• PC Building Thread V6 - "running six RGB controller utilities at once" edition
    999 replies, posted
Some monitors come with displayport 1.1 selected in settings by default. At 1440p that's a 85hz refresh-rate. I think at 1080p it's 120? If the monitor isn't garbage it can be changed in settings to 1.2+.
I have a question about some Vcore troubles. I've put my Ivy Bridge under a decent overclock over half a year ago. To get 4.5GHz, it took me 1.21V and high LLC (that's level 4/6 on my mobo IIRC). Now everything's been running fine, I've tested it to be P95 stable for hours, temps are acceptable, I've had no bluescreens or unusual application crashes, it's just running as it should all the time. BUT, after a couple months of running the overclock, it turned out that it's not exactly 100% stable. I randomly got a singular WHEA error (I think the event log even mentioned it was correctable), but no other signs of instability. So I thought ok, let's just raise the voltage by .01V and not worry about it. So I go on for another couple months, and there it is, one or two random WHEA errors again. I add another .01V and go on with my life. Here I am now another month or so later, and another error just popped up again. But by now, things are looking pretty fishy. I don't want to keep raising the voltage on and on, so I just dialed back the OC a little to 4.3GHz at 1.18V and dropped the LLC down a notch to medium. Here's the relevant sensor readings under a P95 load: https://files.facepunch.com/forum/upload/209687/51d66f14-e0ee-4e5d-8a38-9ca893811512/43g_118_medium.PNG I'd like to note that when I had started overclocking it, the CPU seemed to be stable at 4.4GHz with these settings, but now I had to drop it to 4.3 to get it fully stable. What exactly could be the cause of this? I've been staying pretty conservative with the voltage since I don't really have much thermal headroom with my cooler, and 1.23V is supposed to be perfectly safe. Is it possible that the high LLC setting is causing so much overshoot that it's been slowly degrading my CPU during this time? I'm really not sure how much I can trust these sensor readouts, as with high LLC, both Vcore and VR Vout report a voltage very close to what I've set it to in bios, even under full load. Once I drop the LLC to medium, Vcore still reports a very close voltage, but VR Vout drops drastically by ~.05V under load, as you can see in the screenshot. Another explanation could be that my VRMs are simply crapping out on me. My mobo is 6 years old by now and it's fairly low end, so I wouldn't be surprised if it wasn't built to handle anything more than a mild OC. Nothing seems to be overheating, but it looks like the VRM isn't running very efficiently, if the readings are to be believed, it's losing 15W after the voltage conversion. Do you guys have an explanation for this? Any similar experiences perhaps?
It's hitting ~82 on full load. It's also using two display ports, and my CPU is i7-8700k.
That's a totally acceptable temperature. Something else is going on.
Could've gotten a bad one like Inacio did.
Another thing I just remembered is that some people were saying all the Intel microcode updates for the security vulnerabilities made their OCs less stable. It's likely just a coincidence, but slightly decreased stability every few months does line up surprisingly well with the frequency of new vulnerability discoveries
My Westmere was entirely unstable with the 1E microcode which was the Spectre mitigation update, I fought instability for about a year before I just dumped 1.415v down the vcore and dialed back the Uncore multi. For me though, I found that having either MSI Afterburner or EVGA Precision(this one was much worse) makes my GPU boost up higher if they are simply installed even if they aren't running. It was a nightmare trying to troubleshoot what was causing what for about a year as it would sometimes be stable for months at a time. There's something you can do to windows, overwriting the dll that loads microcodes with a blank file that's read only or something along those lines. I've found that usually you need far more voltage than first seems to be stable in the long term though.
As mentioned, the Vcore setting itself isn't the problem and I still have plenty of headroom there, I was just worried that going overkill with LLC might be degrading my CPU. I wanna keep this system running for at least a few more months before I upgrade, so I'd rather avoid any catastrophic failures till then. I guess I'll either keep the OC dialed back a bit or try pushing the voltage further, but there's not much room left in terms of thermals. At 1.25V it'll easily surpass 80C under full load, I guess you can't expect much better than that from a Hyper 212.
LLC could but unless you're seeing the spikes I doubt it's remotely high enough to damage it. Power draw and heat are far more damaging IMO than quick spikes that you aren't even seeing reported to the sensor. If you were seeing the massive overshoots I'd recommend to not use it but as long as you're thermally okay and your voltages report near to or within spec, I wouldn't worry. When you lower your LLC you're going to have to raise voltage to keep it where the LLC was previously filling in. If you think the LLC is hurting it you can always do that. I think you really just need more voltage because 4.5ghz at 1.21v is pretty good silicon.
1.21V was the seemingly stable OC that I started with, but after months of "oops, guess not really", I'd need to go for 1.24 or even higher with less LLC. The thing that I'm wary about is that one Vcore sensor behaves one way, the other behaves completely differently and I don't know if either of them are accurate or if the real voltage could be like .1V higher. I guess not because that would produce drastically higher temps, but it still feels like walking in the dark.
I mean, I'm getting 144+ in battlefront 2 and such, but destiny doesn't seem able to hit it for some reason. It could be just destiny? I'll do some more benchmarking with other games, but iirc I tried Witcher on completely maxed and only got ~120
That sounds like CPU bottlenecking for sure. TW3 is still pretty demanding on CPUs and my CPU struggles to make it past 110 FPS in it.
What solutions do I have? Can't really believe 8700k is bottlenecking
The 8700k not overclocked isn't actually that impressive, despite the "Intel CPU for gaming" memes. Atomic has a 2600X, IIRC.
Should I overclock then? I also have no idea how
You can generally start out with Multicore Enhancement/MCE and finding which XMP profile works, those are generally two things you'll want to always turn on with Intel platforms. Basically just enables turbo on all cores and runs the memory at the advertised speeds. Pay some attention to what your cpu vcore voltage goes upto during stress tests, you're going to need to increase it up from that generally as a starting point. Mild overclocks are generally just add vcore, increase core multipliers and change the LLC so it doesn't droop down so much that when you smack it with a load it becomes unstable. There's tons of other stuff you can play with but that's going to be the most simple you can get. I've only messed with a 6600k which I'd best describe as a garbage tier bin so I have no advice for skylakes.
Alright so I overclocked it to 4.5Ghz, did notice a slight difference with games now. Still though, wish it was hitting 144 in things like No Man's Sky
No Man's Sky is horribly multithreaded, try out the experimental Vulkan branch though. Enter "3xperimental" for a beta code to unlock the branch on Steam. I got like a 70-80% boost in FPS on my GTX 1080 with my old X5675.
So before I pull the trigger on this I wanted a 2nd opinion. This basically happened when my computer blue screened with a registry_error. After this I couldn't get the computer to "turn back on". What I mean by that is I could turn on the machine. But it wouldn't display anything (can't even get to a bios screen) I don't think the computer is getting to the desktop either because it never plays the windows start up sound. In addition to this the Mobo isn't powering any of the devices attached to it. I've basically tried everything. Removed ram sticks, reinserted the GPU, reinserted the cpu. Same result. I'm really leaning on it being the Mobo. What sucks is that Ryzen gen 3 is out soon so I don't want to upgrade prematurely but I really need my build right now.
Any/all Ryzen boards will work with the 3rd gen Ryzen, but they're not going to have PCIe 4.0 support, which is gonna be big for next gen GPUs. You should probably wait until X570 motherboards become available before picking up Ryzen 3000.
I only have a gen 1 Ryzen right now. So would I be better off picking up a cheapo one to last me till gen 3 comes out?
Have you tried clearing cmos yet? Pop the motherboard battery out for a while and hit the power button a few times while it's unplugged from power. I'm sure any of the highest end X470 boards will be fine. If you really want a useless AM4 cpu, the AM4 Athlons are $30 at Frys. Might hold you over for a month but with all the Zen 2 hype, it might be really rough to get a Zen 2 around launch. There's been a bunch of great deals on 1st gen Ryzen and APUs too.
I forgot to mention I did do that also
I'm currently building in my Thor V2, and so far the trickiest part was installing the damn stock wraith cooler. The screws just did not want to latch onto the backplate at all, but after like an hour of fiddling with the screws and applying horrifying amounts of pressure to each screw until they stuck, I finally got the damn 2600x's cooler to latch on. Apparently its a common defect for wraith coolers to have terrible screws. Now i'm dealing with my computer case's provided screws for the psu, not being long enough to screw the PSU to the case. There's like a soft cushion in the way of where the back of the PSU would sit, and it definitely looks like it should be there, i'm a bit apprehensive to peel it off.
How much do you guys reckon my Ryzen 5 1400 is slowing down my 1070 in real world perf, if at all? I've never really run into a 100% usage scenario in either CPU or GPU outside of synthetic stress tests, but I'm still getting GPU benchmark scores a little lower than they really should be for the 1070. I've had a few people insist I'm hard-bottlenecked in games by the 1400's per core perf, and honestly I believe them, but I can't tell by how much. If I'm gonna upgrade the CPU, I want to try to keep it in rough parity to avoid overspending, while still making sure the GPU is the weakest link so I'm not leaving any frames on the table.
What's your memory speed and resolution? Less than 100% GPU usage means that you're limited by something, its most likely the CPU or a framerate cap.
Memory is 16GB (8GB x2) 2400Mhz DDR4, no timing tweaks or anything. I hear Ryzen really loves fast, tightly timed RAM, but I dunno if it would have a pronounced impact on its communication responsiveness with the GPU.
Oof, yeah, the RAM can have a fairly significant impact under 2666, 2933 or 3200 are preferred.
I've generally had good luck overclocking lower end memory, they usually don't bin the cheap stuff so you can somtimes find like 1066mhz DDR3 which will do 1866mhz where as higher end memory is binned specifically for it's XMP profiles and you won't get that much more out of it. Just run a memtest on it for a cycle or so and that'll generally tell you if it's unstable or not.
what liquid cooling brings to the table is more sustained delta temps. If i were to push a copper heatpipe into a radiator, there would be mostly a big variance in the temps- when i was using air, for instance, my old 3570k was outside 10-15C in variance, so it could be 50c then instantly the next second 57, back down to 45, etc. Liquid cooling doesn't have that issue, though. it tends to have a far more sustained cooling curve due to the consistent flow of heat away from the CPU and the fact that the method of transit (water) is often far closer than heatpipes, as heatpipes tend to go in this direction, baseplate, copper pipes, solder, etc when watercooling can literally be copper-water done.
Sorry, you need to Log In to post a reply to this thread.