|
WanderingKid posted:If XMP works then great. If not, you can look at the sticker on your ram or on the box which tells you the dram frequency, timings and voltage. I have a 2x 8gb g.skill trident z 3200mhz CL14 kit. Recommended voltage is 1.35v. Are XMP settings independent of any changes I may make to the CPU speed, or will I need to go to "manual mode" if I start raising bus clocks and adjust the RAM multiplier as needed?
|
#
?
Jan 19, 2017 16:46
|
|
|
|
| # ? Jun 6, 2024 02:15 |
|
|
RAM multiplier? Raising the bus clocks? I dunno what board you have. ASUS BIOS seems to keep alot of the memory tweaking stuff separate from the cpu tweaking stuff. If XMP worked for me I would have just 1 clicked that, rebooted and happy days. Loading the XMP profile for your memory just sets all the relevant manual options for you so you don't have to. When XMP failed, I went into the BIOS DRAM settings and attempted to do the whole lot manually but it was confusing as hell because you had a tonne of fields for primary timings and secondary timings and I didn't know what the gently caress was going on. Not at all noob friendly. So I just loaded the XMP profile, made a mental note of the settings it changed from auto like DRAM voltage, frequency and the primary timings then switched back from XMP to auto. Then I manually entered the numbers it changed. It was maybe 5 values: DRAM frequency from auto to 3200 mhz DRAM voltage from 1.2V to 1.353V DRAM CAS# Latency from auto to 14 DRAM RAS# to CAS# delay from auto to 14 DRAM RAS# ACT time from auto to 34 This is for a 2x 8GB G.Skill Trident Z kit (3200 mhz, CAS14) I started at a lower DRAM frequency (2800 mhz) which worked, then 3000 mhz (worked) then 3200 mhz (also worked). Its the same poo poo as selecting XMP except when I load XMP, I can't post but when I put the same changes in manually it does. Weird.
|
|
#
?
Jan 19, 2017 17:44
|
|
|
If you whack the memory speeds up using XMP (or however) and it doesn't work, before dropping it down try playing with VCCIO and system agent voltage. Infact check those even if it works fine with XMP - sometimes the board will increase those when the memory speed is increased and it can increase it to fucky levels.
|
|
#
?
Jan 19, 2017 18:23
|
|
|
GRINDCORE MEGGIDO posted:Infact check those even if it works fine with XMP - sometimes the board will increase those when the memory speed is increased and it can increase it to fucky levels. Hmm, you are right. Leaving VCCIO and VCCSA on auto, will make both skyrocket if you whack up DRAM voltage. My defaults were: DRAM voltage = (auto) 1.200V VCCIO = (auto) 0.960V VCCSA = (auto) 1.072V Manually changing DRAM voltage to 1.353V... VCCIO = (auto) 1.180V VCCSA = (auto) 1.280V Thats a helluva bump. I tried taking it down to 1.05V and 1.07V respectively but got mem test errors, followed by a spate of oddities like my USB wifi disconnecting and refusing to re-enable. Can't tell if they are related because I don't know what the effects of too low VCCSA are. This is rapidly turning into a chaotic experiment where stuff is going wrong and I don't know when, why or how. Currently mem testing at 1.10V and 1.10V respectively with DRAM voltage at 1.353V. Its going good so far and my wifi hasn't poo poo itself yet. What are some common effects of VCCIO and VCCSA being much too high or much too low? Googling it gives you alot of pretty vague warnings either way. WanderingKid fucked around with this message at 21:57 on Jan 19, 2017 |
|
#
?
Jan 19, 2017 21:53
|
|
|
WanderingKid posted:Hmm, you are right. Leaving VCCIO and VCCSA on auto, will make both skyrocket if you whack up DRAM voltage. I used a TweakTown guide for the VCCIO and VCCSA that said they should both be set at 1.180v-1.190v for 3200MHz RAM. It got rid of some of the stability issues I was having and reduced the latency quite a bit more, but I've been wondering if that's too high. I haven't noticed any issues, though.
|
|
#
?
Jan 19, 2017 22:18
|
|
|
Its still mem testing fine at 1.100V and 1.100V for me. What kind of stability issues were you having? Edit: nope, mem test error at 1.100V and 1.100V. Took alot longer this time but now trying at 1.100V and 1.125V. WanderingKid fucked around with this message at 23:11 on Jan 19, 2017 |
|
#
?
Jan 19, 2017 22:53
|
|
|
WanderingKid posted:Hmm, you are right. Leaving VCCIO and VCCSA on auto, will make both skyrocket if you whack up DRAM voltage. I had memory test errors when it was too low - there were probably other problems too, but I was only testing RAM stability at that time. If it's too high, it'll probably be stable, but I'm OCD about not using more volts then it needs. I think both of mine are set to 1.2v for 3600 RAMs. (I'm using 1.35V voltage on the RAM). quote:This is rapidly turning into a chaotic experiment where stuff is going wrong GRINDCORE MEGGIDO fucked around with this message at 00:51 on Jan 20, 2017 |
|
#
?
Jan 20, 2017 00:41
|
|
|
WanderingKid posted:Its still mem testing fine at 1.100V and 1.100V for me. I don't remember since that was a few months ago and I uninstalled all of the software I was using for testing and benchmarking, including most of the logs.
|
|
#
?
Jan 20, 2017 01:32
|
|
|
I'm up to VCCIO = 1.1250V and VCCSA = 1.1500V. Actual voltages are a bit higher. VCCSA actually fluctuates around 1.17V and 1.18V. Tried VCCSA = 1.1375V but got 1x unexpected shutdown after posting and then 1x BSOD with bugcheck name: ATTEMPTED_WRITE_TO_READONLY_MEMORY. Wifi issues continued yesterday (though that could still be unrelated). WanderingKid fucked around with this message at 17:53 on Jan 21, 2017 |
|
#
?
Jan 20, 2017 21:24
|
|
|
Is it normal for utilities like hwinfo and cpu-z to report dramatically different voltages than my BIOS? My BIOS shows 1.15v, while the utilities in Windows are showing voltages fluctuating between 1.29 and 1.36 which seems really high. Running 7700k, ASUS Z270E ROG Strix motherboard, Phanteks PH-TC14PE cooler. I'm new to the overclocking thing so maybe I'm fretting about nothing.
|
|
#
?
Jan 23, 2017 05:34
|
|
|
They are almost always different and neither value is accurate. The only way to get an accurate reading would be to measure VCCP from the motherboard with a voltmeter, but I doubt your mobo has a service manual with schematics or test pads on the board itself. Mine doesn't.
|
|
#
?
Jan 23, 2017 11:21
|
|
|
I've got an i7 5820k that I'm overclocking. I set the core voltage to 1.110, clock to 4.20ghz, and stress tested it for a half an hour. Heat never went above 75 degrees, voltage and processor load steady throughout. is there anything else I need to change or should that be ok? does the bridge clock matter? terrorist ambulance fucked around with this message at 06:11 on Jan 24, 2017 |
|
#
?
Jan 24, 2017 03:31
|
|
|
Currently messing with my new i5-7600K. I've got the vcore set to 1.2 (which is stock, yes?), and the voltage mode set to "Adaptive + Offset" in the BIOS (it's an MSI Z270 board), with the Offset mode set to [+] and an offset of 0.175v. Everything appears stable at 5GHz thus far. Absolutely all other relevant OC settings are at stock. Before I proceed to see if I can push it further, I wanted to check and make sure I wasn't loving something up with these settings, and thereby potentially frying poo poo.
|
|
#
?
Jan 24, 2017 07:14
|
|
|
Sudden hardlock where the screen stays on the last image and the sound goes into the last-buffer-loop thing - CPU and not RAM, right? It's bizarre, it only happens when I'm playing around in the NOX android emulator, and happens even if I underclock the CPU  I'm concerned about the RAM because it ran for years at 1.65v -1600, but it's 1.5v 1866 that MSI didn't recognize properly.Unless my RAM is right on the edge and an OS cell (since the OS is always in the same physical place) is the one faulting first.
|
|
#
?
Jan 24, 2017 13:33
|
|
|
Reposting in this thread since the original thread I was in is not for this type of stuff. My problem is that I can not get a stable overclock even at 4.5Ghz @ 1.345V. I followed the instructions according to this video. What I encounter every single time I use the CPU-Z stress test, is my screen going black, not once did I get a BSOD. The longest I could get it to run for the stress test at 4.5Ghz was roughly 6 minutes then I get a black screen again and I have to restart. I've also just now updated to latest bios but haven't attempted to overclock again. WanderingKid posted:Kinda hard to tell with so little information. Whats core voltage? Whats max core and package temperature shortly before it blackscreens? Does it always blackscreen or does it sometimes BSOD? If so, what is the bugcheck name? Agrajag posted:I have the voltage set to 1.345, same as in this video, and an OC of 4.5Ghz. Agrajag posted:Should I just start increasing the voltage in .025 increments till I can get a stable oc at say 4.7Ghz instead of 5.4Ghz? Also is there some kind of resource/table thing that I can refer to and see what voltages most people are using to get their stable oc's? Just checked HWinfo at default everything the temps for the CPU are: Value: 27C - Min: 26C - Max: 31C CPU fan is at 400rpm range I found this website and tey are generally getting 4.8Ghz at 1.35V, what gives? I must be doing something hella wrong yet I'm doing everything exactly as indicated. http://www.overclock.net/t/1621347/kaby-lake-overclocking-guide-with-statistics Agrajag fucked around with this message at 21:57 on Jan 24, 2017 |
|
#
?
Jan 24, 2017 21:51
|
|
|
Different processors perform differently, sometimes you get a bad one.
|
|
#
?
Jan 24, 2017 22:16
|
|
|
craig588 posted:Different processors perform differently, sometimes you get a bad one. if that's the case I have the worst luck in history because my i7 2600k wouldn't stay stable above 4.2Ghz, and even at 4.2Ghz my computer freezes roughly every 2 days on the dot.  I ABSOLUTELY can not believe i have such terrible luck. Please someone help me! Agrajag fucked around with this message at 22:24 on Jan 24, 2017 |
|
#
?
Jan 24, 2017 22:21
|
|
|
Agrajag posted:How do I look for a bugcheck name?  The bug check name is CLOCK_WATCHDOG_TIMEOUT quote:Value: 27C - Min: 26C - Max: 31C These are idle temps. What are your 100% load (stress test) temps just before you black screen? Real Temp will show you the temperature of each core, how far off TJ max you are and has indicators to show when you are thermal throttling. The load temps just before you blackscreen are important because you want to eliminate thermals as the cause of your limited (perhaps throttled?) clock speed, possible cause of instability etc. Do you have XMP settings for your memory loaded? If so, get off that poo poo so you don't confuse memory related instability with cpu related instability. WanderingKid fucked around with this message at 22:39 on Jan 24, 2017 |
|
#
?
Jan 24, 2017 22:37
|
|
|
Agrajag posted:if that's the case I have the worst luck in history because my i7 2600k wouldn't stay stable above 4.2Ghz, and even at 4.2Ghz my computer freezes roughly every 2 days on the dot. These guys try to determine how far individual CPUs will OC and sell them with a slight mark up at that rating. I've never bought from them but it's worth considering if you have a specific clock target in mind and don't want to buy and return CPUs to a store to try and find one. https://siliconlottery.com/
|
|
#
?
Jan 24, 2017 22:41
|
|
|
Rexxed posted:These guys try to determine how far individual CPUs will OC and sell them with a slight mark up at that rating. I've never bought from them but it's worth considering if you have a specific clock target in mind and don't want to buy and return CPUs to a store to try and find one. Yeah that's not much of a markup to make sure you don't get hosed on clockspeeds. Will they give you any sort of help if you need to claim a warranty with Intel though?
|
|
#
?
Jan 24, 2017 22:45
|
|
|
I'm still not totally convinced Agrajag has birdshot for silicon though. Still lots of information to get first.
|
|
#
?
Jan 24, 2017 22:49
|
|
|
WanderingKid posted:
never had screen notification popup. i'll check go check the other stuff and get right back with the info XMP settings for memory were initially loaded but I turned them off and still got black screens
|
|
#
?
Jan 24, 2017 22:52
|
|
|
Reset your bios (reflash)
|
|
#
?
Jan 24, 2017 22:54
|
|
|
Yea, set everything to stock including the RAM to 2133, and overclock it to 4.5GHz with 1.35v. Be interesting if it is temps, Black screen could be the BIOS shutting it down.
|
|
#
?
Jan 24, 2017 22:55
|
|
|
I mean I understand that silicon lottery is a real thing but part of me doesn't want to believe you bought a cpu that came from an off day at the Intel factory when some technician was accidentally overseeing immersion lithography on a wafer of dessicated dogshit.
|
|
#
?
Jan 24, 2017 23:02
|
|
|
When a lovely 7600K is one that "only" gets 4.9GHz, I'd be inclined to believe it is something other than the silicon for one to only get 4.5GHz.
|
|
#
?
Jan 24, 2017 23:07
|
|
|
Oh, I hadn't been following processors so much. Processors are pretty cool now. I thought the average was like 4.8Ghz.
|
|
#
?
Jan 24, 2017 23:53
|
|
|
You could get unlucky. But theres unlucky and theres user error. Until user error is ruled out, Agrajag still has hope in our hopeless post Obama world.
|
|
#
?
Jan 25, 2017 00:03
|
|
|
Ok here it goes and I'm kinda confused: programs used: Real Temp and ROG Realbench (all in celsius) First run 4.5Ghz @ 1.35v ROG Realbench - all the options checked - 1 run @ 15minutes - up to 16GB of RAM selected Real Temp readings - temps at load were around the 60-64 range all cores - TJ max was 40 something (half the window for Real Teamp was coverd by the popups from the stress test program) result: instability detected The test had a crash popup for "LuxMark-x64.exe" but I did not get a system crash/BSOD. I did get a momentary blackscreen and then it went back to desktop. I did a quick google and apparently it has something to do with Nvidia drivers or something, I'm not too sure, but I did close my EVGA Precision XOC for the second run without restarting the pc. Second Run 4.5Ghz @ 1.35v ROG Realbench - all the options checked - 1 run @ 15minutes - up to 16GB of RAM selected Real Temp readings - min temps: 24, 25, 25, 26 - max tamps: 71, 69, 70, 71 - during the test temps around 64-68 Result: No crash this time and ROG Realbench gave a successful stress test result @ 4.5Ghz and 1.35v. What the eff? I mean I'm not complaining here and I guess I'm happy that my 4.5Ghz overclock seems to be stable this time, but what the hell? It cant be that EVGA Precision XOC was the cause this whole time can it? edit: oh yeah and for the results thing for ROG Realbench I can not seem to get it to show up and says I need to do one complete run to get it but I did do a full run. In the box it does say "result hash match". Agrajag fucked around with this message at 00:44 on Jan 25, 2017 |
|
#
?
Jan 25, 2017 00:38
|
|
|
Should I attempt to go above 4.5Ghz?
|
|
#
?
Jan 25, 2017 01:01
|
|
|
Hmm. Black screened and recovered. Sounds like your graphics driver stopped responding and recovered, no? Maybe Precision is the problem. I'd get rid of it unless you really need it, at least for now. Wouldn't be the first time software has done whacked things to my PC. MalwareBytes web protection module (mwac.sys) frequently BSODs my system with bugcheck code BAD_POOL_HEADER. At the time I was memory overclocking so I thought it was memory instability but nope. Disabling web protection module fixed it. Go again at 4.6 ghz. Temps look fine.
|
|
#
?
Jan 25, 2017 02:08
|
|
|
WanderingKid posted:Hmm. Black screened and recovered. Sounds like your graphics driver stopped responding and recovered, no? Maybe Precision is the problem. I'd get rid of it unless you really need it, at least for now. Wouldn't be the first time software has done whacked things to my PC. MalwareBytes web protection module (mwac.sys) frequently BSODs my system with bugcheck code BAD_POOL_HEADER. At the time I was memory overclocking so I thought it was memory instability but nope. Disabling web protection module fixed it. I mainly used Precision XOC for the ability to set my own GPU fancurve. Otherwise, the fans would not even spin until they hit a certain temp, and I'm not really comfortable with that. Ok, I'll give 4.6 Ghz a run.
|
|
#
?
Jan 25, 2017 02:11
|
|
|
Try getting rid of it anyway and just monitor your gpu temps on the next iteration. If you get concerned then you can end RealBench and reinstall it. For now, just try to eliminate anything you don't need that could possibly cause instability. Precision looks like one of them if RealBench is black screening during a gpu stress test and you are able to recover. Or hell, maybe your graphics driver installation is hosed. Dunno yet, but you'll find out soon.
|
|
#
?
Jan 25, 2017 02:15
|
|
|
My friend, a YouTube game streamer by trade, got a 6800k and asked me to overclock it, it has an AIO water cooler with 140mm radiator. I wound up 4.1ghz at 1.3v, I tried for 4.2 and it booted but Prime95 was reporting math errors in the threads. With Blend going it tops out at 60c. Is that good for a quick and dirty stable overclock? I know the 6800k isn't supposed to overclock as high as 6700k and friends.
|
|
#
?
Jan 25, 2017 17:12
|
|
|
Zero VGS posted:My friend, a YouTube game streamer by trade, got a 6800k and asked me to overclock it, it has an AIO water cooler with 140mm radiator. Prime95 is a particularly intense stress test as it really hammers the CPU with new instructions (AVX2) only present in post-Haswell CPUs (e.g., the 6800k Broadwell-E). Broadwell-E CPUs have an "AVX Offset" in the BIOS to help with this - generally an offset of 2 or 3 (corresponding to 200-300MHz lower clocks when running AVX code) - that will help with temps, stability, and getting a higher overclock with the (probably) ~90% of games/apps that don't use AVX. Regardless, 4.1GHz isn't too far from what these chips most often reach - even 4.3/4.4GHz often takes 1.35-1.4V, and that's without testing Prime95 or anything more than RealBench for an hour. (https://siliconlottery.com/collections/all/i7-6800k) X99 has a ton of weird settings that can help in often strange ways, so getting a better overclock can definitely take a lot of time/effort. If high speed RAM is involved (especially 3000MHz or higher) I would double check the VCCSA and VCCIO voltages as they get fed ridiculous amounts if left on Auto.
|
|
#
?
Jan 26, 2017 01:27
|
|
|
I tried stress testing my 6600k with Prime95. 4.5 ghz at 1.26v. Results were good, 0 errors, 0 warnings, but my temps skyrocketed to 85C. When stress testing using CPU-Z and the intel tool, my temps never got above 65C. Is this normal?
|
|
#
?
Jan 26, 2017 03:03
|
|
|
Yes it is. Prime95 will stress your CPU in ways no other app can, so if it's stable and not thermal throttling there, it's good.
|
|
#
?
Jan 26, 2017 04:29
|
|
|
Yeah, that's normal. You might want to bump up the VCore by 10-25 mV beyond the bare minimum needed to pass Prime95 because there are a few programs that can still be unstable even if it's good there (I'm looking at you 3DMark). e:grammar Regrettable fucked around with this message at 02:36 on Jan 27, 2017 |
|
#
?
Jan 27, 2017 02:32
|
|
|
I bought a EVGA GTX1060SC (6GB) for my small HTPC/NAS/Steam-streaming PC and everything is working well. The 1060 really seems to have plenty of performance for the games I'm playing at 1080p, so I'd like to undervolt it a bit so it runs cooler. I'm using MSI Afterburner. The core voltage offset ranges from 0 to 100mV and a negative value doesn't seem to be available. Am I correct in assuming that reducing the powertarget and increasing the clockspeed offset would achieve the desired result? The last time I overclocked a GPU was about a decade ago with fixed voltages and one GPU and one memory clock slider. 
|
|
#
?
Jan 27, 2017 17:53
|
|
|
|
| # ? Jun 6, 2024 02:15 |
|
|
Just try limiting FPS in precision X to 60 and see if that makes it run cool enough for you.
|
|
#
?
Jan 27, 2017 17:57
|
|
