|
eames posted:Does anybody else have problems with the AVX offset kicking in non-AVX programs such as games? That is really weird. Why would it kick off the offset, but then not even adhere to the offset itself? AVX offset isn't a throttle (that I've seen), its usually just pretty simple logic that kicks off a different frequency. What board do you have?
|
#
?
Feb 27, 2018 09:47
|
|
|
|
| # ? May 15, 2024 16:10 |
|
|
eames posted:Does anybody else have problems with the AVX offset kicking in non-AVX programs such as games? Out of curiosity, what temps are you getting in prime95 when you become unstable? I've been tinkering around with similar settings/setup. I am leaning toward being happy with 4.7 and low voltage. It seems like some pretty harsh diminishing returns above that, even with some great silicon.
|
|
#
?
Feb 27, 2018 16:59
|
|
|
Lockback posted:That is really weird. Why would it kick off the offset, but then not even adhere to the offset itself? AVX offset isn't a throttle (that I've seen), its usually just pretty simple logic that kicks off a different frequency. What board do you have? It does adhere to the offset but jumps back and forth between offset/no offset, thus resulting in a lower average frequency. Asus Maximus X Hero with the latest 1003 BIOS. I reset the CMOS and tried optimized defaults, no change. I attached a picture of a test with 5 Ghz and -5 AVX offset. As you can see it holds frequency just fine in non-AVX, steps down for AVX and then does dumb stuff in games such as SC2. This happens in every single game I've tested. I'm starting to wonder if the BIOS/microcode is bugged or maybe the Nvidia drivers execute AVX instructions? It can't be any other application running in the background or it would step down for non-AVX Prime. headcase posted:Out of curiosity, what temps are you getting in prime95 when you become unstable? I've been tinkering around with similar settings/setup. I am leaning toward being happy with 4.7 and low voltage. It seems like some pretty harsh diminishing returns above that, even with some great silicon. 5.0 GHz manual 1.34V ~80°C in Prime95 80K non-AVX. 90°C with 80K AVX (over 200W) but that's too hot and not stable, hence why I'd prefer to use an offset. 4.7 GHz with AVX is around 80°C again. Typical gaming temps are in the low 60s, rarely touching 70°C in unusually CPU intensive games. I could get those temps down with a more aggressive fan curve but  .That's a delidded 8700K with a single-fan Noctua NH-D15S at 22°C ambient. The CPU also does 4.8 GHz at 1.23V rarely ever exceeding 60°C but I bought this big cooler and figured I might as well make use of it, at least in winter when I don't mind the relatively small amount extra heat. The higher Vcores only seem to make a major power consumption difference in stress tests or rendering, not so much in  where the CPU usually hovers at ~75W anyway.While I have this huge run-on posting going I'm wondering what LLC setting is worse for the CPU: manual 1.4V with 60mV Vdroop or 1.34V with the tiniest amount of negative droop. Der8auer has ties to Intel and sells his pretested/warrantied CPUs with up to 1.42V and LLC6 (= zero Vdroop) so I figure either of my options should be safer than that. Nobody seems to know just how bad those "invisible" LLC spikes really are. 14nm seems fairly robust so far. eames fucked around with this message at 18:35 on Feb 27, 2018 |
|
#
?
Feb 27, 2018 18:29
|
|
|
What is the favored monitoring tool now? I read the OP, but still found lots of discrepant opinions on temp/voltage accuracy and spyware addons. I'm looking for something that is free, lightweight, honest, doesn't send telemetry to a server. I would just use the Asus tools, but they were making unwanted settings changes.
|
|
#
?
Feb 27, 2018 19:45
|
|
|
headcase posted:What is the favored monitoring tool now? I read the OP, but still found lots of discrepant opinions on temp/voltage accuracy and spyware addons. HWiNFO64 by a huge margin is the best.
|
|
#
?
Feb 27, 2018 20:32
|
|
|
Wow. That sensor list is overwhelming and cool to see.
|
|
#
?
Feb 27, 2018 21:32
|
|
|
headcase posted:Wow. That sensor list is overwhelming and cool to see. Right click—> hide e: Just got my 980ti Hybrid back from EVGA RMA. I bought it used on amazon probably was a miner card and they still RMA'd it no problem when I started to have artifacting. EVGA are legends. Time to see how this one overclocks. VelociBacon fucked around with this message at 20:58 on Feb 28, 2018 |
|
#
?
Feb 27, 2018 22:31
|
|
|
Trying to solve a bunch of issues with a bunch of variables after a botched delidding job (bent pins on the kaby lake socket) Would instant spikes to 100c and instability during a cpu-z load test indicate a faulty aio cooler ? The cooler is brand new. I would've thought if the pump or fluid was bad that it would gradually rise in temps. Instead idle temps seem normal but I get an instant spike and then thermal throttling. If I stop the load testing the temps instantly return to a normal idle temp. Maybe it's the socket damage causing a misreport on temperature ? Maybe the bent pin is causing a short ? Maybe the CPU is also fried ? Problem is I don't want to swap motherboards and the find out the CPU is shot, and then have to swap the CPU for another kaby lake cpu. Any thoughts ? Maybe I'll reinstall the air cooler and see what the temps do. I've triple checked the thermal paste under the IHS, and also redid the paste under the cooler 3x now.
|
|
#
?
Mar 3, 2018 02:00
|
|
|
I'd try the air cooler. Sounds like it's not contacting?
|
|
#
?
Mar 3, 2018 02:12
|
|
|
Sounds like a bad connection. Don't you have to modify the mounting pins (whatever it is you screw down to hold the cooler on) when you delid to account for the lower height? I think some cooling solutions have something built in where the bolts or screws stop at a certain point that is supposed to position the cooler right at the perfect tension on the heat spreader.
|
|
#
?
Mar 3, 2018 02:17
|
|
|
This cooler (Captain 240 RGB) is spring loaded and i checked the contact patch of the paste, it's spread evenly like it should be. So I just ran cpu-z load test for an hour, it soft crashes something, maybe explorer.exe and thermal throttles but it allowed me to run the test. all 4 cores bouncing off 100c the entire time and the instant I stopped the test it jumped down to 55c and within a minute it's idling at 30c. I'm pretty sure it's the cooler that this point. Also do the radiators usually have some weight to them ? this one is light, as if there was no liquid in it. I figure if it had liquid going through the system, it would take awhile to cool off also. and there is no heat whatsoever at the radiator.
|
|
#
?
Mar 3, 2018 02:47
|
|
|
jonathan posted:Would instant spikes to 100c and instability during a cpu-z load test indicate a faulty aio cooler ? The cooler is brand new. I went through this exact process with the same symptoms on a 7700k I delidded. I thought I hosed up the delid, redid it, and it turns out the brand new Corsair H-series was empty. Strapped a Evo 212 on it and it was fine.
|
|
#
?
Mar 3, 2018 04:02
|
|
|
NeuralSpark posted:I went through this exact process with the same symptoms on a 7700k I delidded. I thought I hosed up the delid, redid it, and it turns out the brand new Corsair H-series was empty. Strapped a Evo 212 on it and it was fine. Wouldn't you know immediately when it's empty because the pump would scream like a banshee?
|
|
#
?
Mar 3, 2018 04:14
|
|
|
NeuralSpark posted:I went through this exact process with the same symptoms on a 7700k I delidded. I thought I hosed up the delid, redid it, and it turns out the brand new Corsair H-series was empty. Strapped a Evo 212 on it and it was fine. Thanks. I guess I'll start an RMA unless they let me fill it myself.
|
|
#
?
Mar 3, 2018 05:24
|
|
|
scratch that, I installed my CM 212 air cooler, same exact results. I even pulled the cpu retainer and clamped the cooler right down onto the cpu/IHS just in case it was a clearance thing. So I put it back to stock frequency, and loaded default bios settings. Reran cpu-z stress test and it sits at 60c across the cores. So I got my thermapen which is calibrated. It reads 25c on the IHS. The heat sink and the heat piping is all cool to the touch. Colder than the room. WTF is going on ? The only thing I can think of is the damaged socket is throwing off the actual temp readings. Or the IHS is not contacting the actual cpu however I've pulled the ihs 3 times and it is spreading the paste nicely. If the cpu was 60 degrees (or 100 with 4.8ghz overclock) would it not heat up the ihs enough to measure on the thermapen or heat up the cooler/heat piping enough to feel ?
|
|
#
?
Mar 3, 2018 08:24
|
|
|
Alright, so I've got myself an 8700K on an Asus Z370I, and it's been pretty straightforward overclocking. I just want to check to make sure these numbers seem alright: 4.8GHz overclock at 1.280v (Adaptive) in BIOS. CPU-Z reports the voltage at 1.264v under load. HWMonitor shows the max VCORE as 1.280v, and the max VID as 1.295v. Temps under my typical load (games, for the most part) are in the upper-60s to upper-70s, with a fairly aggressive fan curve. Cooler is a Corsair H100i. I could probably push to 4.9 or 5GHz, right? Right now I'm just giving this particular clock some time, as I want to see how stable it is under real-world use. It seems perfectly stable, currently. The numbers look good? I'm not killing it with voltage and whatnot? Gonkish fucked around with this message at 11:24 on Mar 3, 2018 |
|
#
?
Mar 3, 2018 10:29
|
|
|
Gonkish posted:Alright, so I've got myself an 8700K on an Asus Z370I, and it's been pretty straightforward overclocking. I just want to check to make sure these numbers seem alright: Looks alright, if it's stable there start bumping clock speed until you reach instability. Apparently that chip is fine under 1.35v so I'd guess temp is going to limit you first.
|
|
#
?
Mar 3, 2018 12:31
|
|
|
jonathan posted:scratch that, I installed my CM 212 air cooler, same exact results. I even pulled the cpu retainer and clamped the cooler right down onto the cpu/IHS just in case it was a clearance thing. So I put it back to stock frequency, and loaded default bios settings. Reran cpu-z stress test and it sits at 60c across the cores. So I got my thermapen which is calibrated. It reads 25c on the IHS. The heat sink and the heat piping is all cool to the touch. Colder than the room. WTF is going on ? The only thing I can think of is the damaged socket is throwing off the actual temp readings. I'd be thinking gently caress it and get a shim from aqua computers, and direct die cool it at this point
|
|
#
?
Mar 3, 2018 14:31
|
|
|
jonathan posted:scratch that, I installed my CM 212 air cooler, same exact results. I even pulled the cpu retainer and clamped the cooler right down onto the cpu/IHS just in case it was a clearance thing. So I put it back to stock frequency, and loaded default bios settings. Reran cpu-z stress test and it sits at 60c across the cores. So I got my thermapen which is calibrated. It reads 25c on the IHS. The heat sink and the heat piping is all cool to the touch. Colder than the room. WTF is going on ? The only thing I can think of is the damaged socket is throwing off the actual temp readings. The 7700K has pretty well documented issue that causes impossible heat spikes, i.e. from 40°C to 100°C to 40°C within a second or two. Last time I checked it was believed to be a sensor issue. 6700K and 8700K don't have this problem. It's one of the main reasons I didn't buy a 7700K back then. Maybe your problem is related? Here's a giant thread on the Intel forums: https://communities.intel.com/thread/110728 The official Intel statement is about as helpful as you'd expect: https://www.extremetech.com/computing/248908-intel-responds-i7-overheating-issue-cluelessly-suggests-stop-overclocking
|
|
#
?
Mar 3, 2018 18:10
|
|
|
eames posted:The 7700K has pretty well documented issue that causes impossible heat spikes, i.e. from 40°C to 100°C to 40°C within a second or two. Last time I checked it was believed to be a sensor issue. 6700K and 8700K don't have this problem. It's one of the main reasons I didn't buy a 7700K back then. Maybe your problem is related? Mine is the 7600k (i5). It's been cool and stable for a year at 4.8ghz 1.4v until I delidding. But now I don't know if it's a damaged CPU or the damaged socket (which IS damaged)
|
|
#
?
Mar 3, 2018 20:14
|
|
|
eames posted:The 7700K has pretty well documented issue that causes impossible heat spikes, i.e. from 40°C to 100°C to 40°C within a second or two. Last time I checked it was believed to be a sensor issue. 6700K and 8700K don't have this problem. It's one of the main reasons I didn't buy a 7700K back then. Maybe your problem is related? My 7700k did that spiking too but it stopped after I overclocked it and put a 280 clc on it. Now its always warm but stopped spiking up to 80c. The spikes are way more reasonable like from 43-58
|
|
#
?
Mar 3, 2018 21:55
|
|
|
VelociBacon posted:Wouldn't you know immediately when it's empty because the pump would scream like a banshee? Oddly enough it didn't shriek, and still registered RPM on the header. Before I tossed it I popped the filler cap with an Allen key and got nothing out of it. 
|
|
#
?
Mar 4, 2018 00:44
|
|
|
Dang it was totally dry. Can't help but think they should test for that.
|
|
#
?
Mar 4, 2018 01:43
|
|
|
that is an issue i have always been wary of so I only buy CLCs that are a new model so I can be sure it had a relatively recent production date
|
|
#
?
Mar 4, 2018 02:03
|
|
|
So tonight I discovered that, for god only knows how long, my CPU has been running at higher voltages (1.24 vs 1.19) while staying at stock clocks. I never saw high temperatures and have been on water so it's more of an annoyance than anything else, but the fact that I wasn't even aware that I was giving my much-appreciated long lasting chip an overshock is  Apparently this is because I had the multiplier manually set to stock clocks instead of choosing Auto for nearly everything (the exception is manually specifying DDR3-1600, which resulted in another "helpful" voltage+ that I caught early and shut down). It's like another example of why I don't OC this thing. Asus in this era was just super aggressive with assisting you by discreetly upping voltage corresponding to other things you are doing. Craptacular! fucked around with this message at 09:06 on Mar 4, 2018 |
|
#
?
Mar 4, 2018 09:03
|
|
|
If that happens even when not overclocking, might as well overclock and check.
|
|
#
?
Mar 4, 2018 09:26
|
|
|
GRINDCORE MEGGIDO posted:If that happens even when not overclocking, might as well overclock and check. The problem is, compared to what I've seen of modern boards (particularly watching Ryzen OC vids), the tuner panel for my old motherboard has so many options that even a detailed guide can't help but be intimidating, doesn't have a big obvious "core ratio" entry like a modern ASrock or something would, and as I mentioned is subtly cranking other options. In my case, I just had Turbo set to By All Cores, but left the manual entry at the Auto (stock) setting, and it was applying a higher voltage than setting the Turbo switch to Auto (aka defaults/not-overclocking) and getting the same numbers for performance.  Some of that is on the shoulders of another default(?), which is the one that sets CPU voltage to Offset Mode and sets the offset to auto, so cranking up the cores would "helpfully" apply even more voltage. I think what I would have to do is set Turbo back to By All Cores and set the multiplier there, then set the voltage to Manual Mode and specifying a voltage. And then praying I didn't gently caress anything up. I guess if you can't tell from this post, I've absorbed a lot of information about overclocking, but I'm scared to dip my toes in because (A) my board's long list of options intimidates me, and (B) I can't really afford to lose this CPU.
|
|
#
?
Mar 4, 2018 09:59
|
|
|
Modern boards get volt happy, my Asus z170 dumped higher voltages then your vcore into vccio because I set the memory multiplier.
GRINDCORE MEGGIDO fucked around with this message at 14:10 on Mar 4, 2018 |
|
#
?
Mar 4, 2018 10:30
|
|
|
Craptacular! posted:The problem is, compared to what I've seen of modern boards (particularly watching Ryzen OC vids), the tuner panel for my old motherboard has so many options that even a detailed guide can't help but be intimidating, doesn't have a big obvious "core ratio" entry like a modern ASrock or something would, and as I mentioned is subtly cranking other options. In my case, I just had Turbo set to By All Cores, but left the manual entry at the Auto (stock) setting, and it was applying a higher voltage than setting the Turbo switch to Auto (aka defaults/not-overclocking) and getting the same numbers for performance. There’s nothing to be scared of. Your bios might look complicated but all you really have to change is your multiplier and vcore unless you want to get fully autistic like the rest of us. Just set a manual vcore if your board allows it to scale down when idle, or an offset/adaptive to get it where you want and play with the multiplier until you find something stable.
|
|
#
?
Mar 4, 2018 10:59
|
|
|
The rule of thumb is don't trust Auto anything even when running not OCing. Put Vcore to offset +0V and manually set the rest of the voltages to stock (Google them for whatever platform you are on) then start tweaking from there. Yeah, I do agree the out-of-the-box OCing/overvolting shenanigans without any sort of prior warning on current mobos is getting too far now. Palladium fucked around with this message at 12:15 on Mar 4, 2018 |
|
#
?
Mar 4, 2018 12:05
|
|
|
Adaptive voltage is simple AF also. Just put in a realistic voltage for the turbo state, and see how high it clocks, watching the voltages in the os to make sure it's not being fully retarded. Everything non turbo can just run at "stock".
|
|
#
?
Mar 4, 2018 14:14
|
|
|
I've actually been able to cut from stock voltage with an overclock to 4.6 on a 4790k so I think it's worth cutting voltage at non-boost clocks but it's definitely more effort.
|
|
#
?
Mar 4, 2018 14:25
|
|
|
Seems like the ideal situation to me, crank the turbo a bit and undervolt the rest. I wish this asrock extreme 4 z170 supported adaptive voltage. Seems bizarre AF not to include it. (It was cheap as hell though. I know it's not great) GRINDCORE MEGGIDO fucked around with this message at 14:34 on Mar 4, 2018 |
|
#
?
Mar 4, 2018 14:32
|
|
|
Palladium posted:The rule of thumb is don't trust Auto anything even when running not OCing. I actually kind of prefer it so long as I can see what it's doing. ASUS AI Suite II no longer works after the Meltdown patch, but I've thought about using Intel XTU, if just to see what results it comes up with before diving into BIOS.
|
|
#
?
Mar 4, 2018 19:20
|
|
|
Palladium posted:Yeah, I do agree the out-of-the-box OCing/overvolting shenanigans without any sort of prior warning on current mobos is getting too far now. I think time will tell, maybe the motherboard manufacturers are right and all those voltages are safe to use for longer periods of time. My Asus Z370 board would dial in over 1.4V VccSA/VccIO with their 5 GHz profile and XMP even though 1.175V is just as stable. Similarly for CPU Standby, Core PLL and some other voltages that I'm not even sure what they're doing. The Asus auto settings tend to work very well but what's the point when they degrade or even kill your CPU/IMC after a year of use. I'd love to know their testing methods and whether these profiles are in any way approved by Intel.
|
|
#
?
Mar 6, 2018 18:29
|
|
|
eames posted:I think time will tell, maybe the motherboard manufacturers are right and all those voltages are safe to use for longer periods of time. My Asus Z370 board would dial in over 1.4V VccSA/VccIO with their 5 GHz profile and XMP even though 1.175V is just as stable. Call me a chicken, but when Intel doesn't specify max safe voltages for Vccsa/Vccio in their own official technical documents I'm ain't going put anything above the stock 0.95/1.05V, especially when all they do is to squeeze out an insignificant +200MHz out of my DDR4 and I also plan to keep my 8700K rig for at least 5 years. I also downvolted my DDR4-3200 to 1.25V because why not if its already stable there.
|
|
#
?
Mar 7, 2018 07:50
|
|
|
What is the best guide for overclocking my 2500K? It seems like everything made back in the day is now outdated. I tried to follow this guide from the first post, but Windows wouldn't boot. I assume that I'm running a later BIOS version complicates my ability to follow these guides, which are almost all written when the CPU first came out.
|
|
#
?
Mar 8, 2018 00:34
|
|
|
Revol posted:What is the best guide for overclocking my 2500K? It seems like everything made back in the day is now outdated. I tried to follow this guide from the first post, but Windows wouldn't boot. I assume that I'm running a later BIOS version complicates my ability to follow these guides, which are almost all written when the CPU first came out. What cooling are you running? Generally if you have some decent cooler you can just set your vcore to 1.25 or so and your cpu to 4.5Ghz and see how it goes.
|
|
#
?
Mar 8, 2018 01:05
|
|
|
Off topic, but I just want to rage in the OC thread about how crap my 6600K is at clocking. Please carry on.
|
|
#
?
Mar 8, 2018 01:38
|
|
|
|
| # ? May 15, 2024 16:10 |
|
|
GRINDCORE MEGGIDO posted:Off topic, but I just want to rage in the OC thread about how crap my 6600K is at clocking. What kinda results are you getting? Or not getting?
|
|
#
?
Mar 8, 2018 01:51
|
|