Overclocking Megathread 3.11 for Workgroups

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > Overclocking Megathread 3.11 for Workgroups

«‹›5 »

future ghost: Dec 5, 2005; Gun Saliva

Alereon posted:

I was thinking of waiting for either news on a possible Samsung 850 Evo or another round of discounts on the 840 Evo, but I'm sorely tempted to just pull the trigger now.

840 EVO's are already going to be ridiculously fast unless you have really specific requirements or something. I can't remember if you already picked up a SSD but moving from a C2Q to that chip should be mind-blowing even if it won't quite hit 5ghz. I'm still coasting on my 2600k thanks to HT making encoding and other highly-threaded tasks a cakewalk.

# ¿ Jul 13, 2014 02:40

Adbot: ADBOT LOVES YOU

# ¿ May 11, 2024 20:19

future ghost: Dec 5, 2005; Gun Saliva

echinopsis posted:

Overclock your pancreas

# ¿ Jul 31, 2014 21:38

future ghost: Dec 5, 2005; Gun Saliva

Martytoof posted:

Not sure if this is the right thread for it, but I had my H60 CLC die on me today (CPU was hitting 80c, BIOS read N/A for pump "fan" speed) -- thankfully my system shut itself down. The sales guy said he had an H60 die on him too. Is this something that's plaguing H60s? I had them replace it with same, but now I'm wondering if I should have picked up something different. I was just about to start thinking about OCing my i7 920 but now I'm worried this is going to donk out again in a week.

I would have though tit was one of those things where if it survives the first day it will run forever. But it took like 8 days to die.

An H60 isn't really going to cut it for overclocking a 920 at reasonable volume anyways so you might have dodged a bullet. You're going to want a higher-end CLC unit for that chip or otherwise look into aircooling. A used Thermalright Ultra Extreme 120 would be enough and not too expensive if you don't mind doing push-pull 120mm fans, or something like this would handle that chip with no problems if you don't want to step up to an H80i or something.

# ¿ Aug 10, 2014 23:56

future ghost: Dec 5, 2005; Gun Saliva

If it's a BFG power supply it's probably too old to rely on as the company went under in 2010. Better off looking into a newer higher-quality unit.

Also from the Hardware thread you're using a 680 LT SLI board - it's possible it's failing due to age (or just nForce loving up), but IIRC a common issue with those boards was the northbridge and southbridge fans failing early and causing lockups when the chipsets overheated. Check the fans to see if they're working and free of dust.

Modern chipsets usually run cool enough to avoid needing active cooling fortunately, but older nvidia chipsets did not tolerate fan failures well.

future ghost fucked around with this message at 17:53 on Aug 22, 2014

# ¿ Aug 22, 2014 17:35

future ghost: Dec 5, 2005; Gun Saliva

It's probably fine to run it a bit longer, but I'd run hwinfo in sensors-only mode while running one of the 3DMarks or something to see if the voltages (12V, 5V, and 3.3V) look within spec before I'd rely on it. Software readout won't be exact but it should give you a general idea. For a PSU recommendation check the OP of the parts-picking thread in SH/SC but in general you'll probably want something made by Seasonic.

# ¿ Aug 23, 2014 08:18

future ghost: Dec 5, 2005; Gun Saliva

>90C is really too hot to be running even during synthetic testing unless you're actively trying to kill the chip really quickly. At most I would accept ~80C max in IBT or ~72C max in P95 -- anything higher than that may mean the chip could go above "safe" levels in heavier normal use.

# ¿ Aug 25, 2014 20:46

future ghost: Dec 5, 2005; Gun Saliva

I probably should have ended that post with ".. on Sandybridge", since Haswell would certainly end up higher, although 90C still seems like alot for only 4.0 ghz on a 4670K.

# ¿ Aug 25, 2014 21:57

future ghost: Dec 5, 2005; Gun Saliva

Don Lapre posted:

Its normal for some heatsinks to be able to twist a little bit, usually the kind mounted by a single bar going through the middle. I wouldn't worry about that.

The last 212+ I installed would twist slightly, and the old Thermalright TRUE-style mount would do that as well. Doesn't sound like anything to worry about if it's just twisting and not sliding all over the place or something.

# ¿ Aug 26, 2014 18:46

future ghost: Dec 5, 2005; Gun Saliva

You're stressing the system more in PS2 since the graphics card is involved whereas it wouldn't be in AIDA64, along with the rapid voltage changes you'd see in games. If you're seeing too much vdroop to remain stable at full load you'd either need to compensate with LLC or a higher vcore offset to keep the CPU stable. I know gently caress-all about Asrock boards, but I assume that there's higher offset options and maybe a setting for LLC if you want to try that. If you're overclocking the GPU as well you should probably revert to stock while you're attempting to get the CPU stabilized.

# ¿ Sep 20, 2014 19:13

future ghost: Dec 5, 2005; Gun Saliva

Probably worth reseating the cooler with new TIM since mid-high 80's might be acceptable but not 100C.

# ¿ Sep 21, 2014 17:20

future ghost: Dec 5, 2005; Gun Saliva

Probably need to reseat, though what's your vcore set at?

# ¿ Sep 26, 2014 01:17

future ghost: Dec 5, 2005; Gun Saliva

Combat Pretzel posted:

Putting two fans in series would be similar to putting them on resistors, right? I'd like to ghetto-slowdown some fans without resorting to buying new stuff (be it fans or resistors).

It's probably easier to make a 7V modded cable if you're looking to do this with a few fans:
http://www.overclockers.com/forums/showthread.php/372297-Guide-to-the-7v-Fan-Mod

Note that you can run them at 5V this way too, but many fans have issues starting at 5V vs. 7V.

# ¿ Oct 1, 2014 21:04

future ghost: Dec 5, 2005; Gun Saliva

I have a couple 7V modded 4-pin cables I stripped from old fans I've used off and on for years without issues. It's hard to cause any damage unless you maybe have a very old and low-end power supply (the posts were written in 2005 where this would be more likely), but most modern units can handle minor cross-loading like that without any issues. Aftermarket GPU coolers like the Accelero line regularly include 7V and 5V adapters these days so there's little to no risk. That said 5V will be quieter assuming your fans start up at that voltage and you don't need more than minimal airflow.

# ¿ Oct 1, 2014 23:40

future ghost: Dec 5, 2005; Gun Saliva

Afterburner's nice since it gives you some basic options straight-out that probably won't do much damage, but if you want you can go into settings and setup manual fan curves, edit voltages, and extend official overclocking limits fairly easily.

Also not having to dip into the registry to disable ULPS/zero core for each driver update. :arghfist:

# ¿ Oct 11, 2014 19:56

future ghost: Dec 5, 2005; Gun Saliva

TheCobraEffect posted:

I'm trying to overclock for the first time, but I think I must be doing something wrong. I have a G3258 and an ASRock z97 Pro 4 motherboard. I put my core voltage at 1.350 and I can only get windows to log in at 4.1 GHz. It's perfectly stable like that, the temperature will hit a max of 65, but it generally stays around 60 C while running prime 95.

Is there some setting I'm missing to help me get up to 4.4-4.5 GHz like I was expecting? If my chip is just limited to 4.1 GHz that's okay, and I'll probably knock Vcore down so it'll run cooler. It was stable at 4.1 GHz with Vcore set to auto, but I'm not sure if I should leave it on auto.

Thanks for the help!

Are you using the latest BIOS version for your motherboard?

# ¿ Nov 15, 2014 23:12

future ghost: Dec 5, 2005; Gun Saliva

You'd probably need some kind of custom long screws to make it work but IIRC people usually use long bolts and nuts to combine standard rads & thicker fans.

There's also universal radiator plates and gaskets available that may help:
http://www.sidewindercomputers.com/radiatorparts.html

Performance-pcs is another good source that might have what you'd need.

# ¿ Nov 17, 2014 15:07

future ghost: Dec 5, 2005; Gun Saliva

With alternating fan/rad sets it'll work fine as long as they're not super restrictive, but you definitely don't want to just stick two rads directly together and you shouldn't do more than two.

# ¿ Nov 17, 2014 16:47

future ghost: Dec 5, 2005; Gun Saliva

GPU core overclock set too high probably, or at least insufficient voltage. What power supply are you using?

edit: Factory Factory makes a good point. If your 280 has Elpida VRAM it's likely to be overclocked too far, and it might not be truly stable at those clocks even with Hynix chips, assuming 280's are even made with those.

future ghost fucked around with this message at 23:11 on Dec 13, 2014

# ¿ Dec 13, 2014 22:37

future ghost: Dec 5, 2005; Gun Saliva

I have a giant tube of MX-4 on hand for GPUs, but I haven't really used it much since I'm still going through the leftover Chill factor III stuff from a review that's been working fine. I usually don't give alot of thought to TIM in general but I prefer to avoid anything like AS5 that requires burn-in or that possibly risks causing a short. Same with weird poo poo like gallium Coollaboratory Liquid PRO or anything that requires more thought than 'squirt and tighten'. Not planning to micromanage loving thermal paste.

# ¿ Jan 9, 2015 19:36

future ghost: Dec 5, 2005; Gun Saliva

That's pretty warm although not unexpected with the stock cooler. The easy-mode overclocking probably bumped up the voltage by alot as well. I'd look into replacing the heatsink with something better, or at least manually set your CPU vcore. The 'safe' upper voltage limit on your chip's going to be about 1.38V but I wouldn't go that high on the stock cooler. Something like a 212+ EVO or that newer Cryorig ~$35 cooler would do the trick.

future ghost fucked around with this message at 19:43 on Jan 17, 2015

# ¿ Jan 17, 2015 19:27

future ghost: Dec 5, 2005; Gun Saliva

Having the radiator fans as exhaust is fine. The only change I'd make is turning the bottom fan into an intake or alternatively removing it.

# ¿ Feb 8, 2015 00:46

future ghost: Dec 5, 2005; Gun Saliva

For a 130W chip I'd want to use something beefier than a 212 anyways if only for noise reasons. It looks like you can get a free 2011 mounting kit for a Black Rock from BeQuiet's site. Alternatively if Thermalright's available where you are you can get a universal heatsink mount for any of their ~2007+ coolers for about :10bux:

if it doesn't already support it. Picked one up for my old TRUE so I could reuse it on a modern board.

# ¿ Feb 14, 2015 00:36

future ghost: Dec 5, 2005; Gun Saliva

Depends on the monitor but you'll probably have better luck asking for specifics in the SH/SC Display thread.

# ¿ May 2, 2015 07:27

future ghost: Dec 5, 2005; Gun Saliva

You should just be able to set the clock speed, voltage, and timings to whatever it's rated and it should just work. Clock speed and timings both matter in different ways, although neither will probably make a tremendous difference for games - you might gain a few FPS going from like CAS11 1333mhz to CAS9 1600mhz. If your modules are rated for higher speeds there's no reason not to set them accurately unless they're rated for 1.65v or something by default. Specific applications that benefit from memory speed or tighter timings would likely see more gains.

# ¿ May 6, 2015 07:25

future ghost: Dec 5, 2005; Gun Saliva

TenementFunster posted:

I have a 120mm Thermalright heatpipe cooler that is a few years old that needs to be reseated with new thermal paste on my CPU due to a slowly increasing temps. It currently has one 120mm on there on a rheostat, but I was thinking about putting two 120mm fans in push/pull config. Is this a more effective setup as far as noise/performance goes than a single beefier 120mm fan? I'm don't have much of an overclock on my old i7 2600k, but I'd like to keep it reasonably quiet.

Is there a favorite 120mm CPU fan out there currently? I have a Scythe fan with some kinda fancy bearings on right now, but I don't know if there is a new hotness now.

Depends on which cooler you have. If it's a TRUE or Silver Arrow then push-pull is good since they're more restrictive. If it's something with more spacing and larger fins like an HR-02 then you can just throw on a single fan. I used to have push-pull but I dropped to a single 120x38mm Sanyo Denki x1011 fan on a rheostat at 5v-7v and it works fine on a 2600K.

If you have a Scythe Gentle Typhoon you're going to have a hard time finding a better 25mm fan, but if it's a Slipstream or a Kaze you might want to look into replacing it. If you go with a single 38mm-thick fan you can get most of the pressure benefits of push-pull without needing 2 fans.

If you want to go with 38mm fans you could look into a low or medium-speed panoflo like this:
http://www.sidewindercomputers.com/pa12ulqu.html or
http://www.sidewindercomputers.com/pa12mesp.html
-Avoid Delta (loud) or Yate Loon (cheap, but lovely sleeve-bearing fans)

For 25mm fans you could grab a couple of these for push-pull:
http://www.sidewindercomputers.com/scge120mmsic2.html

# ¿ Jul 12, 2015 17:51

future ghost: Dec 5, 2005; Gun Saliva

Is it changing the internal bus/uncore speed at all? Try setting it manually to 100mhz. Also 1.34v is really high for 4ghz, although with auto settings it's probably just pushing things harder.

What board is it?

Parts Kit posted:

Also when some more ram gets in and I have the case open anyways should I go ahead and redo the thermal goop with something from Arctic Silver? Got a tube but went ahead and used the bit that came on the Corsair H80i when I put it in yesterday. After pulling enough stuff out to get the H80i's mount in place I was kinda in a 'gently caress it, let's get this together and running' mood. And a stupid ram question -- the stuff in the computer should be 1600, but is running at 1333 according to the BIOS. Any reason I shouldn't just manually bump it up where it should be?

It's fine to raise the memory speeds/timings to stock settings if it defaults to the lower speed. My ripjaws default to CAS 11/2T but can be set to CAS9/1T manually or with the built-in profile. Also 69C-72C isn't overly warm for Prime95 so I wouldn't worry too much about it even putting the clocks back up to 4.3 since it'll almost never hit the same temperatures, as long as voltage isn't higher than the supposedly-safe limit of 1.38V (that said, I've allowed my 2600K to hit up to 1.39v/1.4v rarely with LLC since I got it without any noticeable effects).

future ghost fucked around with this message at 19:58 on Aug 22, 2015

# ¿ Aug 22, 2015 19:54

future ghost: Dec 5, 2005; Gun Saliva

Parts Kit posted:

Thanks. That was 72 degrees after about 5 mins so should I just not worry and give it a little more, maybe 4.5ghz? Normal processing temps have been much lower while playing games/using photoshop and it did auto throttle down when it hit 72 degrees.

I wouldn't worry too much about 72degrees in Prime95, but I'd suggest testing for a longer period of time before moving up clocks. Also you should try to figure out how to use offset vcore settings rather than rely on the automatic voltage option, since alot of boards will pump vcore hard on auto if you go above ~4.2ghz on these chips. Better to have it raise the vcore to a max of ~1.38v when necessary instead of leaping to 1.45 or higher.

# ¿ Aug 24, 2015 16:55

future ghost: Dec 5, 2005; Gun Saliva

You don't need to use the XMP. Just verify that all the timings and voltages are correct for the rated settings.

# ¿ Aug 28, 2015 23:46

future ghost: Dec 5, 2005; Gun Saliva

Sh4 posted:

Yeah, what's your point

How does that still work

# ¿ Sep 2, 2015 20:43

future ghost: Dec 5, 2005; Gun Saliva

gently caress CPU fan curves. Just strap a gigantic fan at 5v-7v on your noctua/thermalright cooler and forget it exists.

# ¿ Nov 16, 2015 04:52

future ghost: Dec 5, 2005; Gun Saliva

You can't really overclock with that system but you could get a more powerful videocard than the 760 and that would make a noticeable difference.

# ¿ Nov 29, 2015 05:09

future ghost: Dec 5, 2005; Gun Saliva

Sh4 posted:

If you look above I posted my hwmonitor with the same board as yours and my cputin value is normal, also don't know your cpu but seems like its similar to mine for Vcore where you can go safely to like 1.45V if you cooling is good enough

I read way too many reports of 1.5V killing 2500k's and 2600k's when they were still sold. No way I'd try 1.45V on a 22nm chip unless you were planning to replace it a year from now.

# ¿ Dec 11, 2015 19:39

future ghost: Dec 5, 2005; Gun Saliva

Sh4 posted:

Those can do 1.5V no problem, maybe its just bad luck or user error, ran mine at 1.6v for at least a month for science and nothing really changed (I had the theory that zapping an old cpu would make it OC more, i think it did something but my 5Ghz dream is still crushed)

No I meant that they can manage it, but it's not a 24/7 thing (or shouldn't be) as electromigration can get alot worse even if temperatures are kept in check. Tons of Northwood-A's died from it before people figured out that 1.8-1.9V was way too much voltage for them.

# ¿ Dec 12, 2015 00:48

future ghost: Dec 5, 2005; Gun Saliva

Check voltages with sensors-only mode in HWiNFO and see how it matches up. 1.4V is pretty high I think, but I'm still on Sandybridge so I don't know how far below 1.4 you need to stay for that chip.

# ¿ Jun 29, 2016 00:27

future ghost: Dec 5, 2005; Gun Saliva

My board (Aorus Z390 Pro WiFi) also has a ton of options that I haven't figured out just yet. All the guides I'm seeing are recommending constant vcore/disabling power options, etc., but not doing a great job explaining the more esoteric settings.

I ended up settling on 4.8 all cores on my 9700k for now since I can run it at 1.2v DVID. It could probably go lower than that, but the minimum vcore at a -.070v offset drops down to .600v at idle and I'm not sure what the lower limits are for the chip. XMP for the 2x16GB 3200mhz Corsair sticks took VCCSA and VCCIO to 1.3v which seemed really high, so I dropped both of them down to 1.2V. I don't know if there are any benefits of running them lower than that but memtest86+ ran all day yesterday without errors.

Temperatures under a HR-02 do not go above 70C during normal use, but max/very high IntelBurnTest and OCCT small data sets will push some of the cores >95C. Is that normal? At 1.2v I'd expect it to be lower but it may be some LLC setting doing this, although HWiNFO does report CPU power at 190w+ so maybe that's to be expected. I have it at 'normal' LLC but I'm not clear on what the differences are between normal, low, and standard. Realbench tests max out at about 60C per core. The massive heat increases are a little disturbing after being on a cool-running overclocked 2600k for years. If it's just going to run hot in stress tests it might be just hitting the limits of this cooler (it was designed to passively cool socket 1366 chips), and I'll just deal with it as it's nearly silent and normal use doesn't go above 72C for any core.

future ghost fucked around with this message at 18:58 on Dec 12, 2018

# ¿ Dec 12, 2018 15:05

future ghost: Dec 5, 2005; Gun Saliva

Intel allows increased rated tjmax to 115C, but tjunction is still 100C so it seems prudent to try and stay below it if it all possible. With water-cooling I've read that up to 1.4v is relatively safe (though increasingly hot the closer you get), but with my 9700K on air I'm finding that staying below 1.3v whenever possible is the only way to keep stress-test temperatures somewhat sane.

Edit: 4.9ghz on this CPU actually runs IBT cooler than 4.8ghz at the same DVID, despite using more power. I'm not sure how that happens, but I'm just going to blame Gigabyte's confusing LLC tiers. 5ghz didn't like this vcore at all so I think I'll just work on lowering DVID for 4.8 some more. An extra 100mhz isn't worth the possible instability.

future ghost fucked around with this message at 03:27 on Dec 13, 2018

# ¿ Dec 13, 2018 02:51

future ghost: Dec 5, 2005; Gun Saliva

VelociBacon posted:

Sorry it's actually VR Vout. Here's a screencap:

This is with HWinfo64 v5.92-3580. Make sure you update your motherboard BIOS also.

I have two listings for VRout on my Z390 Pro WiFi board that both give different values and I'm trying to figure out which is more accurate. Which sensor heading is that using in your screenshot?

Also does anyone know what a (rough) lowest idle before limit is for a 9700k? I've got mine undervolted to .6v at idle which has tested stable, but I'm kinda afraid to try for anything lower.

# ¿ Feb 7, 2019 20:21

future ghost: Dec 5, 2005; Gun Saliva

Just a heads-up, the next one or two times you reboot it may switch back to the other BIOS. My half-assed 9700K 5ghz attempt ended up doing the same thing on an Aorus Pro WiFi board, where it switched to the backup BIOS which I updated and tweaked for overclocking, before it then decided to switch back to the non-backup BIOS but with default settings. I took it in stride as it least it prevented me from rendering everything unbootable.

# ¿ Feb 14, 2019 06:57

future ghost: Dec 5, 2005; Gun Saliva

I didn't know ECS even made X58 boards, or at least ones that didn't immediately self-ignite.

# ¿ Mar 20, 2019 17:24

Adbot: ADBOT LOVES YOU

# ¿ May 11, 2024 20:19

future ghost: Dec 5, 2005; Gun Saliva

LRADIKAL posted:

Was unlocking and overclocking a Phenom II X2 555 that I got from a buddy to use as a Plex server. I settled on 3 cores, after unlocking one of them resulted in bad stability. I boxed it up in the cheapest case I could find on newegg (mistake). I began having stability issues again. Turns out the case had almost no intake space. I pointed my thermometer at the mosfets...

Alot of cheap boards in that era didn't even pretend to put heatsinks on the VRM section. Get an Enzotech MOS-C1 kit and some sekisui thermal tape and you won't have to worry about it. Rub a pencil eraser on the VRMs before applying the tape.

# ¿ Jul 20, 2020 20:22

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > Overclocking Megathread 3.11 for Workgroups

«‹›5 »