overclocking

[kind walrus]

It's A die anyway, aim for like a trfc of 500

Do like 500 400 300

Increase by 50 if you error

[steady lance]

Wait, so is it like tREFI where I want to just keep upping till unstable?

[sudden torrent]

Yes

[kind walrus]

Until stable

The lower you go, the less temp headroom

And the slightly faster you are

Until a theoretical limit per the ic

[steady lance]

So tune all the tRFC as high as I can?

[sudden torrent]

tRFC you want low, tREFI you want high

[kind walrus]

Start at the expected number, increase until stable

[steady lance]

Oh okay.

[kind walrus]

Start at 500 400 300

[steady lance]

Gotcha, now I understand.

Go as low as I can while being stable.

[kind walrus]

I run 720 because anything lower and I can't daily my ram over like 60C

[steady lance]

Do I repeat this for all the tRFC valuies like tRFC1, 2, 3?

[sudden torrent]

The max for tREFI is 65535 iirc

[kind walrus]

I'd much rather having up to 90C headroom

[sudden torrent]

AMD only cares about tRFC1, the other two might as well not exist

[kind walrus]

Mhm

They work in blocks of 4 8 or 16

[steady lance]

for whatever reason my pc did not like to boot at max tREFI but 65k boots and works fine.

[kind walrus]

Which is fine

[steady lance]

So just leave them on auto?

[kind walrus]

Like I said, 64K, 32K or don't bother

[steady lance]

also idk what is going on but somehow my average ram temps keep getting lower with each tighted timing?

[sudden torrent]

tRFC2 and tRFC4 can be auto, the memory controller doesn't use them

[kind walrus]

Not a clue

Probably ambient

[steady lance]

Wait. 🤦‍♂️ just realized why.

I was not running my AC during testing before, been keeping my ac at 20c lately.

[steady lance]

Thanks for the heads up.

[kind walrus]

It's ambient then

Which you should really run your maximum ambient if you plan to run this in an inch of it's life

1.6v will show temp errors quick if it's only a couple C over what's comfortable

[steady lance]

Any specific things i gotta do for the other timings in primaries or do I just lower everything from here on out and turn FCLK up as high as I can?

[steady lance]

I have not seen my ram go past 50c since I installed the fan.

[kind walrus]

That's 1000% A die

Lmfao

[steady lance]

usually around 45c during load

[kind walrus]

Also

[kind walrus]

You haven't done trfc yet so you probably have a lot of headroom

[steady lance]

Is it possible for it to be whatever that B die you guys mentioned before?

[kind walrus]

No

It's not samsung B

Nor M die

[steady lance]

Gotcha.

[kind walrus]

If you're unsure, send a pic of the sticker

[steady lance]

Kingston has a code system?

[kind walrus]

It should end in "ETMH" for hynix

Mhm

No clue what hynix

[steady lance]

Ahh will definitely check that later.

[kind walrus]

But this will show you it's hynix

[steady lance]

sticker on the ram right?

[kind walrus]

Mhm

[steady lance]

gotcha.

Before I go, what can I do to help tune FCLK?

[sudden torrent]

Find the vSOC sweet spot, it's usually around 1.1-1.15v iirc. Exact number varies by bin.

[kind walrus]

Just run 1.25vsoc

Ask @proven canopy for a good fclk test

I used mlc gui and ran 10 passes of each

To check theoretical speed

[sudden torrent]

From what I've seen vsoc too high causes regression

[kind walrus]

Since it's perma desync

Meh maybe

I think basically everyone runs either 1.25v or 1.3v

[sudden torrent]

That might have been because that was the auto setting on the board before the agesa patch that keeps them from blowing up

[kind walrus]

Mhm

A lot of boards just auto 1.25+ to my knowledge

It's a bit different to AM4, fclk just spits perf regression not stability regression

You don't really get wheas just ecc scaling back

[sudden torrent]

Yeah and I've seen the oscilloscope to prove it, silent errors

[kind walrus]

Yep

Forks has a better test that I can't remember the name of

I used mlc gui tho and ran 10 passes of each fclk over 2000, took the fastest of average 10 (2133)

[sudden torrent]

ycruncher?

[kind walrus]

Maybe?

I can't remember which is why I pinged lol

[steady lance]

Hynix confirmed

[kind walrus]

Told u

Now if you manage a low trfc then you 100% have A

I'm like 95% sure from the sub 50C 1.6v behaviour though

[sudden torrent]

low trefi or trfc

[kind walrus]

Trfc sorry

I make a lot of typos lately don't I lol

[sudden torrent]

Happens to the best of us

[kind walrus]

I feel like the old wise man and fal is my aid

"No felix you meant trfc"

"WHAT? WHAT DID YOU SAY?"

"LEMONS?"

[sudden torrent]

https://tenor.com/view/lemonade-national-lemonade-day-lemonade-day-gif-18125253

[kind walrus]

Oh man I'd kill for one of those rn

[sudden torrent]

Enjoying your summer then?

[kind walrus]

We just recorded the most humid day ever in our country yesterday

Yes||no||

[sudden torrent]

Gotta love when it's so humid you can't even sweat

[kind walrus]

It truly was

Good day to just be a seal

Couldn't tho

[steady lance]

Gotcha.

[steady lance]

Agreed.

[steady lance]

Felix do you by any chance know at what voltage A die starts having deteriorating?

Based on what I've read 1.6v is kinda the upper limit so long as it's cooled but I am not sure what voltage A die can handle beyond that.

[kind walrus]

Idk

I'm not really one to go beyond reasonable daily

Highest I've done is 1.65v

[steady lance]

Gotcha.

By de-syncing I can test how high my ran's frequency can go right?

[steady lance]

to my surprise, tRRDS just booted at 4.

[kind walrus]

tRRDL at 8 should too

[steady lance]

I got it stable at 6.

tRRDL 5 did not wanna work.

Just booted tWTRL 23 from 26.

22 refused to boot

Just got tWRLS 5 to boot.

4 was a literal 50/50 shot of boot or not boot.

[kind walrus]

Are you stress testing these changes?

[steady lance]

My strategy is to get a handful of timings to boot, run TM5 if it passes TM5 then I later run OCCT at 90% for a few hours.

If a error occurs I go through each one of the timings I tuned one at a time with same strategy.

got these settings to pass TM5, temps peaked at exactly 50c, averaged around 47c under load.

with that I've tuned all my tertiary and secondary timings at 1.6v

tRFC1 did not boot at 450. Trying 500 now.

tRFC1 500 boots.

[steady lance]

Should I leave tRFCsb on auto as well?

[steady lance]

Gonna try 475 if this passes.

passed TM5, gonna try 475 also tempts went up by about 2c in the left dimm.

Welp it did not boot, reverting back to tRFC1 500.

[steady lance]

tWR 54 booted and passed TM5.

For tRCD is there a formula I should use?

[sudden torrent]

No but it does influence tRAS

[steady lance]

What should I tune first?

tRCD, tRAS, tRC, or tRP?

[sudden torrent]

tRP will have the most effect of those

[steady lance]

Just saw your message and I've been tuning tRAS? Will double check in a second.

[sudden torrent]

I'd do tRP, tRCD, tRAS, then tRC

[steady lance]

Kay.

No formulas required?

tRC = tRP + tRAS right?

Any formulas for the other 3?

[sudden torrent]

tRAS = tRCD(RD) + tRTP
tRC = tRP + tRAS

Obviously if any of those make it unstable, increase it

[steady lance]

Where do I find RD?

[sudden torrent]

It's one of the tRCD ones, one says WR the other is RD

[steady lance]

Weird I don't have that in bios.

Just tRCD.

[sudden torrent]

Some don't separate them

Usually Intel but some AMD boards also

[steady lance]

So just do tRCD + tRTP?

ZenTimings shows tRCDRD

It's the same as WR.

[sudden torrent]

Yeah

[steady lance]

If anyone asks in the future apparently MSI X670-P does not separate.

tRCD 35 booted, trying 33.

32 failed.

33 failed, trying 34.

34 failed.

Tuning tRC now.

got the primaries to boot, current timings.

now running TM5.

Dang it, we got errors.

[steady lance]

Turning tRFC1 up it seems to be messing with boot.

[steel grove]

Hello, I just installed this (https://pcpartpicker.com/product/zcH8TW/gskill-ripjaws-v-32-gb-2-x-16-gb-ddr4-3600-memory-f4-3600c16d-32gvkc) set of RAM on my PC and have docp turned on but cpuz is reading it as cas 18.

Should I try to boot it up like this?

Or do I need to switch something else up? Thanks in advance!

And voltage like this

[steady lance]

Are you trying to overclock or?

[steel grove]

Just set it to default timings since docp is setting it up to CAS 18

And my ram is cas 16

[steady lance]

Trying to remember what docp.

Does your X.M.P settings work?

[steel grove]

Isn't docp just xmp for amd right?

The docp on bios is appearing like it should

[steady lance]

Oh okay, for me it's just called X.M.P on all platforms I've been on.

[steel grove]

But on boot up on cpu z its like this

The timings are a bit higher

Like my previous cas 18 kit

[steady lance]

You have any auto-tuning enabled?

Also when sending picture send whole screen as it's easier to analyze.

[steel grove]

I turned it off because the docp auto tuning set them to 16-19-19-39 like they should but when I booted up they appear higher like on the CPU z screenshot

[steel grove]

This is how it appears when I just installed the RAM

It appears as it should be working

[steel grove]

But when I boot I get cas 18 on cpu z

[steady lance]

Strange, what board and cpu is this?

5600x?

[steel grove]

Yeah Asus Atrix b550-f wifi

5600x

[steady lance]

Cl16 works fine if you tune it manually?

[steel grove]

Haven't tried bc I'm scared I might screw something up.

Is this how you should set it up manually?

[steady lance]

Go to your ram section and manually set the cas.

Worst that can happen is your PC won't boot in which case you clear the close to restore stock settings

[steel grove]

Like this?

And leave the voltage at 1.350?

Or should I change something else

[steady lance]

Yes but make sure you are changing DRAM CAS LATENCY.

Leave voltage the same so we can check if the issue is your ram.

[steel grove]

So this is correct?

Just change the values that I changed there?

And leave everything else at auto and the voltage at 1.350 like it is atm?

[steady lance]

If those are the same timings as your ram's XMP is then yes.

[steady lance]

Correct.

[steel grove]

I used this as a example

The xmp settings for this ram

[steady lance]

Look at the sticker on your ram.

[steel grove]

For 3600mhz

[steady lance]

That shows what settings you are running once in boot which can vary.

[steel grove]

The ram has the same timings on the sticker

[steady lance]

16-19-19-39-58 at 3600mhz

Those are your timings, you set your 4th timing wrong

[steel grove]

It's that I have 6 slots

Instead of 5

[steel grove]

That's why I set 19 three times

[steady lance]

Just occured to me, this is the wrong sticker. Was referring to the one on your ram but this works.

[steel grove]

Oh, let me take the panel off then

Or is it enough with that?

[steady lance]

Odd but yeah try booting it.

[steel grove]

Okay I'll leave everything like in the screenshot incl the voltage

The dram voltage changed by itself should I manually set it to 1.35?

[steel grove]

Or leave it on auto

[steady lance]

Set it to 1.35

We are testing to see if the issue is ram or board

[steel grove]

Okay, I'll boot now to check

[steady lance]

If your ram fails to boot at X M.P settings then it's most likely the ram.

[steel grove]

hmm

so it booted perfectly fine

but i now have 2133mhz of speed lol

I forgot to set it to 3600

[steady lance]

Yeah set the speed.

[steel grove]

the timings are right now though no?

[steady lance]

Currently I am tuning my own ram to figure out what's erroring lmao.

Yeah looking good also for Ryzen, use ZenTimings.

Gives you all your ram info and I think more accurate.

[steel grove]

okay thats an app right?

I'm horrible with ram lmao

[steady lance]

Yeah.

[steel grove]

Can I try to set it to 3800mhz?

[steady lance]

Start with 3600.

[steel grove]

Okay

[steady lance]

Done right you'd do frequency, then timings.

[steel grove]

Ahh okay

For now I just what the speeds I paid for lol

And timings

[steady lance]

Yeah AMD can be annoying with ram from my experience lmao.

Set it to 3600mhz see if it boots, if it doesn't try adding 0.5v to Dram.

[steel grove]

Okay so it booted and I have the full 1799.6mhz

Does this mean that everything is as it should?

[steady lance]

Technically yes.

I would install TM5 (in the pins) and run at least one round to test for stability.

[steady lance]

Much easier to read right.

[steel grove]

Is this the right app? https://www.overclock.net/threads/memory-testing-with-testmem5-tm5-with-custom-configs.1751608/

[steady lance]

Yeah.

Just the default config is fine btw.

[sudden torrent]

No, default is not stressful enough. You want one like anta extreme or pcbdestroyer

[steady lance]

How do I get those configs to work? I think I followed all the steps but they never work.

[sudden torrent]

Make sure to enable xmp even if you're manually setting everything. It does affect a lot more than you think.

[steel grove]

it doesnt let me apply both

[sudden torrent]

What do you mean by both?

[steel grove]

i mean, it doesn't let me manually set and have xmp on at the same time

[steady lance]

Found the wrong timing. It was tRCD.

[sudden torrent]

That... Shouldn't happen

Every motherboard I've seen has let you enable profile 1 xmp and use that as a baseline

[steel grove]

how long does the test last?

[steel grove]

okay, ill let the test finish and check again

[sudden torrent]

I often have trouble with the anta profiles but pcbdestroyer usually just works on mine

[steel grove]

I probably didnt see it since I didnt scroll long enough

[steady lance]

Could you send me the link to the pcbdestroyer config?

[steel grove]

it hasnt found any errors so far

[steady lance]

Test takes about 10m.

[steel grove]

should I stop it, enable xmp, leave the timings as I set them now and restart the test then?

[steady lance]

Nah let it finish

[steel grove]

ah it said test completed no errors found

ill change my settings and retest

[sudden torrent]

https://github.com/integralfx/MemTestHelper/tree/oc-guide/TM5-Configs

[steady lance]

Ty

[steel grove]

I'm sorry yeah I didn't scroll down enough lol

I'm leaving everything else as I had it

Or should I change something else while I'm here?

[steady lance]

Could try seeing if 3800mhz boots.

[steel grove]

Will do

I should leave everything else as is right?

[sudden torrent]

For now reboot and go back into bios so it sets voltages and sub timings correctly

[steady lance]

Btw if you send us a picture of the sticker on your ram, we can help you identify what ram chip (die) you have.

[steel grove]

Okay so no rebooting into 3800?

Just leave it at 3600?

[steady lance]

@sudden torrent, download this one?

[sudden torrent]

For now yeah, you can try 3800 after reboot but I'd recommend finding your memory die first

[steel grove]

Okay only changed docp to on

[steady lance]

I got the same frustrating error.

opens, gives a ton of error messages, then closes.

[sudden torrent]

Are you running as admin

[steady lance]

Yes.

[steel grove]

rebooted and got this

is everything as it should be?

should I retest or should I try 3800mhz?

[steady lance]

On Reddit it says it has smth to do with the config.

[steel grove]

ahhh, let me take out the ram and send you a pic

[steady lance]

Like the "Testing Window Size"

[sudden torrent]

At a glance yeah. That looks like micron memory dies to me.
You can get the other sensor data by updating bios and installing chipset drivers.

[steady lance]

Ram is in binary right? So 1024MB = 1 GB

[sudden torrent]

Sometimes I've seen the program itself be corrupted, and you need to delete it and re-extract it to fix

[steel grove]

Let me get the latest chipset drivers

[sudden torrent]

Technically 1024 MiB = 1 GiB but yes

[steel grove]

And update my BIOS

[steady lance]

@sudden torrent it worked!

I had to change the Testing Window Size in the config because it was trying to test the wrong amount of ram or smth acoording to reddit,

[steel grove]

I should say yes to everything here right?

[sudden torrent]

yes

[steady lance]

formula in config is 0.8 multiplied by RAM AMOUNT divided by thread count

Which config is best for TM5 stress test?

[sudden torrent]

Extreme anta is good for a gaming stable test

Pcbdestroyer is a longer test that can find errors late in testing from heat

[steady lance]

gotcha.

[steel grove]

Omg I had a mini heart attack

I updated the bios but forgot to backup the tpm code

And it didn't let me boot

But I remembered that I had my old bios still downloaded

[sudden torrent]

TPM backup is only needed if you're using bitlocker

[steel grove]

I don't know if I'm using it 😅

I don't remember

I went back to the previous bios to verify if I had it on

[sudden torrent]

If you don't know then you aren't most likely, it's not something you can set up when you install windows. It's done after.

[steady lance]

Fal, any idea why sometimes all of a sudden when I exit out of bios my PC LEDs don't turn on past CPU lights?

[sudden torrent]

The debug lights or the RGB?

[steady lance]

Debug light shows CPU/dram, it gets past that then no lights but everything is spinning

[sudden torrent]

Fast boot

[steady lance]

Turn that off?

[sudden torrent]

If it's a problem for you then yes

Disabling that will force bios to recheck things like the GPU even if it registered as ok on the last boot

[steady lance]

Gotcha.

Is it sometimes called Memory Context Restore?

[sudden torrent]

That is a separate but related thing

[steel grove]

okay its off lol

[sudden torrent]

Then you're safe to update and hit "Y" when it warns you about the TPM reset

[steady lance]

Any idea where I can find fast boot?

[sudden torrent]

MSI right?

[steady lance]

Yes

[steel grove]

Yeah that's what scared me lol

I'm updating again to 3404

I'm thinking of getting a 5700x3d when it comes out, It should just be a simple switch when I get it no? Bc I have the latest chipset drivers and BIOS

[sudden torrent]

In theory yes it should be easy to swap

If it fails then clear your cmos

[sudden torrent]

Looks like they removed it from newer bios versions. They probably rolled it into that memory context restore option.

[steady lance]

Gotcha so I wasn't going crazy.

I tried looking for it the other day when Felix mentioned I should turn it off but could not find it at all.

How long does PCBdestroyer usually take?

[sudden torrent]

Between 3 and 8 hours depending on factors

[steady lance]

My g0d.

So just leave it on overnight?

[sudden torrent]

Yeah that's fine

[steel grove]

finally have everything up to date

[sudden torrent]

Nice now you can see the fclk and such to make sure it's synced

[steel grove]

should I go for 3800 with everything as is?

or do you want me to take out the ram to see the sticker?

[sudden torrent]

What brand is the memory

That looks like a g.skill model code

[steel grove]

https://pcpartpicker.com/product/zcH8TW/gskill-ripjaws-v-32-gb-2-x-16-gb-ddr4-3600-memory-f4-3600c16d-32gvkc

yeah this is my kit

[sudden torrent]

Perfect, g.skill has a code starting with 042 or 043 that tells you exactly what die the kit uses

So yes a sticker pic will help speed things up

[steel grove]

okay no prob, let me shut it down and take out the panel real quick

Do you need a pic of both?

Or just one of the sticks

[sudden torrent]

Nah just one

[steel grove]

Okay

[sudden torrent]

They're a matched pair so it'll be the same code

[steel grove]

Here you go

[sudden torrent]

8821C is CJR

Dual rank CJR so 8Gb

So, the overclocking potential of those sticks will be limited

[steel grove]

Okay

[sudden torrent]

You might be able to run 3800 but don't spend too much time trying

[steel grove]

Should I just leave it as is?

Or should I try 3800 with everything else like it currently is?

[sudden torrent]

Try 3800 with tCL at 20

If that fails there's little hope of it working at that speed

[steel grove]

All three that are 19 at 20?

[sudden torrent]

Just DRAM CAS

[steel grove]

Leave the rest as is?

[sudden torrent]

Yes

If it boots, go back into bios and set dram cas to 18 and see if it still boots

[steel grove]

Okay

[sudden torrent]

You can set up to 1.4v safely to help with higher speeds and lower CAS

1.45v is on the edge of safety for that die

Any higher and it degrades quickly

[steel grove]

ill go back to bios then

you said voltage 1.4 and cas 18?

[sudden torrent]

Awesome. Looks like you'll need to manually set fclk but that's a minor thing.

Yes

[steel grove]

Yeah that's weird it's different lol

Booting back up

[sudden torrent]

It's because Zen 3 is coded to have a ceiling of 1800 fclk/uclk on auto, but most chips can do 1900 with little effort.

[steel grove]

back up

should I try cas 16?

[sudden torrent]

From here you can follow this guide, lots of tips and "why you should do this" https://github.com/integralfx/MemTestHelper/blob/oc-guide/DDR4 OC Guide.md#amd

Or I can walk you through things one by one if you want the TL:DR version

[steel grove]

Hmm, can I test cl 16? And if not try that?

Or should I just go through those steps instead?

Also you told me to set this to 1900 right?

[sudden torrent]

Yes fclk 1900. You'll probably need to adjust other voltages to get that working.

CAS 16 is a good goal but one for later

1.10 V SOC
0.95 V VDDP
0.95 V VDDG CCD
1.05 V VDDG IOD

Asus might call SOC "VDDCR SOC"

[steel grove]

Okay so soc is already at 1.1

I can't see vddp tho

[sudden torrent]

You'll need to change the "voltage control" settings to manual

Instead of auto

[steel grove]

I can set them individually

[sudden torrent]

Oh I think asus calls manual control mode "override" mode

Then the setting appears below

[steel grove]

I can't see vddp though

[sudden torrent]

Ah or that

CLDO VDDP

[steel grove]

Ahh it's right there like that lol

Should I leave the vddp standby on auto?

[sudden torrent]

Yes

These are the voltage settings that are known to work for most chips and are perfectly within the safe margin

[steel grove]

Anything else I should change?

[sudden torrent]

For now no, you want to check fclk and uclk for errors before you do anything else

Boot windows and open zentimings again

[steel grove]

Okay so I should boot it with those values and fclk on 1900mhz

[steel grove]

With these settings right?

[sudden torrent]

yes

[steel grove]

here's the screenshot

[steel grove]

what should I do now?

[sudden torrent]

Do you have hwinfo

[steel grove]

yeah

[sudden torrent]

Ok, have that running and watch the very last sensor. That's the WHEA error one, which is the kind of error the infinity fabric will throw if it's not running right.

A single error means it's not stable

[steel grove]

it doesnt have any errors rn

[sudden torrent]

Run your memory test again and watch it

[steel grove]

will do

already got one lol

[sudden torrent]

A good 30 minutes should tell you if fclk is stable enough

Oof, well no worries we play with resistances now

[steel grove]

I should just stop it right?

[sudden torrent]

Yeah no point continuing

[steel grove]

and go back to BIOS

[sudden torrent]

Yep

Next thing is SOC to 1.15v and VDDP to 1v

[steel grove]

Omw there

[steel grove]

Just that?

[sudden torrent]

For now yes

[steel grove]

Okay I'm rebooting

[sudden torrent]

Better to let voltage handle the instability than tweak resistances up to a point

[steel grove]

Just run the test again right

[sudden torrent]

Yep, watching that WHEA counter again

[steel grove]

Okay

Will just this test be enough to see if it's stable?

[sudden torrent]

Define stable

"Able to run games without losing performance or blue screens" is a common one

[steel grove]

To continue to the next step lol

And that

Haha

[sudden torrent]

There's also "cannot lose any data or reboot under any circumstances" stable

This is enough for the first type lol

[steel grove]

Can I use discord on the PC while running the test? Or should I just leave it running alone?

[sudden torrent]

You can

Just open discord first before the test

The test will use whatever memory is available

[steel grove]

Ah it opened with the pc

done

[steel grove]

What should I do next?

[sudden torrent]

Excellent, next you can do trrds and trrdl to 4, tfaw to 16, twr to 16, and trtp 8

Those should all be easy enough for the memory to do at 1.4v

And they speed up testing time

[steel grove]

Okay I'm going to BIOS

[steel grove]

That's in dram timing control right?

[sudden torrent]

Yes

If it fails set trrdl to 6 and try again

[steel grove]

This and retest?

[sudden torrent]

Yep

[steel grove]

Rebooting

Oof bluescreen

[sudden torrent]

trrdl 6

[steel grove]

Will do

[steel grove]

Okay rebooting

Booted up

Retesting

Stable so far

[sudden torrent]

If it survives this test then I'd save a profile in bios so you can get back to this point quickly

[steel grove]

That's what I was going to ask

If I can just save a profile on the BIOS

[sudden torrent]

100% good practice to do that periodically

[steel grove]

I want to do a profile of just the regular 3600 cas 16 stock profile too

For regular use

[sudden torrent]

Well for that all you need to do is reset to defaults and then load the xmp profile

[steel grove]

Okay

I'll do the test, save

Do default and save again

[sudden torrent]

After you save the profile you can set trfc to 500, which sets it to a little over 260ns at 3800. CJR should be able to do 260ns, or if you're unlucky 280ns which would be setting 530.

Don't worry about trfc2 or trfc4, AMD doesn't use those so you can ignore them

[steel grove]

I'm sorry to ask, but theoretically, how far am I from cas 16 no data loss stable? lol

[sudden torrent]

We'll find out soon

[steel grove]

ayy

[sudden torrent]

You want to get SCLs stable before CAS, since they have a higher impact on performance

[steel grove]

Where do I make a profile

NVM sorry found it

I should be able to reset now right?

[sudden torrent]

Yep

[steel grove]

Can I just default the AI tweaker?

[sudden torrent]

Technically yes

[steel grove]

I can't find an option for this lol

[sudden torrent]

You'd need to reset each option to auto lol

[steel grove]

Can I just manually change everything?

Okay haha

[sudden torrent]

Easier to just "reset to defaults" on the exit menu

[steel grove]

This should be it right?

Just for regular use and gaming while I'm finished OCing

I'm leaving it at this for regular use.

Atleast we did some good progress today right?

Thank you for everything! I won't be able to continue today, but I'll try Sunday. I'll leave it as is meanwhile.

[sudden torrent]

Sounds good. Enjoy your saturday.

[steel grove]

thank you again for introducing me to OC lol

I've been scared of trying it for the longest time haha

[sudden torrent]

Happy to help. It can be scary at first with all the warnings and disclaimers but it's really hard to hurt anything in reality.

[steel grove]

yeah that's what kept me away

[sudden torrent]

The only thing that can potentially damage things is voltage, and I'm keeping you well under the safe limits so no worries there

For example, vsoc is safe up to about 1.3v

[steel grove]

so voltage is the only dangerous thing here?

the rest only should be harmless bluescreens no?

I'm leaving it like this.

I'll be heading to sleep, thanks again.

[sudden torrent]

Yep the only risk with timings is losing data or corrupting the OS, which are generally easy to avoid or fix

[steady lance]

@sudden torrent, my ram survived PCB Destroyer at 52c.

[steady lance]

What speeds can most A-die do de-synced?

[kind walrus]

That's really open ended

Like, 8000+

I'd have no idea for 64gb

Desynced is also a huge push to make it in any way worth a small gain

Bit like PBO in a way lol

[steady lance]

I am just curious how high my ram can clock to.

[steady lance]

I plan on over locking my CPU next once I finish fclk and final primaries. Does 7950x3D also have that 200mhz limitation?

[sudden torrent]

64GB will be the limiting factor in your RAM speed. You'll hit the memory controller limit before you hit the die limit.
You can't overclock X3D, you can only use curve optimizer.

[steady lance]

X3D I can use PBO and CO right?

[kind walrus]

If it's anything like mine, you'll probably have zero headroom on the vcache chip

[kind walrus]

Can, not much point to it tho

[steady lance]

I just can't mess with voltages like I did on my 5700G right?

[steady lance]

I enjoy the tuning process.

[kind walrus]

I can see that since you've spent like a week or two on what I spent a day

[steady lance]

xD yes

[steady lance]

Hoping when I get back home that my ram passed PCBdestroyer.

[steady lance]

I am reading about 7950x3D over locking and noticed is it really abnormal that my CPU stays around low to mid 70c during load?

Every article I am reading shows their cpus hitting 89c under load at stock.

[kind walrus]

No

Mine caps at 73C on water

[steady lance]

What cooler?

The articles I am reading use mostly Arctic LF2 360.

[kind walrus]

W-

Water

Water loop

[steady lance]

Oh custom loop?

[kind walrus]

Yes

[steady lance]

How's the maintenance on a custom loop and when you move the PC do you need to take the whole thing apart?

Custom loops seem super fun to me cause I can picture all kinds of layouts.

[kind walrus]

I don't need to take it apart

It's heavy af for basically no benefit

If I didn't already have the loop I'd be on air cooling now

Maintenance is 6 months to a year but depends on a lot of factors

[steady lance]

What would some of the big factors be?

Weight isn't a big issue for me, I am used to moving my 60lb PC everywhere I move.

[proven canopy]

Custom loop because you want to, not because you need to

Are you running maxed negative co?

[steady lance]

Just got home to check my TM5 run and these timings passed PCB-Destroyer!

Just need to tune tWR and FCLK and I'm finished for ram.

[steady lance]

If you're asking me, then no. I have not tuned my CPU yet, that's next in line though.

[proven canopy]

I would sure hope so at 1.6 vdimm lol. You don't need nearly that much

[steady lance]

xD, the jump from 1.5 to 1.6v made a decent difference for me.

[steady lance]

Do you by any chance know if it's fine to leave VDDIO and VDDQ at 1.6v? My mobo sets those equal to Dram voltage by default. I'm running likely Hynix A.

my temps are 48-52c during long stress tests.

[wheat lintel]

This is A-die? tRFC seems high for it

And jeez 1.6V Vdimm?

[sudden torrent]

He has a fan over it at least so it's at about 50-52C

You are right though, 500 sounds more like M die

[wheat lintel]

It’s safe ik but I feel 1.6V is still high for his frequency and timings

[sudden torrent]

We've tried to convey that to him before

[wheat lintel]

Ahh

[steady lance]

tRFC refused to boot at 450.

And yeah my ram peaks at about 52c with PCBdestroyer and is usually in the mid 30s to low 40s

[sudden torrent]

400 is pretty common for A die so it very well could be M die

Which is still good

[steady lance]

M is the preference right?

for AMD at least.

[sudden torrent]

A is better technically but both are more than enough for AMD

[steady lance]

ahh okay

I will probably go around to tuning tWR and FCLK later tonight.

[sudden torrent]

Even Samsung would be good enough for AMD really, but not as good for the future carrying it forward once DDR5-7200 becomes a common jedec standard that everything can run

The benefit of A/M die is it can hit 8000 on common sticks, but it's not stable on many current memory controllers

[steady lance]

gotcha

anyway I can push my ram further?

[sudden torrent]

Probably not on your current CPU

There might be one or 2 timings you can get down but nothing significant

[steady lance]

what are those?

[sudden torrent]

tWTRL and tWR possibly

[steady lance]

I already tuned tWTRL, only timings left are tWR and tRFC2/SB

[sudden torrent]

Those last 2 you can ignore on AMD, they're not used

[steady lance]

Kay.

gotta run, thanks for the info.

[steady lance]

My new stable timings.

• tWR dropped to 66 from 90.
• FCLK upped to 2200

I will try FCLK 2300 later today.

[kind walrus]

higher fclk ≠ faster performance

you will get corrected errors hindering peak

[steady lance]

I forgot, what does that symbol mean again?

[kind walrus]

does not equal, does not mean, is not the same as

[steady lance]

How do I check for FCLK performance?

[proven canopy]

#overclocking message

[steady lance]

Thanks.

[steady lance]

FCLK 2200 definitely was not 100% stable, I had a idle crash.

[sudden torrent]

If it's crashing then you're way above where you should be. Normally it'll just silently correct the errors, losing a little performance in the process.

[steady lance]

Just incase it was tWR I've returned fCLK to 2000 and ran multiple rounds of pcbdestroyer, it was definitely fclk.

[abstract merlin]

Im thinking about overclocking my ram since it seems like free performance, I guess. It doesn’t look like much of a boost from what I’ve seen but why not

I have 3200 dimms right now. I don’t remember my timings completely but I think its 16-18-18-something

Gotta check later

I dont think my performance increase will be that large because I am not using a high end gpu pushing out really high frame rates

[sudden torrent]

Odds are pretty high that you have a C die kit with those specs, so your overclocking options will be limited but there's still a few things you can tighten on subs, if you're lucky you can get it to 3600 18-20-20.

[proven canopy]

If you have Hynix CJR - check the ddr4 guide in the pins #overclocking message

[abstract merlin]

I was thinking about going to 3600 with timings similar to that too

I will check now

[abstract merlin]

I dont know who makes that kit, but all I can find is sk hynix in software. All I know is sk hynix, micron and samsung b-die

[sudden torrent]

What software are you looking at?

[abstract merlin]

There was a program I dont remember the name of that I am looking for but I used cpuz and just downloaded some pcmark thing

[sudden torrent]

Was it thaiphoon burner?

[abstract merlin]

The program im trying to remember even showed theoretical configs you should be able to run at

[abstract merlin]

Not sure, maybe

[sudden torrent]

Well if the CPU-Z SPD tab says "Hynix" by DRAM Manuf. I'd be inclined to guess you have CJR, aka H8C

So you can follow the DDR4 guide with that in mind when you get to the timings

[abstract merlin]

Ok burner is showing A-die under the die density

[abstract merlin]

Sure

[sudden torrent]

Thaiphoon really only guesses at it. It said my Samsung C die kit was B die

It is good for finding the manufacturer at least

[abstract merlin]

[sudden torrent]

Do you know the brand of memory the sticks are from? After double checking, AFR could be possible but that's a very old die.

If it's g.skill or corsair there's a code on the label that will tell you exactly what it is

Patriot too, to a lesser extent

If you're not sure you can find that in CPU-Z too, it's the "Module Manuf."

[abstract merlin]

I have some 5 year old g.skill trident sticks

During the ram shortage of that time

[sudden torrent]

Perfect, there's a code starting with 042 or 043 on the label. The last 5 digits of that code tell you exactly what they used. You might need to take out a stick to get at it.

[abstract merlin]

Yeah thats a thing I can only find by taking it out

Last 5 are 8820C

[sudden torrent]

Yeah that's AFR which is not good for overclocking.

[abstract merlin]

Brutal

[sudden torrent]

That doesn't mean you shouldn't try, you can always tighten timings even if it's not likely to go above 3200

You can start with the "safe" timings in the guide

[abstract merlin]

Yet typhoon said it was hynix a die

[sudden torrent]

Yes that's what that means, H8A, Hynix 8Gbit A die

All interchangeable terms

[abstract merlin]

Oh is afr and a die the same thing?

[sudden torrent]

Yes different naming schemes but same die

[abstract merlin]

Oh I thought that was a different bin or something

Clearly i am a ram normie

[sudden torrent]

It's not really common knowledge even among an average overclocker so it's fine

[kind walrus]

8gbit hynix CJR

Don't use over 1.45v

[abstract merlin]

Alr. About degradation, how long does ram typically last when not touched? These sticks I have are going towards 6 years

[kind walrus]

Just don't take it above what's suggested

Work with an even lower voltage limit if you want to be sure

[abstract merlin]

True, im just curious

[kind walrus]

Well when someone says observed degradation it's probably a few months or a bit longer in the case of the internet

[abstract merlin]

[abstract merlin]

Brutal dyno pill

There are a lot of secondary timings. Do I have to worry about any of these or am I good focusing on the 4 main timings

We bouta find out

Ima test this with the testmem version you guys linked

I ran it for 3 minutes at 3600 18-20-20-40 with no errors

[abstract merlin]

Now I wanna try 3600 but with the old timings. Ima run it for longer to see if it crashes or something

[abstract merlin]

2 immediate errors

[kind walrus]

The 4 main don't do much

Seconds and terts do a lot

[abstract merlin]

I went from 3200 16-18-18-38 to 3600 18–20-20-40 at the same voltage. I think ima leave it for now because I dont know much about this

[abstract merlin]

[kind walrus]

It's the whole reason XMP doesn't really matter much in a lot of cases unless the frequency difference (MHz/MT.s) is huge

[abstract merlin]

im having frame drops so ima leave it on default and take the L

[sudden torrent]

You'd think that but according to the ddr4 overclocking wiki 8820C is AFR. Granted there's only 2 entries and they're guesses based on SPD so it could be wrong, but that's the only AFR on the table.

[kind walrus]

I used the guide lol

C is the revision

[tall pelican]

thaiphoon absolutely reads cjr as afr

https://cdn.discordapp.com/attachments/328891236918493184/727243761716953189/unknown.png?ex=65b7fc08&is=65a58708&hm=d29a18069f0cf5f8f90e7a279f6d985e662d73451301e1b84804ca746a7d9872&

there's no way in hell 8gbit afr is even posting 4500mts

[faint tangle]

But what if it tried really really hard...

[kind walrus]

It would get an A for effort and a C for lying about it.

[steady lance]

Okay, that was odd. I had a random crash earlier, turned tWR up to 72 and been running same stuff for a few hours, no crashes thus far.

[steady lance]

😆

[steady lance]

Does HWinfi read dram bus read/write in MB or Mb?

[steady lance]

mvm

[deft quiver]

Increase voltage

Put a fan on it

[zenith palm]

No necessarily depends on the die

Increasing voltage can be an awful idea they may be on the max of what is safe for their die currently

[sonic brook]

Can ram overclocking go any amount over what the motherboard is listed to be capable of? and is it good to have a buffer of the motherboard's stated ram speed limit over the overclocking target speed?

[kind walrus]

If it's unlocked for it yes

The sky is the limit in a theoretical sense

[sonic brook]

What can I get up to with this ram kit, this CPU, and this motherboard?
https://www.newegg.com/g-skill-64gb/p/N82E16820374432?Item=N82E16820374432
https://www.newegg.com/intel-core-i9-12900k-core-i9-12th-gen/p/N82E16819118339?Item=N82E16819118339
https://www.newegg.com/asus-tuf-gaming-z690-plus-wifi/p/N82E16813119554?Item=N82E16813119554

[kind walrus]

I wouldn't know

It's early z690 with 64gb of A die

Better than xmp but not sure how much, it would vary board to board

6800+?

[faint tangle]

6600-6800 prolly

[sonic brook]

And I think you said going over 6800 wasn't worth it?

[sudden torrent]

It starts to get unstable above that

Random errors and blue screens for no apparent reason, but it passes every memory test except for the week long test that it gets one error 5 days in

[deft quiver]

Bruhhhh that's to long

That error could of been cosmic rays

[sudden torrent]

No other explanation existed for that result. 7200 - error. 6800 - pass the 7 day test.

And it is repeatable

[faint tangle]

^

[wheat lintel]

Doubt, daily stable? I’d say 6400 at best

6600 was kinda doable and 6400 still was hit or miss with a lot of voltage tuning on my 2x32 A-die with a Z790

[steady lance]

Is it possible for 1.6v on Hynix A or M die to cause my ram to degrade to the point where XMP requires extra voltage to work within a week?

My pc has been having some random crashes here and there so I am trying to figure out what might be the cause.

Crashes never happen during gaming even on the infamous game ark. I notice it during browsing on Chrome or when using Curseforge launcher.

[kind walrus]

Yeah but yours was gigacope

[kind walrus]

1.6 is definitely high and unnecessary for your 6400 tune

That said, degrade? I don't think so

I'd say crashes and bsods slowly occuring is a strong sign your tune isn't stable

Like the slow creep in of a corrupted windows install basically

Been there, done that, many years ago 💀

[wheat lintel]

That same gigacope board can do 7800MT/s daily stable on 2x16

[kind walrus]

That's not translating very well

It should do better on 2x32

The gap isn't that huge

[steady lance]

Is it possible Chrome and Curseforge got currupted on my end?

Cause that tune I did was completely stable on everything except Curseforge and Chrome and when I changed everything back to XMP I was still experiencing issues with specifically those two apps.

[kind walrus]

What did you stress test with

[steady lance]

PCBdestroyer and whatever that other one you guys recommend for TM5 was called.

Passed multiple rounds of both.

TM5 with PCBdestroyer and Anta Extreme profile. I had done multiple rounds of both back to back without issue.

[kind walrus]

Did you do a large test like vst?

[steady lance]

what is VST?

[kind walrus]

Y cruncher

[steady lance]

Is it free? If so, I will test that out.

[kind walrus]

Yes

[steady lance]

I am also gonna reinstall chrome and curseforge before hand on my current XMP profile with slightly extra voltage just to see if corruption is the issue.

[kind walrus]

Also try this

https://www.reddit.com/r/Amd/comments/xz0juq/one_potential_not_ftpm_stutter_fix_for_am5/?rdt=55606

Or also set power supply to "typical current idle" in bios

Those are a couple potential fixes for idle/low power state crashes

Otherwise, I'd go back and start dialling back the ram tune imo

Run xmp for a few days for example and see if the crashes still occur

If they do try reinstall windows and see if it's better again

if so

The tune was unstable

Or rather, even if they don't occur anymore then the answer would be the tune was unstable still

[steady lance]

Hmm okay.

just reinstalled chrome and curseforge, updating gpu drivers while I am at it.

well so far so good.

reverted back to old settings after reinstalling and so far no crashes within the first like 10m of browsing

[proven canopy]

vst in the pins

[wheat lintel]

Definitely not stable if that’s all you tested with

[sonic brook]

Is it better to have a higher limit on ram from the motherboard than you need?

[faint tangle]

I mean, not really but it also doesn't necessarily mean you should max it out

Like b550s support 4400mhz but no way you should try it

[sonic brook]

I was thinking more like going up to 6400 on a 7800 board

Is that preferrable to going up to 6400 on a 6400 board? (I don't know if a board with that exact limit exists, this is for sake of argument)

[steady lance]

Since re-installing chrome I have not had any crashes so far.

Beginning to think the problem was that chrome had become corrupted during the testing phases earlier on.

[sudden torrent]

6400 on current AMD chips may or may not work without desynchronized UCLK. If it does work, unless you're very lucky it will require manual tuning.
Don't desync if you can avoid it. 6000 is generally considered the "safe" spot on AMD that most chips can run with little effort on your part.

You can overcome the latency penalty you get, but you need to get a very high end board to do it by getting the memory above 8000.

[sonic brook]

What about Intel?

[faint tangle]

Intel can do in the 7000s and beyond fairly consistently

Well 13th-14th gen, I can't say that for certain with 12th gen

[kind walrus]

Usually means better compatibility yes but there are way more factors than just that one number

[sonic brook]

Go on...

[faint tangle]

What cpu you have, what mobo, what ram costs, if you're ram tuning, a lot

[sudden torrent]

Intel can get weird instability at extremely high speeds. Exactly what point varies by CPU, motherboard, and pure luck in the silicon lottery. The point of instability is usually 7000-7400 but I have seen cases where 7600 was perfectly stable long term.
When I say instability here, I mean random errors and blue screens. You can still pass benchmarks and many stability tests, unless you get that one random error.

[sonic brook]

I have
https://www.newegg.com/intel-core-i9-12900k-core-i9-12th-gen/p/N82E16819118339?Item=N82E16819118339
https://www.newegg.com/asus-tuf-gaming-z690-plus-wifi/p/N82E16813119554?Item=N82E16813119554

Thinking of replacing with
https://www.newegg.com/intel-core-i7-14700k-core-i7-14th-gen/p/N82E16819118468?Item=N82E16819118468
https://www.newegg.com/p/N82E16813162149?Item=N82E16813162149

[faint tangle]

Get an a die kit and you could likely easily do 7000, I dont know what timings. I dont know A die tuning at all

[kind walrus]

cpu imc, motherboard trace signal quality, ram quality, motherboard training algorithms, bios version, mounting pressure, cooling capability, cosmic rays, alignment of the planets

etc etc

[sudden torrent]

I don't know about "easily," a fair bit of luck is involved for it to be considered easy as xmp. Lots of tuning and voltage tweaking is often required. 12th gen is also weird in that it's very particular about what is stable and what isn't when you're running on the edge.

I assume it's similar on 13/14th since Intel didn't change the memory controller much, just the motherboards got better generally.

[tall pelican]

to 2nd this, I just had a cyberpower prebuilt come in with a 13900k, asus tuf z790 and 7200c34 gskills, and it was instantly unstable in occt mem

[wheat lintel]

3rd this, I’m running Teamgroup 7200c34 2x24 on my 13900k, Z790 Aorus Master on auto voltages and everything and perfectly stable

A whole lotta luck on high freq XMPs

[granite canyon]

I'm absolutely sure you guys get this along. I am upgrading to 14900k and 7200mhz memory. On cpu specs online it says its limited to 5600mhz ddr5

Is this accurate?

[kind walrus]

That's the rated jedec

I.e the base speed it's validated up to

14900K should be able to do well north of 7200

with the right motherboard

*and ram capacity

[granite canyon]

32gb 7200 (2 x 16) nzxt n7 z790

[kind walrus]

I have no idea how good that n7 is but it's asrock derivative and a 4 dimmer so I'd say good luck for trying to get 7200

The outlook is bleak

[faint tangle]

Yeah I'd try like 6800

Nzxt lga1700 mobos have not historically been good for ram tuning

[granite canyon]

I don't need full speed 🙂

[granite canyon]

Dude at the pc shop said i could get it to 7200 but I saw the online info so I got scared

But I get you

I took it in cos my old 11900kf wasn't working properly with my 4080. Crazy bottleneck.

[kind walrus]

Believable

[granite canyon]

Showed him my upgrade plan and he said it should work

[kind walrus]

Is this just for gaming

[granite canyon]

And music production and editing

[kind walrus]

Oop I didn't see that

Ah

Ok

Did you buy a new cooler

Or reuse your old

[granite canyon]

Was saying performance was awful and I got the same fps as my 3060 and 11600k. Old cooler new thermal paste

Temps don't go above 65

[kind walrus]

What cooler is it

[granite canyon]

Aio ftw

Corsair h100x

[kind walrus]

H-

[granite canyon]

H?

[kind walrus]

Ok can I make a couple suggestions

[granite canyon]

Yup

[kind walrus]

Try the nzxt board, see if 7200 works ok

But if not-

One sec

[granite canyon]

I want a 360mm one but I'm stuck on 240for now

[kind walrus]

@granite canyon https://pcpartpicker.com/list/sZtJqR

I would suggest these options

The ram you can swap in, clock up

Or change to a nicer board, one I could ascertain 7200

[granite canyon]

Sick the. My ram is g skill trident z5

[kind walrus]

The aio I would swap to regardless

Z5 is ok

[granite canyon]

Cos it's corsair? Lmao

[kind walrus]

Nah cos it's not up to scratch

14900k is a volcano

Like the 13900K/S/F before it

[granite canyon]

OK OK cool thanks

[kind walrus]

The frozen edge is good

Not the best

But it ships with noctua fans, good ones

[granite canyon]

But thermalright works? I might just get kraken 360

[kind walrus]

And the pump and rad is like $20 on top

Thermalright would be better than kraken

[granite canyon]

Fr?

Bet

[kind walrus]

Kraken uses a new asetek pump but it's so expensive for very little benefit

[granite canyon]

Wanted lcd luke a dunce

Like

[kind walrus]

The thermalright should hit 98% of your requisite for $55

[kind walrus]

Go get an old phone out with a usb cable

And install "superdisplay"

Wire it into your case

Full blown screen for a couple $

Much better than a tiny little aio lcd

[granite canyon]

Lmao. May just buy an old smart watch and 3d print a housing

[kind walrus]

Those you pay too much for

[faint tangle]

If you want "better" go ek 360 at $100-110

But the thermalright edge is plenty

[kind walrus]

Hey that's a great idea

I'll remember that

Keep us updated on it

I would love to see it work

[granite canyon]

Can I send a friend request cos you've been super helpful

Good tonasn

[kind walrus]

Sure

[granite canyon]

To ask

[granite canyon]

Gonna be doing a bunch of cool stuff using lcd and watch quartzes. Some motors too (haptic knobs for software control)

As soon as I'm set up with oc

Pc

[kind walrus]

Sounds really interesting

[blazing blade]

anyone know what die this is?

[sudden torrent]

All I can tell is it's 3200c16 so it could be just about anything

[kind walrus]

Unlikely to be anything too amazing

I don't know how to read hyperx codes either unfortunately

[blazing blade]

alr thanks

seems like c die but yea

if it helps 1.4 volts didnt work

at all

[kind walrus]

If it's c die stick to 1.35v and no more

C die can degrade above that iirc

If you mean

samsung c die

[blazing blade]

Ok

Might be Hynix but I’m not sure

Thaiphoon burner said Samsung but when I search my ram stick online it says Hynix

[sudden torrent]

Thaiphoon reads the SPD data off the stick so the manufacturer's usually correct. It guesses on the die.

[kind walrus]

Thaiphoon is often hilariously incorrect

Ye if you can read the spd then that would be right

Manus often change their supplier

You can double check with cpu-z

Cpu-z I'd trust more

[sudden torrent]

It wouldn't be the first time they used multiple different dies in the same model of stick

Especially something as generic as 3200c16 which could be almost anything

[kind walrus]

Anything and not a good bin of said anything either

Sadly this isn't ballistix

[sudden torrent]

I've even seen the most cursed stick ever. 3200c16 Corsair B-die. The worst bin B-die I've ever seen on the worst PCB.

[kind walrus]

We witnessed this together

It ran like c die

Corsair lpx 5.31

Or was it 4.31

It was def a b die

Either way it refused to go above 3600

[sudden torrent]

4.31 is B die

[kind walrus]

Must've been 4.31

@late topaz iirc had the kit

[late topaz]

I've been summoned

Oh ram

Bleh

[sudden torrent]

Was it you that had the Corsair 3200c16 B-die

[late topaz]

Believe so

[sudden torrent]

Well it's gone down in legend at this point

[kind walrus]

They were black lpx sticks iirc

Confirmed the ramblings of a crazy buildzoid

[late topaz]

I was wondering why I was pinged in a server I was dead in

I still have them mainly because I can't get myself to buy better ram

[wheat lintel]

Corsair also does some funny stuff with DDR5 too

Cutdown H16A

[sudden torrent]

At least the PCB isn't awful like DDR4

[wheat lintel]

True

Still find it so funny that they have dual rank 16gb sticks with cutdown H16A

[sudden torrent]

They probably dug them out of the reject bin like they've always done, and just chopped off the bad part

[wheat lintel]

True, the XMP was like 5600c36 iirc

[steady lance]

Been a few days since reinstalling curseforge and chrome. Since then I've had no issues on my OC.

Software was definitely corrupted I think from the unstable OCs haha.

[steady lance]

Welp, as much as I hate to say it I think my Windows might be corrupted.... Reverted back to XMP and still getting rare once in a blue moon crashes,

If everything is running stock (XMP) and I still experience crashes now and then completely randomly is it safe to assume my windows likely got corrupted during the period where I was overclocking?

[sudden torrent]

Generally you don't use your main OS for overclocking for this reason

[steady lance]

Think it's likely a windows corruption then?

Worst case I don't keep anything important on my boot drive so I am just trying to figure out what the issue is then the solution.

Gonna leave my pc on overnight to see if it crashes cause when it crashed, both times one thing that was consistent is that I had the CurseForge launcher open vs all the time that I didn't have it open and my pc was just fine.

[sudden torrent]

It's entirely possible that reinstalling windows will fix it

But narrowing down common factors is good too

[steady lance]

cursegorge and proton vpn have been common factors

[steady lance]

i think i will just do this tomorrow to wipe out a long list of possible issues

[steady lance]

Left the pc on overnight, no crashes.

[steady lance]

gonna reinstall windows.

[steady lance]

Run the highest frequency you can with the tightest timings, benchmark, then compare between the different combinations to see which performs best.

I'd do at least 5 runs on each freq/timing combination of your choice to get an average to compare with.

[kind walrus]

C die usually caps at about 3600

Run it with 1.35v or less to avoid degradation

If you can see higher, play with it, but posting ≠ stable

P95 large or small

Mmm.... I'd still do a bit of a proper test

VST is pretty solid

Ye

I've heard of vram+large before

Though I don't use it myself, feels more like a more outward test than something I would consider to be direct

Tools I use are tm5 for am4 especially

Vst

FFT*

Occt large

Occt vram I'd use more for a gpu lol

Tm5 anta777 is really good

Absolut too

Tm5 I would make the case has been historically the most thorough for am5

Play with 3733 and 3800 tho, test them on a dummy windows

If they work that's an impressive result for c die

Higher freq. Is the most linear uptick for AM4 due to fclk sync

But yeah you should test each, sometimes you can be stable on a higher frequency but less fast

[wheat lintel]

That’s not specifically for ram, that’s testing your fclk for stability

Actually testing ram for stability you’re looking for ycruncher FFTs and VST, TM5 PCBD and Anta777Extreme and OCCT mem test

[kind walrus]

Oh it was fft's I was trying to think of

Not vtt

I knew there was a second test I just can never remember the name

[short blade]

if you want to be really rigorous with your testing you can run occt 3d alongside ram tests so your gpu is dumping heat into your case

not sure how heat sensitive c-die is though

[sudden torrent]

It's not because it'll rarely be above 1.35v

[misty socket]

Has anyone run into the issue where they cant undervolt in Intel Extreme Tuning Utility even after turning on the microcode selection on MSI bios? Im running the one from january 2024 on an MSI MEG Z690 Unify

[kind walrus]

@misty socket what cpu

[kind walrus]

Use bios

[misty socket]

I7-14700K

[kind walrus]

XTU adds bloat

So half defeats the point