#overclocking
1 messages · Page 39 of 1
Thanks for the offer but i got no $$$ on me atm
Let me know if you change your mind. I want to sell it for $75+shipping. I paid $156 USD new because it is a very high bin.
@dim vale @red terrace
cause i heard it had alot of headroom
i got mine to 3.5 ghz and 20gbps vram with stock cooler
depends on how good your silicon is
not THE LOTTERY
apparently gunnir's photon models have the best silicon so if you want to high OC a b580, get a photon
Taiwan
💀
photon already goes to 2900 by itself
Doesn't exists
why is there so many 13400F
i did multiple tests and got multiple scores
these search results only include 13400f + b580
if you search with any cpu then i wont be on there since people have freaking 14900k's
my gpu core and vram clocks are higher than this, if only I had a 14900k
at least im top 2% in the world
what model btw
the gunnir photon
Okie
as i said here
Don’t worry I’m getting the 9070XT steel legend
Just curious
ok
14654 is average graphics score
no really
lots of clocks to be gained but the scaling is ass
Just beat my old score for 7600x + 6800. 15454 ~> 15816 
This is really frustrating. I am noticing a consistent pattern that when some of my cores try to do +5.75Ghz while the CPU is above 49c that appears to be when freezes happen.
My core and vram clocks are higher, i just need a 14900k breh
What is your current chip?
You can likely scorea decent bit higher by using a stripped down version of windows 11, run the test with WiFi and startup apps off.
Also doing a ram overclock can high raise it a little more albeit minor.
13400f
Ahh definitely a bottleneck.
Surprised you're using an I3.
@dim vale have you tried the tips above?
I don’t have another drive to install tiny11 on, and my ram can’t even get to the max speed without raising the cl rating cause of this cpu
Damn.
I forgot the 400 labels belong to the I5 family. Granted they really should be in the I3 I feel.
They only have 13400 and 13400f, for k models it’s 600
Not sure why they did it that way
how do you stack up sorted by graphics score only?
combined score is kinda useless for GPU benching anyway
full system benching is fun, but on a non K chip I assume you're just going for GPU score
i misread the bench lol

i practically only bench TS, saw the leaderboard and just assumed it was TS
but still, CPU shouldn't really affect graphics score, no?
I just be needing a better cpu so I can beat more scores
It does
Every top score in the world has a 14900k or near that
With lower gpu speeds
Could run the GPU Sub-Ambient.
You should be able to raise your score a lot by doing that.
What that mean
ln2, chiller, etc.
well I would love to get that
I need to contact ASUS or something
but would like help with comprising how I go about it
get a sponsor deal going 🤣
Ikr how does one do that…
After a series of new scores incrementally raised the bar for memory frequency, Intel, Gigabyte, Corsair, and elite overclockers put an emphatic stamp on the world record. Talking Tech goes behind the scenes for the inside scoop, talking to extreme overclockers HiCookie, Sergmann, and others involved in pushing DDR5 frequency to an incredible 13...
@sterile flame
Questions:
- What CPU?
- What Board
- What RAM kit (brand, capacity, speed, timings, and die if you know)
- Is the issue posting, crashes, or errors?
- What are software you testing with
- CPU and RAM temps under load?
- Have you tried changing CPU or RAN voltages?
5900x3d
as rock b850 pro
ddr5 xsky 6000mhz cl30
it might be posting? restarted, posting LED's turned on, then rebooted again with default ram settings
not testing with any software
not sure about temps
no, only changing dram profile through bios.
Just to confirm do you mean 7900x3D or 9900x3D?
sorry, 9800x3d
idk why i typed that 😭
Have you gotten into windows at all?
.
yeah
was gaming on it & everything
everything is fine it literally just won't apply the 6000mhz settings
also sorry if i don't respond right away at work rn
but yeah i was using one of the presets under dram overclocking, it was a profile literally called expo 6000 or something like that
So the PC is working except for the RAM profile not applying?
and i'd hit save & exit, it would ask me if i'm sure if i want to apply, i'd hit yes, it would reboot, take forever, then reboot again & not apply the settings
yeah the pc works great
was playing alot on it
and was monitoring temps via hwmonitor & everything seems fine
Hmmm seems like failed memory training.
Download BenchMate, TestMem5, OCCT, and Y-Cruncher VT3. I can provide you with VT3 if you have a hard time finding it.
When your available I will guide you through how to make a custom ram profile.
.
Also, @sterile flame are you on the latest bios for your board?
definitely yeah. when i first built it 2 days ago i had to bios flashback the newest update
because it wouldnt POST
im unfamiliar with these, ive never went to great lengths to overclock before, im assuming these are all just for benchmarks/monitoring?
They are both a Stress Tests and Benchmark.
If you can spare $10 I strongly recommend downloading Karhu as it's a great CPU and RAM stress testing and benching software.
@abstract merlin @dim vale @red terrace @uneven knot @wooden vigil
I have made significant progress with my co. Been working in large increments to save time.
Currently at:
-75
-50
-25
-20
-50
-100
-20
-20
The cores at -20 need further tuning.
All cores can do 5.75Ghz and I am so far seeing a +200-220mhz increase in All Core clocks under load in OCCT with AVX512 instructions compared to stock.
My limitation currently is due to thermals.
The more extreme the negative offset the colder my CPU cores need to be to remain stable.
Woag
cool
Need some feedback on this. I just bought a Kleev 6000mhz cl28 1.4v ram, but there is a Kleev 5600mhz cl30 1.25v….which is better?
Deciding if I should switch
Keep the 6000cl28@1.4v that is a much better bin meaning you won't need nearly as much voltage for higher overclocks which also means you can cool the RAM much easier.
.
For RAM overclocking you will want a fan to actively cool the RAM.
Ok yeah, i have all the main pc parts, but no case or cooling that’s going to be the last spot
P12/P14 Pro fans and get an extra one for just the RAM.
https://www.amazon.com/dp/B0DPX94PSL/ref=sspa_mw_detail_0?ie=UTF8&psc=1&sp_csd=d2lkZ2V0TmFtZT1zcF9waG9uZV9kZXRhaWw
just get the 5 pack?
ARCTIC, a leading manufacturer of low-noise PC coolers and components, initiated and continues to shape the trend towards quiet cooling systems. In addition to an extensive range of CPU coolers and fans for home use and servers, ARCTIC produces high-quality monitor arms, wall mounts, thermal past...
Also assume I need a good AIO and cpu cooler both?
Yep.
Didn't I already send you one?
Are you getting the case I recommended or a different one?
On the AIO I sent it has decent fans so you don't need to replace them though the P12/P14 are none the less better ones if you choose to replace.
Yeah this one. So just this AIO and the 5 pack of fans should be enough and no CPU cooler?
Strongly considering the case you sent me, just deciding if that’s the look I want and it will have enough room to put all my build plus this screen in it
LEVEL UP YOUR BATTLESTATION Take your build to the next level with this standout 14.5" display. Showcase system metrics, calendars, weather apps, or in-game chats. EFFORTLESS PC CONTROL AT YOUR COMMAND The five-point multi-touch capacitive touchscreen makes interacting with your widgets, apps, and games simple. Use it
Get:
-
the case I listed for a positive pressure system having the AIO being a exhaust in the side
-
two 5 packs of ideally Arctic P12 Pro instead of P14 Pro if possible so you can put 3 TOP, 3 BOTTOM, 3 SIDE, 1 FOR RAM.
-
It should have enough space for the screen if you mount it to where the side exhaust fan usually goes as that will be unused with the config I sent.
.
If you don't want to use Zip ties to hook up the fan for RAM, purchase a 120mm L Bracket for a fan.
Just checked, that should very likely fit in the location I mentioned.
Let you know soon as I choose the case, but 75% sure I’ll use that case you recommend. I want a lcd screen like that one above I linked above. Trying to see if any cool AIOs that have the screen built in worth just getting vs the $100-$120 I’m paying for AIO/screen separate
Purely for aesthetic
Also the possibility that putting this IN the case would make things too hot? Not a single picture shows it in the case just separate or on the outside of the case
I can think of a few that perform as good as what I mentioned (same pump and rad) and doesn't cost a kidney.
A LCD screen won't heat the case up.
Ram tuning
Well this is my default timings and my Voltage is 1.350 I don’t want to add voltage i just want to tune what I can with the current voltage
Also what temperature is stable for ram while gaming?what’s the safe limit?
Thanks in advance
Why don't you want to add more voltage to your RAM?
DDR5 is safe within 1.45v and Hynix A/M can daily as much as 2v with sufficient cooling.
It varies a lot by stick even amongst the same ICs by as much as 15c sometimes, but generally speaking for Hynix A it's around ~50c and ~45c for Hynix M.
I've been running 1.85v VDD daily on my Hynix A die for about 6 months now.
Been dailying north of 1.65v for two years.
My temp is quite high already more voltage sound like trouble
Do you have a fan cooling the ram?
It is strongly recommend when tuning ram as the things that will most the biggest difference tRFC and tREFI both will increase your temps significantly.
What temp should I stay below to keep stable
Read above.
@lilac fog might be worth taking a look at.
Same performance as the one I sent prior but with a large built in screen.
@pure widget
Ok thanks
Nice, it’s basically combining what I was looking at. Regarding this, any chance you can look at the screen I linked with the AIO you recommended and see if there is any advantage of buying this combined one vs the 2 separate? Only reason I ask is because this one is $50 than buying separate. It’s a beast no doubt
Buying them separately will be more of a headache to set up but ultimately may be more practical in terms of screen space and look nicer depending on how you set it up.
Going with the combined option means simple setup, looks decent without needing to mess with anything but no customizability outside of what's displayed.
Well I am extremely happy with my CO, I've got it to be about 20% more power efficient while having +200mhz higher boost clocks. I am currently very thermally limited for pushing CCD0 further so I am gonna settle there for that CCD and start working on CCD1. My peak power consumption went from around ~250w down to ~190w tested using Cinebench R23 and R24, Karhu (Cache+FPU), and Y-Cruncher VST3.
CCD0 (V-Cache)
-75
-40
-60
-20
-50
-50
-20
-20
CCD1 (Frequency)
-10 (to the entire CCD)
@alpine mantle @wooden vigil @dim vale @calm grail @short blade @modern walrus @acoustic crane
Glad you found satisfaction with it
bonus points for lowering power draw
Thank you.
Now I must prepare to torture myself with CCD1 lmao
🗿
Anyways
My findings with AM5 so far is that, the chips can be very efficient but only if the thermals are low enough.
CCD0 is not stable unless I have the air conditioner running which I do whenever I am in the office.
interesting findings
cool
I am not sure how much it varies by cpu since I've only had one 9950x3D and two 7950x3D but I've noticed on all 3 that in order to achieve a flat +200mhz boost on the V-Cache CCD in Single Threaded loads, the core needs to be around 45c or lower under load.
this is just wild ngl, overclocking AMD components now just looks difficult lol
it really gotta be that chilly huh
have you delidded it?
Back. Sorry for the delayed response. @alpine mantle
you're fine
I feel you there. I think AMD overclocking had come a long way with all of its tools which gives the users more control but that also makes it feel a lot more intimidating to do.
Not yet. It's a planned upgrade though.
I want to delid it and purchase a chiller.
Before that though I am gonna get the Corsair SF1000 PSU and upgrade my GPU to the 9070 XT Mercury from XFX.
Debating replacing all my fans with Arctic Pros since my fans are getting old and are very mixed.
better cross your fingers that its a good bin
i got my 9070 xt recently and despite it not overclocking well its a BEAST
Supposedly the 9070xt usually have about 10-15% performance to be gained and the XFX Mercury happens to be the best binned model on the market and has the bios with the highest TDP available.
Apparently you can raise the power limit to 400w on that model.
absolutely insane but incredibly inefficient at the same time lol
How are temps?
its sooo good
usually 48-50c core and up to 70c hotspot
its a very chilly card
this is with a -13% Powerlimit (around 260w)
stock is 304w
I see. What ambient?
i guess as of now 75f?
Oh also, the XFX Mercury comes with PTM7950 which is nice.
AC or room with lots of airflow?
Usually safe to bet on the room temperature to be about 30% hotter then ambient unless it has good airflow.
its windy outside and the AC is off atm but its set to 76f
Gotcha.
I am debating if I should try LM on the 9070xt if it has good OC headroom.
Oh btw, @alpine mantle for AM5 CPUs, additional voltage is still a good way to OC if you can keep temps down.
Prior to my bios update I was running +25mv offset which got me 5.8/6Ghz on my 9950x3D.
That and 14c ambient temp.
thats cool to know
I know someone with a 9950x that does 5.7Ghz all core but with a static OC at 1.475-1.5v which is +100mv.
big voltage numbers 
Yes lmao.
On a related topic, I love the fact that my Hynix A modules are taking 1.85v VDD for daily like a champ.
Ran 1.65v for two years and 1.85v for about 6 months now.
Yo what’s a good score for 5060 ti 8gb stock settings on Time spy?
@alpine mantle Hynix A and M die can daily as high as 2v VDD if temps are under controller. The more voltage the colder it needs to be.
@paper tulip
CCD1 on this CPU is really starting to piss me off.
It's able to clock 6Ghz but only with a large overvolt and the cores refuse to be stable with a negative or positive CO at +200mhz on the new bios.
I am currently trying Scalar 10x with -10 to see if that helps.
-30 all core seems to work pretty well on CCD1 with the override reduced to +150.
meh I'll bother with it when I get 9850X3D (assuming it exists)
but congrats I know that was a ton of work lol
Thank you.
I think I just learned something really annoying. It seems all cores share the Core voltage rather then each core getting X amount because with CCD1 enabled it causes both CCDs to become unstable from lack of voltage.
I could be wrong, it's just my going thinking so far because CCD0 is fully stable when CCD1 is disabled and temps remain nearly identical
I am getting annoyed by the need to keep the ac running for every thing to be 100% stable. Gonna retune the ram at ambient.
@steady lance just relized EXPO is tweaking out and giving me 4800 MHZ mind helpin gme later
?
Sure.
@wooden vigil when you have time, tell me your specs then I will help you with steps.
- CPU
- Motherboard
- RAM (is it Hynix A/M die and what capacity)
- RAM cooling solution
7700
X870 Steel legend
Hynix A Die
Just the heatsink
Setup a fan to cool the RAM because we are gonna go down the RAM OC rabbit hole for this.
Can be as jank as my setup, it just needs a fan and keep that fan at a static speed.
Hmm, results will be very limited without a fan.
At long as it's 6000 it's currently 4800 rn even after XMP
Try this:
-
vSOC = 1.25
-
VDD = 1.45v
-
VDDQ = 1.45v
-
VDDIO = 1.4v
-
SPEED = 6400mhz
-
tCL = 30 (try 28 if you're feeling lucky)
-
tRFC = 420 (try to lower this as much as you can, expect about 120ns for Hynix A)
-
tREFI = 65535
If this doesn't work, raise vSOC to 1.3v (do not set higher then 1.3v to stay within safe limits).
@wooden vigil
If 1.3v vSOC fails, reduce the RAM SPEED to 6000 and try again.
Test using TestMem5 set to Extreme profile, Y-Cruncher VT3, and either Karhu (Cache+FPU) or MemTestPro.
Just did both CPU and linpack testing on OCCT for the 5.3ghz retune, yes the 13600k pulls more power but the effective clocks are great now!
Retune of the 13600k is a success, no WHEA errors (including opening radeon software). Already tested linpack and CPU test on OCCT and now over 2 hours of prime95 small FFT
The only thing that isn't overclocked is the ring ratios but i dont think its that important since my ram is already tuned
stock ring ratio is just 4.5ghz
overclocking it to 4.8ghz brings unnecessary voltage bump and power
I haven't overclocked LGA1700, what does the Ring Clock do? @alpine mantle
just need 9850X3D now lol
Oh snap. That's awesome man, congratulations 🎉.
How much did it cost if you don't mind me asking?
This is confusing
ROG dimm.2, it's for an nvme expansion card.
Weird
It supports 2 4.0x4 M.2 ssds via the included Add In Card.
that third one is for extra nvme ssd storage as a expansion slot card
Get Grammarly today! http://grammarly.com/ltt1
AMD, what are ya doing buddy? You gave us so many good CPUs the last few years, but THIS is what you're showing off at CES this year? "AI" CPUs that are only marginally better than the previous ones, a better binned refresh of the 9800X3D called the 9850X3D, and then they talk about an AI agent to ...
I wonder if it will perform better then my V-Cache CCD.
It's currently doing 5.65Ghz flat with temps in high 50s. I decided to retune my PBO because I didn't want to rely on running the AC 24/7.
Going from +200 with -15 being the worst CO, the CPU is now doing -30 as the worst CO at +100 and boosting without the need for AC.
I wonder if the frequency to voltage scaling is mostly linear or not.
Well unfortunately -30 caused Windows to Freeze giving me a still image.
Hmm gonna try a new idea.
To help with Idle/low power state freezes the bane of my existence,I am gonna use Curve Shaper to run a very minor Positive CO at MINIMUM and LOW frequencies.
Anyone have some good bios settings for a Asus Strix x870e-h gaming with a Ryzen 7 9800x3D
For overclocking. For best FPS
EASY OPTION
- Enable PBO
- Enable XMP/EXPO
HARD AND MORE REWARDING OPTION
-
Enable PBO
-
Enable XMP/EXPO
-
Using your RAM profile as a baseline, tune the ram's frequency and desired First Cas Latency, then do fCLK, then move on to tRFC and tREFI timings before doing Tertiary, Secondary, then Primary timings.
-
Tune the CPU using PBO's Curve Optimizer to reduce temps and improve efficiency resulting in higher boost clocks.
The GPU is simply, crank the power limit, then raise GPU Memory and Core Clock till the app or windows crashes.
Then back it off till stable
With GPU Memory Clock you will notice artifacting beforehand
To test CPU and RAM stability download:
-
Karhu (run it in Cache+FPU)
-
Y-Cruncher (run it in VST3)
-
Lots of games, Streams, and just idle stability.
Idle stability can take days to detect.
@steady lance @alpine mantle all core -10 is not genshin stable
💀
as a reminder, 24/32/30/30/30/30/36/36 passed every stress test under the sun on this chip
back to the drawing board...
yup basically gotta start over from scratch
using genshin as my only stress test lmao
evidently no other stress test matters
kinda funny ngl
what if another game like zenless zone zero or honkai star rail can be more sensitive than genshin 👀
💀
alright nvm something else is going on here
I noticed that every time the game fails to load wallpaper engine also crashes
game is failing to load with stock cpu + jedec ram + stock gpu + freshly reinstalled radeon drivers
im confused
That is not good, either it's a bug with the game or the fact that there's stability issues with your PC
for a game as popular as this you'd think it would be widespread news if it was a problem with the game
You can try ddu and verifying or reinstalling the game.
What about for asus prime X870-P WIFI please?
Same instructions.
finnaly got my ram OC done
Congratulations 🎉
5.814Ghz! Core 0 V-Cache CCD on my 9950x3D on AIO. Testing using SuperPi 32M on BenchMate.
@acoustic crane @abstract merlin @static oasis @dim vale @somber perch @short blade @alpine mantle @barren ridge @calm grail @fervent talon
-60 CO with Scalar 10x

Understood.
+25mv to the V/F curve's maximum voltage by adding additional points and supposedly tells the cpu to try and clock more aggressively.
Right
though no one has been able to provide an answer on how it tells it to clock more aggressive
I know how fit and scalar works for the most part but what gains are made
btw thanks to this I was able to confirm, my msi board really did let me apply -115 co lmao. Cause if that was in millivolts my cpu right now would not be posting
That’s crazy
I am GPU limited though my 1% lows are better and I get higher scores in benchmarks.
Tru…
I am unfortunately also thermally limited right now.
The gains rhat nobody talks about
Notice that the CPU was hitting 60c.
How much higher does the cpu clock now
Compared to 1x
compared to stock PBO (-10co all core and no frequency offset) it's a few hundred points higher now with -20 all core (-60 on core 0) and +311mhz.
Unfortunately all core loads has me very thermally limited now.
I am clocking higher than a 9850x3D with +200mhz 😄
Boosting to 5.865ghz in office loads.
Now this is the weird part
for some reason my cpu is pulling less power then it did on my MSI X670E.
On that board it would've chugged 250w or more with my current settings. Right now I can't get it to pull more than 220w.
Maybe lower curve offset combined with some other limitation is making it draw less
You say you are thermally limited so probably that is stopping it from going higher and use more power
Idrk tho
Hmm guess I won't find out till I reduce cpu temps.
I disabled CCD1 and brought temps down by 3-4c which kept it clocking 5.8ghz more. Also I reduced the max frequency from pbo to 150 from 200 because CCD1 tried to go to 6.1Ghz which cause my BIOS to freak out.
Well... That was fun. Bclk caused my bios to not detect my windows drive so I had to tear my PC apart to reseat it along with a CMOS clear then reset my windows pin....
Maybe I won't be touching bCLK for a minute or two.
overnight run of Karhu (FPU+CACHE). This is +200mhz PBO, -60 CO on core 0 and -10co on the rest. Same settings in SuperPi 32M on core 0.
@abstract merlin so in SuperPi, the difference between -60 c0 and -10 on the rest with +200mhz compared to -60/-15 co with +311mhz is roughly 7-8s.
@abstract merlin I've noticed something weird that I will investigate after I swap my PSU.
Remember the 200w limit mention?
Well for some reason it seems the power limit is hard locked to 200w PPT despite a 9999w watt limit in PBO
There is a setting called HIGH PPT MODE so I will try that and see if HWinfo reports the same limit
Got the new PSU hooked up and I tried +200 with -20 to the other cores. Scored 332s in SuperPi 32M. So 2 seconds less.
Also I enabled HIGH PPT MODE and it raised the power limit to 280 watts.
Got bCLK 102 to work again by just letting the motherboard control the PCIE slots rather than me manually setting them
Thanks to that, my VT3 score has gone from 1.01^10 to 1.05^10. also I managed to see a peak of 5.66Ghz all core in VT3 with an average of 5.15Ghz
Interesting bCLK 101 and 103 do not post into windows but bCLK 102 can.
Hmm bCLK 102 consistently posts but is not stable. Chrome tabs act finicky and sometimes need refreshing.
I think I am just gonna have to be content with +200mhz on the CPU and instead focus on ram.
@acoustic crane @abstract merlin @fervent talon @rose oak @shell dust so after messing around I am scoring about 7770-7800 in Steel Nomad on this card (~7% above average in SN).
In games I am seeing between 8-11% performance increase over stock with +10% power limit.
-
+10% power limit (374w limit with 550w transients)
-
-110mv
-
2940mhz Memory Clock (+425)
-
Fast Timings
I highly doubt that -110 is actually stable
I thought -120 was until I had some crashes so I dropped down to -110
Then 100
Then -90
After a few weeks
Do not use benchmarks, use a variety of games
And test with a framerate cap to make a lower, variable load
Been testing using:
- Helldivers II
- Marvel Rivals
- The Isle
- Hogwarts Legacy
I tested both with and without a fps cap in HD2 and Rivals
I am gonna start working on my RAM to get an idea of what this CPU and RAM can do on the new board.
Trying 6600 1:1 at:
- 1.3v SOC
- 1.65v VDD
- 1.5v VDDQ
- 1.5v VDDIO
- Nitro 1/2/0/8x/8x
DR at 6600 is once again proving to be a pain.
Status update, it seems the more aggressive the undervolt, the less power is available for the memory so it can't clock as high. So far it appears games prefer a aggressive undervolt vs aggressive mem overclock.
Memory clock seems to give better 1% lows while core undervolt seems to yield better better peak frames.
odds I can get this stable? I'm guessing like 3-5% lol
how stable is it currently?
lol terrible ngl but oddly, this kinda works
HOWWW
Give me your secrets please
Apex did all the work I dunno lmao
Hmm.,
Did you adjust any resistance? @modern walrus
all auto tbh
I have DR 2x32GB Hynix A on the B850M which should be in a similar ram oc class as the APEX.
I only set primaries to terrible & VDD/Q to 1.4
I started at 7000 & kept bumping 200mt/s expecting it to crap out around 7400-7600 tbh
Huh interesting.
I wonder if DR Hynix scales linearly to SR Hynix when it comes to 2:1 mode.
I never got >7600 on X670E Taichi
also Ggood mind if I pick your brain about water cooling?
Let's move to #cases-psu-and-cooling
sounds good
I wonder if this could actually be stable ish
Are you putting the timings on full auto?
I set primaries only
Hynix A die?
I think I am gonna try copying your settings with slightly higher voltages.
yes. I can't get thru ycruncher yet unfortunately but did get 7zip at least
supposedly non-X3D have better IMC so maybe it might work if throw enough spaghetti at the wall lol
I think it's likely more difficult than it's worth for me to try to get this daily stable but I might try 8000 or 8100 for daily
tbh 6200-6400 1:1 might still run better anyway
I am thinking of trying it now
I thought 8000 2:1 is considered the equivalanet of 6400 1:1.
Should I disable PBO before trying?
at least when I tried 6400 1:1 v. 8000 2:1, 1:1 ran faster on ycruncher
I don't necessarily think you need to disable it but I would just to have one less variable to consider
I'll do my best
I started with 7000c46-46-46-86 with auto for everything but VDD/VDDQ which I set at like 1.3 iirc
Ran it through Steel Nomad, an hour of Vulkan Memtest, multitudes if games, overnight idle, and Rivals Benchmark
How much SOC? For voltages I am thinking:
- VDD 1.6v
- VDDQ 1.55v
- VDDIO 1.5v
I am not sure how 2:1 scales compared to 1:1.
1.255 for SOC
it doesn't need crazy SOC you could probably even start at like 1.15 or 1.2 tbh
you only need a lot of SOC for 1:1 afik
I dunno if you need that much if you're gonna start with tCL 46 lol
scaling is more with VDDIO from what I've seen
@modern walrus which primaries did you set?
For some reason this B850M doesn't follow the standard Prim/Second/Tert format
tRC defaulted to 200?
199 for me lol
I reverted Nitro to 2/3/1/8x/8x
oh ya I left those auto but 2/3/1 is auto
So let's try this:
-
Your primary timings with auto tRC
-
Auto every other timings
-
1.25v SOC
-
1.6v VDD (can always lower)
-
1.55v VDDQ
-
1.5v VDDIO
Gonna try 8,000 2:1
Gotcha. Would 1.2v be fine if it works?
Just so I have an idea as to how far I can take it.
I think that might be a little high
I mean 1.15v turns purple on Asus
usually like yellow is high-ish, red is high, & purple is like good luck
Ahh I see.
I am pretty sure I was running 1.17v VDDP for the last few months prior to swapping my board.
supposedly vddp is more for training than like runtime but that's just me repeating what I read lol
I kinda miss my MSI board in the sense that it let me set VDD up to 2.5v and it didn't have the -60 CO limit.
which MSI board? I wanted the ACE but they didn't make X870E
MSI X670E Gaming WiFi.
It let me set CO as much as -200 though I never got anything past -125 to post on my best core.
that one looked good. The Carbon was nice too
I have a love/hate relationship with MSI BIOS
Agreed lmao
I really miss pressing the left & right arrow keys to jump like half the BIOS
Better than Gigabyte in my experience so far.
I've really only ever had one Gigabyte board which was my B550 Aorus Master which was alright
I don't like that Gigabyte uses a lot of non standard names in the bios and like the timings use the 1st/2nd/3rd organization.
it had dual BIOS but no way to select one... which was kinda stupid but better than the B550-F Strix board that made you clear CMOS whenever it didn't like your timings
I think 8000 might be a no go right now
It's still training
Clear CMOS and up VDDP?
if it's still on Code 15, it's trying so let it cook as the kids say lol
I don't have a code reader
I wish I did
ah okay does it light up the DRAM light? I'm not sure if that's the same thing tho
Yeah DRAM light is on
does it always do that during training or only when it's stuck there?
Too bad they didn't make MSI Unify-X for awhile
I mean Z690 Unify-X was pretty meh but Z590 was great
Always
All I know is the dram light is on whenever it's doing anything involving memory
ah well if it's taking more than like 90 seconds, it's probably not going to POST
at 8000?
maybe try like 1.35-1.4ish for VDDIO
Anything below 8000 probably isn't worth running
Less voltage?
tbh I dunno if 8k even is
ya it gets really touchy
like I was pushing + or - for like minimum steps
Gotcha. I will set VDDP to 1.15v and VDDIO to 1.4v
oh & at 8000, maybe try tRCDWR/RD & tRP at 48
Seems my VDDP was already at 1.1v
I think auto usually sets like 1.094xxx or something so maybe 1.15ish
wait I think it's 0.94xxx on auto
Kay so:
-
tRCD 46 to 48
-
tRP to 48
-
tRAS 86
-
tRC 200
-
VDDIO 1.5 to 1.4
-
VDDQ 1.55 to 1.5
-
VDDP 1.1 to 1.15
any progress? maybe try 7600 just to see if it runs then step up. or try auto to see what the board comes up with
Holy sh
We may have made progress
Still on DRAM but it's actually training.
Before it was stuck
Dang it
Boot failed
maybe drop VDDIO to like 1.35?
gotta find the IMC's happy place lol
cuz something with like signal traces & noise or something beyond my mind's ability to rationalize
I summon the lucky belly
like people who can turn a little green square with metal on it into games... I dunno how their minds work I can't be on their level lol
Kay it's rebooted and now entering training
The board is a B850M Force V2 WiFi
It's nice that if it fails to get into BIOS rather than being stuck it undos the settings and loads me into BIOS with a pop up.
.
Another post fail
I am gonna try this Memory Target option
does it have memory presets? sometimes that gives you clues on what the board wants
Not that I see so far
hmm could try manually setting the on-die terminations?
I am setting memory target speed to 8,000. Bumped VDDP to 1.18v and VDDIO to auto. Let's see what she does.
Elaborate please.
these crazy things
this is all I've figured out about how (some) of these work
RttNom: Idle/Steady State
RttWr: Writes from CPU to DRAM (tWR, tWTR_L/S)
RttPark: Long sequential writes
RTTParkDQS: read timings alignment
Another boot fail....
Do you think it would be more worth while for me to try getting 6600 1:1 stable since I was able to post with that?
if you can get 6600 1:1 to run, that's probably way better
I couldn't get that stable at all
like it can POST but wouldn't get thru any benchmarks
I was able to game on 6600 when I was in Alaska prior to a bios update
So let's try that
For 1:1 higher VDDIO is better yes?
Hmm we will try 1.45 then rather then 1.5 from the get go
like 1.22 VDDIO & 1.265 SOC for 6200c28
That seems like a lot for 6200
I was running 6400 at 1.24v SOC
I am in Windows @modern walrus
6600 1:1
Poop timings
nice. leave them loose & stress test to make sure voltages are okay before timings iirc
VT3 or Karhu?
Well it isn't staining it's pants so far
Running Karhu GPU+Cache
Minute in
nice nice
I kinda gave up on 8200 getting stable but 8100 is getting thru ycruncher now at least
Like let me put this in perspective
I wonder how high you can run FCLK
I couldn't get past the 1 minute mark no matter what on either of my MSI boards
cuz I think FCLK runs higher at lower VSOC or something counterintuitive like that
I am now 3 minutes in with no stutters or issues
F
It just errored
Still this is major progress
1.65 VDD might be too heavy
You think it's to much?
just drop tCL to like 38 or 36 maybe so it needs the voltage lol
oh yaa that's a thing for sure
I remember on DDR4 with bdie as soon as you hit a specific temp, it instantly crashed
I am gonna put a fan on the ram and drop voltage to 1.6
damn & you don't even have tREFI maxxed out or tRFC super tight yet
which seem to really affect temps iirc
btw MSI tomahawk is a pretty good looking board
the like green & black one
@modern walrus this look good lmfao?
ends justify the means
Running Karhu at the same settings just with the fan to see if the reduced temp makes a big difference
it'll probably run until it hits 50C
Sad it errored at 38.
damn maybe you do need that much VDD
I really hope it's not a VDD limit
Cause this board has a stupid limit of 1.65v
I raised VDDG to +100mv from stock and lowered VDDIO to 1.4v. will try 1.5v VDDIO after
maybe just tighten timings then
Seems less VDDIO causes errors to consistently error faster
well no two IMCs are exactly the same ig lol
plus the motherboard affects things too
I know 1.5v is safe to daily on 9000 series but I am not sure if say 1.55 would be safe if that's how much it takes
damn on VDDIO? maybe I should blast it more
From what I know yeah. I've run 1.5v VDDIO for about a year now and there are other folks in the OC discord that daily 1.5 for their 8300-8400 2:1 setups
7000 series the safe limit is considered 1.45v
I feel like as long as it's not overheating it's probably not thaat bad lol
oh I see
cuz of the way they moved vcache or something
why'd you get rid of the MSI board btw
Two reasons:
-
Downsizing my case to the custom one I am making which I can bring as Carry-On
-
Because I was able to acquire this RAM OC board at half price.
ah I see
I sold my MSI board to make up for all but $5-10 for this board swap
We made it past 3m!
And going strong
Nitros are at Robust, 2/3/1/8x/8x
If we make it to the 5m mark then I will try enabling XMP. @modern walrus you think this gonna work?
We did it!
gotta drop timings now lol
I am gonna try XMP with tCL 26.
damn how much VDD does that need?
We are gonna find out. I was running VDD 1.85v till I got this board with a 1.65v limit. The extra VDD at 1.85 was because I floored a lot of the timings
And got tRFC 115ns stable xD
how do you cool that without like water or LN2? lol
It's called quality stock heat spreaders that use a thermal pad, all aluminum heatspreader, and slotted for air to pass with a 180mm fan at 100%
oh ya that's a lot of air moving there
I bought 15 Arctic P12 Pro fans for my custom case
I plan to setup the rad fans as pull so it blasts extra air onto the ram since the rad is mounted to the side panel
That should help me keep things cooler
oh ya I love those things cuz of the 5-packs lol
Agreed
Btw posted without issue
Running Karhu for 5m then I am gonna try playing HD2 with my pals on the ram tune
@modern walrus you have HellDivers II?
F it erroed 1m in
Gonna try the XMP tCL of 32
I don't
I probably should
At least Resident Evil Requiem comes out in like 6 days
Btw I am finally making the jump to QHD.
Bought a 240hz QHD Mini-LED monitor from AOC.
hot take: I always play 4k lol
I like high FPS. Anything below 120 feels a bit choppy to me.
Argggg
This is being a pain
Errored again
I play 4k even tho I really need to get glasses lol
I wonder if I am actually limited by VDD
that seems like a lot of VDD for tCL 44 lol
Only thing I changed were loose primaries to XMP
Makes me wonder if this means I need to either get a board without the 1.65v VDD limit or get better ram
But THAT IS GONNA BE EXPENSIVE
@modern walrus @steady lance what are some good CPU stress tests that find errors fast while not taking too much power (prime95 - small/smallest fft). I already tried linpack and cinebench r23
My cooler can handle up to 210w fine before it starts to thermal throttle at like 216w+
Karhu (FPU+CACHE) and Y-Cruncher VT3 for all core. For single core I like to to run SuperPI 32M set to the specific thread.
Thanks
Trying to get 5.5ghz to work at 1.2v for the 13600k
OCCT is pretty good & has plenty of testing options https://www.ocbase.com/occt/personal
Ocbase is the home of OCCT, the most popular all-in-one stability / stress testing / benchmarking / monitoring tool available for PC
don't forget load line calibration
still leaving that on auto lmao
gah 5.5ghz doesn't work, guess ill lower it to 5.4ghz or maybe... Lower the ecore clocks
already using OCCT btw
so it managed to pass occt linpack test but not r23

OCCT will overheat his cooler though
I am very frustrated. I am like 99.9% certain I am VDD limited for 6,600mhz 1:1
Cause 6,600 with loose timings works but if I tighten timings it becomes unstable.
.....
G.Skill 6000cl26 kits are now $1,000 usd
Got whea errors with 5.5ghz, that clock speed needs more than 1.2v to be error free. I'll stick with 5.4ghz again
... Can I shunt mod the VDD power delivery?
Any particular reason for not using like 1.225v? I am not to familiar with Intel overclocking.
Too much power draw lol, it gets to a point where it's not worth it
At 1.195v I'm already reaching around 175-178w with r23
And 180w ish with linpack
Ahh so you can't go higher without a better cooler?
Goal was to run this chip behind its advertised turbo tdp (181w). Cooler can handle up to 210w without throttling
Say do you per chance know how to mod bios? @modern walrus @alpine mantle
Definitely not
how bad is it if my vsoc spikes to 65000W
Lmao I love this misread.
Reminds me of when HWinfo thought my V-Cache CCD hit 10.9Ghz
You know
I am regretting not buying the 2x32 6000cl26 ram kit when it was $350
I had instead at the time opted to buy some headphones at $400
Honestly it'd be cheaper now to buy an X670E gene and not have the VDD limit.
i think thats i have simular ram
bought mid 2023 and probably has 2x in price
I paid $230, its now $900 lol
Why is it runny at 5800mhz...
6000 was unstable and too hot
Sff case?
no im in the antec flux pro but I think the dimms are just in a dead spot in the case
gpu blocks bottom intake, qDIMM blocks front intake
@latent wave embrace the inner engineer
good luckk
i transporrt my PC to and from college so its a bit too much jank right now
Mhmm gotcha
I am working on a custom 45 linear inch case right now for travel purposes
@latent wave @alpine mantle @modern walrus at least it seems 6400cl26 is doable at 1.65v
Nice, and I got my CPU stable at 1.21v @ 5.5ghz
Lovely.
Niceee..
I am gonna need to drastically improve my cooling. Right now, tREFI max is causing the ram to get past 50c. I miss my 10c ambient....
Getting addicted to cpu oc again for daily. Managed to get the 13600k up to:
P-Core: 5.5ghz
E-Core: 4.4ghz
Ring (cache): 4.8ghz
I've tried linpack and r23 as stress tests to prevent thermal throttling and those passed. Can't use small/smallest FFT with avx2 because it draws way too much power my cooler can handle.
I feel like increasing the ring clocks up to 5ghz to finalize it.
Interesting that on Intel cache can be tuned separately.
Yep, so far 5ghz doesn't work at 1.220v so I'm trying 4.9ghz for cache
That will be the last tune for the 13600k
I see. One of the biggest limits for OC on AM5 is that the cache boosts with the cores. To hit 5.75ghz, both the cores and cache need to reach that.
That sounds super difficult ngl
Temps become a serious issue. I can't get 5.75Ghz sustained unless the CPU is within 50c.
Past 50c it quickly drops to 5.6ish
Guess it has to be cooled by a fridge lol
Quick update on the 4.9ghz ring clock, it failed. I'll go back to 4.8ghz and lock in the tune
My stable GPU OC so far:
- +10% PL
- -70 UV
- 2900mhz with Fast Timings
Survives gaming, blender, benchmarks, etc.
I am impressed by how much of a difference, reducing tRFC makes.
Y-Cruncher VT3 scores:
- 6400 with auto timings: 1.29^10
- 6400 with XMP and tCL 26: 1.3^10.
- 6400cl26 with XMP and tight tRFC : 1.37^10.
@alpine mantle @modern walrus
I am going to need to come up with a better way to keep temps under control.
0.o this board is starting to show it's OC abilities. Managed to post tRFC 115ns without any issue. On my MSI boards I never got below 117ns and only 120ns or higher was stable.
Hmm I am gonna try to bring this down another 2-4 ns or so.
I won't be able to raise tREFI till I drop my temps significantly
tRFC 350-365 refused to post. tRFC 368 did so if this works I think that's going to pretty much be as low as I can go.
Did not work.
So far lowest that passes VT3 is tRFC 370
@latent inlet I know you're here
🕵️
I saw you typing.
I got tRFC 370 (115.75ns) to post and pass VT3 though tRFC 368 (115ns) posted but did not pass.
Okay so I reduced VDDQ because it was rather high, and that gave me the thermal headroom to max tREFI. The score in VT3 is now 7.2% faster compared to 6400cl26 with XMP plus auto.
Gonna loosen up tRFC to 116ns to avoid freezes.
Indeed, that and trefi makes a noticeable difference in latency/bandwidth
I think I am gonna do fCLK next before working on timings.
Man I swear, most of the folks over in the OC discord have a nasty attitude. All they say is good or bad with no explanation and when asked you get cussed at.
Must be all those hours on ram tuning leaving them cranky XD
Anyways I'm itching for a 5.6ghz overclock now. My 5.5ghz tune is mad stable
Goal is to get it running under 1.25v
Now at 1.240v
Guess that didn't work lol
Gonna try 1.250v and that's it
Thermal throttled, ok I'll just push voltage up to 1.230v just to max out ring clocks
That doesn't work either, guess 5.5/4.4/4.8ghz is the final tune. Honestly not too bad!
All of these pics is the current OC tune passing those stability tests. I'll keep this till my next major pc upgrade, not planning on getting a 14th gen cpu for my socket. This is satisfactory 🙂
[Benchmark Scores]
Use an AIO.
In any case, well done.
I know! TEC (peltier) with air cooler on top for the TEC.
Lol
I really am fine with this tune tho, games feel a lot smoother/fluid now so im thinking the 1% lows have increased
good stuff
@alpine mantle @modern walrus I may've found a way to get 6600 1:1 stable. I noticed that at stock settings my SOC voltage was averaging about 15mv lower then what I set it under load.
So I just raised the LLC SOC level and so far it's passing VT3.
Dunno if that is super tight or not but its good that its staying stable from VT3
1:1 tho sheesh
Most 7000/9000 can't run 6400 1:1 stable. Highest stable for the average 7000 and 9000 series is 6200mhz 1:1.
2:1 is very dependent on the sample and board rather than the ram.
@alpine mantle @modern walrus 24m in Karhu FPU+Cache so far.
passes karhu now i need it to pass multiple runs of vt3
Power limited right now since I am at stock settings.
Without limits it pulls around 250w.
lol nvm then, my cpu would pull up to 230w and throttle hard if i were to run prime95 with avx2
Fair. I need to use Push/Pull on my aio to keep my CPU from thermal throttling when pulling 300w.
Though since my ambient now around 26-30c not -10c the CPU refused to pull 300w.
i see all these poeple with thermal throttle issue on their 9950X and i be sitting here with my cpu at 40c gaming
could just be posts about putting their cpu through a stress test or that their cooling solution is done incorrectly. They can always undervolt it anyways
Gaming is very different from sustained high load.
9950x draws 170w stock. I am talking about preventing my 9950x3D from throttling when it was pulling nearly double that at 300w.
so you guys are ocing and doing runs with the settings
Yup.
I also use the OC settings as my daily tune where the cpu is hammered in 3D apps like Blender.
Did you enable SOC Uncore mode? It's in AMD overclocking settings somewhere. Powerdown disable good too
Decent showing for cpu at stonk clocks ig. FCLK runs 2200 at 1.255 ostensibly well
Delid seems successful so far
What does UNCORE mode do?
It defaults to enabled I think.
Yeah gaming isn't comparable to a synthetic stress test. 170 though is the TDP which is just a number amd pulls straight out of their butts. PPT is usually about 30% higher than TDP on their cpus at stock so seeing 220W+ in stress tests is pretty normal. It's kind of like intel advertising pl1 and boost clocks in the same breath.
Lmao fair enough. Though that's assuming PBO isn't enabled. The rated TDP and the PPT is usually very similar in my experience if PBO is disabled.
I spend some time at a tech support forum and one of the tests we have people do is running prime95 and furmark with hwinfo logging to chart the results. Don't see a whole lot of r9s but seeing the 65W ryzen cpus running in the mid 80s ppt and the 105W ones up around 130-140 is the normal result. The x3ds are generally closer to the advertised TDP though, the 7800x3d in particular is typically well under the 120W TDP in real use. Prime95 is a synthetic test and about as worst case scenario as it gets for power draw so maybe there is a difference there if you're using something more realistic. Our goal for that test is typically mainly to check temps and voltage regulation.
7600x is another one but that probably didn't need the 105W TDP they gave it.
it doesn't let SOC or iGPU power down. helps with stability sometimes
Disable it to keep it static or?
This is making a huge difference in my RAM temps.
7c change in RAM temps on the P12 Pro vs my 180 GP-18 fan blowing directly down on it.
1.6 gets spicy for sure
enable to keep static
Gotcha. Right now I have SOC set to 1.3v with UNCORE enabled and flat LLC for the 6,600 experiment.
I will be damned if it takes north of 1.65v VDDQ to get it stable because I am fairly certain the IMC can handle 6,600 1:1 based on the above.
Significant progress! Made it ~25m before erroring wheras it was erroring between 1-5m in. Changing ProcODT PD helped a lot.
@modern walrus @alpine mantle
Decent progress
I made some changes and it's now 40m in with no issues.
More importantly, I have two P12 Pros blasting the ram from the top and bottom now to keep temps under control.
Is the ram liquid cooled?
Currently air cooled with two P12 Pros at 100%.
So I've tried every ProcODT PD impedance from 53.3 to 480 OHMs.
Going below 50 OHMs constantly errored and 120 OHMs lasted slightly longer then 480 OHMs at around 28-29m vs 26m however 120 OHMs scored between 1.16-1.17^10 while 480 OHMs scored a consistent 1.18^10. Do you think this means 480 OHMs is more stable?
@modern walrus I managed to get 6800 1:1 to post out of curiosity.
Gonna start futsing with ProcODT PU and see if that helps with stability at 6,600 1:1
1h30m, huge progress!
@modern walrus @alpine mantle @somber perch @distant cypress @acoustic crane @abstract merlin
Hahahahahahahaha!
Finally I got 6400 1:1 with GDM disabled and Nitro 1/2/0 to not just boot but also run!!!
This was only like my 50th attempt or so if you include the other two AM5 boards lmao
Man it feels good to finally pull this off after a year of futsing with different boards
Apparently what I had to do was change the On-Die-Termination settings for CK/CA/CS
What I've learned from this Fiasco is that although a good board makes a big difference, changing the ODTs to better match the components makes a much bigger difference.
still going strong
Tuff
What are on die terminations
Oh resistances
Yeah these I have no idea about
Yep.
ProcODT PD and PU control the signal between the RAM and IMC.
All I know is that they exist and are a major factor with things like mcr and memory training
And dialing in overclocks
Those are what helps me stabilize 6,600 1:1
Right
And right now I am futsing with CK/CA/CS drive resistances to get GDM stabilized those to my understanding affect the resistance of the traces within the ram directly.
I am trying to get 6600 1:1 to boot with GDM off and Nitro 1/2/0 right now.
Though right now that seems unlikely
Isn’t your infinity fabric clock supposed to be a little higher?
Or is that just splitting hairs
Isn’t it supposed to follow the other clocks increasing? With the memory clock at 3200 shouldn’t it be 2133?
Stock IF is 1800 though with XMP/EXPO it goes to 2,000.
Right now mine is set to 2000 to minimize variables though my chip can do 2,200 stable.
As for the whole syncing the IF to the memory clock, that's only beneficial if you are chasing a fraction of a ns to get speed records. For real world use, the higher bandwidth provided by running the IF as high as possible yields better performance.
AM5 CPUs have rather low bandwidth so increasing the clock of the IF makes a big difference for the total bandwidth of the processor.
@abstract merlin does this help answer the question?

Oh true that is a really small gain
after trying out my highest daily 13600k oc (5.5ghz pcore/4.4ghz ecore) the gains were very minimal, and even if 1% lows may be improved it was smooth at stock to begin with. I managed to undervolt the processor with stock clocks at a -0.150v offset and it consumes 128w all core on cinebench r23
performance in games remain fine
I guess it really is true that these cpus are already too strong out the factory lol
About a 9.8-10% performance gain by:
-
tCL 32 to 26
-
tRFC1 280ns to 118ns (how fast a cycle is complete)
-
tREFI 12,000 to 65535 (How long data is stored till a new cycle is needed)
-
fCLK (Infinity Fabric) 2000 to 2200
The performance difference between 2000 and 2200 fCLK is about 2%.
Tune the ram to the moon!
not going back to that herculian task
9.8% raw computational performance increase by just tuning the heck out of the ram.
Brought my average FPS in HD2 from 115 to 135ish fps.
i already tuned the ram and i think its fine
Compared to stock 6000cl32 XMP
Seems moderate. tCL seems unusually high for a tuned 3600 profile?
didn't bother changing because the kit i have is not so great
Gotcha
Right now I am debating 6400 1:1 with GDM off and Nitro 1/2/0 or 6600 with GDM on and Nitro 2/3/1. Honestly I think the performance difference would be less than 1%.
6400 GDM off continues to be rock solid.
Currently working on SCLs.
@modern walrus @alpine mantle any progress on your ends?
Also I was right about this. The difference is margin of error so I am opting for 6400 1:1 with GDM off and Nitro 1/2/0 so I use less voltage.
just stuck with an undervolt profile on stock clocks instead of overclock lol
@steady lance reverted back to OC profile lmfao
good
Did 5 runs of stellar blade with this OC profile, memory crashes after 2684mhz it seems. Also had to turn down core UV to -40 but its stable now. Not gonna increase power because the gains are minimal (extra 1-2fps)
What about 1% lows with high pl?
don't think theres a difference tbh, games already running smooth lol
8200 is a no-go. 8100 has been such an uphill battle that I just dropped to 8000 a few minutes ago. I got 8100 thru OCCT's memory test for an hour but it errored out at something pathetic like 38% on Karhu. I've been trying to tune VSOC & VDDIO but haven't been able to find the right combination yet. It's been annoying as hell frankly lol
I suggest starting by lowering vSOC then reduce VDDIO.
1.55v VDDIO is fine for daily on 9000 series.
I've been at pretty much every setting between 1.0-1.2v on VSOC & 1.3-1.45v on VDDIO lol
what's your voltage combo at right now?
for 8000c36-46-46 it's VDD/Q 1.55; VDDIO 1.35; & VSOC 1.2
CPU doesn't like >1.4 apparently for VDDIO
Interesting.
I should probably go back to the drawing board & auto all the timings except primaries
I wonder if it's because you are in 2:1
like it'll POST >1.4 but once it's at or above 1.45, it'll error loading windows
I read where someone had better luck setting RTT Wr to 80 or 60 ohm instead of 120 but it doesn't seem to wanna POST lower than 120. It'll POST at 240 but that doesn't really seem to provide any benefits
I dropped FCLK down to 2000 in hopes of eliminating that as a variable but it'll POST 2200 fairly easily
I'm almost annoyed enough to open Excel lol
I haven't made one of these ridiculous things in awhile
ah well I'm gonna play Resident Evil Requiem on 12900K/3080ti for now ig
@modern walrus solution. Go DD with a chiller to safely run SOC above 1.3v
babg, base and bbse are in the epstein files 😭 😭
just put a radiator outside your window with anti-freeze in it lol
does your board let you set >1.3 VSOC? I think mine might if I enable LN2 mode but I haven't tried
Mine has the mode but I haven't tried using it.
ya I'm reluctant cuz of those stories of CPUs getting hecked on Asus & Asrock boards even with the 1.3 limit
Almost always due to very poor contact and insufficient cooling
I was running 1.31v for about a month or so on my MSI board.
hello i could use some help with my 7600x
so im watching this famous video about "fixing" ryzen 7000, and at 4:55 i get this screenshot for what to manually limit PBO to
ryzen 7600x , msi mag b650 tomhawk what would i do the numbers for
on my bios, I have this
I’m so confused
@kindred spade is your goal to get performance or make the chip more efficient?
If it is performance than set PPT/TDC/EDC to the maximum value that your bios allows and then enable Boost Clock Override, set to Enable Positive, and set it to +200, from there set a Negative All Core Curve Optimizer value of -10 and see if you hit those clock speeds and whether or not it is stable.
Trying to make my chip not hit 98c on HWinfo every time I open a game. Ive been told its voltage issues.
If it is reaching 98c on stock than you have a serious issue with your TiM or a insufficient cooler.
It’s a b650 so I’m probably screwed
-
What is your ambient temp (approximate)
-
What TiM (thermal compound)
-
What cooler
Has no affect on CPU temps.
If the voltage deliciery is faulty (which this b650 msi mag is known for) then yes, it does affect cpu temperatures
I know the for sure lol
I am doubtful of overvoltage. Not impossible but doubtful.
Not trying to be obtuse., trying to learn, why are you doubtful?
If it was supplying a excess amount of voltage you would be running into instability issues, crashing, and a long list of other issues. In addition your bios is reading normal idle voltages for the core.
In HWinfo does your core voltage ever exceed 1.4v?
Please bare with my tone. I am just trying to be as straightforward as possible.
Also haven't slept in about 20hrs.
No no. You’re trying to help, I am infinitely grateful for it. I also haven’t slept a lot lol, trying not to be difficult
I do get some crashes and micro stutters, but that could be caused by temperature issues so I’m not sure.
I didn’t see it go above 1.4
Hitting 98c in gaming at stock with normal voltage readings reek:
- Poor cooler contact
- Insufficient cooler
- Bad mounting pressure
- Bad TiM coverage
I'd honestly take the chip out, wipe the contacts with some alcohol, blow out any dust in the motherboard socket then carefully reseat the CPU and reapply the TiM and mount the cooler with even mounting pressure.
Shotgun approach the possible issues in one go.
If you don't have 100% alcohol, 70% or higher medical alcohol works fine. Just wipe the contacts down with a lint free paper towel (paper kitchen towels are good for this) and blow on the contacts to help it dry off faster. About 15m is sufficient.
Then your voltages are normal.
1.4v is the max voltage the CPU is allowed to be supplied with, without any overrides.
@kindred spade try the above and let me know if it is fixed.
Also what cooler and Tim are you using?
I have 90% iso that I use to remove paste. Shouldn’t be an issue
Good.
Artic fox paste? I think
Peerless assassin
Arctic MX?
There is no Arctic Fox paste.
Not ideal but sufficient.
Better frame timings and lows with a negligable boost to max fps as that is affected almost entirely by the cpu and gpu if the ram isn't the bottleneck.
I currently have a 3600, 9060xt 16gb, and (2x8) 16gb of 3600mhz
I'd expect better frame timing and lows then with a minor uplift to max.
maybe ~5% difference on average if you can get 4,000mhz 1:1 stable.
If you play ram sensitive games like Marvel Rivals or Minecraft, the difference is much larger.
Damn.
So what would be better, going to 32gb of 3200mhz? Or overclocking to 4000 1:1? @steady lance
The higher capacity will yield no benefit unless you're using your current 16GB max.
What’s a good starting point for PPT, TDC & EDC limits on a 7600x?
I have PBO enabled with throttle limit set to 85 & curve optimizer -30.
Are you going for performance or efficiency? If performance just set it to the bios max as the CPU will only pull what it wants.
Well, I recently updated my bios and now cinebench goes to like 80c. I got it down to 70c but on my other bios I wouldn’t go higher than 60-65c.
Which is weird because the only thing I did on the other bios was turn on PBO & my curve optimizer to -30.
What I had to do now is manually set PPT, TDC & EDC.
So it’s almost like the -30 wasn’t really doing anything on it’s own anymore.
What is your score in CineBench
Compared to before.
Sounds like the bios changed the boost behavior
It was about a 600 point difference.
15505 on the old bios
14909 on the new bios
I’ll double check
600 points is a big difference.
Hi guys, i can send images here? I perform my first undervolt today and im looking for opinions
gave my top 10% bin 265kf to my fiancé so i treated myself to this. hopefully it oc's better than my potato second 265kf
Yes
Thank you
Dont judge me if i do something wrong, is my first undervolt lmao. I was reading and looking information about undervolting and get some stability on GPUs and this was my first one. Tell me if i forgot something, thanks guys! (Radeon RX 6650 XT / Ryzen 5 5600)
PD - Power Limit on 10%
Im not looking for something competitive, i only want some stability and better performance to play smoother
Set min frequency 100 less than max.
Turn off fast timing.
As you test you can keep lowering the voltage too see what gives you a better score
Do you have Expo enabled in bios?
I got it back up to 15280
Thank you! I will try this today and let you know the new score.
I perform a factory reset on Adrenalin to start the undervolt fresh and now on every benchmark on time spy adrenalin crash out to default setting. I do something wrong? 😅
try DDU. sometimes adrenaline gets upset from too many crashes and just refuses to work right
Thank you my friend! I do a AMD Cleanup Utility, new drivers and give me a boost on score thank you again 🙌
AMD cleanup doesn't clean everything, DDU does a full wipe. but if your current setup works, i suppose you've done a fine enough job
Yes I thought about doing the DDU right away, but I saw it was working perfectly and left it alone haha
is this enough to be called clock stretching?
average clock vs. average effective clock in VT3
Averages should not be measured over multiple cycles because between cycles the clocks change rapidly.
hmm seemed consistent to me
Startup VT3, reset the monitors in HWinfo and eyeball the Average Clock and Average Effective Clock whilst it is running the test cycle. If the variance is within 50mhz it's fine.
Ideally they should be the same but that's in a perfect world.
yeah it was always within 50mhz was just wondering if 25-30 was too much
You're doing fine.
A perfectionist will say otherwise but that is fine and working. If it's over 50mhz that's when there is definitely a issue.
Even a stock CPU with no PBO enabled will have like 25mhz of variance.
btw, turns out a single run of cinebench 2026 is a better CO stability test than a week of aida64
lol
Aida64 is only good for ST co stability.
I remember cinebench r23 being less of a stability test than opening firefox
I usually run PI 32M as it's super light to guarantee the core can actually reach the desired clocks then run VT3 for all core test.
oh yeah i passed vt3 for days too
with basically no fluctuations between cycles
couldn't even pass 2 tiles of cb2026
Try PI 32M I've found that to be good at detecting how the CO affects the cores.
Interesting I will try this.
CB2026 in MT it ST mode?
What CPU again?
@short blade did a comparison run whether or not my 9070 XT benefits from a vram OC while heavily limited in power draw (-30pl / 213w)
Left: 2518mhz stock + fast timing
Right: 2734mhz + default timing
should be able to squeeze out another 100-150 pts if I dropped fmax to +100 and set a more aggressive curve but I'd rather keep fmax +200 cause the chip easily holds max clock in games
think you should do fast timing on both for an apples to apples comparison
but yeah it makes sense at 213w the core and vram might be fighting each other for power budget
i guess so, 2734mhz + fast timing did a black screen hard crash the last time i used it when screenshotting stellar blade lol
💀
I don't like changing vram clocks on this card specifically because for whatever reason it just wants to black screen
changing timings don't do that
so it gets very annoying to hard restart the computer many times
Always optimize for what you use it for. You could use Curve Shaper but that things very tricky to get right.
I tried messing with curve shaper for a bit as far as I could tell it wasn't actually doing anything
but that was a long time ago on a much older agesa so idk
I wish AMD would enable setting separate fMaxs for each CCD.
is there an agesa changelog for am5? could only find one for am4
agesa 1.3.0.0a finally fixed wifi/bt on my board
I haven't checked in a while.
Last one I had, for the Force V2 board was improving memory support.
buying ryzen always feels like buying into a beta test
better than buying into intel and having your cpu self-immolate though 👍
Seems the Force has had two new bios since
Both are for 9950x3D2 support/optimization and reducing vulnerability with Mixed DDR mode.

