|
jonathan posted:Between die and IHS I used arctic silver. I was reading that the biggest difference is closing up the tolerance between the die and IHS by removing the glue between the IHS and PCB. So I popped the IHS, cleaned all the sealer/glue off, already a thin layer of arctic silver over the die, then reglued the IHS using the weakest JB weld. Then used the motherboard CPU retainer to keep the IHS clamped. I only glued down the 2 side tabs on the IHS. Not the entire perimeter. Let's back up a second. What load tests are you doing? AVX (prime95 small fft) with no offset? Or like, the CPUID stress test? What happens if you run a game or something for 20 minutes, what temp ranges does it stay in? What voltage are you running at? The speed of a spike is not an indication that something is wrong. What is concerning is that your hitting the thermal ceiling doing tests, but without knowing the tests and conditions its hard to say if its really wrong. I could make a perfectly water cooled system hit the thermal limit if I really wanted to.
|
#
?
Mar 13, 2018 20:19
|
|
|
|
| # ? May 15, 2024 16:21 |
|
|
Lockback posted:Let's back up a second. What load tests are you doing? AVX (prime95 small fft) with no offset? Or like, the CPUID stress test? What happens if you run a game or something for 20 minutes, what temp ranges does it stay in? What voltage are you running at? Sorry, that was a 5am post after a lovely work night. I'm using cpu-z's stress test. I was able to repeat it with Asus's CPU burner test within the tweak GPU utility. Then I swapped motherboards and now the core temps are 80c using cpu-z's stress test.
|
|
#
?
Mar 13, 2018 21:32
|
|
|
jonathan posted:Sorry, that was a 5am post after a lovely work night. I'm using cpu-z's stress test. I was able to repeat it with Asus's CPU burner test within the tweak GPU utility. Then I swapped motherboards and now the core temps are 80c using cpu-z's stress test. That is really high, what voltage are you running the OC at? Otherwise to me this sounds a lot like bad thermal contact or maybe an issue with the heatsink itself. I would be very surprised if damage would cause thermal behavior like that.
|
|
#
?
Mar 14, 2018 01:04
|
|
|
Lockback posted:That is really high, what voltage are you running the OC at? Otherwise to me this sounds a lot like bad thermal contact or maybe an issue with the heatsink itself. I would be very surprised if damage would cause thermal behavior like that. Yeah I guess I'll order in some liquid metal and redo the die to IHS. What's the best for IHS to cooler ?
|
|
#
?
Mar 14, 2018 08:56
|
|
|
jonathan posted:Yeah I guess I'll order in some liquid metal and redo the die to IHS. What's the best for IHS to cooler ? In my experience its usually not very significant what brand you use, though I have seen batches that were bad. Arctic Silver 5, or Noctua NT-H1 are both fine. I have heard this stuff is good though I've never used it. https://smile.amazon.com/Thermal-Gr...a-no-redirect=1
|
|
#
?
Mar 14, 2018 13:51
|
|
|
My framerate is dipping below 60 a little bit in the new Rat Game so I finally decided to look into overclocking my i5 2500k. It turns out that I have this thing called OC genie on my old P67A-G45 so I enabled it, and it promptly made my computer unable to reboot. A quick googling with my phone told me that apparently OC genie doesn't like it ram running at 1600MHz or something like that. So a quick BIOS reset later, I'm here to ask you guys: Is there a goon-verified tutorial for overclocking the i5-2500k, hopefully on something alike my motherboard? I'm asking here specifically because there appears to be a plethora of guides for it on the internet, and some of them are telling me to reinstall windows first (seems a little excessive) or tell me to update my bios (I heard that's a bad idea under most circumstances) Edit: Actually I'll try reading the OP first Edit2: the OP links this guide so I'm assuming it's a good guide. I'll try doing what it says the next time I have 4 hours to spend tinkering with this stuff and rebooting until it somehow works. Minorkos fucked around with this message at 19:57 on Mar 16, 2018 |
|
#
?
Mar 16, 2018 19:53
|
|
|
Move your multiplier up and download some stability test tools. You can try something like Intel XTU for quick and dirty software OC if you want to just temporarily try certain clocks and look at voltages and test thermals. It’s rare that the CPU is the limiting factor in games, given that people have a tendency to turn the details up (“gotta make it look better than the console version! PCMR brah!”) until their graphics card wheezes from all the things being rendered. If settings are low and your FPS is still bad then you’re CPU limited. What GPU do you have?
|
|
#
?
Mar 16, 2018 20:00
|
|
|
Craptacular! posted:Move your multiplier up and download some stability test tools. You can try something like Intel XTU for quick and dirty software OC if you want to just temporarily try certain clocks and look at voltages and test thermals. I'm afraid I don't know the terminology yet so I don't really know what moving the multiplier up. Does that simply mean moving up a setting (CPU ratio?) without tinkering with other settings and seeing if my computer starts with it? I'm using a 970 GTX. The CPU and the motherboard are the oldest parts of my PC by a somewhat decent margin (both from like 2011 or something). I haven't needed to upgrade my CPU because yeah, it's not really required for most games. Rat Game 2 however puts an incredible amount of rats on the screen at once, which supposedly is the factor dipping my framerate. I don't really want to squeeze anything crazy out of my already old CPU, just looking to get that supposedly-easy performance boost that this CPU can get.
|
|
#
?
Mar 16, 2018 20:06
|
|
|
2 pages ago someone asked the same question. Please start here and let us know if you have any issues. The conversation gets useful on page 111. https://forums.somethingawful.com/showthread.php?threadid=3465021&pagenumber=110&perpage=40#post481952322 Also, I guess the OP needs to have the 2550k section removed as this is two people in a week it's screwed with.
|
|
#
?
Mar 16, 2018 20:20
|
|
|
Minorkos posted:I'm afraid I don't know the terminology yet so I don't really know what moving the multiplier up. Does that simply mean moving up a setting (CPU ratio?) without tinkering with other settings and seeing if my computer starts with it? Set the CPU multiplier to 42x and the VCore to 1.28V. That should be pretty stable and shouldn't generate too much heat. Do some testing with some CPU Burn-in applications to make sure you're stable. Play with DRAM OCing next, but if nothing is working your RAM just might not be very overclockable, mine wasn't. That should give you a nice boost pretty easily. I assume by "Rat Game" you are playing this, so every CPU cycle helps https://www.youtube.com/watch?v=YnO406cOVmM
|
|
#
?
Mar 16, 2018 21:32
|
|
|
I figure this is the best thread to post this in: if you have temperature problems with the NZXT S340 Elite case you can take off the front assembly, unscrew the metal front plate from the plastic carrier and attach it directly to the case using a few (preferably weak and round) 30mm magnets. The result looks completely stock but allows the intake fans to breathe way better. Previously I had a 8-10°C CPU/GPU temperature delta between front assembly on/off, with the restrictive plastic part removed there's very little difference if any at all (1-2°C). Credit goes to a guy in the overclock.net megathread. 
|
|
#
?
Mar 16, 2018 23:09
|
|
|
Aww man, that’s the case I want. 
|
|
#
?
Mar 17, 2018 00:52
|
|
|
Craptacular! posted:Aww man, that’s the case I want. I have one, it's fine without that mod
|
|
#
?
Mar 17, 2018 01:05
|
|
|
mewse posted:I have one, it's fine without that mod Oh yeah, I didn't want this to come off wrong. It is a fine case but the intake situation is not ideal for removing a lot of heat with positive pressure by default. If you just use the two stock 120mm exhaust fans you won't get any benefit from that modification. I use two Noctua 140mm 2000RPM intake fans and noticed that the resulting positive pressure inside the case vented through the harddrive department louvres back into the front intake, creating a thermal short circuit that raised case temps significantly because the intake fans recycled hot air from the case. My system has no HDDs so I closed that area with some black plastic and tape. This starved the intake fans for airflow (topped out at 1750 RPM). The little magnet mod fixed that and now it is just about perfect. My fans rarely spin over 1000 RPM under full CPU/GPU load and the difference between room and case temperature is 2-3°C.
|
|
#
?
Mar 17, 2018 08:53
|
|
|
Lockback posted:Set the CPU multiplier to 42x and the VCore to 1.28V. That should be pretty stable and shouldn't generate too much heat. Do some testing with some CPU Burn-in applications to make sure you're stable. I did what you said and I booted up A-ok. Is the matter of testing for stability as easy as running CPU-intensive software and seeing if I blue screen or not? I ran CPU Burn-in for a few minutes and it only took up 25% of my total CPU in Windows, and then stopped with no errors. I also ran the stress test on CPU-Z and it didn't do anything especially scary either, besides slowing down everything and occasionally forcing Core Temp to reload its window. My temperatures seemed to cap out at 66C after running at 100% for a while. Is there anything specific I should look out for from now on, besides CPU temps and blue screens? Edit: Actually after a bit more testing, my temps did max out at 73C. Still, a quick googling tells me it's fine as long as it's below 80-85C, so I assume it's okay like that? Minorkos fucked around with this message at 13:58 on Mar 17, 2018 |
|
#
?
Mar 17, 2018 13:42
|
|
|
Minorkos posted:Is there anything specific I should look out for from now on, besides CPU temps and blue screens? If you run https://www.hwinfo.com/ it has a stat for Windows Hardware Errors (WHEA) so if u ever get any of those you're technically not stable.
|
|
#
?
Mar 17, 2018 14:20
|
|
|
You’re right that those temperatures are fine. Stable is relative. You have the guys who insist on running P95 for 24 hours, but you don’t have to do that (you don’t have AVX2 to be worried about, but still, don’t) in order to play your game. How far you’re willing to chase a definition of stable is up to you, but “not as stable as it was before while still being stable enough to play your game” is easily attainable.
|
|
#
?
Mar 17, 2018 14:21
|
|
|
Minorkos posted:I did what you said and I booted up A-ok. Is the matter of testing for stability as easy as running CPU-intensive software and seeing if I blue screen or not? I ran CPU Burn-in for a few minutes and it only took up 25% of my total CPU in Windows, and then stopped with no errors. I also ran the stress test on CPU-Z and it didn't do anything especially scary either, besides slowing down everything and occasionally forcing Core Temp to reload its window. My temperatures seemed to cap out at 66C after running at 100% for a while. Congrats, you just got a cpu ~33% faster and its running reasonably cool. If you get any weird crashes you might want to consider stepping the voltage to 1.3, but I think your in a good spot. Trying to get much higher than where you are will require more fiddling and won't result in much performance improvement, so I think your probably best just to leave as is. I'm a Hard Over Clocking Professional Gamer Dood and I used the 2500k until about 5 months ago. This should buy you significantly more time if you were doing ok at 3.3GHZ.
|
|
#
?
Mar 17, 2018 18:04
|
|
|
Asus Realbench will find instabilities pretty quickly without putting completely unrealistic loads on your system. It uses a loop of open source programs such as 7zip, GIMP, x264 encoder, OpenGL rendering. As always use it at your own risk, etc.
|
|
#
?
Mar 17, 2018 20:38
|
|
|
GRINDCORE MEGGIDO posted:I'd test the overclocking with offset right from the get go. It'll save testing it again after you've found its happy place. Intel XTU passed it's initial benchmark with these temps/voltages:  ...And then on the stress test Windows crashed immediately. Oh well. Maybe I should aim for 4.2 instead. Craptacular! fucked around with this message at 03:43 on Mar 19, 2018 |
|
#
?
Mar 19, 2018 03:27
|
|
|
Craptacular! posted:Intel XTU passed it's initial benchmark with these temps/voltages: According to that screenshot, you were actually running 4.5ghz @ 1.291v? You need a little more voltage for 4.5ghz to be stable on the 2500K. You still have some thermal headroom too, up to 80C is safe for a 2500K during stress testing, since real world temps will be far below that. I'd try 1.34v. Alpha Mayo fucked around with this message at 07:45 on Mar 19, 2018 |
|
#
?
Mar 19, 2018 07:40
|
|
|
Alpha Mayo posted:According to that screenshot, you were actually running 4.5ghz @ 1.291v? You need a little more voltage for 4.5ghz to be stable on the 2500K. You still have some thermal headroom too. I'm using a 3770K. The poster I quoted suggested I just try setting a desired clockspeed and leave the voltage offset alone just to see what it does. That's what it did. I'm not exactly sure if I'm even really CPU bottlenecked in anything I'm doing, but I figured the games I've been playing lately have been of the "16GB of RAM is noticeably better than 8GB" variety, that I should see what I can get out of my CPU without doing anything dangerous. My other thing is, because I'm new to overclocking, I don't know how to tell what true voltage really is. There's the VID figure HWInfo displays under the CPU, and there's the vcore number (along with many others) listed underneath the motherboard.
|
|
#
?
Mar 19, 2018 07:44
|
|
|
Sorry thought you had a 2500K trying to do 4.3, Ivy Bridges weren't as good at overclocking as Sandy Bridge due to the die shrink resulting in a hotter, smaller chip. HWinfo is pretty accurate. A lot of motherboards like to 'help you out' on stability by overvolting beyond what you tell it, HWInfo tells me the correct voltage. Also your C-States are working (downclocking/downvolting when idle). A lot of people have to disable them to achieve stability, especially if the motherboard's VRM can't keep up. Personally I always go for a stable overclock where I can keep them on, the extra constant power usage/higher temps aren't worth the performance gain of turning them off. You could still go up on the voltage a bit, IMO best mode is to use offset mode so you can keep EIST/Cstates working. I would give it like +0.040v in Offset mode, check temps again after 10 minutes of IBT, and if below 80C on all cores, run the stability test again. If it still fails, then the chip just isn't going to do 4.5 and I would go down to 4.4. Stability decreases exponentially with frequency, so the 100mhz difference between 4.5 and 4.4 will give you most of the stability you'd gain from going from 4.5 to 4.3. Alpha Mayo fucked around with this message at 07:58 on Mar 19, 2018 |
|
#
?
Mar 19, 2018 07:56
|
|
|
Craptacular! posted:I'm using a 3770K. The poster I quoted suggested I just try setting a desired clockspeed and leave the voltage offset alone just to see what it does. That's what it did.
|
|
#
?
Mar 19, 2018 18:59
|
|
|
Llamadeus posted:VCore is the actual voltage being supplied, I'd move that to the top in hwinfo instead of looking at VID.  This wasn't even the stress test since it failed immediately ,but if we're crashing at 1.25v, that must mean I own a crappy chip, since some dude's old YouTube video was full of replies of "you need 1.3v to hit 4.5? Sorry you own garbage man, I'm stable at [1.14 - 1.28]." I should try again at 4.4 and 4.2 before making any voltage adjustments.
|
|
#
?
Mar 19, 2018 19:16
|
|
|
obligatory load line calibration video because many people think BIOS/UEFI Vcore = load Vcore https://www.youtube.com/watch?v=NMIh8dTdJwI Setting 1.4V with LLC1 or 1.25V with LLC6 can end up producing the exact same Vcore under full load (this rough example is for Asus, Asrock would be the other way around...). Many people claiming super low voltages on their overclocks only mention the setting in the BIOS but run some horrible LLC settings that give them reverse Vdroop (which is horrible). Those same people usually don't know how LLC works. eames fucked around with this message at 19:33 on Mar 19, 2018 |
|
#
?
Mar 19, 2018 19:23
|
|
|
I've watched that before but the only part that really was understandable to the layman with no electrical background was "set this to medium, avoid higher settings" which was nice because the guides written in 2013 recommended Ultra High. The problem with videos by people like Buildzoid is that he's not beyond sticking an electrometer on his motherboard to see what voltages are. GN has another video where Steve does that while going on about how Gigabyte boards under-report voltages. At the end of the day, part of the reason I'm seeing what can be done without adjusting any voltage-related settings at all is because I refuse to take this to the point where I'm prodding my motherboard with an electrometer for voltage readings. At that point, I'd rather just stick with stock everything and buy a 2700X on Black Friday.
|
|
#
?
Mar 19, 2018 19:32
|
|
|
I like GN's content but as a YT commenter pointed out, he may have accidentally measured Vcore instead of Vsoc. see https://imgur.com/2y72Ng4
|
|
#
?
Mar 19, 2018 19:45
|
|
|
Thanks for reading YouTube comments so I don't have to.  My original plan was "manual vcore 1.25, LLC Medium, set multiplier" but at the end of the day I spend so many hours just shitposting on SA and watching YouTube that not taking advantage of power saving features is unfortunate. I currently have stock voltage, LLC set to Auto, and was testing with that but am back to stock frequencies. Should I set LLC to Medium before doing anything else?
|
|
#
?
Mar 19, 2018 19:50
|
|
|
To be honest I don't know the answer to your question, these days there are many ways to reach a desired load Vcore but there's no data on which method is the best to avoid CPU degradation from either too high voltage or those invisible spikes. In your case (without adaptive) I'd use manual Vcore to find a LLC setting that displays a sensible but not too extreme amount of Vdroop and then use that LLC setting with auto or a small positive offset, making sure it never reaches unsafe levels. I don't know what these levels are for Ivy bridge. For my personal machine I simply settled with an OC that should be reasonably safe all around (adaptive voltage below the "accepted maximum" and LLC with some Vdroop). With my CFL that's 1.35V set in BIOS and 1.30V under full load. That means I'm leaving 100-200 MHz on the table which is fine with me because no application I use shows a difference between 5.0 and 5.2 Ghz. For comparison Silicon Lottery advertise their CFL CPUs as high as 1.425V with LLC6 (= slight reverse droop) but they only provide a one-year-single-replacement warranty. Gee I wonder why. eames fucked around with this message at 20:17 on Mar 19, 2018 |
|
#
?
Mar 19, 2018 20:13
|
|
|
Lockback posted:In my experience its usually not very significant what brand you use, though I have seen batches that were bad. Arctic Silver 5, or Noctua NT-H1 are both fine. I have heard this stuff is good though I've never used it. Update: So I shipped in some liquid ultra and some grizzly kryonaught. Ran a stress test. 98-99c was the core temp spread (at 4.84ghz) Opened the case, popped out the chip, delidded, brushed on the liquid metal, reglued the IHS, clamped it, Dabbed on some kryonaught. Ran cpu-z and... 48-55c after 5 mins of stress test. Oops that's with auto fan on the aio radiator. Let me crank it to 100% so that it's a fair test. 46 - 50c degree spread reported by core temp. Thanks for the advice gents. I was close to swapping for an 8700k. So for the rest of you. Normal arctic silver paste does not cut it under a delid. Even with the IHS closer to the actual die, liquid metal dropped the temps over 50 degrees from arctic silver. Gonna try for 5.2ghz now and see how it works. Without delid and with the 212 air cooler I couldn't break 4.8ghz without a lockup at 1.43v. I'm down at 1.3v at 4.84ghz right now and stable.
|
|
#
?
Mar 27, 2018 01:32
|
|
|
I suspect there was something wrong with the application or just a bad batch of Arctic Silver. Liquid metal will be better, but that spread is WAY more than you'd normally expect.
|
|
#
?
Mar 27, 2018 03:09
|
|
|
Would liquid metal benefit me much with a 2500K? I know my chip can do 4.8ghz but the temps climb into the upper 80s at that point, so I leave it at 4.5. Using a 212 cooler with AS5. Some games are a bit CPU bound with my 1070Ti.
|
|
#
?
Mar 27, 2018 20:24
|
|
|
Alpha Mayo posted:Would liquid metal benefit me much with a 2500K? I know my chip can do 4.8ghz but the temps climb into the upper 80s at that point, so I leave it at 4.5. Using a 212 cooler with AS5. Some games are a bit CPU bound with my 1070Ti. I think that sandy bridge was soldered, so if you delid it you would break it.
|
|
#
?
Mar 27, 2018 20:31
|
|
|
It would improve the temps by a couple of degrees, but as mentioned you need an oven and specialized tools because Sandybridge is soldered. You definitely can delid soldered IHSes safely, it's just a more specialized process. It's not going to be the change of 10-15C from going from conventional TIM either, the soldering is good. Edit: It'll be a waste of time and money to do it and you shouldn't but you'll be everyone's hero for a day for doing something hard. craig588 fucked around with this message at 21:32 on Mar 27, 2018 |
|
#
?
Mar 27, 2018 21:27
|
|
|
It seems likely that the real problem in that situation is probably the Hyper 212; I use one on my 2500K as well and it's great (maxes out around 70C) , but I'm only running at 4.4GHz and I assume pumping out a lot fewer watts as a result. It's probably not going to be able to keep up with a really serious overclock that puts out more than ~130W (30-35W for a 6mm heatpipe, x4) and you'll want either a a larger air cooler with more/larger heatpipes or a large AIO instead.
Eletriarnation fucked around with this message at 22:41 on Mar 27, 2018 |
|
#
?
Mar 27, 2018 22:30
|
|
|
Alpha Mayo posted:Would liquid metal benefit me much with a 2500K? I know my chip can do 4.8ghz but the temps climb into the upper 80s at that point, so I leave it at 4.5. Using a 212 cooler with AS5. Some games are a bit CPU bound with my 1070Ti. I'd get the Thermal Grizzly Cryonaut if you want the best improvement for the buck over AS5, also what is your case? If you can fit an AIO like this $65 dual-120 then you'll drop by like 30c: https://www.newegg.com/Product/Product.aspx?Item=N82E16835181098
|
|
#
?
Mar 28, 2018 00:33
|
|
|
What's the current state of AIOs vs Air these days? I have a 120mm AIO that's five years old but kept running at low RPMs through it's life. I figure I should replace it when I replace the CPU, but it is kind of noisy and I'm guessing Ivy Bridge has to have the worst temperatures of the 4-core Intels because the paste only got better with time. On the other hand, this ancient 2010-ish case only supports 240 on the outside via tube holes, and doesn't support 280 at all.
|
|
#
?
Mar 28, 2018 00:51
|
|
|
Air if you have a big case that can fit it, AIO if you have anything small or awkward. Custom loops can perform better but still need maintenance, I haven't done a 560MM rad because of that, wouldn't want to trade out my D15 after trying a Corsair 280mm AIO though, similar performance but more noise.
|
|
#
?
Mar 28, 2018 00:58
|
|
|
|
| # ? May 15, 2024 16:21 |
|
|
Zero VGS posted:I'd get the Thermal Grizzly Cryonaut if you want the best improvement for the buck over AS5, also what is your case? If you can fit an AIO like this $65 dual-120 then you'll drop by like 30c: https://www.newegg.com/Product/Product.aspx?Item=N82E16835181098 I could definitely fit it, though I've read things that make me worry about AIO durability. Never owned one but can they last for years? If I bought one I'd want to use it with my next CPU when I eventually build a new machine (probably in 2019)
|
|
#
?
Mar 28, 2018 03:32
|
|