#overclocking
1 messages ยท Page 79 of 1
1.6 is fine
I got 3800c13 booted at 1.6 but it errored after 3 mins of tm5
could be possible to get stable with some further tuning but idk
oh I'm gonna drive you guys freaking nuts when I start trying to OC RAM
what levers could I even try to pull to make c13 work
procodt?
assuming I didn't want to exceed 1.6 daily
Are there like diminishing returns on these magnitudes? 3 cores are actually lower once +1 on the magnitude for it
voltage is all you really have
I'll stick to c14
If I put Max CPU boost clock override to 0mhz, isn't this just effectively a backwards all core OC?
but each core can jump to 5.0...
lol AMD is killing me
SOME cores
what do you mean?
PBO
This is where I'm at
what does that means
I just made it up
you know the term "underclock" exists?
lol that's what I meant
I think
I dunno you guys are like OGs at this stuff & I"m on month 3ish of annoying this discord
max temp 68.8
I mean... is this a success or have I just taken a series of facts & strung them together into a giant ball of nonsense?
cuz I'm like magnitude negative 30 on 7 cores rn lol
So I'm gonna bump the Max CPU Boost Clock Override up to 125mHz & bump everything down 2 magnitudes (if that's even a unit of measurement)
which is kinda sitting pretty
1.119v under load is cool but 44.3 ratio is meh
okay prime95 drops the ratios to 37.3 wth
seems like some of the ram errors i got before were due to core 0 being unstable at stock voltage with 200mhz override
and not actually ram errors
i'm seeing a lot of timings tighten that i couldn't tighten before
why is your EDC only 112A?
no wonder it's throttling so hard
you're essentially limiting your 5900x to use less power than a 5800x
soo bump it all the way to 160?
i just let my motherboard decide the limits
140 is the limit
let's see if that'll make prime95 deliver its usually beating
ya I'm gonna change to motherboard forget trying to guess that
or make sense of it
maybe that's what's wrong
I have it on cute little puppy mode
I hate Ryzen Master. I can't change the EDC without restarting to enter Auto-OC mode
if you're just doing pbo, and want something good enough do like 200 ppt, 150 edc/tdc
also for the love of god, uninstall ryzen master
.
lol okay
I think this marks the 4th time I've uninstalled RM
I don't even remember why I re-installed it
It still shows the 112A limit on HWiNFO. I'm gonna restart & see if it goes away. If not I guess I'll manually configure it in BIOS
because you've never changed it in bios, its not going to change in hwinfo
I changed it back to motherboard from the settings I got from RM because I listened to some clown on YouTube
wthhh
I changed it in BIOS
to 150
and it's still 112 on HWiNFO
Ryzen Master ghost
hm
did you uninstall it with windows uninstall or something more powerful like revo uninstaller
maybe there really is a ghost of it affecting something lol
the former
not sure if it would actually affect anything I don't use ryzen master
I'm gonna download revo
still limited to 112A
I guess I have to reinstall RM...
Lol think this will work? https://www.amd.com/en/support/kb/faq/gpu-601
dude if you save your OC profile, the 112A becomes part of it no matter what you change it to. Ryzen Ghost lives in the BIOS settings or something. It took me an hour to defeat that crap
Anyone know if "The application-specific permission settings do not grant Local Activation permission for the COM Server application with CLSID
{6B3B8D23-FA8D-40B9-8DBD-B950333E2C52}
and APPID
{4839DDB7-58C2-48F5-8283-E1D1807D0D7D}
to the user NT AUTHORITY\LOCAL SERVICE SID (S-1-5-19) from address LocalHost (Using LRPC) running in the application container Unavailable SID (Unavailable). This security permission can be modified using the Component Services administrative tool." is a CPU or GPU issue?
pretty sure you can just ignore that
it crashed my game 3x now
& caused reboot both times
wow
and they happen ALOT
I didn't even highlight them all but there's 1059 of them
since 6/22...hmm
are you sure that that's the cause lol
cause even brand new windows installs get thousands of those
with stock hardware
probably not
I see kernel-power errors
Those sound more serious
& have red things next to them
this rig has spent its entire life at stock with ram at jedec 2133
kernel-power just means that the system crashed, on their own they do not indicate anything else
look for things that happened directly before it
I'm just not sure whether it's the mods on the game or the OC that's causing the crashes
I'm leaning towards the mods on the game
because this OC survived prime95 for an horu & AIDA64 for an hour
I know that's not an especially long stress test
but I'd think it would be enough to crash it compared to a game
using prime95 is not a particularly good stress test for curve optimizer
did you try what i mentioned?
which one?
undervolts are more susceptible to crashing at lower loads
and prime95 small ffts is a constant, insanely heavy load
corecycler
oh crap no I didn't know what that was
I did like 5 single-core CBs but I'll try that now
it's a powershell script that automates running a lighter version of prime95 limited to a single core
and it'll cycle through the different cores automatically and also periodically start and stop testing
just to test system stability as it changes clocks/volts
basically simulating normal usage such as starting a game
oh wow ya that's perfect
you can also configure it to use y-cruncher
i did 4 days of prime95 and 2 days of y-cruncher
have not had any issues so far
okay I'm gonna do corecycler rn
the more cores you have, the longer you should test it
since it tests one core at a time
the logic behind is that is that it allows you to test maximum boost clocks for stability
compared to all-core where the clocks will not reach their max boost
it will also report which cores throw errors and stop testing those cores in the following iterations
ah ya that makes way more sense with PBO
any core that throws an error just up the CO value by 1
that's how i realized my core 0 actually needed to be overvolted, corecycler reported errors at +0 and +1
that's good cuz then I can adjust each core instead of taking away 1 magnitude from each everytime
yeah you should definitely do per-core
i showed you my curve right
it's all over the place
oh ya all over the place lol
because i used corecycler to test the individual limit of each core
made me think mine was totally wrong
if i was doing all-core i would have to overvolt every core to +2 and that would be big suck
ya cuz then it's just a guessing game
is this your only system?
i'm taking my sweet time tuning and testing everything because i have another solid system to use in the meantime
which is why stress testing for 6 days straight wasn't at all a big deal lol
no I have this one with the 5900x and another one with 10900K
with the 3060
which should be plenty to play Resident Evil 3
yeah i'd use the other system and make sure you take your time testing
undervolts are a lot harder to get right than manual ocs
I have a surfacebook 3, go2, & two old laptops but I don't think those can hang
ya this has already taken longer & with the manual OC I had no clue what I was doing when I started
also this might sound stupid
but after running corecycler and all that
do some normal web browsing and stuff as part of the stability test too lol
No that makes sense. That's probably like 75% of what I do anyway
I wish these monitors had more inputs so I wouldn't have to swap them lol
i have a monitor inside my pc that's been handy for testing

lets me put it wherever i want
lol that must be a big case
it's a 7" monitor
that's a cell phone lol
more like a small tablet sized screen
but yeah
portable raspberry pi monitor
it's being used for testing now, once i have it hooked up to my main monitor setup it'll just be a decorative panel
does it test logical or physical cores? 6 minutes per core on 12 is awhile but on 24 that's hours
oh okay so like 72 minutes or something
i saw core 0 error for the first time on iteration 26
YMMV
i only have 8 cores so it's 48 minutes per iteration
26th?
yeah
26 iterations on 8 cores is 1248 minutes
one day is 1440 minutes
like i said, ran it for 6 days straight
well it's better than crashing
was checking on the status of my corecycler from 400 miles away
you can also teamviewer without a monitor
when i showed you the distributedcom screenshot from the rig that was running stock
that rig doesn't have any display or inputs attached whatsoever
just teamviewer
does teamviewer work on Mint?
I could throw it on my 2012 HP craptop
and i think mint is a derivative of ubuntu?
oh well that's basically the same thing I think
you can run Gnome desktop on it & it's basically ubuntu
or UI whatever it's called
i've turned on auto access so i don't need password to connect into my own rigs
actually that's probably way more complicated
dunno, i don't have much experience with linux
used mint for a bit just to do some NASM x86 programming for school
it's great for an old computer that's otherwise trash cuz it's so slow with windows
I wonder if they ever did ubuntu mobile
I think android just ruined that opportunity for them
now uses manjaro linux but used to use ubuntu touch
supposedly win11 is going to have android app support
the thing is you can only get it from the Amazon store
but since it supports android apps, you can sideload them
They're doing an Android subsystem the same way they did a ubuntu subsystem
it's not available on the dev release yet
it's like everyone is teaming up against Apple
which is great I hate Apple lol
I let my ex convince me to get one
the XS Max or something
maaan the second I hit 18 months I went right onto att. com and got an android again
It just feels so dumbed down
& like you can't change much on it
Like Apple is like here this is what you get deal with it
The best part of iphone tho is iMessage. MMS has such a low data limit that picture quality is mediocre and video is a joke
just overclock the iphone
can you do that? you probably have to do all that jailbreaking nonsense
I was gonna root my Note 20 ultra but I decided it wasn't worth the mission
I don't think so lol
finally found the real limits of my ram timings
now to see if I have headroom to drop voltage
ayyyy I fixed my mount
Ahh fair enough
Any of y'all looking for 3070Tis, I ordered my 3070 from this place & they're legit.
I was super sketched out handing over $800 to them when I did
It took almost 2 months to get the card but I got it. It was real.
It's in my computer rn
I tried to tell those people over on falcodrin twitch but they all thought I was lying
or ya know, just newegg it
waiting 2 months to then overpay for a gpu isnt really what I call "in stock"
How come you're getting a Radeon card btw?
there was no such thing as the shuffle last year
I don't think it would take two months
because nvidia can suck a fat one
Some regions have stock insted of shuffle
In newegg
Israel store get stock pretty often
I ordered it in mid April & got it beginning of June
& that's before they got their suppliers on point
But hey I'm just throwing it out there for anyone who can't get a card as an option
I make $0 & have no incentive at all here other than when I sent them an email saying thank you they asked if I could help spread the word
They get Radeons sometimes too
Lol nvm
I'm not trying to piss people off so someone else can make money
I've run corecycler 3x. The first two I stepped away from the computer & came back to a crash. No idea which core is crashed on
So now I'm camped out watching the grass grow here
Can you set corecycler to just test one core or do you have to wait for it to do all of them & get to the one you're focused on?
I noticed it doesn't completely isolate the core. Like it says core 5 rn & HWiNFO shows 5 at 48.3 but 1, 4, & 7 jump to to 48+ occasionally
& 0
It's testing Core 11. The ratio is consistently lower than the others & there's a lot of ramping up then ramping down of the fans. but consistently high-ish temps. Think that's an indication of issues there?
You can on jailbreak
Is it really worth it tho? I mean what do you do with an overclocked iPhone?
oh yeah I got that too
I had to run it 2 times
it detected 0 errors once
but then I couldnt open the start menu a few days later
This is where I'm at so far... organization is the key to success right? lol
so i ran sfc /scannow
and it detected errors and fixed em
so I ran TM5 again and 10 errors in 10 mins lol
RAM OC sounds like a nightmare
3dmark/games probably
not that it's possible to oc one
I dont oc mine lmao
its already hanging for dear life
its as worth it as any other oc tho
3dmark so you can say you have the 812th fastest iphone in the world lol
Well maybe if you have an older model but usually mobile games are pretty in-step with the phones they're played on
im almost certain that my chip is water damaged
and that's probably even more true in the Apple world where so many people use the same devices
since I know it is water damaged
how's your battery life?
like I get way lower scores than I used to
not like how long it lasts but on that one setting you can see what the health percentage is
so ya battery health I mean
I suspect a dead core or 2
doesnt matter
Well the phone turns down the speeds to keep battery life more or less consistent
I'm saying if you buy the little $11 kit plus the like $15 battery & swamp it yourself maybe it's worth it lol
not mine
I swapped the battery on the little iphone SE 2016 I gave my 3-year-old & it runs way better
wont make an ounce of difference to my phone
btw I don't think 3 year olds should have phone but I wanted to be able to facetime him whenver I want lol
Im on jail break
I can't believe that thing got an iOS 14 update lol
I have the throttle feature removed
oh
well can the battery still reach high enough voltage to run at full speed even with the throttle disabled? I mean like chemically
considering I most definitely have a few dead cores it should
also its still in the 80s
they might just be damaged
technically your 5900x has 4 dead cores
but then again I think some lower level intel chips are just screwed up higher level chips
with a core disabled
product segmentation is really just dead cores
TL;DR my phone is messed up and I should probably get a new one
I think Apple did that with the macbook air
Hit up asurion haha
so how do I activate those extra cores in the 5900x?
jk
I know I'd need like crazy tools & a microscope or something
it's not like there are tiny little toggle switches somewhere
plus the firmware would be like wtf
it took you 24 iterations to figure out which core was messing you up??
26
you actually used to be able to do this
not for the 5900x
but when amd was making phenoms with 2, 3, and 4 cores
26? holy crap I dunno if I have patience for this. I'm gonna auto-OC I don't care anymore lol
as the process matured and yields improved they had many functioning quad cores but had to meet demand for lower end parts
so they would just disable some cores and sell it as a 3 core
I guess it was cheaper to just manufacture the same thing and disable cores
you could re-enable that core in bios and had a decent chance of it working
well here's the thing
they manufacture 4 core die
at the beginning of the process the yield is not that good
some have 2 working cores, some have 3
so instead of throwing them away they just sell the same package with less cores
yeah that still goes on to this day
have you ever done a de-lid?
the difference is that now they physically fuse off any extra cores instead of just disabling them
yes
oh then I definitely should not do that
you will see much less benefit from delidding 10900K and it's also riskier
Oh @short blade 1usmus is making a new amd cpu OC software
oh no
1usmus's last software damaged rawr's cpu by putting in 1.35v
sooo
idk i'd avoid his stuff
that reminds me, I'm showing 1.456 on these single core tests
that's way more than 1.2. What's going on here?
don't worry about it
when you're using pbo the cpu will intelligently manage voltage
and go down when it's unsafe
that's how it manages those higher single core clocks than all-core oc
Still pissed
with all-core oc and manual voltage you take away its ability to self regulate
which is why you have to stay very low
if you start up prime95 small ffts rn you will see it drop to protect itself
it is recommended to leave LLC on auto for pbo
for all core oc i usually just set LLC to maximum because i'm targetting a specific voltage
Actually I don't want to know. I've already invested too much time into this offset circus
i ran 5.1ghz 1.35v maximum llc on 10700K
so it just stayed at 1.35v regardless of load
ya if you leave it on Auto on the 10900K it'll try to do like 1.46 or something crazy for the half a second before you hit TJ
maximum llc can behave differently on different boards though, some boards have maximum llc actually go above what you've set
which ofc is dangerous
lol wth then why do they even let you set it?
If PBO so brilliantly manages voltage, why do these crashes keep happening?
Why isn't there an OC where you just set voltages for every ratio? like at 47.2=1.171; 47.3=1.174 or something
then it can be auto outside every parameter you set
like who cares what the voltage is at 32
(within reason)
or maybe you could set it like the fan curves
or something
I mean it's time-consuming as hell but so is 26 iterations of corecycler lol
sorry got into a boss fight
because you are messing with it
I wish corecycler somehow graded the core's performance on each test too. I dunno what the metrics would be but if it could somehow say "Completed the test on Core 2 (CPU4). 90/100"
or something
this is basically what curve optimizer does
the cpu by default has a table of voltage - frequency pairs
oh okay. I mean something that's just like an excel sheet meaning if x then y
for example,
1.3v | 4.2ghz
what curve optimizer does is allow you to shift that table
if you apply an offset of -25 you are moving the voltage column 25 steps lower
Ya the dude in the video I watched explained how it shifts.
and attempting to have the cpu do 4.2ghz at that much lower voltage
But he also said find the right amperage before you start & you saw how that turned out
if it can't do it, then it crashes
but is that always like linear?
yes
well then it's not really a curve lol
this does not really make sense because corecycler does not test performance
ah I dunon what I'm talking about just ignore me
you are optimizing the curve, it's already a curve
& well ya it doesn't test performance but can't it like detect if it's close to error? or is that beyond its capability
you know you can track which cores (really threads) causes a crash right?
nope, it either errors or it doesn't
in event viewer??
yeah
yes, core 0 reported its first error at iteration 26
before that there were no errors whatsoever
ooohhhhh mem test iteration
because the curve was almost stable
no, corecycler
it was 1 offset away from its stable point
oh
which is why it took a while to find the error
very intermittent errors can only be found intermittently
it didn't crash either, prime95 just reported a rounding error
Resident Evil 3 can find it everytime I play...
got 0.5 when expecting 0.4 blah blah
Then Event Viewer gives me all that 10013 error nonsense
and nothing conclusive
I mean I'm not even sure if the error is GPU or CPU
I suppose I could jus tturn off GPU OC couldn't I?
i'd recommend ocing one thing at a time yes
Do the tests done by each iteration vary? i.e. if I take an iteration intermission after 4 finishes will I lose out on the benefits of 5 v. 1
no
I thought I had a stable OC for the GPU but now I'm beginning to question it
which reminds me
should I use nVidia's optimized settings for the game?
ya true it never crashed it on my other system
just crashed the game at worse
worst
nvidia optimized settings: "oh, you're getting above 60 fps? let me turn on 5K resolution with 4x TAA"
lol
saw a post where a guy was wondering why his 300 fps in siege dropped to 40s
they're trash if you move the little bar to performance instead of quality?
lol it says I should select "auto" for one setting
auto isn't even an option
I'm gonna try playing wihtout mod manager running at all & see if the same crap happens
@south sky done now thanks for all your help
yeet
Anyone here ever overclock a Cyrix CPU as a complete joke?
I did and I got to 0.19 Ghz at 3.2 V before frying the thing
Funniest thing Iโve done in a while imo
Finally. Core 4/8. 3.5 hours of CoreCycler. I set my phone to record it while I slept. it crashed on 4/8 lol
I set the magnitude back 2 points % it's still maxing 51x on HWiNFO (Core 4)
& not %
I'm gonna keep dialing it back until it doesn't hit 51x
If I'm getting the same ratio on, for instance, core 9 whether the magnitude is 19, 20, 21, 22, 23, or 24, is it likely the most stable magnitude is the lowest? or in this case, 19?
yes
What's strange is, (using that same example of Core 9) if I further lower it to say 22--since it wasn't stable lower than 24--my multi-core CB will be like 25-50 points higher. Is that because less wattage is allocated to Core 9 under load so that wattage can be used somewhere else?
Like maybe to slightly increase the output on a core requiring higher voltage? I dunno
Or maybe that's just inconsistency in CB
Because I'll also see slightly lower voltage
but this kind of fine tuning seems like I'm out on margins anyway & it's kinda not worth it
Like I saw 21,502 v. 21366 on CB by changing the magnitude on C6 from -21 to -22. That's a .994% difference. Worth it? eh
I wonder what the margin of error is on CB lol
Then again, that's the type of thing that leads to an ostensibly stable OC that'll crash later & you won't know what the cause is until you do CoreCycler for 5+ hours
But my bs continues. 1 magnitude increase from 20 to 21 on core 7 just netted me 77 points lol
(back to the 10900K) How come when my core ratio is set to 52, I'm not seeing anything higher than 51 on HWiNFO?
anyone know why my bios is freezing like 2 seconds after entering it now?
only clearing cmos fixes it
but after applying anything it will do the same
using previously stable profiles
I can boot to windows no problem but it doesn't seem to recognize my curve
my boost clocks are lower than before
I have 2 more 6x86s lying around
yes
that sounds like a lot of error
avx offset?
might be unstable is some strange way
Could also be tvb shenanigans
manual core ratio though
Yes
Asus and gigabyte 100% need tvb stuff disabled when doing manual overclocks because it'll drop the 100mhz. Not sure about asrock and msi
If it were that presumably it wouldn't always sit at 51x since it'd be below 70C under idle
No AVX offset but I abandoned the effort because stable below 1.35/no TJ Max appears unobtainable
It'll gsod
gsod?
Green screen of death

51 all core is already pushing it since temps can hit 86
try a heavy load
Lol ya prime 95 will probably make it burst into flames
that voltage for daily is too high anyways
It was stable enough to play 9 hours straight of Resident Evil 2 so good enough for me
What's degradation v on 10900k?
The 51 all core is 1.28-1.31 under load
Max temp 86
(on CB)
R23
Lol sorry I ask anyone who knows what they're talking about
It's not like I'm doing a survey or something. Just interested to see how much variation there is to the answer
it really depends because if you do something like disable HT you can literally run like 10% or higher voltages with the same amount of deg
so there is no fixed voltage that if you go over you'll get deg
What's HT?
Oh right
degradation is mainly temps though
why would anyone do that?
Better performance
Without HT??
Yes
Hmm interesting
Internal silicon temps are really important, and in something a 9900K your internal to junction delta is about twice something like an 8700K
There are just so many damn factors when it comes to degrading cpus
I think llc is medium (aorus motherboard so pretty serious droop)
If you are below both the max current limit and TJmax you will be good for at least 5 years of very heavy load
for a 10900K that's 245A
Current as in amps? I saw that thing pushing 200 today
current limit's can be hit at like 1.25v load
For safety it's good to know current in a max intensity workload, because if you run a few heavy avx instuctions you'll get brief jumps to those levels
Should I set the limit to 240? I think motherboard let's you put 256 as max
no, the limits work in a strange way on most boards
Reason: Bad word usage
I trust the aorus board way more than the msi it was on.
Throttling without an L is a bad word? News to me lol
different word
I enabled something enhancement & it let it above 51 but it was pointless because voltage was really but still nowhere near stable
So I'll stick to 51. Hopefully my last delta fan for the arctic freezer 2 will be here tomorrow lol
43 cfm to 113cfm x2 on 125x25 & 155 on 125x38mm
Use life is 13140 hours at full load 70C (193A) and 30660hours at 30c 30A. Intel has this deg data but pretty much all the chips that die in this time are defects
Doesn't help much tho tbh. The pump can't get the liquid out quick enough
Oh wow maybe I should just drop back to 50 for longevity sake
Or torture myself with another curve optimizer like my 10 day crusade with the 5900x
you'll probably be fine
that's not of 100%
The top of that graph is like 3% failure rate
yeah which is the spec for a 9900K
That's for a 9900K
it's 245A for a 10900K
and 10700K for that matter
10700K's should be tanks
I wonder how much better a basic custom loop with 1 360 rad would cool it v arctic freezer
any recommendations for bios freezing?
reset settings fixes it so it really sounds like instability
instability in bios?
it's my old pbo + curve optimizer profile
i guess i'll try redoing my curve from scratch then
pain
Fitz said something the other day about trying to get 1900mhz FLCK. Is that for CPU OC or RAM?
Btw my 3070 ain't so bad
161 from best. I tried to squeeze more but it just wasn't there
FCLK is the CPU infinity fabric clock
That's the acronym I was looking for. Mine shows 1800mhz on HWiNFO. Should I try to set it higher then adjust offsets?
you'd want it the same as MCLK or memory clock
actual memory clock not transfer rate
Unless you can clock well above or quite a bit below it's better synced
yes
Ah okay & um... Why?
Stability?
Increased speed?
Sorry I can Google it
Or like better communication w/ ram?
Will Google now
Sync all 3? MCLK, FCLK and UCLK
performance
UCLK will be synced no matter what but you want it synced in 1:1 for performance
managed to boot with xmp and original curve for cpu
running corecycler again
if I get any errors I'll know it's degraded
not sure why my ram oc wouldn't work now though
degraded IMC? I kept vsoc at 1.105
idk if that would be possible
Thoughts on the next secondary/tert timing to work on? giving up on primaries for now
tRC for sure
try 60 if no boot try 62
trfc looking high as well
tried 600/446/185 for tRFC a while back and that wasn't stable, could try tRFC again tho, gonna do tRC now
Hm
Forgot what I went by, but I went by a ratio for 1/2/4 that I found somewhere
Currently running testmem5 for tRC 60
About an hour in, first cycle still, no errors yet
Is that vsoc right or is that wrong? 1.1313 is kinda specific lol
Also a bit on the higher side if it is right
trfc is a huge one, but twr, trtp and scls
oh, and leave trfc2 and 4 on auto, they dont do anything
trfc chart
It's prob Mobo doing something weird I think I have my vsoc on either 1.1 or 1.15, forgot.
Hm
Thank
it would be 1.15v soc, there's some droop to zentimings
That makes sense, my zentimings shows like 1.0875 or something when it's on 1.1v in bios
nice
I hope replacing my cpu isn't going to change anything on the ram front
generally just rcd and rp might change
I managed to get 11th place on the 3DMark Port Royal Hall of Fame tonight https://www.3dmark.com/pr/1113088
The 3DMark.com Overclocking Hall of Fame is the official home of 3DMark world record scores.
Also 11th on Time Spy extreme graphics score. https://www.3dmark.com/hall-of-fame-2/timespy+graphics+score+extreme+preset/version+1.0
The 3DMark.com Overclocking Hall of Fame is the official home of 3DMark world record scores.
Using a 280mm radiator submerged in a cooler with ice and water.

Nice work, do you hwbot?
3rd place on HWBOT https://hwbot.org/benchmark/3dmark_-_port_royal/rankings?hardwareTypeId=videocard_3311&cores=2#start=0#interval=20
I don't really think core clocks are a fair 1:1 comparison but yeah amd have better raster perf this gen for gaming just lack the Nvidia features such as nvenc and dlss
Yeah I know I just found it interesting
It seems that AMD GPUs have better raster performance in DX11, but what about DX 12?
How no one broke the record with the 6900xt weird
people have for hwbot
Oh ok
Any recommendations for overclocking on a 10850k with a z590-f and 360mm aio?
firestrike at that point is a cpu benchmark, you'll see all of these tests were run with a 5950x
I heard ram was happening here
BASED
Based fitz recommending not to tune the useless timings
Reason: Bad word usage
SCLs make a massive impact to bandwidth. tune them.
tRFC down to 300ns
also you should do higher mem clock
3800 at least if its Rev E
you'll bneed to increase tRCD and tRP and tRC if its DJR//CJR
It's cjr
hm
@orchid flame How would you tone SCLs, just go down one at a time, or are they linked?
Overclocking recommendations for a 10850k on a z590-f with a 360mm aio? I'm not very familiar with overclocking
check known safe voltages, set to the highest safe, start with 50.0 all core and tweak from there
or undervolt
either works
Due to my 3600 performing worse than a 9880H, I'm going to OC it, any reccomendations on how to start? And my Temps are fine
PBO->enabled
guessing thats in bios?
Yep
for gaming pbo is still slightly better imo
for multi core manual is necessary
6c is going to hit all core loads in most modern games
Is samsung c die and samsung b die simmilar cuz i was gonna do my ram timings but in ryzen dram calculator there is no option for samsung c die
no, they're almost polar opposites
c-die is almost the worst possible
b-die is the best
only a-die and possibly hynix mfr is worse than c-die
actually no hynix mfr can do 3200c16 1.35v pretty consistently so it's probably better
No way
There are many chips worse then it
almost the worst possible that people buy today
not considering stuff like early 2015 ddr4
I'd rather C-Die over CJR
too many reports of c-die not hitting 3200c16 1.35v xmp
user error
Shhh ur making my wam feel bad
i mean i could be confusing it with a different die
but whatever they're putting in all the corsair vengance lpx that's failing to run xmp
C die can comfortably do 3600 c18
then i must be thinking of something else
Time to go faster clock
maybe it's a-die after all
Some CJR (3200 bin stuff) doesn't even pass 3600, and the subs are crappy
my a-die gets negative voltage scaling above 1.25v
oh boy
I replaced my 5800x and I'm still getting freezes in bios on the new one
any idea what could be causing that
reflashing bios didn't fix and windows boots fine
I wonder if it's a problem with my 3080, this didn't happen with my 5700xt
I've also had intermittent crashes with my 3080 installed
maybe i was wrong to suspect the 5800x
guys I have a question on how to overclock
how much do I put my power limit at
and how much do I put my memory clock at
For max performance max the power limit but see if the temps are good
my cpu hovers at around 75 under full load, if that. id be fine to oc basically as much as i want (slowly) right?
honestly idk why it even gets that high with a 180w cooler on a 3600
but yknow
lost the lottery which is why im ocing in the first place
single core is at 7700k level, multi just below 9880H
Start with +100 on core clock and test it with 3d mark timespy, if its not crashing and no arctifacts on screen add +25 to the core clock and do it again till you have arctifacts /time spy crash and then make it -50 from the value that crashed
How much do I put power limit at
Same with memory clock but start with +200
has higher heat density than intel 14nm
What load
ID se-224xt
Basically 100%
Streaming and playing warzone
Used cinebench to benchmark
How many watts is it pulling
Uhh
On r23?
Ok yeah 75c on r23 is fine yeah
I think 1231 on single core
9231 on multi i think?
And multi is supposed to be at like 9485
Guys i need help
1235 so thats not bad
Whats up
Offir said
-_-
oh I thought u werer talking to him
.
.
bread
im playing fortnite rn
Its free on steam
ok
Demo
demo is free
Is free
but memory clock it says current clock at 200MHz
Ah
you sure it doesn't say +200mhz
What gpu?
1070
Screenshot
And dm me
I'm currently reinstalling windows, this is so much fun
Dont go above 1.25V
Switch to manual?
Oc tweaker I figured
Cpu frequency and voltage I'm guessing switch to manual?
And do I change the processor frequency?
Sorry for the ping @lavish tundra want to make sure I don't break smthn if I save
what are the options under that menu
like auto and?
PBO is usually under AMD overclocking
Which is a submenu that's usually in a weird place
you can but that wont OC the thing
put it there
It's on manual
Normally it doesn't show the mhz or voltage
What about ratio
Once I select manual those drop down
Not that I see
Oh I found PBO
Should I enable it and just leave the voltage alone?
Alr
Done
Booting now, ty
Gonna run cinebench and see how that goes
And my windows crashed
FUN
Not even relating to the oc
This has been happening for days
Just reinstalled windows and nothing
What the hell is happening microsoft
average of where?
Basing off of this https://www.cpu-monkey.com/en/cpu-amd_ryzen_5_3600-927
FWIW the cpu-monkey score is using PBO as well
guru3d achieved 3575 with the r5 3600 in cinebench r20 at stock
while cpu-monkey reports 3689
PBO increased it by 11 single core
guru3d reported 462 single core in r20
cpu-monkey reports 472
i've found cpu-monkey scores to be fairly consistent with overclocked results
Alr thanks
it reports 15245 for 5800X in R23
My mc is still much lower than cpu monkey overclocked
mine gets around 15400 with PBO on
did you see what week your 3600 was manufactured on the IHS?
week of manufacturing is incredibly important on ryzen 3000 chips
my ryzen 5 3600x from week 39 of 2019 could only do 3.95ghz all core
It came in a pre built, but its a custom build now
my ryzen 5 3600x from week 31 of 2020 could do 4.2ghz all core
Only parts left are gpu and cpu
the only way to check is to uninstall your cooler and wipe off the thermal paste
not really worth it
my first 3600x got around 3300 in R20
for comparison i was able to push a newer 3600 NON X to 4181
but anyways it's likely you just got a worse bin than whatever data cpu-monkey is using
are you using the stock cooler?
Se--224xt
ah
yeah i'd chalk it up to silicon lottery
i don't think you are necessarily doing anything wrong
Yeah I know
I've been thinking that for a while
Good while actually
Idle cpu is sitting at 3.9ghz
the good news is that core clock barely matters for ryzen 3000
ram and infinity fabric speed is much more important
not ideal but not bad
ram speed has absolutely zero performance impact on cinebench
but it makes all the difference in gaming
i've benchmarked 2133c15 ram vs 3800c14 ram in cinebench, all other variables equal
the difference was 7 points
i mean
consider yourself lucky that it's working
i'm currently sitting here with an rtx 3080 that's crashing at stock
๐
I've dropped over 1800$ on this pc
Bc skytech screwed me over
For basically a 7700k 3060 ti build
you have no idea how blessed you are to have working hardware
i've been getting an absurd amount of defective parts lately
The blood pressure this gave me disagrees '-'
How much was your 3060ti
1275$ for the build from skytech
your cpu is already working as promised
Oh
Cant you sell the other parts tho
Whereas my friend built a full 5800x 3080 build for 1750
Sold the ram and psu
Rest was trash
i could have done my 5800x 3080 build for sub 1500 if i didn't decide to spend so much money on making it pretty
Nope
:(
but hey.. worth
They absolutely screwed me over
No not really
I could've done a full custom off the start with a 2080s and it woulda cost less
i spent over $300 on aesthetics and i regret nothing
Im mining now to fund future upgrade when ryzen 7000 series comes out
Lol i bought corsair aio and fans before i knew there are way better options for less money-_- could have probably save 100$ and have better thermals
Was going to stick with this but now I might trade my 3600 up to a slightly better cpu that isn't the losing side of silicon
But its fine i guess-_-
I just need a decent cpu since I stream
I do
i tested cpu encoding on a 5600x
Still lags a bit when I play warzone if I do max settings
it was completely unusable
over 50% dropped frames
nvenc was perfect
and the 50% dropped frames was only at 1440p 30fps
Now idk how to test if my gpu lost lottery
Honestly wouldn't be surprised if skytech was buying less than par products for cheaper pcs
They legit sent a 400w white instead of 600w gold plus
Biostar movo
Case incompatible with other mobos
test it by overclocking it lol
I have
Can do anything I want with it
But idk how to test how it compares performance wise
3dmark time spy, time spy extreme, fire strike, fire strike extreme, fire strike ultra, sky diver, cloud gate
bought the whole 3dmark suite on sale for $4
Oof
with how much i use 3dmark it wasn't even worth thinking about
i do client builds, 3dmark is my second most played game on steam
the demo gives you free unlimited time spy and fire strike though
Ima just do unigen and im gonna compare online
