#overclocking
1 messages · Page 9 of 1
Is he tryna cool a cpu with ice?
yea thats dry ice

he put it in a metal cup with no interface fluid as well
Need that for my i9-13900ks with moderate speed DDR5 (since I need 64GB minimum for my workflow and ideally the new 2x48GB GSkill kits coming out)
me wen no acetone
is it acetone or am i confused
perhaps confused
Acetone works best, but you can use iso
I’m trying to overclock my ram from 3200mhz to 3600mhz I already know my ram can do it but I need like step on step how to do it
some other good info in this as well
What is the highest possible overclock while undervolting a 13700k
Depends on the bin
All chips are different
And handle voltages/clocks differently
I see I'm still kinda newish on it
Fair enough
How does VRMs affect OCing something, could someone explain? Thanks.
The basic principles of how a motherboard or GPU VRM works.
Patreon and Shirts: http://cxzoid.blogspot.co.uk/p/paypal-safer-easier-way-to-pay-online.html
The Twitch:https://www.twitch.tv/buildzoid
The blog: http://cxzoid.blogspot.co.uk/
The Facebook: https://www.facebook.com/actuallyhardcoreoverclocking
You'll have to tune out the random parts of nonsensical rambling though as you listen 
the rambling is the best part
It shows just how unscripted and stream of consciousness it is. Which a lot of computer people can relate to lol.
I'm guilty of going off on somewhat relevant tangents and ramblings occasionally myself
When I want to learn, I don't like convoluted, borderline useless and sometimes confusing communication. Utterly inefficient
Overclocking server has a fake ping icon as a joke btw. 

Don't even try this
We do know that the 3D cache is very voltage sensitive so that might be what died
Well people are gonna be curious and try this not understanding that load and stock voltages are entirely different things
(both ways)
Ofc 1.3V died
1.3 is high regardless
Yup
3dv or not
I'm totally not running my 5800X at 1.3v daily
💀 yeah but you probably know the risks
Leave it to once again MSI having a means of not giving any damn about the rules
i mean i daily my chip at 1.2v, but have i put 1.55v through it? yeah
was that also at 5.4ghz? yeah
1.55 is unironically not even degredation level
X3D I'd maybe start at 1.1
I've tagged 5.15GHz on mine, 4.8 all core
I'd 100% measure fit sv12 and load vid
I dailied my 5900X on 1.2V for days, but that was cos the ai overclocking was auto enabled for static oc
Thanks asus
11th is probably the most braindead best for newbs platform to get into if they have a little bit of dabloons
Agree
(for oc)
Best for DDR4 at least
does really good at low voltages, and really good at high voltages
I mean zen 4 is supposedly pretty braindead too
fclk decouple disagrees
lo
Ok that I have heard too
Or be me and set -15 all core and call it done
I do have a good sample yes
i have a market segemented 11700k so sugma
My best core liked positive co
Did 5.138 effective
And I remember that number cos I spent ages pushing it 
Ya it was like, almost +200
yeah, 5ghz zen was kinda like what to expect right?
Ya
if you had a high bin
Even a regular bin
A lot of 5600x's you could dial +200
Then slap on some half baked co
4.85 all day
There would be potentially a small factory increase to the boost bin if you get a good sample
So technically the max boost on a 5800X is 4.7 but mine would do 4.9 out of the box, +200 gets me to 5.1
CO gets me the rest of the way
I never really did any extra push with bclk
It's possible there was more
I did up to like, 101?
Got scared for my ssd went back lol
I know mine crashes if I try to push more than 2 cores past 5.15
I'm at my limit for ambient
i just sit at 4.9 1.2v
I know you don't mean all core but that would be awesome if it was
thats all core lol
not zen
11700k
103+ hurts ssds
You can't compare clocks across different architectures tho

+100 core +1000 memory a very dirty oc
I finally took the time to do a per core CO.
I think I got a good sample. This passes 12H of corecycler, P95 smalfft, and linpack.
Ive yet to take the time
I'm really not looking forward to that.
😭
It has taken the better part of 2 days including testing
have you left it at idle?
3 days at idle revealed some wheas that corecycler couldn’t catch for me
Yes
Core 0 is my "best" core too so if I wanted to push for max clocks on single core I have a leg up, my best core can be the only one active with the others disabled
Have I mentioned that I LOVE this cooler? This is during R20 all core
yes air
isnt 2040mhz stock for most cards?
that’s not real boost lol
stock 2055mhz
Melty Ti?

Melty stonk
2055-2100mhz is fairly easy at 1093mV, not sure how much so without power limit bypasses tho never tried that
What's the highest you can go on the memory?
my mem is pretty bad, below average
7900Mhz is max i can fully stabalise without weird graphical issues
average should do 8000mhz
my core does 2175mhz stable on water, 2205mhz benchable in everything.
my 2070s doing 8200 in its sleep near 16 hours a day 💀
Quote the gpuz clocks instead, less confusing
fair yeah
Somehow squeezed in a top 10 for the 2.5B stage in the gskill comp benching last minute. https://hwbot.org/submission/5244373_
Nice
micron? 💀
no just very doodoo samsung
Rip
had my cpu jump to 2889 mhz for half a second lol
gonna start to try and manually OC it. timespy score is lacking 😢
Manual oc all core for TS and most benchmarks, CO negative offset + positive fmax for daily.
I did manage to get my CO tuning within 75 points of my best R20 run but all core still wins
Timespy likes flat clocks
ye for temp stuff like benching it's fine
Trident Z5 RGB series DDR5 memory are the latest G.SKILL flagship memory kits designed for ultra-high extreme performance on next-gen DDR5 platforms built with 12th Gen Intel CoreTM processors and Z690 chipset motherboards. Featuring a sleeker and streamlined aluminum heatspreader design, availab...
ratings?
Meh
This is the kit I'd get
Both will do 6400
Can't imagine that'd be anything but m-die, yeah?
I think you have a chance for a-die at 6400? and higher gskill if buying new
Can tell from the serial
While we're on the subject, a lot of people have told me G.Skill no longer uses m die, would you know if that's true?
If they do that's an m die sku
Ye
I've only heard about what kits are guaranteed a-die vs maybe a-die
Hm ok
If m die still exists that's 100% an m die sku
If I was looking for A die I look for skus released post like october 10th
Possible that as time goes on, no skus are m-die
Ye
There's also ripjaws s5 K1 and K2
K2 is new and basically just a die afaik
I've only seen it on amazon listed that way tho
Is overclocking ddr5 worthwhile? Don’t know if it gives meaningful performance uplift
Past XMP :P
Gonna look into it tomorrow then! If you have any resources that are useful, would love them :)
Trident Z5 RGB DDR5-6000 CL30-40-40-96 1.35V 32GB (2x16GB) Intel XMP Trident Z5 RGB series DDR5 memory is the latest G.SKILL flagship memory kits designed for ultra-high extreme performance on next-gen DDR5 platforms. Featuring a sleeker and streamlined aluminum heatspreader design, available in metallic silver or matte black, the Trident Z5 RGB...
This is the kit I have now
Good kit
Is ram like CPU/GPU where there’s the “lottery” or not? I know that there’s different memory chips(?) that are better than others depending on brand or company but that’s the extent
Yes and no
You're mostly board and cpu limited with hynix ddr5
But easy check which type of hynix you have - take a picture of the barcode sticker on the ram and send it here
There's something called an 042 code on G.Skill sticks and if yours ends in M or A it will depict which version of hynix chips your ram has
On 7700X this kinda doesn't matter much? But there are subtle differences
Ends with M!
This kit was marketed for “intel XMP” lol. Funny
Meh, xmp are just simple numbers
Intended for a wider range of compatibility
Properly tuned ram will stomp any simple xmp
9 times of 10
Is temps ever a problem? I’ve seen mine get to like 65-75C and with actual like manual I expect it’s higher lol
RAM waterblock ⁉️
Hynix m yes, gets pretty hot
I'm not sure for AM5 though, cos can really only push to like, 6200/6400 really for some sensible daily
CL28? It's not much of a leap from the xmp for things that might need voltage
I'm not really the best source for AM5 tbh sorry
Lol ur good. I’ll do some research tomorrow, not sure how big of a change it’ll be since after manual overclock CPU vs PBO and CO the difference was marginal
Just more stuff to check out 😄
11 time out of 10 (because there's always that 1 kit that has to be RMA'd)
you overestimate xmp if you're only saying 9 times out of 10 
Well I've never tried it but I would wonder if 5333 Kingston ddr4 xmo would be slower than 2133 tuned
Similarly any horrible micron A ddr5 5400 tuned vs like, 8000 g.skill xmp
Just keeping that margin there for some insane extreme I mightn't know about 
2133 tuned 💀
I interpreted it as manual vs tuned on the same kit
then again
my goal for this ddr3 I'm working on right now is to beat ddr4-3200c16 xmp
should be easy
That would be dope
i'm pretty sure i can do it
or at least get very close on bandwidth
but much better latency is guaranteed
although i suppose it's not really a 1:1 comparison since different cpus
this ddr3 is now doing the thing where it’ll pass 128 cycles pcbdestroyer, 8 hours VST, and 3 cycles of absolut but then the exact same profile will inexplicably fail pcbdestroyer in 30 seconds the next day
I love ram oc
i buy random crap ram to oc for "fun" and complain about it when i'm supposed to be having "fun"
i fell off the edge of stability long ago
Missed the gskill comp deadline by a few minutes, but I'm happy with a #5 wr using a mem pot I made from scrap copper sheet.
https://hwbot.org/submission/5247306_
Probably not.
Ah sadge
We'll see how many of the germans get dq'd for the r15 shennaigans
I got in a 5143 sub before the cutoff, but that's 14th place in the stage
5413 would've been second
Oh you still subbed? Oh that's at least got you running for a ram kit
Ez moneys
Hope you do well tho, massive props now that I've seen the amount of time and dedication that gets really put in to this stuff
(It's a comp ofc but well, still surprised a bit lol)
Might as well, it's a #5 WR and 107 globals.
cheapaz chipz when (i am broke)
Yos seems to want to go with 5870, or some other 5xxx card
that'd be fun, cheapaz is my fav comp
hope they pick something like tropics like 260 comp, purely gpu bound
There are far better choices. Heaven, vantage GPU, ts GPU for dx12, etc
fair
i just wana purely do gpu oc and modding, i tend to find gpu modding more fun than the actual benching
@mental turret I think 2800 was the problem
I'm actually getting way better performance at 2666
testing CR 1 and tRDRD 4 right now
was 2 and 6 at 2800
tRRD/tFAW passed overnight at 5/20 instead of 6/24
each cycle now takes 2.75 minutes, down from 3.61 at 2800
goddamn i just manually tuned ram and its like 7% faster vs xmp 😭
only 7%? you have more work to do then
Keep going 
i just followed someones guide and they said it was pretty safe + they had A die
even the most bottom of the barrel kits should be able to do 20-30% better than xmp depending on the workload
so i can try to make it a lil better
Well you had what, 6000 C30 before right?
yep
That's a decent increase for ryzen
i didnt adjust the speed just timings
even in kits where the primary timings/frequency can't move, most of the performance is in the subtimings
Timings is about all you can do
ill just uhhh move each one down a teensy bit one at a time
That's the idea
Slowly, carefully
Keep a backup of your windows install while you do this in case something goes way wrong
Yeah it's fun seeing genuine improvement
gigabyte bios is so funny
ive seen like 8 typos, 5 or so "no help string"s, and 3 options that lead to nowhere
Bro mine wouldn't even do 2400 with loose timings
Gotta get more dimms

😭
It's purty tho so that's all that matters
my 4770K couldn't go above 2400 either
somehow I got my XTX to idle at 30W? vram speed isnt like always maxed lol
oh wait
i unplugged a monitor so its just one lol
💀
Ya and if you run that second monitor plugged in at a low enough bandwidth that idle will still stay low
Like 1080p 60 or something basic
Otherwise it's a toss up with vblanks as to the idle draw with a lot of people
And ye higher hz means more memory bandwidth
ill get a second gpu for my other monitor
oh yeah daily i just run all core CO pbo +200
Ye
and it never gets over like 65 in gaming
You'll need to ask someone what zen 4 static voltage is good for benching, I wouldn't know
(goes double when I have a 3d chip lmao)
You probably can
didnt even change Tcl Trcd or Trp
but those are super sensitive i think
not quite sure
I had fan speed 100% on everything 😢
ill point my fan towards my pc lol
my CPU seems to not wanna go over 5480~ sustained
Leave ac on overnight at 16c, turn on pc as frozen monolith
Freeze self
But succeed in oc

ill get some blankets
itll be worth
energy bill will ruin me but for the OC its a must

It barely crashes at 1020 mV…
if i just keep running it…
I’ll get lucky eventually
1020 seems a bit too low for it but I might get a middle ground
Default, 2770
Haven’t played with VRAM yeah
Def play with it
Dial it up
Big gains in vram
Try 2784 normal timings
Specifically this number
Why so specific
It works in steppings
Ah fair
2784 clicks up the vsoc to 1.5 from 1.0 iirc
Vsoc? Memclk?
I can't remember the specifics
Potato potato 😉
I think I won the lottery with this card tbh since I haven’t had to do some insane tuning
I’m running that rn
And fast timings are useless correct?
Correct
I can already see a minute FPS difference
On air that would be INSANE
It's possible
mfw overclocking is more fun than actually playing games
I mean tbh these cards are basically waterblocked with their abilities
Yeah I’m just running it at max fan speed rn
Coolers thick af
And it stays under like 65
Yeah, red devil is particularly thick
Core probably gets to like 80
Got an insane open box deal on it
Had to pay like 50 extra vs a XT
Niceeee
Graphics score didn’t change much but imma just keep going
🙏
2804 seems to be stable
Keep going but add a touch of voltage
You can also be stable but perform worse due to artifacting/minor ecc
Yeah
Play with it a bit
You might be tapped out on vram
2770 is decent
If not 2776
Iirc that was the step
I’m gonna fiddle with it and see
Mine likes 2784 mem, 1.031 vcore
Normal timings
And spikes to like, 550w constant lmao
I think I got 34.3K-ish too but I haven't really properly played with the card yet

Gonna make a custom loop someday just to beat this record 
Dunk rad in ice water 
Genius
How’d that work out
I want to do this so bad

Watch out for condensation if you do
Seems like I am winning compared to my friends 3070ti and 3500x
Barely, if any lol
That's a non-zero chance of killing your CPU still if you don't watch for it
water collecting on the board from condensation isnt even that dangerous, unless you board is super dirty or something
not like there'd even be enough condensation for that to be a concern 🤓
Idk, some people live in the swamp
If a bead of even pure water bridges a few volts, stuff will corrode quickly
yea it depends on environment for sure
Yea ik lol fans and shop towels
My rooms dry tho so we'll see how it turns out
Thanks tho : )
my z690 corroded overnight
Sorry what
Did you soak it in chlorine wtf happened

im pretty sure it was already somewhat corroded from lilene owning it, just it was under heatsinks so never noticef
Oh
my loop leaked lol
Nooooo
gpu stuck in 8x tho, so i think some corrosion under pcie area
Omg
im just too lazy to cleanup atm coz 8x dosent really affect 2080ti
So it's still good? Or concerning
works fine
done 12100f, g6900 ram and ur 12900kf oc on it so far
950mV really makes non k ram oc a snooze
That KF sounds a little degraded
it probably is, lilene was using 1.55v and up
Might still be possible to run the same clocks tho
idrc, it's a daily im gonna run stock lol
gonna degrade lilene's 13900k as payback 

i don't think this 4790K is a very good bin
maybe needs a delid/better cooling, dunno
seems to not be capable of 4.6 @ 1.3 stable
hey guys question here I have a day one 3080TI FE what mm size thermal pad do I buy
thanks
I am gonna try my hand at Over clocking tomorrow, should I try CPU or Memory overclocking first?
CPU: R7-5700G
RAM: Oloy DDR4 3000cl16
Memory would be downright boring. If you're lucky you get 3600 c18. Unlucky, 3200c16.
The CPU is fairly easy to overclock too. There's the basic PBO enable that most would be good with, but PBO advanced tuning gets the most gain.
so cpu first ram second?
any good tutorials on OC you recommend?
If you know your way around the bios, https://www.reddit.com/r/Amd/comments/qik4t3/zen_3_pbo_and_curve_optimizer/
CO tuning is VERY time consuming, but might be necessary to get the full benefit of setting +200 boost override.
Only way to know if you need to is to test though
@mental turret time to reboot and see if it passes again
if it does then these are my final settings
tRDRD 4/5/5
tRDWR 12/12/12
tWRRD 18/5/5
tWRWR 4/5/5
tCKE 1
tREFI 24576
think that's all the timings that aren't visible there
good enough
what’s this r20
i imagine i’d be close to my 2700k cause of the clock diff
4.6 vs 5.2
but avx2
13600k 5.2ghz 1.17v on p95/occt
also 85c but tbf i turned my fan curves down
1.17 bios set or load?
load
👍
40w idle? damn that's a lot
i have a yt stream with hw accel off, discord and a vm using 2 cores rn so i wouldnt really say idle but normal use ig
mb
mm yeah that's not idle
Hey, just OCed my Asus Dual OC 3070 to a 3d Mark Speedway score of 3905 (from 3426 stock). +240 core, +1210 memory. Does that seem like what I should reasonably expect from this card?
With my CPU and any 3070, the listed top score is 4010, avg 3500.
in r20? sounds like r23 lol
even with avx2, the 4790k is not a 3x increase from the 2700k 💀
Well don't stop now, keep it running and see how long until you calculate the final digit of pi
Not going for actual benchmark scores, but haven't been able to get 1T to work quad channel high freq yet
The s2d is definitely easier to work with than the gdie though.
I'll do secondaries and terts as a kit of 4 eventually. Kinda insane to waste the time doing them 1 by 1
You just need more power
Did it... get worse at 1T?!
Probably just a fluke, I'm retesting it rn compared to 14tras at 1t
Okay, 13 tras is the issue
Not surprising
tRAS? More like trash!
Gottem
I really need to go to bed I'm starting to worry myself
ah ok that makes sense 💀
What chip is this on
this is just on x79
Sandy bridge 🗿
no
No?
ivy
yes
You didn't sleep last night huh
He did
I saw the sunrise this morning 😭
Listening to katy perry
y
Did you listen to california girls
The vid I posted
Oh most definitely
nope

Far I'm setting up a rl craft server
Now that I got the port forwarding figured out
when will it be playable
Can I have ip
Sometime today
baller
Maybe
What's the specs of the server going to be
Haswell🗿
?
Or I can run it on my 5900x but I don't have enough ram for that lmao
How much ram u got
Oof
With that much lol
i used to play and host with 8gb ddr3-1333 jedec
Fire
How
20 player server
Lol
i5-2410m or something
Probably not rl craft
Apparently rlcraft needs like 8gb alone
this was like 10 years ago
Lmfao
To play smooth
back in middle school with my school laptop
Mods weren't that big back then
also @ancient bridge i'm subbing my 4790K scores now and damn they suck
bad bin and bad cooling
😭😭
ok this is not true
i remember back then still
my server was still very very different from vanilla
I don't realt remember any big mod that early tho
just didn't install too many performance heavy mods
maybe no single big modpack but mods were def still a big thing
Main thing with rlcraft it's textures are rough
Lol
whoa, how is this worth globals???
did i never run y-cruncher on any of my 7700Ks?
I don't understand what's the point of overclocking
most of my subs on this 4790K are literally worth 0 hardware points so i'm surprised that this is worth globals
fun
well, ram and gpu oc are practical for gaming
cpu oc is basically just for fun
Is it a lot more performance I was told it tears your cards up usually
you were told wrong then
Oh
if you know what you're doing then overclocking has no impact on your hardware's lifespan whatever
there's even a type of overclocking that increases hardware lifespan
When you OC are you just giving your card more voltage?
not necessarily

i cut my 3090's voltage from 1.1v to 0.825v
while maintaining the same performance as stock
what makes you think it's a good bin
bad cooling is not debatable i literally slapped a $20 air cooler on it
Well that would be undervolting
undervolting is overclocking
Don't you run the risk of shorting your card then
...what?
shorting is something that occurs strictly as a hardware level as a result of hardware damage
i don't know where you are learning these thing but do not go back there
Probably thinking of so.ething else if your parts ain't getting enough juice
if your voltage is too low it will just crash
then you up the voltage or lower the clock and try again
I'm only in my second semester of IT at college so I'm still learning
i found the stable clock at 0.825v to be 1800mhz
this has nothing to do with IT
your boss will never ask you to overclock or undervolt a dell
?
linda from finance will never ask you for a liquid nitrogen refill
Why not tho lol
💀
It's a joke
We never touched on it in class and my dad tells me not to do it because it tears the equipment up
So I find this stuff interesting
I overclock everything I own
even my laptop has overclocked cpu, gpu, and ram
no issues with any of it
I was curious what I could do with the OC version for my 7900xtx
there are a lot of good guides out there to get started with
Which one? I might have the same model
And I can help either way
Ah ok well if it's not built don't worry yet
But I have a 7700X and 7900 XTX, overclocking can do a lot to help performance in whichever way you'd like
Better temps, better frame rates, and/or better noise levels are the main changes from it
The XTX is fairly easy to overclock
From my experience at least
Have you had any driver problems. I know at the release there were some idk if it's been fixed or not
From what I've heard major issues are fixed but overclocking is still touch and go
Oh okay
At least adrenalin isn't automatically overclocking your cpu any more 🙃
What's adrenalin
AMD GPU drivers
In certain games there’s minor visual bugs(one game)
But that’s it
Otherwise super simple and easy to use
The best part of that
Technically they can't prove you overclocked it but it's still warranty breaking officially
You can achieve some crazy benchmarks for overclocking but day to day it’s minimal improvements
💀💀💀
What a funny story
I didn't hear about anyone being denied warranty service because of it at least. It's extremely hard to prove the CPU was overclocked.
CPUs rarely die even with PBO on to begin with
I am gonna try overclocking and I noticed this is for the 5600-5950x not R7-5700g, would it work the same?
More or less the same concepts yes. Just slightly different actual values.
Nope
Oh, so setting it to high and I might fry my chip?
It does have a max temp value but you can absolutely fry it with too much voltage
Got it, so I need to be very carful with voltage then.
CO will be undervolting it so no worries there, as long as you don't manually control the voltage
The GPU market is STILL a disaster BUT there is still hope! Buying an APU CPU as a place holder while the graphics card GPU market issues boil over is a good option!
✔ Yes you can build a $600 Gaming PC in 2022!!! https://youtu.be/O6zRmpJrwDg
🔔 Subscribe to PC Tech Hustle: https://www.youtube.com/c/PCTechHustle
HERE is a guide to squeeze ou...
Any issues with using a video?
I tend to be able to follow along better with videos.
Haven't seen it so I can't say either way
dont follow this as hes setting a static oc on it
thats not good with ryzen unless you're benching
just turn pbo on +100mhz or so depending on your chip
my 5700g was able to do +200 but they all differ
So keep PBO on?
I am about to start overclocking as I just got done installing OCCT, CPU-Z, and HWINFO.
cool
Would you mind guiding me through this process as you've used a 5700g before?
I'd really appreciate it!
yea sure what brand motherboard do you have so I can get an idea of the layout
B550M DS3H and I believe I was told it's the v1.4
Got my phone on, now opening up bios.
In bios.
does it look like this?
Extremely similar.
good good
yea no prob lol
okay in the tweaker tab click on precision boost overdrive and makes sure its set to advanced if you want to try and tune it
Done.
otherwise if you just wanna "set it and forget it" make sure its just set to enabled
I want to tune it.
okay see where it says pbo limits?
Yes
try +100mhz for now
Set it to manual?
yes
PPT LIMIT?
leave the "precision boost scaler" as is
gotcha.
show me your screen
I
oh I see lol set it back to enabled or auto
You should be able to adjust your max clocks in boost clock override
I'm not familiar with this bios so just got confused for a second lol but we're essentially allowing the cpu to boost to a higher frequency for longer
below curve optimizer
all good
K.
try negative 15 on all cores for now
undervolts the cpu so lowers temps and power draw and sometimes increasing performance at the same time
Oh, I thought you needed more volts for higher clocks.
So -15 means minus 0.15v I am assuming?
okay also is xmp/docp enabled?
Let me double check.
sometimes yes but not always
the equivalent of xmp just allows your ram to run at the rated speeds
open occt
set these settings
Yeah
start and check for errors
Small
Extreme
Variable
1
Auto
Auto
yuh
Started.
My temps are so low.
good lol
Probs due to the Push/Pull LT720 with P12s.
How long should I run OCCT?
Average core temps are about 40c.
Oh just noticed that it's an hour long test
So far 6/8 of the cores have managed to hit 4.7ghz.
dont have to run it for that long
im sure @short blade has better ways to test stability
Is 4.7ghz on 6/8 cores good?
yea not bad, might be able to push further
I am assuming the reason the other two didn't hit 4.7 is due to bad silicon?
Think 20m should suffice?
I'd prefer very stable as later on after the next upgrade I am gonna reuse this CPU for a server.
Thought games were considered a bad way of testing stability?
Ahh okay, would it be fine if ai try synthetic benchmarks then once I hit what seems to be the CPU limit then run games for stability?
yea thats fine ig lol
How will I tell if a core fails the stability test? At the end of or during the session?
Where will I see the errors?
Where it also says when it started?
yes
sweet
like I said my 5900x was able to do +200 fine
so we'll see what yours can do
Hopefully I can hit +4.8ghz all core.
if you're using PBO, you should be mostly testing with corecycler and leaving the PC idle
if you're using manual oc, stop using manual oc
hes not
i made sure of that lol
Thanks again for helping me out Danny.
OCCT/multicore loads in general are not very useful for PBO stability testing
So I should use CPU-Z or Core Cycler?
cpu-z is not a stability test
I was gonna tell him to do single after but you know me if its stable in game its stable to me lmfao
2 days of corecycler and 2 days of idle overnight, check for any wheas
you can just do it overnight and close it during the day
i did a lot more than 2 days but

you should probably do more than 2 days if you prefer "very stable"
Oh okay, that's a lot more reasonable. Thought I had to do that like For 2 days straight.
I did it for a month straight lol
I wasn't in any rush to swap pcs
Okay.
In the mean time what are quick tests I can run while I check what frequencies I can boost to?
with pbo you'll see the highest frequencies at idle
just check frequency in whatever your most common workload is
gaming, encoding, whatever
Any synthetic tests that can do the trick?
/s ?
nah I was pretty much just too lazy to swap my 5600x rig with a 5800x rig right away
I don't understand how those two are connected
I was using a 5600x rig and the 5800x rig was downstairs running corecycler
basically just bought the parts for the 5800x because I found some really good deals
I ended up selling the old PC for more than I spent building it
buyer offered me more than my asking price, not sure why
anyways, 2 days is a reasonable amount of time to call it good
just make sure you check for WHEAs, not crashing isn't enough
Just did:
+200 MHz
Reverted to default voltage
Time to run it through another round of aoCCT while I eat.
+200MHz doesn't always net maximum performance on zen 3 btw
What does then?
Why is that?
you'll generally need smaller offsets to be fully stable at higher overrides, which will lower performance in heavier loads where the +200 override doesn't really matter
because you won't actually be hitting +200 but you're still stuck with the smaller offsets
if you're going to be running all-core workloads using it as a server, you'll probably get better performance out of a lower override
What do you mean by offset?
curve optimizer offsets
Oh. So you meant that it won't always hit those increase in clock speed and still be stuck at the given voltage?
how do i put this
my 5800x needed +5 to core 0 to be fully stable with +200 override
it would hit a maximum of 5.05ghz at idle
but all core clocks would only be 4.55-4.6ghz
if i used an offset of +0, i could have run more aggressive offsets and gotten an all-core clock of probably 4.65-4.7
So extra voltage is needed to be stable at higher clocks for single core but it would sacrifice all core performance and if I used a lower voltage it would help for all core?
So I should Undervolt for better all core?
curve offsets = undervolting
you don't actually want to apply a negative voltage offset on zen 3
you want to use curve optimizer to do it per core instead
Even if it's for an all core boost?
yes, because each of your cores have different silicon quality
for example, my curve was
Offset: +200
PPT: 160
TDC: 140
EDC: 140
Core 0: +5
Core 1: -18
Core 2: -23
Core 3: -13
Core 4: -23
Core 5: -15
Core 6: -18
Core 7: -16
Ahhh okay.
So I should try +200mhz then configure each core's voltage till it hits the extra +200mhz boost?
how much time do you want to spend on this
Do +200 and run corecycler. Any threads that crash, do a positive offset, they need more volts. The threads that don't crash start doing a negative offset. Start around -10 for the good cores and test again. If they crash, start edging back to 0 until it's stable. Otherwise you can keep going into the negative, testing each time.
This does take a LONG time. It took me about 2 days including testing time.
A single pass of corecycler takes 10 minutes per core. Once you think it's stable you'll want to run it overnight.
Got it, so each session will take 80m for all cores to pass correct?
If you let it test each core each pass yes
Is there a faster way?
If you get a couple threads that crash then you can just stop the test and adjust in bios before running again
While you're finding the stability point you might just do a 30 minute run and get 2 threads that crash, that's when you stop and adjust
Maybe +5
And if it's still not stable go higher
this one?
yes
it still says starting?
Where?
top left
Oh okay, how will I know if a core passes or not?
Will an error message show up at the bottom?
It'll tell you yes
That's fine, it's still running
right as I said that new lines showed up
22:59:17 - Completed the test on Core 5 (CPU 10)
22:59:17 - Set to Core 4 (CPU 8)
Running for 6 minutes...
So no issues then?
on Core 5.
No immediate issues at least
Stable enough for now on that core until the long test later
Aight.
After +200 passes on all the cores then I run it over night and if no issues occures I should then try +300?
There is no +300
You'd go into curve optimizer and tell it to do a per core setup
Negative offsets for cores that don't crash, positive for those that do
When you're entering numbers in the fields, it's always as a positive number. The negative sign is applied automatically for a negative offset.









