#overclocking
1 messages ยท Page 78 of 1
Intel mlc
I use the MSI MAG X570 Tomahawk and apparently on that board "Game Mode" locks the CPU into a 600MHz overclock
Even stranger, when I ran CB R23 at "Default" on Ryzen Master (PBO settings enabled on MB, which I think Master must have overridden once Auto-OC selected) it says max 5150
the window just closed and I'm not sure if it just finished and closed itself or it crashed
gotcha I'll put it in a batch file so it doesn't close automatically
But yeah can't use aida since the latency test is reads (also not that useful pulling like 2.5GB/s)
Buuuttt CB scores at Default, Auto-OC, & game mode respectively: 21463; 21485; & TBA (3:30 left)
what program is that?
I see
But yeah, dark blue is write, lighter blue is read
Th other thing that's probably not good is Game Mode is showing an average core voltage of 1.422, which exceeds the 1.2 degradation voltage I learned in here
Using read-only traffic type
how do I change that
whoa y'all called it, 11254 CB down over 10k! wth
command line argument?
I might be dumb but none of the provided command lines look like they'll change that
anyways tCWL 9 finished but definitely looks like there were errors
let me try it
well it didn't complain invalid argument that's a good sign
it's still saying read-only
documentation says you're right
okay after adding the --loaded_latency argument it works
tCWL 9
tCWL 10
tCWL 12
and finally tCWL 14
what do you make of those numbers @south sky ? i have no idea what inject delay is so it's hard for me to say which is the best
tCWL 9 looks the best but the variance is very high for some reason
Well should i try tcwl at 18 anyway as it crashed while running tm5 at 16, but i had just changed trfc from 630 to 600
arshia suggested changing those values
but changing em gives errors
any other values I need to change beforehand?
yeah the high variance confused me as well maybe i should run more tests
how high can you typically get memory clock? for gpu
for 30 series cards, i guess how high is too high i wouldnt wanna mess anything up lol
how does this look and what can I improve?
dont increase the core voltage. go +50 on core till crash
then go back to the last good value
try +25
if that works use it otherwise stick to the last good value
same for mem
question on the mem, i was doing that last night to my 3070ti and it was stable up to +1700 i think, is that normal?
it seems too high but i could be wrong lol
monitor what?
voltages
did you raise them over stock?
well how do you make sure theyre safe, just dont mess with voltage or?
is buying the full 3dmark worth it?
nah not really
demo has firestrike and timespy
dont mess with em imo
is is 85% off so $5 Imma get it I think
if you need it sure but theres really no need imo
I guess I won't get it if it is probably pointless for me
Two more delta fans. Vroom vroom
It sounds ridiculous in here
There's like an EF-1 in my case rn
Maybe I should swap the front intake fans too. Does delta make 200mms?
I could run two plugs out of the case & connect some 200s at 120v. 370cfm lol
When you get that high use 3d mark timespy to bench, haven is too light
Just don't touch the voltages unless you are undervolting
And then it's a graph not set voltage
Nah heaven is heavier mate
from experiance
My extreme OC profile hangs on heaven
works flawlessly on timespy
What? Nahhh heaven is just an easy one you can use and on the fly oc
My card disagrees
If you're on nvidia, ts will pl throttle down more than heaven because heaven is lighter
Im on amd
rx570
still holds true
@lavish tundra
all cards that have power limits it applies to
(that hit the power limits)
if you don't have anything limiting clocks but you are getting that, well that's weird
Ill test that out with hwinfo and let ya know
guys i put XPM on profile 1 in the BIOS but still when i start my PC and open the task manger it shows that my RAM speed is 1600 while in BIOS is shows 3200
is there any other setting i should do to make my RAM run on 3200?
Check what it says with CPU-z
it likely is at 3200mt/s which is the same as 1600mhz
DDR stands for double data rate meaning there are two data transfers per clock
so 3200 xmp ram is technically 1600mhz
@south sky CPU-Z says the same as task maneger
it's 1600 MHs
CPU-z reports the correct speed
ram manufacturers ratings for memory technically aren't correct
damn, 1600MHs is some nice ram, generally its MHz, which is 3200 speed ram
kk thanks guys
Haha those motors are more like 750w
How would I go about Overclocking a RX-5500XT just to get a little better performance out of it
where is a good place to start learning the basics of ocing my cpu
@pseudo hull See Pins in the channel there is a OC guide for Zen
More like a chart of tested settings. Its helped me when I had my R5 2600X
For a 5600x just do manual pbo settings
And/or there's a curve customiser thing
Msi afterburner is easy to use
does pbo stand for something I haven't found that
For gaming the core clock is more important than memory, so its better to oc that if I'm correct. Just make sure you have a good psu and extra power to use for overclockimg, so you don't fry your pc
And that it doesn't overheat
I am gonna wait till I get my extra fans that come today
What temp does your cpu run at ?
37 with discord and ryzen master
Ryzen master isn't good to use for ocing I've heard
You should use bios
Also I meant under full load, or atleast under a good bit, so gaming or something along those lines
What cooler are you using?
fuma 2
Yeah it def shouldn't be running that hot, do you have any case fans next to it? If so make sure you don't have them going the wrong way (intake vs outtake), I did that and my 3600 was getting like 87 bc of it
Fixed it and it dropped to around 70
You'd have to ask in #old-cooling that I forget which way they're supposed to go
Coofor?
Tf happened there
I have the 4000d airflow with the two included fans
There we fo
Wow I cannot type
Yeah I'd definitely wait for your fans before ocing
they are going the right way
And check Temps then before ocing too
Precision boost overdrive is pbo
Auto oc is worse than pbo btw
Sounds right, 5000 series is v toasty
5600x is supposed to be the low power usage one tho
And with the fuma 2
All 5000 series are v hot
@zenith palm
If you go advanced you can set some sub options for pbo
@zenith palm
Advanced
Try set limits to advanced too
I would do manual if ya wanna push it get max perf, if not do mobo
imma do mobo
Should I change this?
Or this?
Ehh +200 probs, wait what mobo is it?
Ehh the throttle limit is probs fine auto can't remember off top of head
prime b550a-ac
is 200 good for my mobo
It can probs do 200, sometimes 150 etc or something performs better tho iirc
how about 175?
split the difference
Honestly you're gonna have to test +100 and then go up
If your perf drops go back a step
Ye, then go back and change the + number up/down and see what performs best
nope
default is whats set in bios
@pseudo hull
as long as PBO is enabled in bios ur good
ok
I
am running a multi core cinebench test
in the muti core test my speed isn't going above 4300
ish
debating if its worth overclocking my 3600 with a 3060 ti
with my ram at 3600 its been performing as it should, not having issues streaming warzone anymore like it was before
will you have fun or are you not happy with your preformance?
idk, with it running as it should im happy but if itd be worth it i wouldnt mind doing it
also interested in how itd perform
I mean I am doing it cause it is fun for me but it is up to you if you want to put some hours in
at 4.748 boost clock my cpu is siting around 56
how is this oc?
looks kinda average or maybe even below average
really on the single core it was boosted to 4.75
so?
that is .15 higher than the advertised boost speed and I haven't really manually done anything just done the basic stufff
cpumonkey is saying 1572 for r23 single core, and 11268 for multi core for 5600x
when I done with this 3dmark I'll make sure my ram is runing full speed
@faint tangle I forgot I could do it from the Radeon Software and did it that way. Thank You though. MSI afterburner works really well for Nvidia cards since AMD has included their own OC utility in the Software Package
didnt know amd had that, glad you got it working
Yeah Its been something they added for the RX cards
yeah idk too much about amd gpus, i prefer nvidia
I've used both, amd gives much better control over everything
Yeah and it seems to work better when Both CPU and GPU are AMD cause the OC tools kinda work with eachother
the oc tools dont work with each other though
I though Ryzen Master and the GPU tuning did.
1: ryzen master and radeon drivers dont communicate with each other
2: you shouldnt use ryzen master to overclock a cpu
this is the only part that somewhat communicates with each other
Ok maybe thats what I was thinking of then. But yeah Ryzen Master Sucks for OC its better anyway from the BIOS
When I do PBO on 5900x then try to set RAM clock, it doesn't POST
what clock are you trying to set?
the default 3800
everything else on auto
by default I mean the speed the RAM says it is
I have to do CMOS reset everytime
does it post at the right clocks with pbo disabled?
hmm I don't know I'll have to try that
I think also the auto-overclock on Ryzen master messes it up
but I had to reinstall Ryzen master because I was only showing 6 cores on HWiNFO
this most likely
okay I'll have to figure out how to uninstall master and still show all cores on HWiNFO
I couldn't figure out why I got about half my usual score on CB then I realized there were only 6 cores showing
I wonder if ASUS RamCache III is to blame
I'm gonna uninstall that junk too
I ended up using one of those extender wires that you usually connect to the reset switch when it's used as a RGB control. I attached it to the clear CMOS header so it's easier to clear
I'm gonna try 3200 first since that's officially supported or somethign I think
no POST
I miss the simplicity of manual OC on 10900K
okay I changed autoclock mode to like CPOC or something and now it seems to be working
but HWiNFO is showing 1900mHz for Memory clock
heck
man I really want to all core OC but this processor isn't good over 1.2v if I remember you saying
Why does voltage matter if temps are fine. CPU Core voltage (SV12 TFN) shows 1.231 but my max temp is 71 degrees on CB23
Because temps are the only driving factor in degradation. Voltage and current alone can have big impacts
Too much voltage can allow the current to blow straight through the gate oxide
Whilst too many electrons can literally break the material in the chips
what cpu?
It's better to just go off current
See what current you get with PBO and stuff and match that if you want to be safe
If you want to be bold you can go over that
But that's your choice
how does boost clock override work with zen 3 and curve optimizer?
i have it set to 200 but i'm getting boosts 350mhz higher than boost clock
but some people claim they are getting the same speed with +100 and better stability
it's confusing me what it actually does
it was very straightforward on zen 2... set +200 get up to +200 period
it raises the Fmax which is the maximum clock the cpu can go to, what chip?
5800x
boost clock 4.7
i'm seeing 5.05
another person said their 5800x hits 5.05 with +100 override
that's because the 5800x boosts to 4.85ghz stock
AMD just claims lower so people see it
Stock max clock for the 5800x is 4.85ghz
hmm i see
Well they are wrong or lying
The AMD Ryzen 7 5800X is built using just one CCD, which eliminates a lot of latencies and bottlencks in the multi-core topology. We also saw it boost close to 5 GHz regularly, out of the box, without any overclocking. This one-two-punch combination helped it beat the 5900X in gaming and several other tests.
what ever works
Core 0: +2
Core 1: -19
Core 2: -25
Core 3: -15
Core 4: -27
Core 5: -16
Core 6: -21
Core 7: -19
also i was told that using LLC with CO is a bad idea
i just left it on auto and it set level 3 which is the middle for asrock board
i'm seeing maximum 1.488v with around 1.25v at full load
should be ok?
i really don't want to have to redo my curve it's really annoying to test stability for curve
it's ok
yeah the chip accounts for it, so you just get worse transients
transients? so like higher voltage spikes?
all cores show maximum of 1.488v on hwinfo after 72 straight hours of benching
sweet
it's single core benches so it's not hitting 1.488v all core
that would be quite a problem
Intel rates their chips to 1.92v transisents which just shows how much transisents don't matter
yeah 1.92v virus mode and 1.79v TDP
i still need to tighten tCWL but i am currently 400 miles away and teamviewer does not work in bios lol
...would asus memtweakit work?
maybe i should try it
no
that high on laptop cpu?
it's for both s and h line
u line only 1.72v
virus mode is intels name for above tdp
My 5900x 3D Mark score (w/ RTX3070 OC at mem:8176MHz/clock:1815MHz) with PBO is mediocre. It's like 150 pts more than my 51/48 10900K w/ RTX3060 (OC also). Why tho?
pbo?
Precision boost overclock?
I think fitz is asking for pbo settings
Oh lol
CPU Score
5900x was way higher w/ 45.3 AC than PBO
I must be doing something wrong
rather 46.3 all core
I'm gonna try PBO again
What pbo settings you running?
I had it at advanced +200mhz
I have to run Ai OC Tuner at DOCP or it won't run the RAM at 3800
it'll be stuck at 2666 or something like that & if I change it, it won't POST
I'm sorry, but in all honesty I don't really know what that means
imc volts
for RAM?
yeah
Sorry man I know it's frustrating to try to help someone clueless
https://cdn.discordapp.com/attachments/610602203849162772/859545328486187018/20210629_172252.jpg on this page, vddcr soc to 1.15, vddg ccd and vddg iod to 1.05, cldo vddp to 1.0
on here, you have memory clock at 3800 already
Well yeah, you can't just set the frequency you need to enable XMP (docp on Asus)
That sets voltages and timings to match
I changed the settings from the all core 46.3 so those numbers reflect that
Let's see if that posts
Okay it POSTed
HWiNFO shows 1900 mem clock which I assume means 3800/2
Rn all cores are showing 44.8 ratio. Weak
Well maximum on two cores is 48.5
I guess RAM is okay but pretty blah performance from the CPU
Maybe CB isn't the greatest way to measure performance?
3.-7. are all this computer at different settings
The problem with those settings tho is that just about all of them run at 1.26-1.28v which to my understanding, is too high
If that's the voltage under load it's a tad high yes, but as an idle voltage it's fine
that's under load
level 3 LLC if I remember correctly
4.67 & 4.7 made it through CB but on AMD64 OCCT stress test caused reboot
4.63 made it through OCCT AMD 64 1 hour. Max temp was 77.4 if I remember right
prime95 would probably murder it
So I'm back to auto everything & PBO +200
I do pbo offset as the last thing, maxing that out isn't always best depending on how your bin is
except for the voltages Fitz told me to put in
Putting more voltage into the SoC will introduce a little more heat to the CPU, keep that in mind too
I mean aside from maybe a higher litre/minute pump I don't think I can do much better cooling wise
2 360s & 1 140 rad
I'm down about 250 points on CB
Yeah that's no concern in your case then
okay let me find where to put offset
Personally what I do on the core voltage is set it to offset mode and let the motherboard decide what's best. I have a -0.0275v offset on my core which in my testing was the ideal balance between heat and performance. I hit over 5GHz on all core loads. I could probably go higher with a better cooler.
You'll have to test what works for you though. This is specific to my personal system and testing.
That's pure PBO, no static settings
I had to play with the curve optimizer too in order to get there
oh shoot I thought that's where I was supposed to go
On this screen you've got it on offset mode already but on auto so there's no change
man 10900K is sooo much easier
and that's where I should try -.0275?
It jumps to .03125 because the increments are .00625 apparently
You can start with that (or -0.025 if your motherboard doesn't like to narrow it down that far)
My board allows very fine voltage control and I love it for that
ya this motherboard came with the RTX3060 from the shuffle
in hindsight I'd put an AORUS board
Well it POSTed. Let's see what HWiNFO says
49.3 on two cores
how do you make it do all core?
Make what do all core? Cinebench?
wait jk well here:
How do you make PBO do all core?
CB doesn't seem like a great way to test. I runs all the cores at the same ratio
PBO is all core but dynamic based on load
Max effective clock is what you want to look at when doing CB
that's a dismal 4,396
Single core I'm guessing?
Oh you're looking at the clock not the score I see now
4396 MHz isn't that bad, but does leave room to improve
Open up Ryzen Master
ya the score (I hope) will be at least 225xx+
which it did w/ all core 4.55
& basically every other all core
Was that at a safe voltage yeah?
haha I got rid of Ryzen Master cuz I thought it just made a mess
Ryzen Master will give a lot of useful info, I just don't use it to overclock with
well if safe is 1.200 then no. It was like 1.26-1.28ish
I saved the profile in BIOS. I'll load it & tell you exactly
Iirc 1.25v and below is safe or 1.35v mb i can't remember actually lol
1.25v or so under load is good according to AMD, any higher and it gets iffy
1.35v is Intel @zenith palm
Yeah 1.25v, 1.35v is some ram ๐คฆ
I was going based off of SVI2
Ohh or that too
Isn't it better for gaming if it boosts to like 5.0+ on individual cores v. 4.625 all?
You're right about the RAM too though, a lot of dies won't scale past 1.35v
I emailed G.Skill & they told me 1.4 for this 32x2
Wish the CAS was better than 18 tho
Yes, games rely heavily on a single thread, depending on game possibly up to 4 threads
Ya that's what I understood from people telling me PBO is better just don't bother with all core
Once you load up Ryzen Master I can show you which cores are your "best" for the per-core curve optimizer, so the "best" cores will boost higher
The one with the star is the best overall, circles are above average
It depends on the die for ram but ye
It shows 1.281 for a fraction of a second
I'm running my ram at 1.52v on the daily so yes it's highly dependent on die
Not a concern
It's sustained load that degrades the CPU at unsafe voltage
& 1.231 is unsafe voltage right?
According to AMD 1.25v is safe
because that's mostly where it holds steady
Well that & 1.225. Maybe it's about even between those two
Without the voltage offset that's around the 1.25v so the motherboard clearly knows what's safe
Lol max temp is 74.3
My cooling system is finally not the limiting factor (unlike my 10900K)
Yeah AMD uses less power and generates less heat than Intel in current gens
All that bending 16mm pipe was worth it I guess
In hindsight I should've done the loop on the 10900k
because I've hit a wall where even with AF2 with delta fans I can't get below 87 max temp
you think I'd do better with Ryzen Master?
& putting the started cores up to like 50 or 51?
How'd u get an all core 50 & stay below 1.25?
I always get the blah bin chips. Heck.
Not using RM to overclock, just for it to read the detailed system specs to find best cores
My reservoir is crazy rn
Curve optimizer
I don't even know how to use that. I guess that's the next project
22918 v. 22565 ac v. PBO
what do you set the ratio to when you use core optimizer?
curve optimizer I mean
I'm installing Master for the 3d time lol
just leave on default right?
I got 11 on windows dev insider btw
If anyone wants to try that out
As soon as i get win 11 that bar is going right back over to the left where it belongs :)
Where do I find the cores with stars? lol
haha that's an older picture
I moved it on both computers that have win 11
If you go advanced should be in the homepage iirc
It might be collapsed in a tab
Click where it says CCD0
It's a drop down
Same for CCD1 if you have it
And yes you leave core ratio on auto when using PBO
ah okay I see it
sorry had to take a bathroom break
so there's only two with stars
should there be more?
or should I do the built in stress test?
doing the stress test made something override something cuz max is now 5150 so it probably changed the core ratio
I guess stars are the best and white dots are the runner-ups?
well no it wouldn't have changed anythign cuz then it would've had to restart
Yes. Take note of which cores are marked.
okay I took a picture
Head into the curve optimizer in bios. You can start by setting it to per-core mode, then set negative ( - ) sign on each core and set the "best" cores to 25 and all the others to 10
even the runner-ups get 10?
Your bios might start counting at 1 instead of 0 like Ryzen Master
The second best ones will get 25 too
okay awesome I'm getting there now
Just to start the process going for testing, might change it later
ya mine start at 1 also
okay just did it
rebooting
what do those numbers mean? all it says is magnitude
It's shifting the boost table so the best cores can boost higher
oh okay
so it's like a ratio then I guess?
oh shoot I think I left the ratio on 46.3. I'm gonna reboot again
Kinda. It works with the motherboard sensors and logic to determine the best place on the table it should be. If you shift the table left (negative on the axis) higher boosts are selected.
You've got plenty of thermal headroom so you should be able to hit higher boosts. -25 is a conservative start. My best cores are -30. Depending on how you're binned you could go higher.
I hope I'm binned better than my Intel chip
how should I test it?
run a game with HWiNFO in the background? CB does like all core
Keep doing CB since you have results already in there from before
It has a single core test too
boot loop
I can't make it past enter password screen
once I enter password it resets
It's not stable there then
Do you still have the +200 pbo offset on?
Mine isn't stable there either. Maxes out at +175.
Maybe drop it to 100 and see if it passes stability testing, then you can go higher
okay
okay rebooting at 100
little bit of screen flicker action but I think I'm in
It seems to have upset Armourycrate because the rgb colors have gone bananas
multi-core is running at 43x per core
Idle or under load?
You're sure you did the right cores in the curve optimizer? Core 2 in Ryzen Master would be core 3 in bios.
voltage is 1.156
I'm pretty sure ya
I just subtracted 1 from every core number
1=0 2=1 etc
Should be right then. Try setting core voltage back to auto. That negative offset is taking it a bit low.
Oh actually it looks like your version of RM starts counting at 1
Or is that not your computer again in the screenshot?
Ok, and your bios started counting at 1 also?
Ok so core 6 on RM is core 5 in bios
Cores 1, 4, 6, and 9 in bios should be 25 in CO
If that's the current setting and it's not stable still, try setting 4 and 6 to 20
If it's still not stable drop all four to 20
it was stable with all at 25
I just didn't understand why an extra one was set high
It's the lesser cores being at 46 I'm worried about, you can also try doing 15 on the not-best cores
This makes no sense. Like it doesn't listen to me lol
I made a spreadsheet
Riddle me that lol
sorry yellow is actually 20
let me fix it
It's nothing but trial and error to find your best settings tbh. I feel like that on my system too, I know I can play with it for another couple hours and get a couple hundred more points on multicore.
ya for sure
that's why I do the excel sheets
you see how C6 is at 20 but has the 2d lowest ratio?
wth
it just straight up ignores me on C10-11
Look at that tho
is pref short for preferred as in the order the system thinks the cores are best?
I'm failing to see a pattern here lol
I wouldn't worry about that too much honestly, there's always going to be some variance. It'll probably change if you just reboot and run the test again.
I guess I can try putting the numbers in again based on the ratios it shows on HWiNFO
From an hwinfo author
They're all high quality still but some are just slightly better according to the hardware
The slightly better ones are the ones that will be "preferred" and used first when threads are being assigned
For a deep dive on the inner workings you can check this thread out: https://www.reddit.com/r/Amd/comments/dzjs7i/fastest_cores_is_windows_right_or_ryzen_master/
654 votes and 261 comments so far on Reddit
strange that 7 is a higher preferred
I'm gonna read that for sure
you think I should assign the higher magnitude based on preference or based on quality
?
Higher quality cores should boost to designated targets at a slightly lower voltage, so in theory they should boost to a higher max value at the same voltage across all the cores
Give it a go and see how it turns out. Trial and error. Some cores will actually boost lower if pushed too far. You'll have to watch behavior closely.
What's clear from that list though is that CCD0 is stronger than CCD1
Thinking of trying to push my R7 2700 further than 4 GHz, what voltage should be my cap?
safe generally would be like 1.3v for zen+ afaik, but you can most likely go further, some testing I found here:
https://www.reddit.com/r/Amd/comments/amvfi4/ryzen_generation_2_safeunsafe_voltages_tested/
137 votes and 88 comments so far on Reddit
I'll take a look, thanks.
it's not great testing, but it's a start to some info
I'll try pushing my 2700 a bit furthur, but if I can only gain 50 MHz while staying under 1.375v, I won't bother.
I got like 4.15 when I had my 2700 max with 1.375v
If I'm remembering correctly, it runs nice and cool at 4 GHz with 1.25v. I probably should double check.
Yep, 4.0 @ 1.25v
If 4125 MHz is stable (still testing), I will have passed 10k in Cinebench R23.
But it might look more like 4.1...
I hear ya man. I'm banging my head against the wall trying to figure out why it'll show various core ratios & they're all at least 3 points higher than what the cores will run in CB R23
I like overclocking intel a lot better
same
Intel is just like LLC, voltage, ratio, then ring ratio
None of this PBO Curve Optimizer stuff
I mean, that stuff is really beneficial for daily - I meant I like benching on intel better
ya I guess you're right. I just enjoy the process a lot more but that's probably because I spent a lot of time learning it & I know it better than I know this
Not that I know that much about either one but ya]
I think what I dislike the most is trying to stay right around 1.2v
that extra 1.5 on Intel gives you lots of leeway
Overclocking a 32 core Threadripper must be such a freaking mission
I just have to click a couple things and wait. Then reboot, lower the core clock be 25 MHz, click a couple more things, and continue waiting.
I do like that AMD lets you do like 50.3 whereas (at least with my MSI board) you either did 50 or 51, no decimals
watching it do CB single-core rn is kinda cool cuz it keeps bouncing from core to core
Just like Windows update!
My board doesn't let you change CPU voltage for some reason, so I have to go through RM. ๐ฌ
That's cool & all but I care more about multi-core lol
I wish I'd known more about how important a good motherboard is to overclocking before I bought my motherboard
Same here.
I had a gaming carbon wifi that got messed up during BIOS update So I got a Aorus z590 ultra. The MSI could barely do a 50 core ratio. The Aorus does a 51 with better thermals than the MSI's 50
Yeah, I could probably work on that. RM lets me tune each of my 8 cores independently.
It will probably be my summer project or something.
I didn't expect that kind of difference lol.
neither did I tbh
some random stuff can make a big difference tho
I cut out the top of my H500P Mesh because it was choking the exhaust from the AIO so bad
it was like a 5.4 degree difference
on average and max
Nice
I'm a little disappointed tho because I already switched 2/3 fans on the Arctic Freezer 2 from the stock fans to delta fans & it's only gotten me like 2 degrees
I removed it then cut it so that there wouldn't be open spaces lol
The liquid freezer AIO?
yes
apparently my picture won't go thru
I unplugged all those top fans. The deltas push so hard that they do the spin them anyway
do the spinning for them anyway I mean
They're just decoration now
Otherwise when the deltas kick up it'll cause the fans to like hit the sides. I had one jam
Nice. I just have an Arctic Freezer 34 Esports Duo. Handled the 30 min CB run nicely. 4.1 GHz @ 1.375v, sucking just under 150W, and it never reached 70. Of course, the room is somewhat cooler than normal.
UHHHH
Manual OC?
on a ryzen 7 2700
@sour vortex
Yes, manual oc is best on Zen+ because crappy first attempt at PBO had few options
Zen 2 and 3 PBO is better
1.375V on a 2700 mate
Probably not that under load, I'd hope at least
he ran cinebench for 30 mins like that
it was very much under load
1.35v is ok on Zen+, 1.45v is around where degradation occurs more
Source: Buildzoid (Actually Hardcore Overclocking)
@south sky tCWL 9 1 error, tCWL 10 0 errors
any less-than-straightforward ways to make these able to tighten? they are all the same as auto value
1.35 is the hard limit imo
1.325 is where i stopped
tWRRD 1 might work. Otherwise those other ones can be ignored
Setting them to auto, setting them manually, setting them 1 or 100, doesnโt really matter
tWRRD 1 errored before but I can retry now that I changed tRDWR
it works now
@modern walrus How should I test my per-core tuning? Multi or single thread?
Multi
Alright, thanks.
this is probably a stupid question
but is there any simple way to tell if you have a good bin on your cpu or not lol
hmm interesting tightened a ram timing and tm5 no error but it made gpu start crashing again
depends on the cpu
i hav 3600 XT
determine max FIT voltage first by running p95 with pbo on
then see how high you can boot at that voltage
once you can't boot anymore dial it back down one step and start stress testing it
shouldn't take more than 2 hours to find something with a 90% chance of being daily stable
after that just leave the test running until confident that it's 100% stable
I would recommend just using pbo instead if you're gaming though
being able to get 1900 fclk stable is more important for determining a "good bin" for gaming than core clock
@tall pelican if something was wrong with that please correct me I don't know everything
FIT voltage is only relevant at that recorded temperature, at that reported current
say you get a FIT of 4.0ghz and 1.3v, but then you clock it up to 4.4 and 1.3v, you're increasing current draw by 10%
more current -> more power -> more heat (hotter temps)
turn down the fan speed during that, and you'll see fit go down as temps creep up
manual oc did not increase that temp
that was 100% pump with fans around 50%
I was mostly just curious though and didn't actually manual oc
to keep the same level of safety, you have to limit FIT based off the current draw
so get stock edc/tdc, run the test off that, then when you manual oc, you cant go over tdc/edc
can amperage really be different at the same voltage and wattage
thought it was volts * amps = watts
if you're at the same voltage and wattage, that means you arent clocking higher
was 4.2 all core vs 4.05
that's 4% more current based off clocks alone
I see
well thank you for correcting me
@normal ether you should probably read this
I'm sticking to pbo and curve optimizer
appreciate you taking the time to explain
So it'll show like 48.3+ as max on all cores then when I run CB it'll usually run 44.5 during render on all cores
I dialed back PBO to +100mHz to get higher clocks on all cores
Plus I figured 50.5 was a pretty good single core anyway
Anyway, why does it usually only run 44.5 during the renders?
it can't be temperatures, those are maxing out below like 74
heckin Ryzen
that is normal
heavier all core loads clock lower than lighter and/or less threaded loads
this is a bit of an oversimplification but imagine you have a power budget of 200 watts
hmm but I can get an all OC at 46
if one core is blasting through a light workload it might take 25 watts by itself to boost higher
then lose the single jump to 50.5
but if all 12 are running they are going to divide that budget more evenly
of course there's more factors like voltage and current
which I assume 44.5 all core with the jump to 50.5 sometimes is better than 46 all core.
but the concept remains that all core loads will clock lower
Though it isn't on CB lol
it depends
if you primarily game then yes keep pbo
if you never game and only do things like render then keep the manual oc
I see
Yes that's what I think I'm going to do (keep PBO)
So it's safe to say CB isn't really a great benchmark for gaming then?
I'm pretty sure I did better on 3dMark w/ 46 AC but I'd have to run it again
a 64c/128t threadripper 3990x would crush your 5900x in cinebench
it wouldn't even be close
but the 5900x would be better in all games that aren't star citizen
3dmark CPU test is also an all core render load
ya I noticed on single core CB that the 5900x was killin it
if you want to bench your cpu in games then bench it in games
there are games with built in benchmarks that measure your system performance for that game
lol what a novel, obvious, yet genius idea
such as AC valhalla
I dunno why I didn't think to do that
r6 siege and csgo may be good tests for primarily testing cpu bound scenarios
horizon zero dawn has a benchmark that scales well with high threads
Well I only have Plasmorphia, RE2, RE3, & RE7 presently lol
What's "good" in 3dMark?
like 14k+?
(bias lol)
overall or cpu
I tried the new one they added. I forget what it's called but it tests CPU at different thread amounts
CPU
but I first asked overall
cuz 14k cpu is insane
but ya CPU
I was not seeing a significant bottleneck with my 3080 at 1440p with a cpu that scored 7-8k
like I said 3dmark CPU test is still a synthetic test that is not necessarily representative of gaming performance
My setup is actually kind of puzzling
because I can't tell whether the 5900x or the 3070 is the bottleneck
no, in game settings
ohh okay
csgo at 480p? 5900x will be the bottleneck
I do all my gaming at 4k
control at 4k? 3070 will be the bottleneck
I figured I might as well do 4k cuz my monitor is only 60hZ
cuz it's kinda technically not really a monitor
if you're gaming at 4k (and presumably high-ultra settings) the 3070 will practically always be the bottleneck
but it's 43" so I like it lol
I observed instances of my 3080 bottlenecking my 3600x at 1440p high so
okay so for sure in games I'm bottlenecking at the GPU
Which I haven't even played any games on the 3070 tbh
the 3060 w/ the 10900K did fairly well on RE2 with some pretty high settings
it's weird cuz the 3060 has so much more RAM
vRAM
I see why it was kinda pointless to watercool the 3070. It'll run out of capability way before it'll hit any thermal limits
the 3060 does that with air cooling lol
3060 having 12gb vram is quite simple
the core has a 192-bit memory bus
memory chip is 32 bits wide
so you want to have 6 chips
which are going to be 1 or 2 GB
they didn't want to make another 6GB x60 card because people were already complaining about nvidia having lower vram than amd
therefore they slapped on 12GB and called it a day
even though the 12GB really does nothing for the card
the 3070 on the other hand has a 256-bit bus so their options were 8GB or 16GB
16GB would drive much higher prices and 8GB isn't nearly as precarious as 6GB in 2021
the 10900K got a 13,162 score on 3dmark for CPU & the 5900x got a 12,562
I'd like to see a 3050ti that's essentially a 3060 with 6gb vram for ~$260
isn't there a 3050 for laptops??
I should probably run 3dmark again tho now that I've messed with the curve so much
heck let's do that now.
yeah there's a 3050 4gb that people are complaining about
even tho I'm so tired of watching that same sequence lol
I dunno why anyone does laptop gaming
It's just heat and disappointment
I'm gonna run Timespy rn see if it got better
I mean shouldn't the 5900x be getting a better CPU score?
The 10900K is all core 51 tho
I've been pleasantly surprised with my laptop
I have a surfacebook 3 so ya heat & disappointment for gaming
If you watch HWiNFO that thing hits TJ max constatly
stayed over at gf's place for long enough that the 5 year old laptop with a 60hz screen was killing me
dropped 1700 on a new gaming laptop
I should've done that
slightly edges out an i5-10400 + 2080 super in synthetics
but I thought the detach the screen & have a tablet thing was cool
in some games it gets scarily close to my desktop 3070
I've seen my laptop 3070 boost to 2100mhz core 8800mhz mem in the outer worlds
I doubt I'd use my laptop for gaming if I could really
my desktop 3070 tuf with 112% power limit does 2130mhz core and 8200mhz mem
I have an old ASUS gaming laptop lol
well I have no other options when my desktop is 400 miles away
all things considered it's a very nice machine
damn your girl lives that far away?
your TUF is choppin up my EVGA then
I'll tell you in about 6 more minutes
I'm amazed this thing is doing 110 FPS rn
wait jk
an error just occurred
what the heck man
unstable oc
GPU or CPU? probably GPU no?
your mem clock doesn't look high enough to be problematic
try dropping core by 15
GPU yes
should I do time spy extreme or regular?
which one are you comparing?
I should probably do extreme so I can see 4k but I'm gonna do regular for now
I could never play this game. It's so orange I dunno how else to describe it
hahaha I get what you mean
I personally do 1x time spy stress test and 3x fire strike ultra stress test
with one stress test being 20 passes of the graphics tests and skipping the CPU test
which uni are you going to? I did FSU for undergrad now I'm at FIU
I'm also almost 33 tho. I wish I learned more of this computer stuff when I was younger. I might've done engineering or something more useful than Political science
originally NYU, hated it and left, back at community college for now working on a transfer somewhere else
off topic though
Back in my day we were happy with our 833mHz Compaq that could run Star Craft Brood War pretty well
and DSL got us on battlenet no problem haha
my first computer was a gateway all in one with a pentium 4 that my parents bought for me because i won some math competition
I used it to play maplestory and nothing else
lol nice. I don't think there were all in ones when I was playing games a lot
okay CPU score is down to 11,995 tf
Graphics at 15034
graphics score up, cpu score down. go figure
15330 graphics, 9235 cpu
here's 18447 graphics with 7930 CPU lol
jeez why am I not getting a better overclock then? lottery?
they're both 3070s
probably lottery, the air cooler is doing a fine job of keeping the 3070 down
you can see avg temp is 56C
must be cuz I'm pumping a ton of water onto that GPU and the water is coming out of 360 rad
I live in a poorly ventilated room in california where ambient temp is usually almost 30c
your cpu and gpu sharing a 360 rad?
Lol I got 54 but the ambient temp inside my apt is uh whatever 74 is in celcius
no they each get a 360mm rad
interesting
& then theres a 140 to make sure the water going into the pump isn't too hot
I just have an aio on the cpu and stock cooled gpu
I kinda maybe uh overdid it
room temp usually mid 80s F
I don't think this Thermaltake case gets great airflow
which case?
Level 20 GT aRGB
this 3070 is sitting in a td500 mesh with 3 front intake, 1 rear exhaust, and aio configured as triple exhaust
all that glass means not much air movement
yeah it's probably case difference
That's closer to my H500P build. AIO as top exhaust seems to be the best
yup my new build is using aio top exhaust as well
but that thing has terrible ventilation up top so I did a little bit of cutting haha
I made the rear fans blow in & top fans blow in
it's nice for cpu heat to be directly ejected from the case
I was gonna cut out the bottom a bit & throw a couple 60mm fans to draw from the bottom too
but I'm not sure it's worth the headache
ehh your temps are fine
54c is not a thermal issue you're just hitting silicon limits
don't feel too bad, my 3080 is one of the worst in existence in terms of silicon lottery lol
That case I was gonna cut in 60s has a 3060
just saying my 3080 is a steaming pile of garbage in the overclocking department
so don't feel bad that my 3070 is a fraction of a percent better
I just can't get over how mediocre that 5900x performance was
I guess I should go back to the curve enhancer & see what I can squeeze out of it
I don't think I've hit the limit yet
I just got tired of doing it
pbo/co is going to show you worse performance than manual oc in 3dmark
but better in gaming
I wouldn't worry too much about it
hmm okay
I just like lived & died by those numbers when OC my 10900K
but I guess it's apples & oranges
at the end of the day you don't play 3dmark right?
Ya true
& I wouldn't play that weird TimeSpy game anyway
I dunno why but this ASUS board puts the pump at 100% by default
it's raging so I gotta go check that
jeez
makes no sense
Do you watch any of that Twitch stuff?
like Robey or any of that?
nope
eh it's kinda cool. I watched Thermaltake's this afternoon. They finally replaced that TT RGB Plus app
well just 2.0
what fans do you use?
id-cooling xf12025 argb
I bought them at a pricing error
5 3-packs for $50
each 3-pack has an argb controller, pwm splitter, and 1-4 argb splitter
very nice value for money lol
basically a bit under $3 per fan
heck ya I think the 3 of them was like $18
combine that with the $33 lian li galahad 360mm aio
and i filled up my o11 with 9 fans for very very cheap lol
ya I bought a bunch of thermaltakes on ebay for a decent price
$50 for a 360mm aio + 9 fans 
but the 200mm Riing Trios cost a fortune
They look nice at least
I'm gonna see if 27 magnitude will run stable
odds are no
Core 0: +2
Core 1: -19
Core 2: -25
Core 3: -15
Core 4: -27
Core 5: -16
Core 6: -21
Core 7: -19
that's my curve
you dont use optimizer?
oh wait that is optimzer?
Mine are all -
what translates to more performance - or +?
I don't even know how this ridiculous magnitude system works
well
i have a +200mhz override
and my core 0 is not stable at +200 with stock voltage
so i have to give it more voltage
hence the +2
the other cores are all fine with undervolting
in general you want to make all of your cores as low as they will go
core 0 was throwing errors at +0 and +1
oh so the more negative the magnitude value, the less voltage the core will require at the various speeds respectively, the higher the system can run the core while remaining under 1.2v?
well frequencies not speeds
will any core exceed 50.5? It doesn't seem to ever push beyond that amount even if I add more magnitude
Also how did you isolate which core causes issues?
This is useful
have you set PPT, EDC, or TDC to anything other than Auto?
I imagine auto for EDC is 1.2
corecycler
whats that
I watched a video on tuning the EDC
kind of a pain but I think I did it
I guess 112A is my sweet spot
do you ever run sfc /scannow on cmd? I try to do it often in case all this crazy OC reboot 800x corrupts windows
wouldn't lowering the max boost clock override give you the ability to put higher voltage offsets? I mean isn't the main cause of crashing for curve optimizer when one or more cores tries to go beyond the frequency limit? (I think that's 5.1ghz if I had to guess?)
no, it's when it doesn't have enough voltage to reach the target clock
yes i have sfc + dism in a batch file
i am planning to a clean windows install once i'm done ocing regardless
oh ya? should I do that? lol
well I'm doing manual ram oc
so it's a good idea to
if you're only doing cpu oc it's probably not necessary
oh ya I don't even know where to start with that
unstable ram can cause wrong reads/writes everywhere
I've nearly gotten every core to show a max ratio of 50+
yet CB runs at 44.5 lol
whatever
I'm kind of confused as to what I'm even accomplishing at this point
Somehow I dailied unstable ram for 6 months and only had sfc /scannow fix stuff twice!
I'm still using the same install