First off, very sorry I've been inactive for past month, now that I'm back :
Originally Posted by TwoCables
Please don't buy used PSUs unless you are absolutely 100% certain you can trust the previous owner to have taken good care of it and that they aren't selling it because it's not working anymore.
Also, you can't just say "I'm going to buy [brand name]" for PSUs. That's asking for trouble. Corsair doesn't make their own PSUs, and neither does EVGA. They use other manufacturers, which we call "OEM" manufacturers. Even SeaSonic is starting to do this.
Before you buy any PSU, please ask us if the choice is safe. The quality of a PSU cannot be judged by the brand name.
I was on very tight budget, and all I could afford at that moment was EVGA BQ600W Bronze+, I've bought that one, used, previous owner gave me warranty (from shop that he bought it), still have something like 1 year of warranty I belive.. and for month of use, it haven't failed me. I've should ask here first, but simply I didn't have enough money to be picky... I know that you're gonna say PSU are not something to be cheap on.. and I agree fully on that one.
Originally Posted by shilka
EVGA has so much rubbish these days that the number of good series is much smaller than all the rubbish
Yeah, mine falls under "unknown" category sadly.. but it's working as intended, haven't failed me for month of usage.. and PC was almost 24/7 on.
Originally Posted by ColonelBlimp
Did that first post say you need 8gb to run Apex Legends with Ultra textures?
Is that true? The 2060 cannot use Ultra textures?
Yeah, I meant Ultra as far as settings can go, so everything on ultra + textures on "Very High" because of 6GB VRAM. RTX2060 is also 6GB VRAM so it will be same, "very high" textures.
Originally Posted by ssateneth
I didn't read most of your post, just the symptoms of your crash.
I will agree that it is not the GPU. Your GPU is also much faster, putting more strain on your other parts to "keep up".
A bad GPU would throw out either artifacts on screen, or DXGI_DEVICE_REMOVED crashes from games and the device driver resetting in order to recover (Basically the GPU would be locking up and device driver reinitializes to put the hardware back into a working state).
I do not think it is the CPU core either. CPU Core issues would usually spit out rounding errors in Prime95, or CPU_WATCHDOG_TIMEOUT errors. Old P95 is not likely to spit out errors, and there is no reason to use old P95 anymore. Grab a copy of the latest P95 version 27.1. You can disable AVX2 and AVX in the new P95's torture test, but for the purpose of seeing if it spits out errors, at least have AVX enabled. It can't hurt on stock CPU settings. Do "Small FFT's".
It probably is not your memory controller either. Memory controller will spit out MEMORY_MANAGEMENT, SYSTEM_SERVICE_EXCEPTION, or some amount of memory related STOP codes like PAGE_FAULT_IN_NON_PAGED_AREA or ATTEMPTED_WRITE_TO_READONLY_MEMORY. As you said you don't get BSOD's and instead full freezes, we can skip this one.
By proxy, it's probably not your RAM either. They would spit out the above STOP codes, plus your memtest86 should've spat something out.
I do think it is most likely either related to your Uncore (L3 Cache) or System Agent (VCCSA voltage or related SA settings). Having too high uncore clock or bad uncore voltage can cause lockups without errors (It'll just seize the CPU). The uncore/cache voltage is driven directly from CPU core voltage on mainstream skylake+ CPU's. Also, some uncores can react poorly to switching between different power states/C states, causing lockups. An 8700k of mine at high cache clocks consistently locked the system up when coming out of a heavy load when certain c-states were enabled. If I lowered the uncore multiplier OR removed lower power c-states, it no longer froze up when coming out of a high load.
The same goes with an unstable system agent (It'll straight up lock up the CPU). System agent has its own voltage rail and defaults to 1.05v on 9000 series CPU's (not sure for 6000 series, but it has to be close). It also has 3 different clocks (400mhz, 800mhz, 1000mhz). This is sometimes called FCLK. I believe 800MHz is the default. 1000MHz can be set to get a marginal performance increase since it's indirectly linked to memory controller, and 400MHz is for when doing high BCLK overclocks.
It seems like a lot of people saying "power supply this" and "power supply that". Personally, I think your power supply is sufficient, and replacing it likely will not fix your problem. But it can't help to rule it out as a cause. If you really want to splurge on a new PSU, I'll throw my lot in and recommend Seasonic brand. You can't go wrong with almost all of their recent stuff, but they generally cost more that any other competing brand. Then again, you get what you pay for.
Thank you very much for that comprehensive response, and again, very sorry I didn't answered it month ago when you replayed.
As for what you said.. Uncore (L3 cache) was default 39x, with default voltage min 1.152v ; max 1.248v (voltage is VCore which is same for L3 cache), I could agree that it can be culprit, but... month ago I switched to this RAM (Kingston HyperX Fury 2666MHz CL15 4x4GB) and haven't had single issues since. PC is basically rock solid.
As for C states, I didn't messed with any of that, everything default in BIOS, and Windows 10 was set to "high performance".
You had problem where it got locked up coming from high load.. mine was completely random, sometimes it would run CSGO for 2 hours, not a single error, then I close the game, run another game, and at very first loading, pc freezes. So I can rule out C states then?
VCCSA voltage was same as it is now, and now it is not causing any problems - min 1.240v ; max 1.264v
Just to be sure, when lockups with old RAM happens, it was on "full default mode", so :
- Factory default BIOS
- Fresh new Windows 10 Pro with all drivers
- Not a single OC software
- Not a single change to default BIOS values (except for AHCI instead of RST, so I can boot my system)
Now, thing is, since I replaced RAM memory, lockups stopped (thank god), but still I'm unsure of why they were happening to begin with.. If it was uncore aka L3, it would be same on new RAM. If it was VCCSA, again, new RAM, no lockups (and it's even more load on IMC, since my crucial was 1x8GB and my kingston is 4x4GB) I guess if it was some voltage problem, most likely would be now with 4 DIMMS instead of only 1.
Now, I borrowed that Crucial ram to a friend who has ryzen 5 1600, and he has 1x8gb hyperx fury 2400, and he put mine Crucial along with his kingston.. so that's 2 different brands, with 2 different speeds, he had no problems at all..
So, at the end, replacing RAM worked for me, but that same RAM that I've replaced in my friends PC works without any lockups... Which makes me wonder... what the heck was the problem to begin with?