[FIXED] High clock speed but too little voltage for the chip to handle safely.
[FIXED] High clock speed but too little voltage for the chip to handle safely.
Hey everyone, it has been a while since we saw your post. Here is my setup: Thermaltake Chaser MK-1 case with an i7-5820k processor (speeds between 33 and 36), a Thermaltake Ring Silent 12 Pro cooling fan on the CPU, X99 Deluxe motherboard, 40GB of DDR4 RAM running at 2400MHz without enabling XMP features, a Corsair HX1000 power supply unit, a GTX 1080 graphics card, two 120mm fans blowing air in from the front, two 200mm fans pulling air up from the top, and a 140mm exhaust fan on the back. I need someone to check if my overclock is working properly.
I was being very careful with voltage, staying under 1.25 volts (Vcore) because I found myself tired when using Mod Edit sometimes haha. Yesterday, I changed the thermal paste to Arctic Alumina and thought why not? I had been sitting on temperatures of 45, 45, 44, 43, 42, 42 all at Vcore 1.22v (using hexi-core) for a few days without any problems. So I tried hitting every single core at 46 and bumped the voltage up to 1.30 volts while keeping VCCIN on auto set to around 1.920-1.936. Looking back, I already knew that 45 @ 1.3v was pretty fine. It wasn't a problem. So I tried going higher again at 46 cores with 1.290 volts. Again, no issues (from what this average person can tell). At 47 cores with 1.280 volts, everything stayed stable. But then at 48 cores with only 1.275 volts it wouldn't even boot up, so I guess we are almost at the wall limit, but I don't think I need to push higher than 1.3v for 48 or more cores since temps started leveling out at that level.
My other settings include turning off C-states in my ASUS MCE and keeping EIST enabled. I also use Balanced power mode in Windows, and all PLL (processor clock) and LLC (memory clock) are set to auto with no other changes made to the BIOS. Here is where I admit some fault. I used to do stability tests regularly but just don't enjoy them anymore. Sometimes I game, otherwise I make websites or handle general bookkeeping work, which isn't that heavy-duty stuff. So now I skip these stress tests entirely. I've tried RealBench, AIDA, P95, IBT, and plenty of others before, but I don't like how they torture my system when I know it won't hit anything near those limits with real usage. However, I still use AIDA and Prime to check for maximum temps.
I noticed something interesting with AIDA though. When I test at stock settings (36 cores @ 1.07v), the timer counts up four numbers before skipping one (like counting 1,2,3,4, then pausing, then going to 6,7,8,9, and pausing again). This happens every single time without fail when using stock settings. So now I use that pattern as a benchmark. If I push the overclock too far, this timer starts slowing down; however, if I open AIDA, start it up, and get temperatures level out just like at stock (taking 4 seconds then a small delay), then consider it good while gaming with games like GTAV or FH3 for a few hours without worrying about instability appearing. Have I finally hit the jackpot here? Is it 47 cores @ 1.275v running fine on air? Or have I missed something, or am I just massively unstable and haven't figured that out yet?
My PC works like lightning fast; I wait for nothing at all because everything loads immediately right away. Temps max out at 75c under stress while gaming drops to 65c. At least, I feel good enough to drop the voltage a tiny bit more while keeping my stability intact. Is my AIDA theory wrong? Basically, if the CPU can tick that AIDA timer along exactly like it does at stock speed even when getting hammered by 100%, then that's a clear sign of stability. Thanks for reading this and for any help with confirming where I stand on this overclock check. Cheers!
Sounds good, voltage looks great too, and the temperatures seem okay right? Just a few things: Have you checked the voltage while running a stress test? You might set it down at 1.275 volts, but the motherboard (for some reason) pushes higher voltages. Maybe you can get by with lower voltages on other parts like vccin and system agent; auto settings are always very careful and if you try that, you could end up with a cooler temp and less power use. For stability testing? It really depends on you, so just use the PC normally if you don't see an BSOD or any weird behavior (checking system events is fine too). If everything runs smoothly for your normal work, then it's mostly okay. But you should find out...
So yeah, the voltage looks good and the heat is fine too. Just a few quick notes: Have you checked if the voltage stays right when you run stress tests? You might set it to 1.275, but sometimes your motherboard pushes out more than that on its own. Sometimes other parts like VCCIN or the system agent can handle lower voltages without breaking things. The default settings are usually super cautious and will keep things cool and save power. As for stability testing, you should probably just use the PC normally if it doesn't crash with a BSOD or act weird. Checking system logs isn't bad either. If everything works fine under normal conditions, that means it's stable for your usual tasks, which is mostly okay. But to be sure, try seeing if things break on demand instead of waiting to fail later. Maybe take a rest period for stability checks instead of just running Prime95 all day long. I don't think you really need 24 hours of Prime95 unless you want. In my experience, running the realbench stress test for about half an hour and passing is a great sign that things are stable. Honestly, AIDA and Prime95 aren't super good at checking stability by themselves; they're mostly just looking at max temperatures. I had some overclocks passing Prime95 for 24 hours but crashing right away when playing games or doing realbench. Hope this helps!
@ tshinhar thanks for your reply mate. It is definitely not as stable as I thought because it booted fine this morning but had a WHEA error on the first restart. So I lowered the clock speed down and also tweaked the voltage slightly, now it's 46 @ 1.265. The restarts are smooth so no problems yet. Good questions.
I have been watching temperatures but not voltages under load, although in saying that I do check max voltages after testing and they are always very close to what I set in the BIOS (so about 0.016v off). That's probably just because of LLC/PLL doing its thing and it is still well safe.
I use HWInfo for monitoring. VCCIN I've tweaked, it seems to sit between 1.8 and 1.94 no matter what. So I am thinking I will set it from Auto to 1.94 and just leave it alone. At least I know I'll be slightly over max VCCIN at load so should be good.
SA i'm a bit unsure about this. I have quad channel, 8gb x 2 and 4gb x 6 (I got all these parts for free in a shopping bag and just built it with all the RAM there, should take some out) but not sure of a good range for SA or exactly how it will affect my RAM. Got any advice?
Pretty sure from memory SA hovers just under 1v. I keep an eye on event viewer, nothing unusual thankfully. Will do exactly that.
I like Realbench, just don't like my GPU getting hammered at the same time...but I guess that's true for testing. I do half hour runs and check. Thanks again for your time, much appreciated!