Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
spunkshui
Oct 5, 2011



VelociBacon posted:

I don't think anyone in this thread will have ever had a GPU stop working from overclocks that you do in-software like MSI afterburner.

I killed a MSI 4870

There is a button in MSI Afterburner for MSI cards that has a warning and unlocks more poo poo.

DO
NOT
CLICK
THAT
BUTTON

For a brief moment it was glorious, it was stomping the crap out of heaven with like 30% more fps. Then... it crashed and its clocks dropped to like 500 and never left.

I tried everything I could think of to get it to work but I crippled something.

To be fair it had very aggressive warnings.

Adbot
ADBOT LOVES YOU

ufarn
May 30, 2009
Can someone explain why DRAM Calculator seems to barely require anything for jumping from 3200 MHz to 3600 on my G.Skill RAM? For some reason, it seems to require less than the 3333 overclock.

Only thing that goes up significantly is SOC voltage from 1.0125 to 0.05V.

Is there any reason not to try this (aside from it appearing to be too good to be true?




The current 3200 SAFE profile works just fine, even though I don't think I specified the manual voltages since my RAM is 1.35V by default.

ufarn fucked around with this message at 00:43 on Jan 28, 2021

Xaris
Jul 25, 2006

Lucky there's a family guy
Lucky there's a man who positively can do
All the things that make us
Laugh and cry

ufarn posted:

Can someone explain why DRAM Calculator seems to barely require anything for jumping from 3200 MHz to 3600 on my G.Skill RAM? For some reason, it seems to require less than the 3333 overclock.

Only thing that goes up significantly is SOC voltage from 1.0125 to 0.05V.

Is there any reason not to try this (aside from it appearing to be too good to be true?




The current 3200 SAFE profile works just fine, even though I don't think I specified the manual voltages since my RAM is 1.35V by default.

Reading this, G.Skill does seem to overclock pretty well. Maybe the trick is to click in multiples of 400 or something, but that's beyond my understanding.
Samsung b-die is good yeah. I don't know why 3333 would be different than 3600 but it is largely anecdotal-aggregation based afaik, of which DRAM Calc is actually great for because it's almost all absed on samsung b-die examples of which most people do 3600/3733/3800/etc. No reason not to try it, worst case is you have to do a CMOS reset which isn't a big deal, use a paper clip or pull the battery. Sometimes mobos have BIOS reset buttons or somethings they auto-reset after 2 boot-up failures. Not really much skin off your back just always make sure to backup your current "safe" mobo settings so you don't have to re-enter them by hand. Y ou aren't going to fry anything, the "danger-zone" for voltages is like DRAM Voltage >1.45 (more realistically like 1.5+) and VSOCs of like 1.15V+. Looks entirely reasonable and not too good to the true.

And in addition GSKILL Samsung B-die is a very popular ram that a lot of people have OC'ed. You can take a look at the Zen3 DRAM OC spreadsheet and see what others with your ram are doing with voltages/frequency/timings/etc. Of course half the fun is pushing each particular sillicon lottery up there so maybe don't use the BEST you see because it may not work - so maybe choose something lower-to-middle of the pack to start with. https://docs.google.com/spreadsheets/d/1dsu9K1Nt_7apHBdiy0MWVPcYjf6nOlr9CtkkfN78tSo/edit#gid=321590489 You can filter by RAM type or something close that matches yours. It's populated virtually by any Samsung B-die which doesnt matter much if it's GSKILLZ or Patriot or TeamViper RAM.

Btw there's no real reason to black out the memory, it's just a model number. I think you have this ram so someone else has like these results for example;



The bottom of the two examples is probably a little more relevant because it's only 3733 and also running at VSOC of ~1.05v whereas the upper one pushes it up to 3800 which is entirely possible but probably more MOBO depdent for success >3733. Not that you necessarily need to copy it, but should give you a feel for various timings/voltages/etc. You can achieve quite a lot with relatively modest voltage increases for b-die like the bottom example has.

Xaris fucked around with this message at 01:01 on Jan 28, 2021

ufarn
May 30, 2009
Well the 3600 OC was a bust. Didn't even boot at all. Pretty weird how instantly it failed, compared to when I once tried a 3333 profile that took a few days to show its weaknesses.

What's weird though was that BIOS showed the stick as 3700 in the right-hand info panel. What's up with that?

Also, what's a safe voltage window for memory? I don't wanna fiddle with it otherwise, I just used the 3600 here because the recommended value was on 1.35V anyway.

SuperTeeJay
Jun 14, 2015

I had my overclocked 8700K professionally delidded about two years ago and, following a recent case swap, the temperatures of a single core are averaging about 10c above the other cores under normal loads but still staying below 70c. If I run the Prime95 stress test, with or without AVX enabled, the second test in the sequence will cause that core's temperature to soar to 95-100c and all fans go to full speed. The other cores stay at about 75-80c.

I've cleaned/remounted the NH-D15, which I left in place during the case swap, and am still seeing the same results. Am I right in thinking that the thermal material between the cores and heat spreader is probably buggered?

ufarn
May 30, 2009
Two years is pretty quick for thermal paste to need repasting. Maybe the case move separated the thermal interface between CPU and heatsink, which is why it can't cool it more?

It's also possible the paste pattern or job didn't cover the core in question properly or something?

Thermal paste itself expires pretty quickly, especially Noctua's, so if you're using the same old tube, it's probably past its expiration date. I just looked at my box, and it says max three years of storage time.

SuperTeeJay
Jun 14, 2015

ufarn posted:

Maybe the case move separated the thermal interface between CPU and heatsink, which is why it can't cool it more?
I had to flex the motherboard a bit to get it out of the case so maybe the heat spreader shifted slightly with the cooler. (The delidder used liquid metal and reattached the heat spreader with a material that he described as elastic.) It's actually a bit of a nuisance that this is only affecting a single core and the cooling is still mainly effective as it means that sorting this out isn't worth the risk of bricking the CPU.

mewse
May 2, 2006

SuperTeeJay posted:

I had to flex the motherboard a bit to get it out of the case so maybe the heat spreader shifted slightly with the cooler. (The delidder used liquid metal and reattached the heat spreader with a material that he described as elastic.) It's actually a bit of a nuisance that this is only affecting a single core and the cooling is still mainly effective as it means that sorting this out isn't worth the risk of bricking the CPU.

There's a good chance the heat spreader is ludicrously easy to remove now. Most people just use RTV silicone after delidding which is way, way less adhesive than whatever the stock goop is.

If you get some fresh liquid metal and re-do the application, chances of bricking the cpu are pretty low because the mechanically stressful part (delidding) is done already.

SuperTeeJay
Jun 14, 2015

mewse posted:

There's a good chance the heat spreader is ludicrously easy to remove now. Most people just use RTV silicone after delidding which is way, way less adhesive than whatever the stock goop is.

If you get some fresh liquid metal and re-do the application, chances of bricking the cpu are pretty low because the mechanically stressful part (delidding) is done already.
The delidder has suggested that the heat spreader can be removed with a playing card, so I've ordered some Conductonaut and will give it a go. RIP in advance to my processor.

VelociBacon
Dec 8, 2009

I've always worried about liquid metal since we generally mount motherboards vertically and it just seems like it would 'run' down. I know tons of people use it I just.... ehhhh.

SuperTeeJay
Jun 14, 2015

Trip report: the heat spreader is still firmly attached by a glue/whatever on all sides so I decided against trying anything stupid to get it off. The idle temps for all cores are around 25-30c and the normal load temps are 50-60c. Not ideal that small FFT testing with AVX is raising 1 core to about 95c (20c above the lowest core temperature) but I haven't seen any dramatic temperatures in regular usage. Still, I'll use this to help justify what will most likely be a sidegrade to an 11700K and have it professionally delidded again before selling it on.

SuperTeeJay fucked around with this message at 01:02 on Feb 4, 2021

fletcher
Jun 27, 2003

ken park is my favorite movie

Cybernetic Crumb
I've got a Asus Maximums XII Formula, 10900k, and 16GB (2x8) G.SKILL TridentZ DDR4 4600 (F4-4600C18D-16GTZR) and trying to squeeze the most out of it. Seems stable in benchmarks and gaming but I haven't been able to run the RAM at full speed, 4200 is the fastest I've been able to get without crashes. CPU temps have been great.

How do I run this RAM at full speed? If I wanted to get another 16GB kit am I able to run 4x8GB @ 4600 or is the faster RAM speeds only when you have 2 sticks installed? I know I probably won't even notice the difference in memory speed but eh what not try and go for it.

BIOS settings as follows (anything not listed is default/auto):
  • Ai Overclock Tuner: XMP II
  • XMP: XMP DDR4-4608 18-22-22-42
  • Asus MultiCore Enhancement: Enabled - Remove All Limits
  • AVX Instruction Core Ratio Negative Offset: 0
  • ALL-Core Ratio Limit: 52
  • DRAM Frequency: DDR4-4200
  • Min. CPU Cache Ratio: 47
  • Max CPU Cache Ratio: 47
  • CPU Core/Cache Voltage: Manual
  • CPU Core Voltage Override: 1.340
  • DRAM Voltage: 1.45V
  • CPU VCCIO Voltage: 1.300
  • CPU System Agent Voltage: 1.300
  • CPU Load-line Calibration: Level 7

VelociBacon
Dec 8, 2009

fletcher posted:

I've got a Asus Maximums XII Formula, 10900k, and 16GB (2x8) G.SKILL TridentZ DDR4 4600 (F4-4600C18D-16GTZR) and trying to squeeze the most out of it. Seems stable in benchmarks and gaming but I haven't been able to run the RAM at full speed, 4200 is the fastest I've been able to get without crashes. CPU temps have been great.

How do I run this RAM at full speed? If I wanted to get another 16GB kit am I able to run 4x8GB @ 4600 or is the faster RAM speeds only when you have 2 sticks installed? I know I probably won't even notice the difference in memory speed but eh what not try and go for it.

BIOS settings as follows (anything not listed is default/auto):
  • Ai Overclock Tuner: XMP II
  • XMP: XMP DDR4-4608 18-22-22-42
  • Asus MultiCore Enhancement: Enabled - Remove All Limits
  • AVX Instruction Core Ratio Negative Offset: 0
  • ALL-Core Ratio Limit: 52
  • DRAM Frequency: DDR4-4200
  • Min. CPU Cache Ratio: 47
  • Max CPU Cache Ratio: 47
  • CPU Core/Cache Voltage: Manual
  • CPU Core Voltage Override: 1.340
  • DRAM Voltage: 1.45V
  • CPU VCCIO Voltage: 1.300
  • CPU System Agent Voltage: 1.300
  • CPU Load-line Calibration: Level 7

AFAIK adding 2 more sticks won't do anything that would allow your current ones to run at faster speeds than is stable currently and I'm not even sure that quad channel would be supported. Have you tried running it in XMP I (not II) at 4600?

Sorry this isn't about the RAM but from those settings I would think that your system is never downclocking/volting. Pretty sure with the ASUS mobos and manual voltage even with normal c-states it won't let it idle. Might be worth checking.

fletcher
Jun 27, 2003

ken park is my favorite movie

Cybernetic Crumb

VelociBacon posted:

AFAIK adding 2 more sticks won't do anything that would allow your current ones to run at faster speeds than is stable currently and I'm not even sure that quad channel would be supported. Have you tried running it in XMP I (not II) at 4600?

Sorry this isn't about the RAM but from those settings I would think that your system is never downclocking/volting. Pretty sure with the ASUS mobos and manual voltage even with normal c-states it won't let it idle. Might be worth checking.

I tried running XMP I but wasn't able to find any stable configuration, seemed like BSODs on boot no matter what I tried.

For the downclocking it seems like that is happening from what I see in hwinfo:

Internet Explorer
Jun 1, 2005





Hello everyone! Just a quick note to help out the folks who browse by bookmarks. We've started a SH/SC feedback thread and would love it if you stopped by to say hi and let us know what you think.

https://forums.somethingawful.com/showthread.php?threadid=3961558

bagual
Oct 29, 2010

inconspicuous
Regarding RAM overclocking, got a GeIL 32G kit to run with an asus x570 and 5600x, initially D.O.C.P set it to 3000mhz CL16-18-18 1.35V.

went into the bios to set the auto overclocking for the cpu after getting a better cooler and thought hmm why not also just set MCLK/FCLK to 3600/1800 and let everything else go auto on ram to see what happens.

Apparently it's undervolting to 1.2V and running at the set frequency, i've noticed no instability and computing workloads seem to have had a slight boost

decode-dimms spits this out in linux

code:
---=== Timing Parameters ===---
Minimum Cycle Time (tCKmin)                      0.938 ns
Maximum Cycle Time (tCKmax)                      1.500 ns
Minimum CAS Latency Time (tAA)                   14.061 ns
Minimum RAS to CAS Delay (tRCD)                  14.061 ns
Minimum Row Precharge Delay (tRP)                14.061 ns
Minimum Active to Precharge Delay (tRAS)         33.000 ns
Minimum Active to Auto-Refresh Delay (tRC)       47.061 ns
Minimum Recovery Delay (tRFC1)                   550.000 ns
Minimum Recovery Delay (tRFC2)                   350.000 ns
Minimum Recovery Delay (tRFC4)                   260.000 ns
Minimum Four Activate Window Delay (tFAW)        21.000 ns
Minimum Row Active to Row Active Delay (tRRD_S)  2.950 ns
Minimum Row Active to Row Active Delay (tRRD_L)  4.925 ns
Minimum CAS to CAS Delay (tCCD_L)                4.875 ns
Minimum Write Recovery Time (tWR)                15.000 ns
Minimum Write to Read Time (tWTR_S)              2.500 ns
Minimum Write to Read Time (tWTR_L)              7.500 ns
I'm really dumb and haven't overclocked for years, are these good enough timings? Everything's running smooth at the set frequency now, seems like diminishing returns trying to go manually mess with it.

hobbesmaster
Jan 28, 2008

Primary timings are excellent (cl14), secondaries are kinda wacky (tRFC and tFAW) but I wouldn’t touch it.

If that’s truly stable I bet that kit could go over 4000mhz easy. Kinda wild they sold it as 3000 cl16 which is kinda garbage.

bagual
Oct 29, 2010

inconspicuous

hobbesmaster posted:

Primary timings are excellent (cl14), secondaries are kinda wacky (tRFC and tFAW) but I wouldn’t touch it.

If that’s truly stable I bet that kit could go over 4000mhz easy. Kinda wild they sold it as 3000 cl16 which is kinda garbage.

Went to look again at the spec sheet for this kit (GeIL Orion AMD ed.) and no CL14 anywhere.

I guess the silicon lottery paid out, funny enough since i got it from a stock clearing sale.

Wasn't there something about ryzen infinity fabric being bad over 4000mhz? Next time im in the mood to gently caress around with it for fun i'll try going higher, but right now it's stable as a rock, thanks magic AI.

Helter Skelter
Feb 10, 2004

BEARD OF HAVOC

bagual posted:

Wasn't there something about ryzen infinity fabric being bad over 4000mhz? Next time im in the mood to gently caress around with it for fun i'll try going higher, but right now it's stable as a rock, thanks magic AI.
It's not really that it's bad, but a lot of Ryzen CPUs have a hard time pushing the FCLK that high and you can lose performance if it and your RAM clocks are mismatched (ex: 3600/1800 is matched, 4400/2000 is not).

If you're just messing around seeing how far you can push the numbers on some sticks you got on sale and aren't as concerned with actual performance, it doesn't really matter quite as much.

Spacegrass
May 1, 2013

I have a Dell Optiplex 790 and my bios doesn't have the overclocking ability. I've used just about every overclocking app i could find and none work on here. Is there a modded bios you could recommend for this system? My processor is a sandy bridge intel 2400. My motherboard is a generic stock from dell (0D28YY). Or should I not even not attempt it?

VelociBacon
Dec 8, 2009

What cooling solution do you have? Stuff like one off Dell 'solutions' are built to purpose and are designed to minimize cost of manufacture, I would expect not to naturally have the cooling headroom for any additional overclocks, and likely has a lovely power supply and weird non-standard connections or fittings that make it even more of a headache trying to get another cooler on there or to have stable power delivery.

Spacegrass
May 1, 2013

VelociBacon posted:

What cooling solution do you have? Stuff like one off Dell 'solutions' are built to purpose and are designed to minimize cost of manufacture, I would expect not to naturally have the cooling headroom for any additional overclocks, and likely has a lovely power supply and weird non-standard connections or fittings that make it even more of a headache trying to get another cooler on there or to have stable power delivery.

I have a stock fan. The power supply looks weird. Anyway I just want to overclock it to maybe 3.4/3.5. I have been overclocking my GPU and the temperature(s) looks fine.

mewse
May 2, 2006

Spacegrass posted:

I have a Dell Optiplex 790 and my bios doesn't have the overclocking ability. I've used just about every overclocking app i could find and none work on here. Is there a modded bios you could recommend for this system? My processor is a sandy bridge intel 2400. My motherboard is a generic stock from dell (0D28YY). Or should I not even not attempt it?

Trying to overclock with a non-K cpu and a OEM motherboard like that is a total crapshoot but you might be able to pull something off

From here:

quote:

What's this? The good news is that the rumors of Intel killing overclocking for non-K Sandy Bridge processors were slightly exaggerated! If you've got a CPU with Turbo Boost functionality it can be overclocked!! The bad news is that it's pretty limited. Our Core i5 2400 is rated at 3.1GHz with a typical Turbo speed of 3.2GHz with four active cores, with an absolute maximum of 3.4GHz under a single core load. By changing the Turbo multipliers to the maximum allowed either in the BIOS or the Intel Extreme Tuning Utility and raising the power consumption limits to 135W we were able to get the Core i5 2400 to run at a solid and steady 3.6GHz with a full 4 core load. 2 and 3 core loads are allowed to run at 3.7GHz and under single core loads Turbo was allowed up to 3.8GHz. These are noticeable increases in speed and only added about 15w of additional power use. It was on default voltage as well.

So you might be able to bump up turbo multipliers using intel extreme tuning utility to get some extra MHz. I severely doubt it's configurable in the Dell BIOS.

Spacegrass
May 1, 2013

mewse posted:

Trying to overclock with a non-K cpu and a OEM motherboard like that is a total crapshoot but you might be able to pull something off

From here:

So you might be able to bump up turbo multipliers using intel extreme tuning utility to get some extra MHz. I severely doubt it's configurable in the Dell BIOS.

I've used that program, it doesn't work on here. I've tried to download older versions, but they always update to the latest version.

I must have a locked chip. I give up. God I suck. I could have bought a Intel 3rd or even 4th generation chip/computer for the same price I spent on this Dell box. I just didn't know anything about Intel chips. What a tragedy.

Spacegrass fucked around with this message at 23:34 on Jun 24, 2021

Coxswain Balls
Jun 4, 2001

I started getting crashes to desktop and the odd bluescreen when playing PSO2:NG recently despite it not being a resource-heavy game. Someone told me that it's because it uses AVX which isn't common, and that would explain why I haven't run into issues with the handful of other games I play that are definitely more taxing.

Ran IBT for the first time since the initial overclock and sure enough it says there's stability issues and stops the test after the second step. Bumped my 3570K down from 4.5 to 4.4 GHz and now the test runs fine with the crashing problems disappearing. Hopefully this isn't a sign that it's on its last legs, but I definitely got my money's worth and then some from it. Thanks for all the extra math you've done for me over the years, little buddy.

VelociBacon
Dec 8, 2009

I'd leave the overclock up and look for an AVX modifier in bios nearby where you can tell it to downclock by xMhz when performing avx instruction sets. That way you retain the benefits of the otherwise stable OC.

Lockback
Sep 3, 2006

All days are nights to see till I see thee; and nights bright days when dreams do show me thee.

VelociBacon posted:

I'd leave the overclock up and look for an AVX modifier in bios nearby where you can tell it to downclock by xMhz when performing avx instruction sets. That way you retain the benefits of the otherwise stable OC.

I don't think those existed for the 3000 generation. That was a 7000/8000 thing I think.

VelociBacon
Dec 8, 2009

Lockback posted:

I don't think those existed for the 3000 generation. That was a 7000/8000 thing I think.

Alright, that's too bad. I use it on my own OC, I think it's a great tool.

Dogen
May 5, 2002

Bury my body down by the highwayside, so that my old evil spirit can get a Greyhound bus and ride
I upgraded from a 6700 to a 12900 and am a little baffled about how different the overclocking stuff is with the two kinds of cores and the way turbo boost works now and etc. Should I even bother? Anyone have any guides? I tried reading an asus specific one and I caught some of it, but a lot of it was Greek to me.

LampkinsMateSteve
Jan 1, 2005

I've really fucked it. Have I fucked it?
I think this thread being dormant for nearly a year maybe answers that. I know very little, but I feel I've heard the gains are not really there for this generation. And that chip is crazy power-hungry and hot as stock. What cooler do you have, out of interest?

Dogen
May 5, 2002

Bury my body down by the highwayside, so that my old evil spirit can get a Greyhound bus and ride
I finally replaced my long serving Archon with a NH-D15 because I didn’t want to gently caress with any potential socket incompatibility. It’s been with me since my 2500K so it’s a little sad.

Honestly I haven’t found anything to throw at the CPU yet to see much difference anyway from my 6700K, I just can’t resist the urge to tweak things…. But maybe it’s not worth it in this case.

VelociBacon
Dec 8, 2009

Dogen posted:

I finally replaced my long serving Archon with a NH-D15 because I didn’t want to gently caress with any potential socket incompatibility. It’s been with me since my 2500K so it’s a little sad.

Honestly I haven’t found anything to throw at the CPU yet to see much difference anyway from my 6700K, I just can’t resist the urge to tweak things…. But maybe it’s not worth it in this case.

If you're not needing more performance you can always tweak your voltages down for energy saving.

I keep my old coolers and use them on builds for other people, server builds, etc. My cooler master hyper 212 Evo has been on maybe 5-6 PCs.

movax
Aug 30, 2008

I'm pretty sure I've been using / applying my Thermal Grizzy Kryonaut incorrectly, but temps are "OK" so far, or so it would seem. My usual method is thin coating on top of the die or IHS -- the Kryonaut stuff is really really goopy / hard to spread though (IMO), compared to the AS5 I usually use.

I have a few builds right now that are in easy positions for me to replace the grease on.

* Threadripper 3960X + HEATKILLER IV Cu Waterblock
* Lenovo ThinkStation mini-PC (i9-11900T + Lenovo copper laptop-style cooler)
* GeForce 1080 + HEATKILLER Waterblock

I went w/ the grizzly for most of them, and temps seem sane -- had bad RAM in my TR build, so haven't gotten to PRIME yet, but it hovered around 39 C in Memtest in a 20 C room. Haven't been able to check GPU temps yet, and for the mini-PC, its a laptop cooler -- peaked around 60 C in Memtest but haven't been able to boot into a 'real' test environment yet (what's a good bootable / LiveCD option for temp stress test?).

Conductonaut looks interesting. I would very much prefer a 'set it and forget it' type of paste, as I'm past the point in my life where I'd tear down my machine once a quarter, scrub it clean, tweak and reassemble. I've also heard/read varying reports of type of paste vs. application (IHS vs. no-IHS).

Also, this is dormant... thoughts on merging w/ water-cooling, maybe? Or one cooling/overclocking thread? I know we get specific OC-chat in the Intel/AMD threads.

VelociBacon
Dec 8, 2009

Nothing against a merge here.

If your PC isnt overheating I would think you actually did fine. I use kryonaut also. Conductonaut scares the poo poo out of me, imagining it getting on the exposed circuits around the cpu mount etc.

kliras
Mar 27, 2021
maybe “overclocking and cooling” just so we can nerd out about thermal paste and fans in general?

Lockback
Sep 3, 2006

All days are nights to see till I see thee; and nights bright days when dreams do show me thee.
39C in memtest seems really good to me. So I think your doing better than fine, no reason to think something is wrong.

Stink Terios
Oct 17, 2012


I'm trying to undervolt a gainward Ghost 3060 Ti and failing miserably. The voltage curve refuses to flatten in MSI afterburner, so the undervolt doesn't actually do anything at all and the GPU still uses it's 200W of power in demanding tasks.



This is after I set the highlighted square and hit apply.

Warmachine
Jan 30, 2012



VelociBacon posted:

Nothing against a merge here.

If your PC isnt overheating I would think you actually did fine. I use kryonaut also. Conductonaut scares the poo poo out of me, imagining it getting on the exposed circuits around the cpu mount etc.

As a first-time user with my last build, I felt a lot better doing some prep beforehand. Since I was doing both the GPU and delidding a CPU, I practiced on my old 970, using the "fancy nail polish" TG sells to mask the SMDs around the die and get my application right. I also cut out a "surgical shroud" to drape over the rest of the card to ensure only the GPU socket/die was exposed. The last part may have been more security theater than anything, but it felt a sight bit more secure than moving an applicator across the entire exposed PCB.

Quaint Quail Quilt
Jun 19, 2006


Ask me about that time I told people mixing bleach and vinegar is okay

Stink Terios posted:

I'm trying to undervolt a gainward Ghost 3060 Ti and failing miserably. The voltage curve refuses to flatten in MSI afterburner, so the undervolt doesn't actually do anything at all and the GPU still uses it's 200W of power in demanding tasks.



This is after I set the highlighted square and hit apply.
I used this on my first undervolt.
https://www.reddit.com/r/nvidia/comments/tw8j6r/there_are_two_methods_people_follow_when/

It's still a bit confusing for a guide but I figured it out eventually.

Adbot
ADBOT LOVES YOU

Stink Terios
Oct 17, 2012


Quaint Quail Quilt posted:

I used this on my first undervolt.
https://www.reddit.com/r/nvidia/comments/tw8j6r/there_are_two_methods_people_follow_when/

It's still a bit confusing for a guide but I figured it out eventually.

Lowering the entire curve by using shift+dragging before setting the desired value is what I needed, thanks!

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply