overclocking

[sudden torrent]

I see no issue with using those D30-140 fans in that case

It's typically things like RAM clearance you have to worry about but no such problem exists on the Torrent

[steady lance]

Should I swap out the 3 140mm i have in the bottom with the 2x180?

till I buy new fans

[sudden torrent]

Definitely wouldn't hurt

[steady lance]

I gotta go over and redo my cables at some point

It mentally bugs me how I did it the last time

HOW ARE THESE FANS SOLD OUT EVERY WHERE

[sudden torrent]

Because they're popular

We have a bit of a fan shortage going now too

Those interlocking fan designs are all the rage right now

[steady lance]

mhmm I can imagine

daisy chaining 10 fans with a lack of cable length is a nightmare

[sudden torrent]

These are always a good bet. Be Quiet makes great stuff in general. https://pcpartpicker.com/product/wBNYcf/be-quiet-light-wings-lx-755-cfm-140-mm-fans-3-pack-bl130

[steady lance]

are phantek's measurements for airflow and pressure reasonably reliable?

[sudden torrent]

Not a single brand does it consistently

[steady lance]

damn

[sudden torrent]

You can't even compare the same brand to different models of the same brand

[steady lance]

Seriously???

[sudden torrent]

It's a completely useless spec because no one standardizes it except third party reviewers

[steady lance]

yikes

But I am really liking these fans on paper so far.

Hits every criteria I wanted

• Better than Arctic PWMs
• Main lighting
• Side lighting
• Interlocking vs daisy chaining f ing cables
• clean modern design
• Doesn't cost $36 a fan

I have a hate relationship with those cables lmao.

@sudden torrent are there any 420mm aios that perform as good if not better than my LT720 that I can fit into the torrent without removing the front panel?

[sudden torrent]

In short, no

[kind walrus]

Especially on that cpu, no

[sudden torrent]

Don't worry about the temp in cinebench, it's a totally unrealistic worst case scenario type of test

[steady lance]

I did this to myself but it kinda bugs me. Since I upped the TJMAX from 89 to 95 my overall temps have increased by about 8c under load.

This has allowed for higher boost clocks though while never going into the 90c range

[sudden torrent]

Which is still completely fine. Ryzen 7000 boosts even at tjmax.

[steady lance]

thought it boosts a lot more while running cooler?

Or did I miss something

[kind walrus]

Try and stay under 89 tho

Freak loads like UE5 might spike it over tjmax

[steady lance]

Highest ever recorded was 81.4c in Cinebench 24. 12 hours in prime95 small ffts right after 4 hours in VST3 landed me a max temp of about 66c

[sudden torrent]

AMD changed that algorithm so now it basically does max boost all the way up to tjmax, and if it starts to go above tjmax then it throttles back

[kind walrus]

Again freak loads

Not synthetics

[steady lance]

Interesting. I might try loading the TJMAX back to 89 to see if it still makes a difference.

[sudden torrent]

Marvel Rivals has been proving to be a good stability test. Lots of people coming in here complaining about crashes when it turns out to be the GPU or memory.

[kind walrus]

Even with my 73C temp cap under all test conditions a random game will push it to the limit for brief moments

[steady lance]

Not sure how but by doing a negative CO it immediately caused every core on CCD0 to go up by about 100-150mhz under load on average

[steady lance]

i have been playing a lot of Fortnite, Marvel Rivals, and Hell Let Loose

[sudden torrent]

If the scores stayed the same then that's clock stretching

[steady lance]

scores went up by .3 consistently

[kind walrus]

Same/minimal gain yea

[steady lance]

from an average of 1.48e10

to 1.51e10

@kind walrus @sudden torrent

[kind walrus]

Effective clock >>>> Immediate clock

[steady lance]

after having posted that screenshot I had let VST3 run till 130 iterations with consistent scores of 1.51

[steady lance]

Please explain.

Effective is the maintained clock under load?

[kind walrus]

Effective is average clock over the period

Immediate clock is current clock at the measure

[steady lance]

ahh okay.

[kind walrus]

i.e the sensor can poll 4.6GHz every interval

[steady lance]

Effective was 1.473 roughly with some being 1.476

[kind walrus]

But the actual speed for the entire interval

Is more like

4.4GHz with bumps to 4.6GHz

And the sensor is just detecting the 4.6GHz

(mostly)

[steady lance]

before it was at 1.458-1.46ish on average over time

[kind walrus]

Hwinfo has a sensor to check effective clock

Use that

If it's out over 50MHz you're stretching

[sudden torrent]

The average on the effective clock is also kinda useless, since it's an average of averages

[kind walrus]

Eh

I mean, yesno

[sudden torrent]

So it's really just peak and current effective you want to look at

[kind walrus]

It can still kinda help

[steady lance]

my scores in VST3 and CR24 both went up

if that's a good indicator

[kind walrus]

Though on R9 X3D I guess it makes sense that avg effective instead of per core is a bit of a bad indicator*

[kind walrus]

Again, up but out by 50 is still stretching

[steady lance]

CR24 used to say I had an effective clock of 5.1GHz accordiing to BenchMate. Since change it usually says 5.25ish

[sudden torrent]

That number isn't going to be super reliable

[steady lance]

Ah.

[steady lance]

what do you mean by up but out

[kind walrus]

You need to look at effective clock lol

[sudden torrent]

If effective clock vs actual clock is greater than a 50MHz difference

[kind walrus]

Yep

Reported being* 50 higher than effective is definitively stretching

[steady lance]

Understood

[kind walrus]

Out by 25 can be considered normal

[sudden torrent]

This is all assuming a workload with consistent 100% load on a single core ofc

Your VT3 test is not that

[steady lance]

This is starting to sound like no matter what I do it won't be consistent...

[sudden torrent]

P95 would work

Now you're getting the idea of what overclocking is really like lol

[steady lance]

Here is what I have come to understand.

It's always a rabbit hole

and the best test is the one where you are actually doing stuff

[sudden torrent]

There is no "best test" if you want to talk rabbit holes

[steady lance]

In which case I have noticed a minor increase in gaming performance and synthetics have noticed a increase as well

so I am gonna call this one a win lmao

[sudden torrent]

You can have a processor that passes literally every stress test, 100% load for a week straight, then it sits idle for 15 minutes and blue screens

[steady lance]

Preach.

Sounds like my previous 7950x3D experience

anything else i can do to improve ccd0?

[sudden torrent]

Fun fact, testing for that idle instability is why corecycler implements random pauses into it's testing

Sudden load spike instability testing too from LLC being bad

Now as for your CCD0, I think you're pretty much done with it

[steady lance]

should I just leave my pc on overnight for a final idle test?

I leave my pc on 24/7 unless there is an issue with it

or run CoreCycler over night

[sudden torrent]

Since you leave it on and idle anyway I'd go for corecycler

[steady lance]

K.

Going to test cores 0-7 over night

thank you guys for always helping and I will be back tomorrow for CCD1

good night.

[sudden torrent]

g'night

[steady lance]

went to bed and noticed the system shutdown, adjusted core 7 and core 6 (I had bumped core 6 up without testing prior)

notched core 7 from -10 to -8 and core 6 back down to 20 from 23

I also reduced core 4 from 23 to 21 because I recall it being a problem child when core 7 became unstable

so did that just incase for the future

[steady lance]

I undid the Co on core 7 to test for instability that isn't caused by that boost quirk from core 7.

Core 3 popped an error over night from CC so I lowered it from 28 to 25.

Applied the negative 10 Co back to core 7 so now to test for overall stability again

[sudden torrent]

The end of the log will tell you what it was doing when it crashed

[steady lance]

I forgot to start a log.

No logs were made it seems

[sudden torrent]

It should make one by default unless you disable it in the ini

[steady lance]

Where do I check?

[sudden torrent]

It should be in the "logs" folder, and it should have asked you at least once to enable a custom view in event viewer

[steady lance]

Where do I find this?

I found the logs folder

but the dates on them are all very old

@sudden torrent we have made progress! It's successfully passed a entire round of CC on all cores for CCD0 since reducing the CO on core 3.

downside is that CCD0 the all core boost has dropped from 4.75 average to 4.7 average.

[steady lance]

Hmm I am slightly annoyed.

It appears I have stabilized CCD0 but I had to lower core 3's CO down from 28 to 23 because when core 7 has a co core 3, 4, and 6 don't react well.

Core 1 appears to be my golden child despite having a co value of 38 compared to 40 on core 0.

Core 1 on average does about 100mhz higher than the rest of the cores under load on CCD0

I guess it's because it is #1.

[sudden torrent]

CO value represents each core relative to itself and is not comparable to other cores

It's an offset based on the built in voltage/frequency table

Which is determined by a silicon fitness value by the factory

Higher ranked cores have a lower voltage from the factory in essence

[steady lance]

Why does having a negative co on Core 7 cause the entire ccd0 to boost higher but as a result also causing those cores to be less stable?

[sudden torrent]

Because the cores are boosting higher, they're reaching a different part of the v/f table. If you add 100MHz that can take a non-trivial amount of extra power, and when you're on the edge like this it can cause instability.

[steady lance]

Understood. But why does having core 7 set to a negative co trigger this increase in clocks for all cores?

Is it because the entire CCD at that point has a negative co?

[sudden torrent]

That part I can only guess on. My assumption is that the core was holding back the others in some way, like increasing a temp value for a neighboring core or possibly a quirk of the voltage regulator.

[steady lance]

gotcha

In anycase, @sudden torrent cross your fingers that this stays stable I reduced the CO on Core 3 which errored in the last CC pass and it has now passed 2 rounds of CC.

For CCD1 should I do a negative or positive CO to increase clocks?

[sudden torrent]

Negative still

[steady lance]

Welp -10 on core 7 caused a idle crash.

Trying -5

[manic helm]

Wow, Wegg has moved up in the world since the last time I checked in

[steady lance]

@sudden torrent before I move onto CCD1 should I reduce the Co on CCD0 or leave them as is?

[sudden torrent]

If it's stable, then it's stable

[steady lance]

Understood

[steady lance]

Start with a -20 co?

[sudden torrent]

You should have an idea of what methodology works for you at this point

What I do is start with -20 on the CCD and let corecycler run overnight to tell me which cores crash

That gives me a starting point for all the cores on the CCD

[steady lance]

You run all the cords every time?

I have been testing each core overnight individually then running CC through all on the CCD once finished.

[sudden torrent]

No, I have it cycle through all the cores in question when it's doing a longer test like 12+ hours

But if I'm adjusting one core at a time I just test that one core

[steady lance]

Oh okay.

I recall Fortnite crashing 1 time earlier I am trying to figure out if it's due to CCD1 being unstable from CCD0 or otherwise.

Because FN won't let me isolate it down to just CCD0 when playing it makes it a bit of a pain to figure out the problem child

[sudden torrent]

That game will definitely be an issue since it just ignores everything that's screaming "USE THESE CORES" at it

Side note, that's why the single CCD chips still outperform the dual CCD, some games just flip the OS the bird when it tells them to use only a few cores

Or some background task (windows) decides to run on the parked CCD which creates a lag spike

[steady lance]

Mhmm yeah very true

[steady lance]

Status update. -35 on core 8 failed miserably. Instant BSOD every time I posted.

Now testing -30. Now BSOD so far

[steady lance]

Well -30 crashed

so -25 it is

tried -30 on core 9 it also crashed, now testing -25 on core 9

@sudden torrent, is there a way for me to lock the max frequency of a individual core?

[steady lance]

CCD1 status update:
Negative COs

• 25
• 25
• 30

[steady lance]

So I went on a bit of a tuning spree last night.

Did CO values for cores 8-14

Fingers crossed I return home with no idle crashes or errors

[restive cargo]

Yes afterwards

[steady lance]

Just got home.

All cores passed CC and boosted above stock albeit slightly

5,850 seems to be the limit so far

[steady lance]

@sudden torrent @kind walrus I need help, why is it my PC does not always post past the Red and Yellow light at the start when I power the PC.

I often need to repeatedly shutdown and spam the power button to get it to boot up. This has been an issue since I unpacked the computer.

I just spent the last 20m to get my PC to boot past the CPU/board light

Is my bios potentially corrupted?

@faint tangle any idea as to the cause?

I am still trying to get the computer to post. Just cleared the cmos I believe.

[steady lance]

I just did the process to clear cmos with the battery.

• unplug power cord
• remove cmos battery
• wait a few moments
• reseat cmos battery
• replug power cable
• boot pc

Still no post

[steady lance]

@manic seal, @neon notch any thoughts on the cause of this issue?

I am still trying to get it to post

RED and YELLOW led on the board light up but it's stuck on that phase

[neon notch]

What led is that referring to? Like vga boot or what

[steady lance]

The motherboard.

• RED = CPU
• YELLOW = DRAM

[neon notch]

It could be training, how long have you waited

[steady lance]

I have waited upwards of 10 minutes before.

Should I try and just be more patient?

[neon notch]

Yeah for now. After that I’m afraid I don’t know what would be wrong with it

[steady lance]

I am freaking out right now.

[neon notch]

You can always go one ram stick at a time, reduce components to bare minimum the typical

Make sure all your connections are fully seated

[steady lance]

Board, GPU, CPU are plugged in fully.

[neon notch]

CPU has onboard graphics if I’m not mistaken right?

[steady lance]

Yes.

[neon notch]

Bare minimum reduce factors for now

Just my opinion

[steady lance]

It's also a 7950x3D so my beard doesn't even allow me to set soc voltage beyond 1.3v so I doubt it exploded.

[neon notch]

Can always inspect pins after boot testing more

Currently freezing my ass off waiting for my car to heat up so you got me for like 10 minutes

[steady lance]

DRAM!

Yes

Jesus fkin christ

[neon notch]

Let it train for up to like 20 minutes lol

It’s ridiculous sometimes

[steady lance]

Seriously bot

Read the damn room

My gosh I have been stressing out like a maniac

For 3 hours now I have been trying to get this to post

[neon notch]

Yeah ddr5 has thrown my heart for a loop a few times

Typically one run of 30mins then a quick on and off again works out

[steady lance]

In this case the issue was it kept getting stuck at the red/yellow light

Passes training just fine

[neon notch]

Do it booted?

[steady lance]

Getting stuck on red/yellow boot light has been happening ever since I unpacked my PC after arriving in Alaska.

Could it be an issue with my case's power button or PSU?

CPU works perfectly and no damaged pins on the socket

[neon notch]

I’ve seen memory doing a retrain on every boot on a gigabyte board before

[steady lance]

GPU albeit old and sht at over locking is by far the most reliable thing I have ever owned

[neon notch]

Couldn’t figure out a fix for it

[steady lance]

Hmm so you think it's the ram?

[neon notch]

I think it might be the mobo

[steady lance]

May I ask why? I wish to know as much as I can regarding this situation

[neon notch]

I don’t see why it would be anything else as of this moment. Getting stuck in boot loops has happened to a few boards ive seen

Just a gut feeling pretty much

I’m no real techie

[steady lance]

... I hope it's not the board because that'd be an expensive fix.

I am on a MSI X670 Pro WiFi model

[neon notch]

For now leave it in sleep when you get off I guess

Leave it running to avoid booting it again

[steady lance]

Just saved my settings and enabled memory context restore

[steady lance]

Thankfully I leave my PC on 24/7

[neon notch]

Well glad I helped a little 👍

[steady lance]

Thank you.

Read the room robo

[neon notch]

*stuff is scary

[clever folio]

I could never as a windows user

[steady lance]

Yes especially when an 800 dollar chip that I already had to deal with warranty once is on the line

[steady lance]

?

[clever folio]

At least for me it has a tendancy to go a lil haywire if it doesnt get its restarts and what not

[steady lance]

Ahh I haven't had that issue.

I do a restart needed or not about once a week when I am not over clocking to clear the cache

[clever folio]

I just turn my pc off every night

[steady lance]

Mhmm I see.

Btw how's the 7600X so far?

Iirc you got yourself an upgrade quote recently

Now anxiously waiting for the PC to post again

I think I am gonna have to replace my AIO soon.

I can bubbles from it often while starting up

Still no post

[neon notch]

Left it running for a while?

[steady lance]

About 10m.

[neon notch]

Do 30 set a timer then turn it off if it doesn’t post

Turn it back on

Either magic happens or you’re potentially screwed

[steady lance]

I have an idea. When it posted I had loaded up my saved co values

I am going to try clearing cmos and disabling the last co I did.

@neon notch do I need to turn the PSU off every time I clear cmos using the pins?

Cmos should be cleared.

I am going to try posting and booting on stock settings

Tried clearing once more.

Going to give it time to train

@kind walrus go you have any idea why I am stuck on the CPU/DRAM lights. I have managed to only get into bios twice in the past 4 hours.

[clever folio]

Got it around september/october, havent had any troubles and its definitely a high end cpu

[steady lance]

Nice.

[clever folio]

And by upgrade that means going from uhd 620 graphics laptop to gre

8 gig ddr4 beast

[steady lance]

Yikes

Solid upgrade

.

[clever folio]

Just a smidgen

I ran that laptop for 6 years

It chugged in terraria above 720p

💀

[steady lance]

Currently anxious as all hell as I wait for this thing to train

Started at exactly 4:00AM. Now 4:09AM.

[clever folio]

Good luck man

[steady lance]

I swear you can't trust either Delivery Companies or AirPort loaders at all anymore.

I first noticed this issue about three weeks ago when I unpacked my PC after having travelled to Alaska.

Wasn't neatly this bad however.

[steady lance]

Still no luck

Got to bios

Going to try booting with just my ram tume

The next motherboard I buy I am buying one with a damn error display

[steady lance]

@neon notch I got the system to post and it's working again. Issue seems to be due to the board re-training itself in a loop so I enabled MCR for now.

I will be actively running tests to check for errors just in case.

On a positive side note, I applied all my oc settings and they posted and booted without issue. But also my all core effective clock on CCD1 has gone up from 4.65GHz to 4.95GHz

[steady lance]

1.51e10 in VST3!

[steady lance]

Decided to remove the Boost Clock Override and just focus on Curve Optimizer

[sudden torrent]

Which means you need to start over

Because changing the override changes the boost table which affects the target frequency which is what determines what voltage it runs at and that's what CO changes

[kind walrus]

Gentle reminder that adding the extra +5 or +10 buffer to your work fixes so many issues

[steady lance]

Forgot to provide a status update.

I found the issue was the same as CCD0 the final core when all are negative causes a massive boost to the all core and single core clocks under load

so I turned core 15 down a couple notches and I have been bug free all day.

Survived synthetics, idle, blender, gamig, and work

[steady lance]

@kind walrus I am taking your advise and adding some voltage to the co. I am doing +3 to all cores bringing me to:

Negative CO Values from Core 0-15:
CCD0 (vCache)

• 38
• 35
• 25
• 20
• 18
• 25
• 18
• 3
  CCD1 (Frequency)
• 15
• 8
• 28
• 25
• 25
• 30
• 33
• 30

[kind walrus]

Lol GL

[steady lance]

So far so good.

[steady lance]

@sudden torrent, when you get a chance could you send me the profile for CoreCycler and the download link? Thank you!

[wide vigil]

Hello- If I just want a mild OC for my i7-10700k on Maximus Hero XII. Will the Asus AI Suite be more than sufficient? I'm just trying to get better frames on Foundation or other RTS games cause I'm falling under 30 FPS on 1440p Medium Graphics Settings.

[dull ginkgo]

What GPU are you using? And what ram are you using?

Generally GPU and mem OC will get you more performance than CPU OC, especially memory.

Also, have you tried testing if you get more performance with disabling hyper threading?

[wide vigil]

I'm using Asus Tuf OC RTX 3090 and Corsair Vengeance DDR4 3600

[wide vigil]

Never tried it. Is that on BIOS?

I have my RAM set on XMP profile and I did some OC for my 3090.

[dull ginkgo]

Yes, it's in the bios, it sometimes provides better performance in RTS games if they're not very well optimized

[wide vigil]

Will try that Thank you. Foundation just fully released this week so it might just be the case.

[dull ginkgo]

For performance, ram OC is probably still the best idea, but it takes the most time and it'll be very hard to get it stable, guide here:
https://github.com/integralfx/MemTestHelper/blob/oc-guide/DDR4 OC Guide.md

[wide vigil]

Thanks. I'll disable Hyperthreading, Use AI Suite for CPU OC and look into manually OC on my RAM if the 2 doesn't work.

[dull ginkgo]

I don't know how Asus implements AI suite, I wouldn't really recommend it if it's still like how the old MSI game mode was, where it just bumps voltage up without doing much

[wide vigil]

I believe it changes 5 settings to achieve stable OC. They call it 5-way Optimization. I'll give it a shot and update this thread.

[faint tangle]

It's really meh

Afaik they haven't had any issues of killing parts but not actually affecting performance much

[dull ginkgo]

Hey I don't think MSI killed any parts, it just degraded with high voltage lol

[faint tangle]

Lmao

[kind walrus]

That software, the bios one right? Especially during the 9th and 10th gen it was known to degrade cpus

Unless this is a software thing in windows which you'll lose almost as much fps running the software dialling it up instead of using manual bios settings

Less programs open = more fps

[acoustic crane]

I think it's a windows program not bios

[kind walrus]

Then it's like an XTU moment

You'll clock the crap out of it then lose at least half of it to a 2% "XTUservice" process

Bios>>>windows

[wide vigil]

I ended up just using the BIOS AI OC since I plan to upgrade in the near future. I can happily say I'm playing Foundation at Normal Quality over 30 FPS 🥰

[kind walrus]

AI OC is the one that degrades those CPUs

https://youtu.be/cbGfc-JBxlY

https://youtu.be/qQ_AETO7Fn4

[dull ginkgo]

wtf I got a 50 minute ad on this

[sudden torrent]

Those are to catch people that fall asleep watching videos. They'll be intentionally boring ads too so they don't wake you up.

At least that one is skippable

[kind walrus]

Tf

[faint tangle]

I thought i was crazy for thinking this

[clever folio]

Nope, just genuine strategy

Although all it does is spread engagement with the ad still being boring

[kind walrus]

Idk why they bother trying to advertise to sleeping people

I wake up and forget everything

My buying history doesn't reflect anything from youtube for like the last 6 years

[steady lance]

I am running Tuned ram and PBO with Motherboard limits, does this mean my ram tune is unstable or my CPU is degrading? Looking for advise as I am concerned.

VST3 profile on Y-Cruncher

[manic seal]

i've only bought one thing and it was from a sponsorship, which the product i use daily now (safety razor from henson shaving)

outside of that though, i havent really bought anything that i saw on youtube

well, asides from games but i dont think those count, since most of the time they arent even sponsored games

[sudden torrent]

Because they still get the ad revenue whether you watch it or not. It's based on watch count and watch time, not eyes on screens.

[manic seal]

what companies would want to pay for an ad placement for a sleeping person though

[sudden torrent]

The economics work out where they pay less for the people that watch for 10 seconds or less or something like that

It's probably changed a bit since I last checked

But that's still the general idea

[manic seal]

i see

[steady lance]

Is 1.45 VDDIO cause for degrading with Zen5 X3D chips?

[sudden torrent]

They're too new for degradation to show

[steady lance]

Any ideas why VST3 threw an error?

Only thing that changed in my system as far as I am aware is the installation of Creative software like the Adobe suite, obs, outdated and stuff like MSI Afterburner.

[steady lance]

This booting problem is going to drive me insane.

Despite clearing cmos I am having extreme difficulty loading into even bios. Stuck on CPU/RAM light.

[kind walrus]

Have you tried going back to default for a few boots to see if it clears up

If the long training is a result of instability then yeah it's gonna be an ass to train every boot even with mcr on or off in the equation

[steady lance]

I believe I did. I can try that again just in case.

[steady lance]

Is running VDDIO, VDDQ, and VDD at 1.45V high enough for degration within a month???

I can't seem to find any info on that.

Tried multiple reddit and Google searches

[kind walrus]

I don't believe so

But I could be wrong

I often forget those numbers

[steady lance]

Okay so I got my pc to boot into windows after a few cmos clears and like 4 hours later.

I am going to run all stock and no PBO or XMP profile till friday to see if anything happens.

[kind walrus]

If you still have issues by then I'd suspect something's up

[kindred harness]

Will a 4080s from pny be able to achieve an oc to hit 3000 mhz on core clock? Seems like anything past 2850ish it's not liking, could this be due to no fan adjustment or this the silicon lottery reminding me I'm a loser?

Or is it also due to the fact that imma dummy and play on 4k lol

[steady lance]

Silicon Lottery and Thermals. That is all it comes down to. My 3070 last I checked can barely oc for its life.

You could look into adjusting voltage and power limits though I can't tell you what the max safe is.

[steady lance]

So far no issues. But my gosh my frames are horrid.... Currently 25-30% slower on average compared to tuned ram with PBO set to enabled.

[kindred harness]

Gotcha... yeah mine gets like stability at 80 on core and 1200 on memory. I see others get higher. So figured id ask. I'm still trying to figure out how to adjust fan curve too

[steady lance]

I know that pain. I am gonna try overclocking my 3070 again and pray that for whatever reason it overclocks better.

I should probably also replace the pads on it tbh. Thing is 4 years old now.

[kindred harness]

That would probably be nice for it too lol I use velocity x for my oc

[steady lance]

Can't say I heard of it.

I am old fashioned using MSI Afterburner.

[steady lance]

If you haven't yet I suggest doing ram oc.

In my experience so far it's the least problematic. Just time consuming.

[kindred harness]

Pny said it was recommended software. I use afterburner for my overlay lol

[kindred harness]

I just put on xmp, still need to learn how to adjust timings 😦

[steady lance]

The only good software MSI makes is their Screen Calibrator, MSI Afterburner, and their bios.

Recently made the mistake of installing MSI Center...

Been clearing out the bloat

[steady lance]

Yeah definitely look into ram OC then. Will yield better results across the board.

What CPU and ram?

[kindred harness]

I accidentally downloaded msi center on the first pre built i bought to try... instantly un-installed lol

[steady lance]

Lmao

Good hardware

[kindred harness]

7800x3d and ddr5 g skill 32 gb 6000mhz cl32

[steady lance]

Bad software

Oh nice.

[kindred harness]

Yeah i luckily decided not to wait for 50 series and wanted to get on pc gaming asap when I got my build. It's just in 4k where it struggles a bit

[steady lance]

7800/7950 X3D CCD undervolts quite well.

That's what I was doing till my PC started having horrible boot problems. Looks like it's fixed but not entirely sure.

[kindred harness]

I did try an undervolt with pbo offset negative at 30... was cool until my pc took a fat dive and started running slow dialed back pbo a bit, seemed fine but was still unstable at times so I axed that

[steady lance]

To much of an undervolt causes performance loss

[steady lance]

You did per core undervolt?

Cross your fingers

[kindred harness]

I believe so. I found a video, said to adjust voltage down a bit, then to offset the pbo cores negative 30.

[steady lance]

I am gonna try restarting my pc

[steady lance]

You'd know if you did per core.

Takes a lot longer

You likely did a CCD undervolt which applies the same undervolt to every core available

[kindred harness]

This sounds right

[steady lance]

I have been doing per core....

Which with 16 cores

Is a nightmare

[kindred harness]

I'm gonna try to find the video lol

https://youtu.be/IbV9tTgPNeo?si=ztgGUtYutvY-uVre

I really wanna say this is it

But i swear I remember it being a short

[steady lance]

Lmao 5m?

I spent 3 weeks tuning my CPU.

[kindred harness]

Yeah the undervolt/oc was new to me lol

[steady lance]

Check out CPU and RAM oc. But properly.

For reference I was getting about 160 fps in Halo Infinite with tuned settings vs default with jedec ram I am getting 10
10ish fps.

[acoustic crane]

Yeah dog not everyone wants to do this lol

[sudden torrent]

The extra CCD and core count adds to the time requirement exponentially

Plus there's no way they did curve optimizer in that vid

[steady lance]

True lmao.

Very true.

I am probably one of the few people who OC and enjoy it enough to wanna spend money doing it.

[steady lance]

@kind walrus, @sudden torrent why does this read a lot like what I am experiencing.

Also it had occured to me just now that when I arrived in Alaska one of the first things I did was update my bios.

[sudden torrent]

It's not, unless it's hanging when you're already in bios and you click the hardware monitor

[kind walrus]

Basically every msi board does that btw

[steady lance]

Would it be worth trying?

[sudden torrent]

Updating? Sure, it can't hurt at this point

[steady lance]

I don't mind redoing my ram oc, either way my cpu oc wasn't finished so I can always restart

[sudden torrent]

The "New OC rules" part might be good

[steady lance]

?

[steady lance]

@sudden torrent IT WORKED!

It booted on it's own without issue!

[steady lance]

Status update.

So I decided to put the new bios to the test.

The issue is not fixed. However it's been drastically improved.

Used to take me an hour of non stop cycling with context restore enabled to boot. I have managed to restart and boot twice in about 30m without context restore.

It has mostly resolved the issue with the bios getting stuck on yellow light after saving to exit bios.

Overall, the bios update has proven itself beneficial which leaves me to think that what I feared was a hardware issue is actually a issue with the bios.

[fervent spoke]

Hello I'm new to pc building and I was wondering if there is any videos or resources I can use to learn/get an idea of how to overclock a pc I'm building 😆

[steady lance]

The pins have some good sources. Although they are a bit old they still teach the main principals.

What parts are you using out of curiosity?

RAM overclocking is my recommendation for getting performance without going down the nightmare that is testing for stability with CPU overclocking when things go wrong.

[fervent spoke]

Ahh I'm using a 9800x3d, x870e aorus wifi7, gskill ddr5 64g 6400mhz, aio lian li gallahad II 360mm and my old psu corsair HX1000i(non-atx 3.0) as for gpu I'm not sure yet I got one ordered in back order of rx 7900 xtx or trying to hunt down any rtx5080

[steady lance]

Solid build!

[steady lance]

My suggestions are:

1. Tune your RAM (Start by figuring out what IC (Integrated Controller otherwise known as die) that it uses. You often hear names like Micron, Hynix, Samsung etc being thrown around. that's what people mean by that.

2. For the CPU start by enabling a manual PBO, do a core override to increase the max boost clock and then undervolt it as much as you can.

Make sure to take before and after measurements for each step of the way to ensure that you aren't making the component slower rather than faster due to it having to correct the errors it makes.

[fervent spoke]

Been due for an upgrade, current build already 7 years old

[steady lance]

Good news is that you upgraded.

Also from what I hear the 9000 series X3D is a lot easier to overclock.

thermals aren't much of an issue.

[steady lance]

Make sure to do your ram oc first.

[fervent spoke]

And then cpu?

[steady lance]

Using overclocked ram will make it harder to overclock your cpu so by doing it the otherway around you are more likely to run into stability errors in my experience since it's way harder to find the problems on a cpu.

[steady lance]

Mhmm yes.

[fervent spoke]

Ahh I see gotcha

[steady lance]

G.Skill has a numbering system to find out what ICs you have.

I forget how to check them but I suggest looking into that

[fervent spoke]

Also the gpu market feels like we are back a pandemic times lol

[steady lance]

Agreed.

to a certain extent we have the similar problem of not enough stock

[fervent spoke]

Would it get better once tsmc Arizona starts producing?

[steady lance]

Currently dealing with medical bills and such again though I plan to save enough for a 5090 in the coming months.

[steady lance]

I haven't heard about this but I'd assume so.

[fervent spoke]

I hope so

[steady lance]

My goals rn are to re-tune my cpu and ram so when I get a 5090 I can utilize it as best I can.

With CPU and RAM tuning prior to updating my bios I gained about 20-25% in average fps and less stuttering.

Granted I tuned the living hell out of this thing.

[fervent spoke]

Down the road I want to upgrade my gpu to the 9950x3d and to a 5090 but when there is stock, would my motherboard still be good then?

For power delivery i mean

[steady lance]

easily overkill lmao.

Though note, 9950x3D won't be any better for like 99% of games

and also dual CCD with the hybrid architecture still has some quirks.

You will want to install ProcessLasso for apps that aren't optimized for such systems to tell it "Hey, only use this CCD".

[fervent spoke]

Ok so let's say want to do productivity down the road like video editing, 3d modeling and such what should I be looking at?

Oh and also gaming lol

[steady lance]

9950x3D.

Just tell thhe finicky games

"Use my V-Cache CCD"

and your work apps, just let them do their thing

Honestly the quirks of dual ccd are much better now compared to launch

I used to have to tell every app what to use, but for the last month I haven't used ProcessLasso and only like 2/10 games did I want to tell it to use only the v-cache ccd

Long story short, @fervent spoke if you do more than just gaming than the 7950x3D/9950x3D is a easy decission.

Just understand that it can at time be slightly finicky.

Do note that the information I am relaying regarding the 9950x3D is pure guess work based on the 7950x3D.

[fervent spoke]

Thanks! I will keep it in mind, so in summary:
1.oc ram
2.cpu
3.gpu?

[steady lance]

Rumors suggest that it is essentially the exact same decent just with a minor clock bump and much better at thermal transfer.

[steady lance]

bingo.

I am happy to help as you go through the steps.

I plan to upgrade to a 9950x3D if I get the opportunity.

So I'm quite curious where this all goes.

Oh, @fervent spoke I recommend using PTM7950 on the cpu. it will help a lot with thermals which are essentially to my knowledge the only major change with the 9800x3D and the reason that it can overclock.

you can buy it for like $15-20 for a pad. Made about a 10c difference on my 7950x3D.

[fervent spoke]

What are those?

[steady lance]

Long story short. It's a thermal pad that at high tempertures basically turns into a semi-fluid like state that fills out all the air gaps between the cpu and cooler making perfect contact over time as it goes through repeated hot and cold cycles

better heat transfer than most pastes and it does a much better job at covering the targeted spots

https://www.amazon.com/dp/B0DB84CQW6?ref=ppx_yo2ov_dt_b_fed_asin_title

$13

that's what I bought.

IT IS NOT ELECTRICALLY CONDUCTIVE

Best performance without going liquid metal and you don't deal with the problems that comes with liquid metal

[fervent spoke]

Ah I see! Also would a thermal contact frame be ok or nah?

[steady lance]

note that I used it in with Thermalgrizzly's contact frame which I took advantage of so I could just apply an entire pad to my coldplate for better heat spread.

https://www.amazon.com/dp/B0CQKDJYNK?ref=ppx_yo2ov_dt_b_fed_asin_title

my exact combination

[steady lance]

note it's a theory but it has shown itself to work so far.

[fervent spoke]

By the way what does it mean when they say it's under 1% load?

[steady lance]

My idea was that that pad by using a wide enough spread it will capture the heat from the cpu and spread it across the entire cold plate.

[steady lance]

?

[fervent spoke]

I've seen that term when youtubers are running game benchmarks

[steady lance]

Oh

1% lows

[fervent spoke]

That

[steady lance]

It basically means your lowest noticable framterate

you know that sudden drop you get on occasion in games that's way below your average fps?

[fervent spoke]

Oh I see! So that's what it is! Like I said I'm a newbie, my last build my cousin build it 🤣

[steady lance]

1% or 0.1% lows are what I personally use to decide what settings to run because I am extremely intolerant of sudden frame changes.

I prefer a consistent 100 fps vs a inconsistent 144 fps.

though most folks are far more tolerant regarding frame drops than me. I am very very VERY picky in that regard.

[fervent spoke]

Me too specially since I mostly play fps games nowadays

As demonstrated by my pfp

[steady lance]

gotcha. I play a lot of ranked fps games in Halo, Fortnite, etc and more than once I have been killed due to a sudden spike from 144 down to 80 fps because my screen teleported. Had my friends spectating through discord to verify this.

I have been capping my frames and trying to ensure I get a 100% stable frame rate for the past like two months because of that.

Fortnite this season has had horrible stuttering

.

Long story short, I recommend getting a the PTM thermal pad I sent if you are doing overclocking, and a contact frame wouldn't hurt. much better than the standard latch the board comes with so it'd be a bit of an upgrade in that regard but not necessary

[fervent spoke]

Literally the reason I'm upgrading, and low vram I got on my poor rtx 2080

[steady lance]

PTM is $13
Frame is $24

double check that the frame supports 9000 series.

I bought it for 7000 series

[fervent spoke]

Will do, thanks for the help!

[steady lance]

quick peak suggests that it does but as always, trust but verify.

[steady lance]

start by figuring out what IC your ram kit has. That determines your max safe voltage for ram tuning. G.Skill has a numbering system for figuring that out.

You will find the numbers on the label located on the ram sticks

[fervent spoke]

Ok so it has samsung ddr5 ic

Apparently

[steady lance]

Does it specify what samsung ic?

for example Hynix has A and M.

[fervent spoke]

Trident Z5 RGB 64GB (2 x 32GB) DDR5-6400 PC5-51200 CL32 Dual Channel Desktop Memory Kit F5-6400J3239G32GX2-TZ5RK - Black

I have no idea what that all means except the first part

[steady lance]

what does it say on the lable located on the ram?

picture is best if possible

because that will have the IC numbers

[fervent spoke]

Umm I would have to get back to you on that since I am at work. Is it ok to mention you later today?

[steady lance]

yeah feel free to ping me any time.

I will get back to you whenever I have a moment

Don't take this as set in stone

but generally speaking to my knowledge all DDR5 ics can tolerate up to 1.4v DRAM voltage without risking killing the stick or causing fast deterioration.

@fervent spoke you will be looking for a short bar number something like this.

0T48KRS821M

that's the part number, it will change depending on what IC you have

[fervent spoke]

Gotcha I will look that up

[acoustic crane]

Not sure if it's legit but Ali has a thermalright contact frame for $5 or so

https://s.click.aliexpress.com/e/_mqvaEKX

(sale price $5.26)

[steady lance]

not worth the risk in my opinion

[fervent spoke]

@steady lance been burned in ali before, indeed

[manic seal]

Contact frame isn't needed or beneficial for am5

[acoustic crane]

I was just posting cause alpha was talking about it

[acoustic crane]

^

[manic seal]

Ah

Problem with contact plate is that it can make things worse or not work if improperly torqued down

[steady lance]

true, just gotta be careful with it.

[steady lance]

true, as stated above.

nice to have but not necessary.

[proven canopy]

https://x.com/gskillgaming/status/1889298346847682703?t=0XTUQqIyvIoCs-cdIVS4Lg&s=19

G.SKILL (@gskillgaming) on X

Feeling that overclock itch?
The G.SKILL OC World Cup 2025 is HERE, with a massive $40,000 USD prize pool up for grabs! Think you have what it takes to push hardware to the absolute limit?

Online Qualifier: February 15 – March 19, 2025 (on HWBOT)
Live Battles: Computex 2025 in

X · Feb 11, 2025, 1:00 PM

[sudden torrent]

Are you going this year? Maybe not to compete but cheer Seby on

[fervent spoke]

How a out the phasesheet?

[sudden torrent]

Do you mean PTM7950?

Because that is good stuff

Ultimately it would only net you about 3 degrees though

[fervent spoke]

Any degrees is good degrees?

@steady lance go it

is that it?

[sudden torrent]

Not really when it gets you exactly zero extra performance

[sudden torrent]

That looks like Hynix A die to me, fantastic for overclocking

Assuming g.skill still does the same coding as DDR4

[kind walrus]

H16A

[steady lance]

Should be fine up to 1.45V then if it's Hynix A as the others suggest.

[steady lance]

Less heat is always better, while not a big difference it is good for a little extra headroom.

[steady lance]

could go higher but not recommended as you will experience diminishing returns and risk deteriorating your ram faster.

I have Hynix 2x32GB sticks most likely A based on what we've learned and discussed here in the past.

So I can help provide reference numbers.

[kind walrus]

You can technically go way higher

Though I don't advise this for daily, due to the limited information on H16A degradation

1.6v seems doable for extended periods, more than that I'm not aware of

(There might be more info as of today but I haven't checked for a number of months now)

[steady lance]

What do you mean by extended?

[kind walrus]

Like 2 months

[steady lance]

Because I need to retune both my ram and cpu due to the bios update

[kind walrus]

I've similarly also seen degraded cases after 2 months

So at least to my knowledge it's not very assured what the limit is

[steady lance]

is 1.5v safe?

for daily

[kind walrus]

Probably

1.45 definitely is

[steady lance]

Gotcha.

Have you come across any news regarding max safe VDDIO?

I can't for the life of me find any information regarding that

[kind walrus]

1.3? I think is generally regarded ok

Definitely haven't heard of anything over 1.4

It's been a while man I do not remember these things well

[steady lance]

my board by default sets VDDIO = to DRAM

It makes the VDDIO turn red past 1.43

[kind walrus]

Couldn't tell ya

[steady lance]

I've been running 1.45

[kind walrus]

Also gotta go

[steady lance]

have a good one.

[steady lance]

OMG, @sudden torrent, @kind walrus this new bios update is awesome!!!

6,600Mhz synced actually posted!!!

[sudden torrent]

That'll be nuts if that's actually stable

[kind walrus]

Post vs stable is a big difference

[steady lance]

So far no issues

Running VST3 right now as the first test

[kind walrus]

My gene posts 6600

[steady lance]

Shush shush don't jinx me

xD

This is still a major milestone though.

Before to get post when I first got this board I had to do 1.6V on my previous 7950x3D

Ram training pasted near instantly

Currently doing 1.5V VDD/VDDQ/VDDIO

1.3v SoC same as before

VST performance went from 1.21E10 to 1.27E10

I have a serious question to ask you two. How risky is it for me to leave my VDD/VDDQ/VDDIO voltages at 1.5V if that is what's needed to make this work?

@kind walrus, @sudden torrent

.

Also on a side note

I reseated the ram and changed the sides that the sticks were in.

Booting is still a bit of a pain but slightly better

[kind walrus]

https://tenor.com/view/8ball-bart-simpson-shaking-shake-magic-ball-gif-17725278

[kind walrus]

Everything except dram voltage (vdd typically) 1.35v

Is safe

Higher idk rn and it's not something I'm actively reasearching

[steady lance]

Understood.

VDDIO does it affect the cpu or the ram?

[kind walrus]

I don't-

I don't know the answer to that 💀

All I can say is it's the voltage to the IO

To what extent I do not know

[steady lance]

Ahh alright.

I wish there was like a engineer's manual I could read for these chips

[sudden torrent]

I mean

There kinda is

How are you with schematics

[steady lance]

Not sure. Though I am pretty good at figuring stuff out.

Is it like a 2D graph outlining where stuff connects to?

[sudden torrent]

I found some documentation from an agesa update that says mem-vddio is the IMC voltage rail, so it'll be CPU side

[steady lance]

Mind sending me the documentation?

I'd love to give that a read over when I get a chance.

[steady lance]

.
Do you think I'd be risking my CPU by running vddio at 1.5V? My board by default sets it equal to VDD unless otherwise specified.

[sudden torrent]

Uh hang on, let me see if I can find the original source

[steady lance]

Awesome! Thank you.

[sudden torrent]

Right so it looks like it's a signal voltage line going between the memory and the APU, based on overclock.net people

If you're not using the APU you could disable it and never worry about long term affects

[steady lance]

That's the integrated graphics I assume?

[sudden torrent]

Yes the integrated GPU

[steady lance]

So just disable that and I should be fine?

[sudden torrent]

If you want to play it safe yeah

[steady lance]

Understood

Are there even any benefits to keeping it enabled?

I use Blender and the Adobe Suite mostly.

So things like Photoshop, Premiere, Dreamweaver, and Illustrator

[sudden torrent]

Nothing in particular

I doubt it's being used to accelerate any of that since you have a full discrete GPU to do the work

[steady lance]

Gotcha yeah I will disable it then.

I wanna keep this running 6600 with all my might if I can.

Been trying to get 6600 stable on this thing since the first 7950x3D almost two years ago

What tests should I use to make sure that it is completely stable at 6600? I have OCCT, TM5 (Anta777 and PC destroyer), VST3.

pasted TM5 with AntaExtreme777!

[kind walrus]

Wtf

Thats not the normal test

The normal one has 0-15 tests and 3 cycles

18m is way too short

[sudden torrent]

That's not synced

UCLK is 1650

[kind walrus]

Oh yeah good spot

Yeah that's desync lol

Correct anta777

[steady lance]

Ahh f

[kind walrus]

Incorrect/some other anta not normally used

[steady lance]

Are you kidding me. Rip

[kind walrus]

Should look like this

[steady lance]

Man I got overly excited

😢

[steady lance]

Weird. The profile file is labeled antaextreme777

[kind walrus]

Not the one we use

[steady lance]

Where do I find the one used here?

[kind walrus]

In the download files

[steady lance]

Also time for me to go into bios and make sure it doesn't do that damn deaync again

[kind walrus]

Load different config

[steady lance]

I feel like an idiot right now...

[kind walrus]

Everyone does this at least once

[sudden torrent]

This happened last time too iirc lol

Easy thing to miss

Even felix didn't spot it right away

[steady lance]

UCLK=MCLK is synced if I recall

[kind walrus]

I only saw the test and flagged thatblmao

[kind walrus]

It resets every update

So does numbers and settings

[sudden torrent]

Yes I have no doubt you would have spotted it after correcting the test

[steady lance]

Well lesson learned

Every thing gets reset when updating.

[sudden torrent]

Yep it's like a brand new motherboard again

[steady lance]

I am still gonna disable the APU though, big thank you to Fal for the tip.

I really want to know what's causing this miserably slow training.

[kind walrus]

I feel bad for pointing it out now

Haha

[steady lance]

I have managed to reduce it with the bios update and possibly with reseating but it's still not entirely gone

How long realistically should I be waiting for cpu training?

I am wondering if I may just be to impatient.

[sudden torrent]

Mine isn't completely tuned and it takes a little under a minute with memory context restore off

[steady lance]

Yeah just remembered it happened on the faulty 7950x3D.

[steady lance]

From power on to windows?

I have been sitting on this darn red cpu and yellow dram light combo for like two minutes now

[sudden torrent]

No from power on to my mouse turning on, which is my indicator that POST finished

[steady lance]

Gotcha

You have a 7900x3D iirc?

[sudden torrent]

yes

[steady lance]

2 minutes now. No mouse or keyboard light.

No*

[sudden torrent]

It can take 5-10 minutes with more difficult settings

[steady lance]

Just to get to the dram training phase?

[sudden torrent]

yes

[steady lance]

Is it just me or does this seem rather obnoxious time wise?

Weirdest part is that it takes forever even on stock settings with jedex 4800 ram speeds.

Just forced it to restart

It immediately trained

[sudden torrent]

Got stuck again then

[steady lance]

Any idea what is causing it to get stuck?

[sudden torrent]

not a clue

[steady lance]

This look right to disable the APU?

I swear I hate autocorrect on phones.

[sudden torrent]

yep

[steady lance]

Interesting my board allows me to adjust the base clock.

That just raises the minimum clock speed I assume?

[sudden torrent]

Everything operates as a multiplier off the base clock

Increasing the bclk increases everything else, including PCIe slot speed

[steady lance]

Interesting. How far can one push it and what are the risks vs rewards?

[sudden torrent]

If you're running the CPU at a x50 mult then increasing bclk by 1 will increase the frequency of the CPU by 50MHz

SATA drives absolutely hate changes to bclk, they can just die randomly

Overclocking your ssd basically

How far you can push it depends on your luck

102 should be doable if you don't have sata drives

[steady lance]

I have one sata drive but I can remove it.

Holy heck

[sudden torrent]

101 max with sata in

[steady lance]

6600 synced past training

Please work!!!

IT'S SYNCED!

or do my eyes deceive me again??

@sudden torrent

[sudden torrent]

yep that's synced

[steady lance]

Hell yeah!

I just installed the new tm5 configs

Which one should I use?

[sudden torrent]

extreme1@anta777

[steady lance]

Welp

Blue screen.

....

[sudden torrent]

Figured it was too good to be true

[steady lance]

Immediately into bios though

What could I change to stabilize it?

Looser timings or higher voltages

Because that was surprisingly far more stable than previous attempts in the past

.
Trying to see if 1.6V is stable or not as a quick test to see if it's voltage limited

[sudden torrent]

No that's not it

You can run 8000c34 at 1.45v

[steady lance]

Holy, 3 times in a row it's posted without me torturing myself by force restarting the pc

Welp 1.6v didn't stabilize it.

[steady lance]

Really?

Thoughts on what I could do to stabilize 6600mhz?

[sudden torrent]

You could play around with lowering voltages and tweaking resistances but based on the instant blue screen I don't have much hope

[steady lance]

Resistances?

Is that the same as LLC?

The true master of Tuning is here.

[proven canopy]

Definitely going, probably not for the comp, maybe for asus demo day or gskill demo with the "we have roboclocker at home" but 7995wx on auto ln2

[sudden torrent]

niiiiice

[proven canopy]

Not nearly good enough on RPL/ARL ddr5 to be competitive top 9 live

I can do ARL valid, but that's about it, barely

[sudden torrent]

No, these

[proven canopy]

Also - if it's anything like last year, there's a random drawing prize, and last place in the comp is still a decent chunk of change, so look into it

[steady lance]

Fall I think I finally found the fix to the darn training issues!

There was 1 specific setting I changed that has made my last 5 boots painless.

[proven canopy]

Oh, also can you pin that original link to comp? I accidentally some of my perms last time I was messing with the discord

Or just give "funny role" manage message perms

[steady lance]

MCR on my board has Disabled, Enabled, and AUTO. Both disabled and enabled caused freezing and stuck in training

While Auto has enabled me to boot pain free while still doing CPU and dram training

[proven canopy]

That doesn't make sense.

Likely it's defaulting to the setting that's inconsistent

[steady lance]

It's been a mess trying to figure out what's been going on with my bosrd

[proven canopy]

Go with MCR enabled, power down auto

[steady lance]

Will do. What does switching Power Down Auto do? I have been facing issues with getting stuck in training for the past month with the only fix being to forcibly restart the PC over and over

[proven canopy]

AM5 daily is as much fun* as you want it to be

64c30

[steady lance]

Note I have tried 3 bios versions and running stock (no pbo) and jedec 4800

Same issues.

At what speed does desync beat synced?

I am running 2x32 Hynix I believe A die.

[sudden torrent]

Depends on the workload

[proven canopy]

More a question of core count + workload. Latency vs bandwidth need.

[steady lance]

Gaming, Blender, and pretty much the entire Adobe Suite.

[proven canopy]

16c am5, easy answer, find highest stable fclk, and that's it, 1:1

[steady lance]

Power Down Enable was already set to auto it seems

[proven canopy]

Unless you want to go for sweaty avx512 heavy benchmark scores, or actually have a use case in daily

[steady lance]

What do you think my chances are of getting 6600 synced to be stable?

It can post and do light workloads but crashes during anta777

[proven canopy]

It's a daily, if 2200 works, great, if not, drop it down until it does.

[steady lance]

Do FCLK before ram frequency?

I am confused by the 2,200 number.

[steady lance]

oh my gosh this feels so good being able to actually restart my pc when I want to

[steady lance]

I am tempted to try adjusting the resistances and voltages to get 6,600mhz synced stable. @sudden torrent do you have any resources you suggest I look at to better understand what these resistances are and how to adjust them?

[sudden torrent]

Every motherboard and every CPU reacts differently to changing those. You'll just have to change each one, one at a time, and see what gets you more stability. It will probably take hours.

[steady lance]

Is there a risk of killing the ram with changes those?

Or is it just increase/decrease till I no longer error

[steady lance]

Did some reading up regarding that, I just can't seem to find out if there is anything that I like shouldn't do / avoid doing.

I see a lot of people tossing numbers around regarding resistances but none with explanations regarding what the resistances do.

[sudden torrent]

Precisely, because most people have no idea what they do. Most people don't even know what the role of resistance is in a circuit. It's also very poorly documented exactly which parts are affected by which resistance values.
There's no risk to the CPU or RAM messing with them, but keep a profile saved in case it fails to post.

[wide vigil]

Is it worth OC the 9800x3d for gaming?

I have a stable OC at 5.3hz, 1.12V and average of 67C

[sudden torrent]

If you can get it to an all core frequency that's higher than the boost clock you would get using PBO (5.4GHz) then yes it can be worth it. I've seen people get 6GHz using just an AIO.

[wide vigil]

What's the difference between Expo I vs Expo 2?

[kind walrus]

About 1

[wide vigil]

https://tenor.com/view/confused-computing-counting-math-problems-gif-14678592

[manic seal]

real

it's just 2 different timings (or both the same timings) that the vendor added to the ram

[steady lance]

Understood. My biggest concern was regarding if something would harm or even kill the ram or cpu. Do you have any sources you could refer me to?

[clever folio]

https://tenor.com/view/armstrong-made-it-up-gif-25653819

[steady lance]

all core override can be higher than pbo?

[sudden torrent]

The 9800X3D has proper support for overclocking. It's not unusual for people to get 5.4GHz all core on air cooling and 6 GHz on AIO. You just need a bit more voltage to the cores.

[proven canopy]

It's hard to kill ddr5 even if you try, never heard of someone succeeding. Same for b-die ddr5.

6g all core for any load? Haha

[sudden torrent]

Just gaming stable, doesn't need to be perfect

[steady lance]

Dang.

Still major improvement over my 5.25 on CCD0 with PBO.

[steady lance]

Gotcha.

[steady lance]

.

Is there a way for me to force my motherboard bios to default to the AUTO setting for Memory Context Restore when I clear the cmos?

This is freaking obnoxious, I am stuck doing power cycles for about 2 hours whenever I need to do a clear cmos because of no post.

[steady lance]

It took me two and a half hours to get into windows just now.

[sudden torrent]

The default is the default. Only a BIOS version change can change the defaults.

[steady lance]

r.i.p

The problem is that unless MCR is set to AUTO instead of ENABLED or DISABLED it keeps getting stuck in the training process.

It defaults to DISABLED which is even worse as far as getting stuck goes.

for some reason when on AUTO it still trains like it's DISABLED but it won't get stuck

[sudden torrent]

My issue with MCR is it gets unstable when I reboot if it's enabled. It borked my OS after I updated BIOS.

[steady lance]

Do you think this is a board or software issue?

[sudden torrent]

More likely firmware

[steady lance]

I am genuinely becoming infuriated by this issue. If it wasn't for finding this bandaid I'd be yelling and screaming at MSI right now.

Last week I spent around 11-13 hours of my week BOOTING.

[abstract merlin]

I check what ram I have and it shows that I have hynix j die

i am at 3200 cl16. could i get it to 3600 with the same timings or nah

or would tighter timings be better for my case

im on a 5600

[sudden torrent]

JJR is basically CJR. Don't expect too much from it. 3600 c17 or 4000 c18 if you're lucky.

[steady lance]

So an update on this whole booting ordeal.

If Context Memory Restore is set to AUTO my pc can boot up normally.

However if my pc was powered off over night it doesn't work meaning it only works if I am restarting or powering the PC up within a short period since powering off.

@sudden torrent, @proven canopy, @kind walrus

Any thoughts on what the cause of this is?

[proven canopy]

Buy an alienware

[steady lance]

Why?

Board is a MSI X670 Pro WiFi on the latest bios.

[proven canopy]

Actual answer - this is the first thing I'd google to learn more / find answers. https://www.google.com/search?q=msi+x670+pro+memory+context+restore+site%3Areddit.com

[steady lance]

Problem is that it can't even reach memory training.

It gets stuck in the CPU/DRAM light phase.

Power Cycling on and off it eventually gets it unstuck from it.

Should I switch to an old bios version?

[kind walrus]

Once again, if switching back to normal default settings mostly fixes these post issues. You have a stability problem with your overclock

Meaning something you've done skates the edge of stability still too much

To what extent that is? No idea, usually something that is often unique to the overclock

[steady lance]

Doesn't fix it.

[kind walrus]

Not something I can advise on other than dial back settings until you find it

[steady lance]

Same issues occur on completely stock settings with no pbo and jedec ram speeds.

[kind walrus]

Are you using a contact frame or not

[steady lance]

Board gets stuck on CPU and DRAM light.

[kind walrus]

Training is what this is

[steady lance]

Yes though it's been happening since before installation.

[kind walrus]

Go back to ILM

Contact frames are infamous for stability issues

Every. Single. One

Lol

[steady lance]

Though this has been happening since before installation.

I can remove it if you think that will fix it though.

[kind walrus]

I mean, sometimes the cpu mount can do it?

But contact frame is always suspect

You've also been using high i/o voltage settings, I can't say if it's safe or not but it's definitely more than stock

[steady lance]

So switch to the stock frame and try again? What do you mean by CPU mount?

[kind walrus]

I mean the cpu mount

And yes I would

Like a ram stick

You insert it

It doesn't sit properly

Reseat it works

[steady lance]

Oh so reseat the cpu.

[kind walrus]

Ye

You can try again with the frame itself

For consistency I would do ILM tho

[steady lance]

Weird thing is tht it's been happening since before applying the IO vtage

[kind walrus]

AM5 never needed a frame

[kind walrus]

Well maybe there's something in the system, usb or storage, that's causing this long post issue

[steady lance]

The one thing that comes to mind is that I updated the bios right around the time I experienced this problem.

So I am wondering if it would be safe to revert to an older version.

[kind walrus]

Not older than june 2023

Older than that, it'll work until the cpu explodes

(vSOC unlimited)

[steady lance]

I don't think I am going to be using an MSI board after this experience....

[kind walrus]

Anyway you've tried a few things, I would keep investigating with an open mind, something you could either have done or something you've plugged in could be causing this

It's hard to just know the answer without an obvious hint so I can only suppose on past ideas

[steady lance]

Been happening since I installed the new SSD so I don't think that's it.

[kind walrus]

It could be actually anything

Even the ssd

[steady lance]

Could my Sata SSD cause it?

[kind walrus]

Yes

A sata ssd killed my 7950x3d

[steady lance]

It was in the pc during transit

I think I will try unplugging that and a couple other things. If issues persist I will reinstall bios again to a older version.

[kind walrus]

Consider anything sus

Even psu or cables

And you'll likely find no change until you find the exact reason for the issue

[proven canopy]

Like pebcak?

[pliant lava]

Hey all question for you trying to do some OC to my new 9800x3d it was giving me some issues, but I seem to have been able to resolve that and stop the crashes when trying to run cinebench. However, I am using ryzen master and HW monitor and I seem to be getting two different readouts for my CPU speeds. Just curious if anybody had any insights to this.

Also if anyone could help guide me on how to do a light over clock or undervolt that would be appreciated.

It looks like all core test my cpu is hitting about 86c and is only boosting to 5080-5120

That’s during an all core test.

Score after 10 mins was 21448

[sudden torrent]

They poll at different rates and look at different sensors

[pliant lava]

Yesterday before I messed with anything it scored 22000. It’s like 2.3% different prob just a temp difference?

[sudden torrent]

A light overclock would be setting precision boost overdrive to enabled and calling it a day

[sudden torrent]

That's normal run variance

And yes the score will go down a little as it runs longer due to heat soak

Until it hits equilibrium

[pliant lava]

Ok cool that’s all really helpful. When I tried to undervolt it kept crashing after like 7 mins and then it wouldn’t even start cinebench. Temps were only hitting like 62 which was crazy

Not sure what I did

[sudden torrent]

Too little voltage will cause it to be unstable and crash

[pliant lava]

Is that in bios?

Is undervolting any good on 9800x3d?

[sudden torrent]

And If you're letting it boost instead of locking the frequency you should be letting the voltage stay dynamic as well, and undervolt using offset mode or curve optimizer

[pliant lava]

How do I do that? I tried in ryzen master but it’s forcing me to do it manually and I don’t know what it is

Or what I’m doing

[sudden torrent]

You can get a little bit from an undervolt but not much. It'll run cooler but you won't get any additional performance generally on that generation

[pliant lava]

Ok would you recommend just leaving it alone?

[sudden torrent]

Yeah even if you spent a full day doing nothing but tuning you'd only get about 2% more performance over just setting PBO to enabled

Good for making funny big number get bigger in cinebench, but that's it

[pliant lava]

Ok so I think that’s set to auto rn but I didn’t change that. Does that sound right?

Sorry I’m sure this is annoying

[sudden torrent]

auto is disabled for that setting

[pliant lava]

Ah ok. Soon as I get my daughter down for bed I’ll run down and check it out. I’m thinking the temp is gonna be to hot either way to really boost no?

I’m using a scythe fuma 2. Peeps said it would be fine

[sudden torrent]

It will boost until it hits 95 C

And then it will hold whatever boost frequency it was at when it hit that

[pliant lava]

Could an undervolt help that?

Ok so pbo will force it to hold 95c?

[sudden torrent]

No, it will let it boost higher which might cause it to get closer to 95C

Either way, if it does or does not hit 95, it'll still be boosting higher

[pliant lava]

Ok cool, would you recommend trying to mess around with curve optimizer or just leave the voltage alone?

Just trying to to see what would be reasonable. I know all silicone is different but I’m curious as I’ve never done it before

[pliant lava]

Turned pbo on and actually scored less again lol I assume it’s just variance but it’s still funny

[bright bramble]

So for a ryzen 9 9900 x I was wondering what kind of BIOS stuff I can do to oc right now when my computer is running sims with all cores I’m getting around 4.8GHz, and running around 60 degrees C

Figure there’s a bit more juice I can squeeze out of it

[steady lance]

PBO enabled with a undervolt using curve optimizer.

enable PBO with Motherboard limits, and maybe a clock override (tells the CPU to try and boost higher accoording to PBO)

measure your performance than start doing a per core undervolt till you find the most stable point without losing performance.

[steady lance]

I am thinking of trying to tune my GPU.

Suggestions on what software I should use to stress test?

Currently game stable for 3 days with +1000 mem.

[acoustic crane]

I've honestly crashed more while playing a game/running an in-game benchmark than I have doing standalone benches

But, at least for undervolting, I usually use kombustor and then 3Dmark

[steady lance]

Is undervolting on a gpu like undervolting for PBO in a cpu?

I never really got to deep into gpu overclocking.

If you don't mind, could you send me the links for Kombustor and 3Dmark? Thank you.

[acoustic crane]

https://msikombustor.com/

https://store.steampowered.com/app/223850/3dmark/

Steam

3DMark

PC gaming is a quest for performance, a never-ending cycle of upgrading, overclocking, and tweaking your system for faster frame rates, higher resolutions, and better quality visual effects.3DMark is essential equipment for this quest. For more than 20 years, 3DMark has been the first choice for benchmarking, testing, and comparing PC performanc...

Price

$34.99

Recommendations

14700

▶ Play video

(3dmark has a free version you can use)

[steady lance]

thank you!

I plan to OC GPU, CPU, RAM. Should I do the CPU and RAM before the GPU?