|
My board is also a gigabyte with the two vcore readings. I was graphing both of them and was pretty sure I knew which one was real but then under sustained prime95 testing one of the values halved itself lol
|
#
?
Oct 23, 2018 22:18
|
|
|
|
| # ? Jun 4, 2024 20:41 |
|
|
mewse posted:My board is also a gigabyte with the two vcore readings. I was graphing both of them and was pretty sure I knew which one was real but then under sustained prime95 testing one of the values halved itself lol Could be that you dropped a few threads in testing and the vcore that halved itself is an average to all cores.
|
|
#
?
Oct 23, 2018 22:32
|
|
|
n.. posted:i know i know, gooncensus is that gigabyte is a garbage fire I know this is hyperbole, but honestly the only concrete and substantiated criticism I can remember hearing of Gigabyte on this forum is that they downgraded components after rev. 1 on a few old budget boards. My main desktop has a ten year old Gigabyte board still chugging along happily with a 45% overclock on a processor it never claimed to even support and no issues whatsoever, and while that's not necessarily exceptional (what with the 15 year old ASUS board I have that still works too) I'd tell anyone asking to consider Gigabyte boards alongside other top tier manufacturers' on their apparent merits. Eletriarnation fucked around with this message at 23:02 on Oct 23, 2018 |
|
#
?
Oct 23, 2018 22:58
|
|
|
TheFluff posted:Huh. Odd. Do they seem to follow each other but with an offset from each other, or is there something weirder going on? Looks like it's the former.  Top one is under "gigabyte z390 aorus pro-cf (ITE IT8792E)" and the bottom one is under "gigabyte z390 aorus pro-cf (ITE IT8688E)" They both droop the same amount under load, but they're about .034v offset. e: had them reversed, fixed CheddarGoblin fucked around with this message at 23:39 on Oct 23, 2018 |
|
#
?
Oct 23, 2018 23:08
|
|
|
n.. posted:Looks like it's the former. The one I ended up trusting on my z370 gaming 7 was the one under IT8792E. My other one is above it and labelled ITE IT8686E, so slightly different than yours.
|
|
#
?
Oct 23, 2018 23:17
|
|
|
Eletriarnation posted:I know this is hyperbole, but honestly the only concrete and substantiated criticism I can remember hearing of Gigabyte on this forum is that they downgraded components after rev. 1 on a few old budget boards. My main desktop has a ten year old Gigabyte board still chugging along happily with a 45% overclock on a processor it never claimed to even support and no issues whatsoever, and while that's not necessarily exceptional (what with the 15 year old ASUS board I have that still works too) I'd tell anyone asking to consider Gigabyte boards alongside other top tier manufacturers' on their apparent merits. Gigabyte is perfectly capable of making good boards, they're just inconsistent. The Z370 Gaming 7 for example has a perfectly good VRM, but on many boards it runs way hotter than it should because the heatsink wasn't tightened down sufficiently from factory. Their BIOS'es are also kinda janky and weirdly laid out, but that's not exactly a huge problem - you can get used to it. As far as this monitoring issue goes, the ITE IT-numbers designations are part numbers for sensor controllers made by ITE Tech Inc, but I have no idea why Gigabyte saw fit to duplicate the monitoring. A quick googling turned up some complaining about the 8686E on overclock.net, but it's overclock.net so take it with some salt. TheFluff fucked around with this message at 23:35 on Oct 23, 2018 |
|
#
?
Oct 23, 2018 23:31
|
|
|
Ok here's a more complete picture of what's happening. This was 3 Cinebench runs followed by 3 POV-Ray runs, screenshot was taken during the last run. You can see that the second column does run hotter, but there's also more voltage under load. My question is, why can't I lower the voltage anymore with LLC at 'turbo' and have it remain stable, when it runs at even lower voltage using 'normal' LLC with droop and is stable then?
|
|
#
?
Oct 24, 2018 00:14
|
|
|
n.. posted:My question is, why can't I lower the voltage anymore with LLC at 'turbo' and have it remain stable, when it runs at even lower voltage using 'normal' LLC with droop and is stable then? Because despite showing the same voltage value the electricals are operating differently.
|
|
#
?
Oct 24, 2018 00:50
|
|
|
mewse posted:Because despite showing the same voltage value the electricals are operating differently. Any way to mitigate that with settings or is it a lost cause? I seem to have gotten a pretty crummy 9700k, seems others are getting past 5ghz at lower voltages no problem, but I'm pretty much stuck at 4.9 so far.
|
|
#
?
Oct 24, 2018 01:02
|
|
|
n.. posted:Any way to mitigate that with settings or is it a lost cause? I don't know, I don't really understand what you're going for. LLC is to counteract vdroop at higher voltages so that the proc stays stable at higher frequencies, it seems like you're undervolting so maybe you don't even want LLC. The only really solid advice I got regarding LLC was following along one of gamersnexus's livestreams where he was OCing an 8086k (I've got a 8700k) where he said if you set LLC to the second highest setting that should get you closest to bang on your manually entered voltage. I think he was using a gigabyte board as well so the second highest LLC setting has worked for me nicely. For a better idea of voltages check out silicon lottery's tables, they were pushing 8700k's to over 1.4v to get them stable. They don't seem to have data for 9700k available.
|
|
#
?
Oct 24, 2018 01:21
|
|
|
mewse posted:I don't know, I don't really understand what you're going for. LLC is to counteract vdroop at higher voltages so that the proc stays stable at higher frequencies, it seems like you're undervolting so maybe you don't even want LLC. Yeah exactly, which is why it doesn't make any sense that my proc is stable at lower voltage with lots of droop than with more voltage and no droop at all. I dunno why you say I'm undervolting, this is only at 4.9ghz (all cores). Anandtech's was running 1.125v at 4.9 and at much lower temps than I'm seeing, and I'm using a 280mm AIO cooler.
|
|
#
?
Oct 24, 2018 01:28
|
|
|
So I've been struggling to get my 4.7ghz OC stable on my 3770k still. I've got it pretty much on the knife edge of being stable, temps are way more than acceptable, so I feel like with just the right bit of tweaking I can get it there. Interestingly if I increase the load line calibration it seems like I get less stability, not more. I kept increasing it trying to get the OC stable but the tests I was running kept crashing, but as soon as I turned it back to auto on my board things suddenly got way more stable and I was able to run tests for a couple hours before they eventually crashed. I've got an asrock Z77 extreme 4 board specifically if anyone knows of a good setting in the UEFI to help with stability. I've gotten an i5 3570k stable at this clock before so I know it's possible at least.
|
|
#
?
Oct 24, 2018 16:16
|
|
|
cat doter posted:So I've been struggling to get my 4.7ghz OC stable on my 3770k still. I've got it pretty much on the knife edge of being stable, temps are way more than acceptable, so I feel like with just the right bit of tweaking I can get it there. Interestingly if I increase the load line calibration it seems like I get less stability, not more. I kept increasing it trying to get the OC stable but the tests I was running kept crashing, but as soon as I turned it back to auto on my board things suddenly got way more stable and I was able to run tests for a couple hours before they eventually crashed. A few possibly-not-placebo things I've used on Z370 are: - set a higher VRM switching frequency (500kHz is usually max, default is usually 300kHz) - disable VRM spread spectrum - disable BCLK spread spectrum (not sure if this was a thing on Z77) The first one will probably make the CPU VRM run a good deal hotter, so you should probably keep an eye on VRM temperatures (hopefully your motherboard supports monitoring them - if it doesn't you can just skip it). Newer boards with high end VRM's use power stages optimized for 500kHz or higher switching frequency, but I doubt that was common on Z77. The latter two are harmless unless you live in a signals processing lab.
|
|
#
?
Oct 24, 2018 16:26
|
|
|
Increasing the VRM frequency on my P67 was the magic that got me from 4.2GHz to 4.6GHz. I think I only needed to raise it to 350, any higher didn't help.
|
|
#
?
Oct 24, 2018 16:31
|
|
|
Any long term concerns with VRM freq adjustment?
|
|
#
?
Oct 24, 2018 16:59
|
|
|
VelociBacon posted:Any long term concerns with VRM freq adjustment? None that I know of except for the temperature concerns. If your VRM stays below 90-ish C under load you're probably fine. The power stages themselves are almost always fine up to 125C, but you don't know where the temp sensor is (if your board even has one) and things like capacitors (which tend to be nearby) are usually more sensitive. If you're really curious you can possibly find a data sheet for the power stages or MOSFET's used on your board and try to figure out what higher switching frequency does to their efficiency, but at least to me those graphs are almost completely incomprehensible. Not enough of an electrical engineer, I guess.
|
|
#
?
Oct 24, 2018 17:09
|
|
|
Looks like my board doesn't have a VRM frequency setting, or I'm stupid and can't find it. I'll have another look later. But for the spread spectrum thing there's only one setting labeled "spread spectrum" without being specific to BCLK or VRM or anything. I've disabled it though and I'm running tests now. For some reason heaven benchmark absolutely hates any kind of instability in overclocks so I use that to validate them, which is weird since it's not exactly taxing the CPU that much. That and Nier: Automata which helped me identify an unstable overclock that I'd had set for like a year. The weirdest things can help identify instability sometimes. Anyway I'll leave it running all day and see if it crashes at all. Hopefully this is the magic bullet.
|
|
#
?
Oct 24, 2018 17:41
|
|
|
Welp nevermind it crashed after about half an hour, a hard crash this time. I'll have to give up on the 4.7ghz dream and drop down to 4.6 I guess.
|
|
#
?
Oct 24, 2018 18:27
|
|
|
cat doter posted:Welp nevermind it crashed after about half an hour, a hard crash this time. I'll have to give up on the 4.7ghz dream and drop down to 4.6 I guess. Look for LLC level. That controls the VRMs.
|
|
#
?
Oct 25, 2018 19:54
|
|
|
redeyes posted:Look for LLC level. That controls the VRMs. It controls the voltage, not the switching frequency.
|
|
#
?
Oct 25, 2018 20:27
|
|
|
cat doter posted:Welp nevermind it crashed after about half an hour, a hard crash this time. I'll have to give up on the 4.7ghz dream and drop down to 4.6 I guess. Do you have a BIOS option for CPU Cache/core current limit?
|
|
#
?
Oct 25, 2018 20:58
|
|
|
redeyes posted:Look for LLC level. That controls the VRMs. As I mentioned in an earlier post, I tried LLC, my board has levels 1 to 5 and I tried each one, and if anything it seemed to make my CPU less stable which is kinda weird. VelociBacon posted:Do you have a BIOS option for CPU Cache/core current limit? There's a few current limit options that I'm not sure about, it's probably easier to just show you guys what options I have. https://imgur.com/a/mU1uASN I was doing some quick googling and found some stuff about ivy bridge CPUs liking PLL voltage of between 1.5-1.7v for overclocking, is that worth trying? I don't wanna mess with something unless I know it's safe really.
|
|
#
?
Oct 25, 2018 22:06
|
|
|
I lowered my Sandybridge PLL voltage to 1.6 or so, it lowered temperatures, maybe. Someone said it's also the voltage the temperature measurement uses for a reference, so there's less power going in, but it might make the temperature readings inaccurate.
|
|
#
?
Oct 25, 2018 22:19
|
|
|
I don't need to monitor temps that much since it's a delidded CPU with liquid metal on the die at the moment.
|
|
#
?
Oct 25, 2018 22:52
|
|
|
From memory, yes, on Ivy Bridge you can try lowering PLL to around 1.5v with no harmful effect. Might or might not do anything for stability. I'm pretty sure that on Kaby Lake and later you shouldn't touch it though - I remember reading somewhere that mucking with it can make the CPU temp sensors measure 15-20 degrees too low on that platform.
|
|
#
?
Oct 25, 2018 23:04
|
|
|
Well PLL doesn't seem to have done anything for stability, oh well. Also I think I misread the load line calibration setting on my board, it looks like the maximum setting is level 1, not level 5, so obviously me "increasing" the setting to level 5 was causing instability issues. I've left it at level 1 manually now and it seems pretty solid, but I'm still not at that 100% stability I want yet.
|
|
#
?
Oct 26, 2018 02:46
|
|
|
Have you checked it since you changed it to 1 because that might be pushing more voltage than you'd expect.
|
|
#
?
Oct 26, 2018 04:52
|
|
|
VelociBacon posted:Have you checked it since you changed it to 1 because that might be pushing more voltage than you'd expect. I have, yeah, it's within safe (though on the upper end of "safe") limits. It's why I'm edging towards just dropping it to 4.6ghz and giving up.
|
|
#
?
Oct 26, 2018 06:23
|
|
|
I'm going to be upgrading my system from a 6700k to a 9700k, so new mobo/ram/psu the works. I've currently got a Noctua NH-U14S on my 6700k, but that will probably be getting sold on with the rest of the stuff. Was thinking of sticking with Noctua and going with a NH-D15, but at that price I'm getting into AIO water cooler territory. I've got a Corsair Carbide 600 case, which has room for a front 280mm and bottom 360mm radiators. I haven't really looked too hard into water cooling before, so with a budget of say $200ish and the goal of max OCs without breaking my ear drums, which route would you all take?
|
|
#
?
Nov 6, 2018 19:43
|
|
|
AIOs will be louder but fit into smaller spaces. I went with a D15 over a H110 for noise reasons.
|
|
#
?
Nov 6, 2018 20:05
|
|
|
I second the D15(S) recommendation unless you're chasing those last 2-3°C under full load for 100 extra MHz or move your PC around a lot (shipping/traveling). It's quieter and less headache overall. Idle pump noise, pump failures, leaks, liquid permeation, badly programmed control software... all rare problems with AIOs but simply not a concern with the Noctua. eames fucked around with this message at 21:04 on Nov 6, 2018 |
|
#
?
Nov 6, 2018 21:01
|
|
|
After reading some reviews, I think the D15 definitely makes the most sense. That said, I ordered a NZXT Kraken X72 anyway because I hate money! It's only a few degrees cooler and a few db quieter than the D15, but I've been wanting to try water cooling for nearly 20 years now so what the heck.
|
|
#
?
Nov 6, 2018 21:43
|
|
|
I've never heard my h115i.
|
|
#
?
Nov 7, 2018 00:12
|
|
|
Only a few hours of testing down, but seems like I've got a pretty stable 5.2GHz @ 1.375v on a 9700k with max temps in the low 70s. Was able to run some benchmarks at 5.3GHz, but it required pushing voltage to 1.5 and temps got up to the 90s.
|
|
#
?
Nov 11, 2018 08:23
|
|
|
Enos Cabell posted:Only a few hours of testing down, but seems like I've got a pretty stable 5.2GHz @ 1.375v on a 9700k with max temps in the low 70s. Was able to run some benchmarks at 5.3GHz, but it required pushing voltage to 1.5 and temps got up to the 90s. Is that with the x72 you said you ordered?
|
|
#
?
Nov 11, 2018 18:56
|
|
|
B-Mac posted:Is that with the x72 you said you ordered? Yeah, using an x72 on an msi meg ace motherboard. Also have some RGB g.skill ddr4 3600 ram because I have no self control https://i.imgur.com/58xLTWO.mp4
|
|
#
?
Nov 11, 2018 19:36
|
|
|
Enos Cabell posted:I'm going to be upgrading my system from a 6700k to a 9700k, so new mobo/ram/psu the works. I've currently got a Noctua NH-U14S on my 6700k, but that will probably be getting sold on with the rest of the stuff. Was thinking of sticking with Noctua and going with a NH-D15, but at that price I'm getting into AIO water cooler territory. Be quiet silent loop 280. Or 360  Swap the fans for vardars or ml140s, or nf a14 pwms. GRINDCORE MEGGIDO fucked around with this message at 14:10 on Nov 12, 2018 |
|
#
?
Nov 12, 2018 13:58
|
|
|
Question for the overclocking thread; I've got a 9900k in a gigabyte Z390 aorus master, and 32 GB of g.skill (F4-3200C14D-32GVK). G.skill has the motherboard in their QVL, however gigabyte does not have the memory in theirs, but they do have the quad channel kit of the same memory. In order to get it working in XMP I had to override VCCIO and VCCSA to 1.3v each, 1.25 managed a POST but locked up, anything less than that wouldn't POST at all (I didn't test them separately yet). There is a fair amount of conflicting information online, some topics say "1.25 max for daily use" others say "1.35 is the limit of the Intel spec". My previous 7700k system in an Asus board ran both at upwards of 1.33v automatically when in XMP mode and never experienced any issues even after a couple years of running 24/7. So is 1.3v going to cause issues down the line, or is it safe to set it and forget it? Should I try lowering one or the other to see which one actually needs it the most?
|
|
#
?
Nov 25, 2018 18:21
|
|
|
Just like with Vcore, nobody (but perhaps Intel with their extensive simulations) really knows. People agree that higher voltages increase the chance of failure (degradation or complete defect of the IMC) but "safe" is relative because it depends on so many unknown factors (silicon quality, temperatures, power quality/VRMs, expected lifespan of the system, your usage pattern, etc). OC/enthusiast motherboard manufacturers are generally not known for "safe" settings on Auto because higher voltages yield higher performance and translate into better marketing/sales. If your board/CPU suddenly dies then good for the manufacturers because you're going to need a new one. Some people have observed that manufacturers even step down their Auto voltages a few BIOS revisions after launch when the review cycle is over.
|
|
#
?
Nov 25, 2018 19:36
|
|
|
|
| # ? Jun 4, 2024 20:41 |
|
|
Indiana_Krom posted:Question for the overclocking thread; I've got a 9900k in a gigabyte Z390 aorus master, and 32 GB of g.skill (F4-3200C14D-32GVK). G.skill has the motherboard in their QVL, however gigabyte does not have the memory in theirs, but they do have the quad channel kit of the same memory. In order to get it working in XMP I had to override VCCIO and VCCSA to 1.3v each, 1.25 managed a POST but locked up, anything less than that wouldn't POST at all (I didn't test them separately yet). There is a fair amount of conflicting information online, some topics say "1.25 max for daily use" others say "1.35 is the limit of the Intel spec". My previous 7700k system in an Asus board ran both at upwards of 1.33v automatically when in XMP mode and never experienced any issues even after a couple years of running 24/7. So is 1.3v going to cause issues down the line, or is it safe to set it and forget it? Should I try lowering one or the other to see which one actually needs it the most? You could try disabling Xmas and manually entering the primary timings and see if that helps, plug in your 3200 speed and enter the main timings the kit is rated for, for example 16-16-16-36 or whatever they may be. You could try playing with the dram voltage as well and see if it lets you back down the VCCIO and VCCSA. I had to manually entree my 2x8 3733 kit to get it to work with my Z370 taichi/9900k combo. Can use it at VCCIO 1.15 and VCCSA 1.2
|
|
#
?
Nov 25, 2018 19:49
|
|