock and make sure your RAM isn't making errors. Once you are sure about the RAM, test it with aida64. When everything looks good at stock settings, then start overclocking the CPU. A new BIOS might force you to start all over again. That's why you should only update the BIOS if there is a real big problem. The problem with your old stress test was that you didn't check the RAM with memtest when you set it to its highest speed. Sometimes, as you raise the clock speeds, something in the memory controller gets upset. With my Ryzen setup, I adjusted the timings on the RAM and overclocked it to 3800 MHz. It looked fine at stock core frequencies. But once I turned on PBO EDC bug, the whole RAM overclock went unstable. So I reset to stock settings and ran memtest again. Every time I did memtest, I got random results. Different settings would sometimes pass or fail depending on what they were set to. What I learned was that I had left the Termination block and CAD_BUS block timings for the motherboard to auto-detect as they were exactly the same as the DRAM Calculator for Ryzen version 1.6.2. Sometimes when I turned on the PC, everything worked fine at first, but then the RAM would fail the test with errors. It took a lot of testing to find the right stable settings so that every time I boot my computer it stays safe. I could even get much tighter memory timings afterward. Now I can reach higher frequencies at very low vcore values. At first I thought I had ruined the memory controller or the cores, but it was just the settings from DRAM Calculator for Ryzen 1.6.2 that were wrong. The latest DRAM Calculator for Ryzen agrees with my fix now.
Unstable settings: Stable settings Issue appeared to be fixed by changing CAD_BUS ClkDrv from 24 to 40. This is power in one single setting.
Thanks so much for your reply! My system was unstable anyway, so I didn't mind updating the BIOS to get a fresh start. Yes, back ten months ago when I first overclocked (OC), I never tested my memory properly; at least not in a way that guaranteed they would stay stable. However, recently I ran into some instability again. So now I turned off my OC and tested the RAM both with and without XMP enabled using Memtest86, HCI Memtest, and GSAT. All of them show no errors whatsoever. I haven't even tried testing my RAM right next to my CPU's overclock yet. I guess PBO EDC is like AMD's version of XMP?
Here are some tests I did all night:
- Multiplier set to 50
- XMP enabled
- AVX Auto
- MCE Auto
- LLC set to 5 (low latency)
- Manual voltage at 1.390v
- VCORE at load with voltages of 1.350v through 1.359v
- VCCIO at 1.250v
- VCCSA at 1.150v
For thermal, I ran Prime95 v26.6 small FFTS for 15 minutes and it passed. The max temp was 97C. For stability, I ran Realbench v2.56 for 8 hours and it passed with a max temp of 92C. Basically by increasing the voltage by just 0.005v, my system passes all tests without WHEA errors or BSOD crashes. What would you suggest moving forward? Should I test my RAM using the CPU's overclock plus XMP? Or should I reconfigure the timings myself like I did before?
UPDATE: I've done an hour of GSAT now with the CPU overclock and XMP enabled, and there are 0 errors! Since I read online that GSAT is stricter and faster at finding faulty RAM than Memtest, I decided to use it. Thanks for the advice!
PBO EDC is just a fancy way to say "overclock my AMD CPU." With my 2500k, I had to raise the core voltage after some time passed. Everything has been working well so far. The silicon lottery keeps eating up power; it's running on 1.387V on the core. [1] [2] If you're worried about things getting worse over time, back off a notch instead. A 1.39 Vcore seems to be safe for now. [1] It might be that your overclock was just barely stable enough at first but got unstable later and needed more voltage to fix it (a new BIOS could cause this too). Hopefully, from now on, it will stay stable. I use Aida64, prime95, and memtest to test things. I don't want to push VCCIO much over 1.250V for 24/7 [3] [1]. https://forums. [2] https://siliconlottery.com/pages/statistics [3] https://www.tweaktown.com/guides/8481/co...ndex5.html
Thanks so much for the info. You can see that my OP max temp was 97c. I thought my fan curve in my system was completely broken, and some of my fans were only running at 40% while the CPU was hitting 97c. After I adjusted my fan curves and ran more tests with 5ghz @ 1.390v (at 1.350v under load), I passed two realbench sessions. In realbench and p95 small ftts, the max temps are now 87c. That gives me a 10c drop under load, which was really cool. But I didn't know this yet; thermals can also cause instability. I thought thermals would only throttle you by slowing down your processor, not causing weird crashes or BSODs. So I started wondering if my system could be stable with lower voltages thanks to better cooling. Besides that, I lowered VCCIO to 1.20v and VCCSA to 1.15v - no errors in GSAT. However, it won't pass IBT when I lower those voltages anymore. So I'm not sure what to do about that for now. Also, is there a free version of Aida64?
Cool post on reddit about safe vcore levels. View: https://www.reddit.com/r/overclocking/co...erm=dso1b8
Thanks for giving me the link. After realizing my fan curve was totally bad, I fixed it and now get much better results. It turns out that too much heat on my CPU made the system unstable even when running at low voltages. By cooling it down better, I managed to lower my voltage while keeping the same speed (5GHz).
I tested using Prime95 small FFTs for 15 minutes and Prime95 large FFTs in-place for one hour, plus Realbench for 8 hours. The results are:
- VCORE @ Load Temp (max/avg): 1.385v / 1.350v -> 87c / 85c
- VCORE @ Load Temp (max/avg): 1.375v / 1.341v -> 86c / 84c
- VCORE @ Load Temp (max/avg): 1.365v / 1.332v -> 85c / 83c
- VCORE @ Load Temp (max/avg): 1.355v / 1.314 - 1.323v -> 84c / 81c
In both tests, my power went in fine at the lower voltages. However, Realbench failed after about 6.5 hours with a BSoD (Blue Screen of Death). So now I've increased the voltage to 1.360v and am testing it too. Hopefully it will pass Realbench for 8 hours. When/if that happens, I will check my RAM using GSAT, TESTMEM5, and Aida64 Extreme.
I'm amazed at how much I can get my temps down just by adjusting the fans, which increases stability. I never thought heat would make your system unstable before testing to be sure you stay within the limits of tjmax while running long stability tests without risking damage. But lowering my CPU from 97c to around 85c has actually brought stability with lower voltages now.
Heat makes both your CPU and DRAM unstable when you push them too far for speed. It also breaks other parts like VRMs, which causes the whole system to shake or crash. I spent a lot of time trying to figure out why my PC wouldn't stay stable while tweaking timings in the BIOS until it clicked: it was all about how well heat got out from around the VRMs and DRAM. The solution was just adding a fan right at those spots. Even moving the whole computer to a room with no air conditioning could ruin things if you go too hard on overclocking. RAM timing, especially tREFI, is very sensitive to heat.
Cool idea, so what type of fan do I need to stick near my VRAM or RAM for the best speed boost?
I taped a 120mm fan tight to my PC case and put it above my memory sticks so air always flows through them.