#overclocking
1 messages ยท Page 32 of 1
I have to agree, at least when it comes to modern boards and based on my limited experience
asrock rgb settings are miserable, everything else works for me
using anything other than asus bios
mfw
^^^ this. I've used them all, asrock is the absolute worst, msi next worst, gigabyte you need to bug CS rep for custom bios, and asus crosshair/maximus is the best
And yes, asus has different bioses from prime, strix and C/M
Asus prime BIOS is THE worst tho
I can attest, the prime SUCKS
That's what you get for a sub $150 mobo though
Because the internet told you, or first hand experience?
They've been actively working to improve that as far as anyone can tell. If you had a bad experience in the past like they made you pay for warranty repair or they just refused to, you can reach out to them now and they'll retroactively fix what went wrong.
Even if the hardware in question isn't in warranty any more
Tragic
I got a world record score in a 3DMark (hardware specific)
http://www.3dmark.com/sn/2045050

They're still selling *poopy "refurbished" monitors
Granted they're priced pretty cheap
But I'm 2 for 2 on getting scratches/dead subpixel/debris stuck in the panel
I mean, people dont send back good panels
you say that, but people might buy it then use it once then return it
costco had that problem with their tvs during past superbowls and they had to destroy the tvs since they dont sell returns
Yeah, and costco also revokes memberships if you do that with any sort of frequency 
Costco has great customer service and return policy, but you also have to be a decent human being and not abuse the costco overlords
Costco is one of the few good companies all around
can you say in detail?
great deals and programs from what ive seen
dont feel scummy lol
i think its more customer orientated tho versus employee
mainly their customer support
Thanks for your reply.
Im a software engineer.
okay, this is my last OC/UV run, managed to get the cpu stable on such a low voltage while keeping it slightly overclocked on pcore/ecore/ring
This is hitting on a thin line of stability though as anything lower will cause it to shoot Whea errors
really satisfied with the result tho
P: 5.2GHZ
E: 4.1GHZ
R: 4.6GHZ
-.165mv offset
CPU Lite load: 2
IA CEP: Enabled
OCCT LINPACK: 139w max
R23: 131w max
WHEA just means "WHEEEEEE A chance for more overclocking fun!"
nooooooooooooo
overclock
What in g0d's name am I looking at...
phase change cooling
aka they strapped a refrigeration unit onto a cpu
oh cascade as well
they're refrigerating the refrigeration unit via the black cube thing which is a heat exchanger so that it can get even colder
It's a 3 stage cascade, fugger is his username on hwbot if you want to see more
Also the founder of Xtremesystems, he built that unit in college 20 something years ago
oh i missed that rotary or scroll compressor under that
lmao i was wondering why there were 2 other condenser units asides from the one big one
oh was this a reciever tank? looked like a rotary compressor or smth lmao
Probably, or a funny looking accumulator
cant see the 3rd piston compressor lmao
I am selling my 3-stage cascade to make way for my next build, costs are becoming a factor to my goal of -200c. Fully functional, no leaks or issues. R507/ISCEON89/R1150 -110C http://www.xtremesystems.org/cascade/c3.jpg 3x Danfoss SC12 1/2HP Compressors http://www.xtremesystems.org/cascade/c4.jpg...
If only I wanted to invest in one
You can build that under 2k
Whats the point of getting the temp so low?
When it gets below a certain temperature, the silicon itself undergoes a change that causes less resistance to occur inside of it. This results in better overclocking scores because it will need less voltage to get higher clock speeds and will be more stable at those higher clock speeds. The trick is keeping it below that point for the duration of the test but not going so low it causes problems elsewhere, so you generally target the lowest temp you can get without those problems (cold bug) and let the temp rise a little during testing.
Id think the parts would become pretty fragile under those kinds of testing conditions?
If you hit it with a hammer sure
It becomes brittle but that doesn't mean it'll break instantly
Aw man but i love hitting my rig that costs 10k+ with hammers 
If anything the paste cracks from thermal stress
From the #1 rank on team wegg himself
So when does splave start passing that stuff out to wegg team?
Already has ๐
help how to how to optimize How to optimize the CPU correctly in Ryzen Master. This is how I have it. I want to know if everything is okay or if I need to adjust it better.
Space heater moment
First time overclocking, i can push to 2680 on gpu clock without crashing but i tuned it to 2650 to be safe, good boost from default 2500
Cut out 15mV as well
And get a stable 2400MHz with the vram
Its a nice boost in performance, some stuff its like a 4 to 5% fps boost but most games i can pull about 10% extra perf
(On parts in my user btw)
this is using adrenalin
Getting out anything from the VRAM is kinda good on a GRE
Most GRE's use minimal binning I've found
Couldve definitely been worse, cant say i won the sillicon lottery but not half bad
Rate the OC
Looks pretty good, do you have the compare link? It's hard to see what clock frequency and such it's running at from a chart
What do you mean by compare? I have the rest of the graphs for everything in another screenshot, it's on my computer so can't post rn
The button that says "compare" under the score
And I only have the demo 3d benchmark so a lot of features are locked
It'll bring you to a webpage where it uploads the result
Gotcha, ill run another when I get home to see
I was pretty happy with results
My cpu is now clocking 5050mhz constant for the most part, def saving me in modded minecraft lmao
Time Spy is included in the demo, and has a CPU test if you want to compare that too
Yeah i was hesitate on running that because I didn't know if it was bad to undervolt my cpu and overlock my gpu
But I was not impressed with the underclocking gpu had to offer.
That's perfectly fine to do
You'll want to run a test to make sure the undervolt is stable though
Yeah everything is running smooth like i didn't do anything so I've been wanting to run it now
That is fair, I saw a couple of post about negative 30 being a bit to much on pbo, but it's been fine so far for me. Hogwarts is running pretty solid
Unstable undervolt is one of those things that you won't know it's a problem until you get a random blue screen or the computer hard locks
I was more worried about games than benchmark score, but when I finally got a good score on the "main" one i tested out my games and everything seems stable
Appreciate that, will definitely run the test first thing in the morning when I'm home from work
I recommend y-cruncher for that. Drop this file into the folder and run it.
Assuming it crashes, just restart and turn back curve a bit in pbo?
It'll run until you stop it, and it'll keep track of any errors it encounters in a log file
It will tell you which core crashes
Is that an app?
Sorry this is a bit new to me ๐
Appreciate that, will look into this as well
I recommend letting the stress test run about 12 hours. It'll test one core at a time by default so it knows if it crashes.
If it does crash a core you can increase the offset (closer to zero from negative) on that core
Damn lol but okay, good thing is I'm bout to have a long weekend
You can use the computer while it tests, fps might be impacted if you try to game though
Fair, appreciate that
first one is spy nomad, 2nd is nomad
I meant the address for the webpage (it's public by default anyway) so it's easier to look at but that works
And that appears to be 2 of the same screenshot
Seems like there's room for improvement even if you can't get the core clocks higher, but that's still a solid score. Possibly some driver tweaks would be needed, it mentions LOD being modified in drivers. https://www.3dmark.com/compare/sn/2481046/sn/2354910#
Interesting I'll have to look into that
Crazy you just pulled my sheet up like that lol
I see, yeh on nomad I couldn't get it to go into boost or my score would've been better
Def gotta look into it cause buddies ran 10ยฐ cooler and more better
Assuming he has silicone lottery on his lol
I deliberately picked a result that was running near the same frequency as you
Both on CPU and GPU
9 degrees isn't going to be 800 points worth of difference
Your video memory clock is a bit lower but nothing major
Yeah steel nomad will not clock at 2850, my games will
Yeah my settings are maxed for oc besides messing w fans
It's probably more down to the driver settings like I mentioned
Ill look at that when I'm back home
Also that software for the cpu is pretty cool stuff I read about it, slightly scared to download lol
But if it does pi indefinitely on a core/thread how will it know when to stop to go to the next one?
It switches cores after a set interval of time
There is absolutely no way to damage your system with it
It's just a program like any other. It uses even less of the CPU than an average modern game.
I wasn't worried about damage, just malware, but im sure some of my 2b2t minecraft mods came from sketchier places lmao
I and several other people use the same software from the same source
I have 3 different versions of it even
Where do I mess with my driver settings?
I managed to pull a lower voltage thanks to MSI's cpu lite load (MODE 2). Last image is the highest score achieved with this oc.
P-Core: 5.4ghz
E-Core: 4.3ghz
Ring: 4.7ghz
Voltage: Target - 1.140v (-.100mv offset)
I'll probably keep this as my daily from now on. What do yall think?
overclocking new processors aren't generally worth it
but it doesn't hurt to try
just becareful when tuning it
don't wanna fry it

then dont do it
overclocking is just pushing your component past their manufactured speeds
โ ๏ธ
with that 3dvcache i doubt it for gaming
no bruh
because i crave that 1 extra frame?
๐ฟ
um
you can control fan speed via bios
fans dont have a clock speed
it really isnt lmao
overvolting isnt exactly good for fans anyways
if you want faster/higher cfm fans, get the fans that is made for that kind of usage
note that some more industrial fans may draw more current, or even require more voltage than conventional pc fans
Not even remotely close
what did you expect
Then you definitely don't want to overvolt them to get them to spin faster because that will burn out the motors. And you need special equipment to do it anyway.
hello, i recently changed my max Freq on my 7900xtx and lowered it 5% to avoid direct X crashed in call of duty and so far it has worked but for some reason i have to change it every time i boot up computer is there a reason why my settings aren't sticking? all this is being done on AMD Software.
it could be that you have another software conflicting it, by any chance you have msi afterburner installed?
if so make sure no profiles are created and keep startup option grey. Make sure these boxes are checked the same way
There is no way to keep settings fully intact for a reason, lets say if an overclock was unstable which led to a driver timeout it will reset everything back to stock. You can create OC profiles on adrenalin though, if you still have issues either revert to an older stable driver or DDU (reinstall)
No.
Only way to make it more than 3-5% faster is using a custom VBIOS and or LN2 both of which are not recommended.
Also keep in mind that you need to take silicon lottery into account. So the gpu may not even like any oc at all like my 3070.
you havent been able to make custom vbios since pascal days
NiBiTor my beloved 
What do you mean? I am pretty sure people have modded the 4090 with a custom vbios here before.
You don't use a custom vbios, you cross flash from a different model, usually the HOF in the case of the 4090 since they have a 1kW bios that disables all temp limits
Oh, I thought a Custom One was used. So the idea is to take the Bios from another model to tell the gpu to avoid all limits?
Specifically for the 4090 yes
That 1000W vbios is dangerous to use unless you know what you're doing and using sub-ambient cooling
Other GPUs you pretty much just pick the model with the highest power limit and use that
gpu oc kinda dead anyway
dont we manage to hit cpu limits in some games with 4090 still 
(at like 4k)
Even with a 7800X3D at 1440p ultra yeah you can be CPU limited
Not as long as I am still alive haha. But yes.
5090 gonna be so cpu limited lol
need rkl with arl ecores 
and maybe arl imc 
slap in more l3 too 
still loses to 9800x3d
9800X3D likely gonna be cpu limited in many 1440p titles lmao
5090 is gonna be a real 4K card
4K ultra native with ray reconstruction on
Just hope Devs don't decide to turn it into the 4K60 baseline for their games
So annoying
It's quite frustrating when devs decide that 60fps with DLSS on Ultra Performance is considered acceptable.
ASA being a good example though it's gotten better it's still very unoptimized.
@celest notch What's a VCore?
@hollow aspen you're not getting a bedroom with that attitude
I'll accept the basement again, hopefully this one doesn't leak!
๐จ
Send help, Rommy forces me to live in the basement, it has no windows!
Well of course it has no Windows, it was advertised as having no storage so where would you put the OS?
She uses Linux, she chases me with a bat screaming she's an Ubuntuer
Lmao it took me a moment to realize the joke.
She's gonna Ub"punt"u your ass later
how can i overclock my intel i5 11400h in my lenovo legion 5
I don't think you want to overclock that
You can't, but you can try using throttlestop to undervolt it
I was able to undervolt my 11400h in the acer nitro 5 by about -70mv at the most
Hey @kind walrus you around?
Henlo
So i just built a pc with my friend
Ggood recommended us a 16gb hynix die ram stick so we can add a second one later (budget restraint)
Ok
Sorry my brain disconnected half way through
It's a jedec stick, do you have a screen shot of timings i could stick on there?
Well before you do that
Go find the ram details in bios
Confirm the manufacturer is hynix first
It is
Ggood has the ram and it's marketed as hynix so if it isn't ill fight teamgroup myself
Tyvm
Ye just google a hynix kit
Copy that xmp
Copy the rest from there
Test
Ur good
Ez tune
That should be easy enough to carry over with two sticks but you'll need to reload and retest when the second one shows
Does that look right for voltages? Took then from an sp 6000c30 xmp
Some of them anyways
I am not sure tbh
No izz ok
It's up and running
I did some googling and made sure
Oh ok
I was trying to find my old settings file then I got dragged off by the phone haha
Ok yeah that looks correct @faint tangle mb I took so long
No no worries, we actually went out for lunch we finished with the ram a while ago
I appreciate it v much
Ah alr
You shouldn't need more than 1.35v if you're only running at 6000, even with tightening timings
I have managed to hit the point where 64GB of ram isn't enough sometimes. I am considering looking into a 2x48GB kit of ram. How well can my 7950x3D handle those sticks?
Im scared to ask how 64 isnt enough 
I have work apps that I need to keep open 24/7, Chrome (8-20 tabs), and Discord those take up about 20ish and when I game my usage spikes to around 40GB. When I am hosting a MC Server or doing CAD or Video Editing I am hitting that limit.
I also terminate all unused apps using Task Manager.
Haha yes
But yeah the idea of getting 2x48 is becoming quite appealing and eventually I can reuse my 2x32 for a server which would reduce my mind usage.
The true ultimatum is to use all 4 sticks for maximum gigs
As long as the minimum highest ram speed isnt too far from the highest shouldnt be an issue
Am I able to adjust both kits to be the same speed?
Yes but it almost certainly won't run at the rated speed of either kit running 4 sticks mixed
Best to just buy a kit and fully swap then
Btw if I recall the new 9000x3D chips can be overclocked right?
Correct. They flipped the die so now the vcache is on the bottom, and the cores can be more properly cooled. That solves the problem that made AMD lock the voltage on previous gens so now they can be properly overclocked. I've seen a 9800X3D run at 6GHz on an AIO.
awesome!
How much does your 9800x3D run at?
I don't have one
oh wait, I think I am mixing people up lmao
ahh I was thinking of Felix
I wonder how good the 9950x3D will be
Hopefully they let the drivers cook a bit before launch this time so we don't need to worry as much about core parking
Hello all,
My buddy just expo his ram to 6000 and noe the computer is coming on just no display does anyone have an idea what could be going on?
Either the CPU can't handle those speeds or the ram cant run at its rated profile.
I assume he managed to use this ram without the expo enabled?
.
If the heat spreaders on my ram are using thermal pads that means I should be able to remove them pretty easily right?
Also I forget how do I check what chips a Kingston stick uses?
Those are hynix, likely M going by the xmp
need help on boosting my ram, it's currently configured at stock speeds and XMP 1/2 doesn't work because of whatever reason. If I try it, it just doesn't turn on my display and I have to reset cmos. Help !
What's your ram config
It can take 5-10 minutes to train the memory after changing any configuration related to it
Updated BIOS versions usually help with that
2666 if I recall correctly, it's rated 3200mhz. I have 2x16 fury beast ram
Ah DDR4
15-20 minutes of waiting, and also did a BIOS update yes
5600 non x, just OC'd it last night with PBO,
MSI B450m-A Pro II (someone told me VRMs in this mobo are better than the rest of my options I had before
Pro Max II?
yes
do enlighten
Kingston was known to hate micron for some time
So that's if I'm not mistaken an unusual pairing
But also a very good sign
You likely have 1 of 4 dies
sticks in slots 2 and 4?
my mobo only has 2 dimm slots
oh awk
what does it mean?
3 out of the 4 are quite good iirc
Even the bad one of the 4 isn't that bad
oh the actual dies okay i get it
dang, i tried to hold off not needing to learn manual ram ocing lol
takes a while but interesting to me nonetheless
so do I just start learning a tutorial now on manually changing their frequencies
since I'd be doing that might as well boost over 3200mhz if i can
did xmp-2 3000Mhz from 2400mhz lol
booted up first try but I kinda want to try 3200 again cus it might have to do something with ocing my cpu, thoughts?
i nono wana :feelsweird:
Just do it for a day lad
Process of elimination
I mean I did try last time before with stock
had a nightmare with that for a few hours trying
You can save your current overclock as a profile so you don't need to totally redo it
I don't know what the benefit is with running stock cpu speeds though cus I'm happy with my boost and undervolt
That last one is the potential issue
The undervolt
Too little voltage to the IO die can cause it to fail to run the memory
how do I test it?
In real world performance you will gain less than 5% anyway, just try it
Very simple, if what wasn't working now works, we got somewhere
Since this is a recent change, a simple flip back will speak volumes
Generally speaking I'd recommend doing memory before CPU since an overclock on one can affect the other
And memory has a bigger impact
Esp. on ryzen
okay that's understandable I'll reset back to stock and try 3000mhz on the ram
Memory is like the biggest impact by a mile on ryzen
Vcache is a good example of why memory does a lot
so if it doesn't run, I lower ram speed until it boots? only then I go back to my cpu boost?
If you can get the memory to run at 3600 at stock cpu settings then that's a good 15% improvement on CPU performance thanks to the fclk
Vs the 5% you gain from PBO
thanks very much
- so use stock cpu speeds
- oc ram, and lower speeds if it doesn't boot, and until it's stable (open a bit more tabs, run discord in bg, play and do some editing for the day)
- if stable, then I resume my cpu oc?
can you rephrase please lol
If the memory is stable at 3200 then something you did with the CPU caused this whole thing, so you need to start over with the OC on the CPU
ah, but essentially everything else is correct
Pretty much
coolio I appreciate it very much
If it's stable at 3200 btw, we can try to push that even higher. 3600 would be a good target.
but knowing my pc doesn't boot with stock cpu and ram running at 3200mhz do I still do it?
Stick with stock and xmp for now
If it's fine like that, after about a day
Start over
And as we said earlier, start ram first
Bigger impact and more fickle
yupp gotcha
btw to reiterate the ram out box runs at 2400mhz, and it's rated 3200mhz (which what xmp-1 & 2 would run at) which already didn't work
yes, it's why I asked about how to remedy this in the first place lol
With the CPU at stock settings, preferably cmos cleared, and nothing but xmp 1 on, it failed?
yessir
That's a bad memory kit then
need help on boosting my ram, it's currently configured at stock speeds and XMP 1/2 doesn't work because of whatever reason. If I try it, it just doesn't turn on my display and I have to reset cmos. Help !
this what I said
well it ran before at 3200 xmp-2 completely fine with my R5 2400g
Right but then we went on that CPU overclock tangent and got sidetracked
sooo.. i manually do the ram oc?
1 out of 4 dies can't run X speed studf like that so I have to adjust
Nope
Warranty the ram
Unless you had a static oc on the cpu I doubt it's the cpu at all
1 year warranty mate
Kingston DRAM Warranty Statement
perhaps it's still under warranty, I remember fully upgrading my pc within the year
Tf not lifetime???
I bought it July 1 this year
Warranty then
Ah you're golden then
I really hope it's not much of a hassle besides the waiting
1 year standard warranty, beyond that you only get a "credit" towards a new kit based on depreciated value
Pfff
That's an instant turn off from me wtf
They're like the first company I've heard that does that
Still better than corsair
how does a warranty claim work for you guys?
never needed to return anything before
Similar to returning something
Contact the place you bought, tell them what's happened and you're looking for a warranty claim
They will probably tell you to contact the manufacturer if US
yikes, this might take a while then
I don't have any other ram to use if I return it
You can ask if they'll cross ship so you minimize that time between
Found the phrase in the warranty btw, all they have to say is they don't have replacements in stock
If Kingston is unable to repair or replace the product, it will refund or credit the lesser of either the current value of the product at the time the warranty claim is made or the purchase price.
1 year standard warranty, lifetime depreciated value warranty ig you could call it
lol "lifetime"
**Product lifetime is defined as the normal time expectancy for the use of products in the industry. But the lifetime warranty may be subject to definitions as set by different countries. For Russia, the lifetime warranty is determined to be for a period of ten (10) years from the date of purchase by the original end user customer.
Ok but if it says lifetime that counts
This is legal definition
10 years is long enough really
G.skill meanwhile still replaces DDR2 if you kill it overclocking
Sheesh, I'm digging more into what the US defines "lifetime warranty" advertisements as acceptable and it's quite the rabbit hole. It's basically a useless nonsense phrase you throw in to make it sound good. You can arbitrarily set whatever time you want based on your own definition of "expected useful lifespan".
that's predatory as hell
Pretty normal if I recall back the laws we have in my country for lifetime purchase
It can be considered predatory depending on the context
Since in this case it'd be a "computer part" I would expect that to fall under somewhere of about 5 years
Over here we used expected normal life of products as an agreed definition up to a maximum of 16 years
So if I buy say, a tire
The expected life of a tire is normally 2 years
So a lifetime warranty based on the normal use of a tire would be, as you'd expect, 2 years
But on the other hand (I'm gonna refer to this because it's an easy example that I remember) you buy a phone
The normal expectancy of a phone can range between 2 and 10 years depending on the context
And we have rules that permit a lifetime responsibility beyond that of a standard warranty (Note responsibility as a specific word, it's not repair or replace necessarily)
You buy a tv, the accepted standard would be over 5 years
The ambiguity is (at least here it is) in fact rigged in favour of the consumer, because every business would love to give the minimum period then bugger off if they had the choice
So this would be a civil case, as a consumer complaint, and the objective for your case to the court (let's assume the extreme example for a moment) would be to prove ram is intended for as long as the computer (as above, ~5 years as a part)
Depending on the case and scenario, it's very ambiguous and that's intentional to prevent clear cut examples especially in like tech where there's a huge catalogue in every category
seller's tryna deny service or smth
This is why I buy almost exclusively from Amazon.
If there are is even the slightest issue so long as I bought it within 30 days it has always been no questions asked guaranteed returns or replacements in my experience.
no one really buys amazon where I live
i live in ph
shopee is the amazon here basically
Gotcha.
This is why I try my best not to, amazon is being flooded everyday with more garbage and shady third sellers that take advantage of this to keep their own stores at the top of all searches, not to mention selling literally dangerous and harmful products
The shady sellers sounds like most retailers honestly. I usually keep my eyes out for who the seller is and because of the return policy I have never had to worry about that.
Just got my PC back up and running.
I have updated the bios and I am finally getting around to tuning this 7950x3D and the ram for it
.
First try I got 6600mhz to post and I started with 1.4v Dram, VDD, and VDDIO
first try timings, I left every thing on auto
Is that with UCLK=MCLK on? It defaults to off at that speed
Ah, no it's not
You want to have that on for sure, it's a major penalty if it's off
Thanks for pointing that out.
I will restart and see if it posts.
@sudden torrent, what are the safe daily voltage limits for Hynix A/M die VDDP, VDDQ, VDDIO?
The most mine will do is 6200 with it on
My previous 7950x3D did 6400cl32. I am not sure about this replacement yet.
I am hoping using this new bios version helps. My previous bios is nearly 13 months old.
As for voltage limits, let me refer you to this screenshot of one of the top overclockers doing 8000c32 https://hwbot.org/imgs/3083389
Wondering so I can just set those voltages and keep pushing the ram till I know what the limits are.
Your vsoc is too high, I can tell you that right away
That alone can cause instability, being too high
Consider me a complete noob, I forgot what some of these voltages are.
What is vsoc again?
The one that kills ryzen cpus when it's too high
1.1-1.2V on CPU SoC is typically good
0.95-1.00V on VDDP
yep
Set it to 1.2?
Do 1.15V to start
What is the difference between override mode and AMD over locking?
Also thank you for helping me.
Does what I need it to that way, which is hold a constant voltage
Your VDDP is fine, you can drop mem VDD and VDDQ to 1.3V and call it good
Oh and I wouldn't touch VPP, just leave it at 1.8V
VDDP is on auto right now.
Understood.
Huh. Well it's not supposed to be lol. Maybe some sort of auto OC going on in that bios.
Should I manually set it?
Nah just keep it in mind if you have stability issues
So:
- 1.3V VDD
- 1.3V VDDQ
What should VDDIO be?
Can you give me a rundown as to what these voltages do? It'd mean a lot to me to understand what it is I am changing. Because the bios info section for them is useless.
VDDIO scales up with frequency and CL, and is the biggest reason the sticks get hot
Gotcha. I am not worried about heat. I have a fan over the ram and room temp is 16c
The guy running at 8000c32 has VDDIO at 1.43V so there's no shot you need more than 1.38V
I'm running 1.35V myself
Was that synced or?
You mean UCLK? No
Yeah
But at that speed it doesn't matter anymore because UCLK is 2000 again
Gotcha I was like how the heck did he get that to sync on a 7950x3D.
So VDDIO scales with frequency and primaries?
What about VDD?
This seems like a halfway decent guide from a cursory glance https://www.tomshardware.com/how-to/overclock-ddr5-ram
DRAM VDD: This voltage is one of the most critical as it powers the memory chips. For everyday use, we recommend keeping it below 1.4V for safe measure.
DRAM VDDQ: The voltage feeds the memory chip's I/O. VDD and VDDQ typically go hand-in-hand. However, in some cases, desynchronizing the voltages and running a higher VDDQ (50 mV - 100 mV) can help stabilize memory overclocks.
Current voltages by the way:
- VDD @ 1.3V
- VDDQ @ 1.3V
- VDDIO @ 1.38V
- CPU NB/SoC @ 1.15V
- UCLK=MCLK set to true
Try posting with these settings at 6600mhz?
Yeah try it. If it doesn't work try 6400 next.
If 6400+ boots immediately run a memory test
I really hope 6600 posts. I spent so many hours trying to get my previous 7950x3D to post at 6600 stable.
Hmmm.
Not a good sign.
I am still on the dram light
Reset as in restart?
I was not aware of this. I had previously forced it into safeboot by force shutting down the oc multiple tumes
It's not foolproof and there are times where it won't reset properly
But it was pretty decent when I had my MSI board
Guess I will give it a couple more minutes.
If this fails what could I do to try and increase the chances of 6600mhz working?
.
Still no post
Just force shut it down and clear cmos?
tbh with UCLK it either works or it doesn't, or you spend days figuring out the exactly correct resistances and voltages
You can clear if you want, or put it into safe boot
Ooh you have a safe boot jumper, that's even better
?
I see the dram light indicating it's cycling
I forced shutdown the PC three times back to back but never got into bios
You can just touch those jumper pins while you power on and you don't need to do the whole song and dance
I can touch those pins while the PC is on and powered?
While it's off, touch them with something conductive and press power
Gotcha
You can let off when it posts
Yikes 16c air feels like 2c when being blasted by the fans
Probably more so after getting used to the warm equatorial area
Lmao if I didn't heat the home I could have some serious cooling.
-16c right now
So I tried the jumper pins but it didn't safeboot.
Try again? I used a metal wrench
Also Dram is cycling a lot
It might have a coating on it, try something else
Nickel plated steel is pretty common for tools to prevent rust
If you wanted you could wire the reset button to the jumper instead, so you can just press that while you power on
Tried a different wrench hopefully it works now.
@sudden torrent, what could I do to improve my odds of getting 6600mhz to post?
I can't really suggest anything specific there, just pure luck and trial and error with getting the exactly correct settings
Rip.
Also this is becoming frustrating.
3 different tools in a row. Not luck getting into safeboot.
Guess you could clear cmos with the jbat1 jumper
What are the maximum safe voltages for VDD/VDDQ/VDDIO on a Hynix IC?
Jbar1 jumper?
Right next to the safeboot jumper
Oh nvm took me a sec to realize.
Next time I buy a motherboard I think I will get one with a clear cmos button.
That's another thing you can wire to the reset button
In fact that's exactly what I did
By reset button I assume you mean the power btn on the computer case?
Yes
FINALLY
But not the main power button, most cases have a second one intended to trigger a soft reset
I am in the damn bios
I see a small button on the fractal torrent
Which wire is that one? I am gonna set it to the cmos like you are suggesting.
It'll be in with the front panel wires
So I bumped up the ram vtages to 1.4 and soc to 1.2
I managed to get to the bios screen but no post yet.
Blue screened.
Giving 8000mhz desynced a attempt.
VDD 1.38V and VDDIO 1.43V
Failed attempt. Also I just rewired the reset button. Hopefully it works.
The button does not appear to be working.
@sudden torrent is it bad if I jump the Clear Cmos while the system is on (not posted yet)
Because for some reason if I try to use the reset button while the PC is off it does nothing.
Nvm now it's functioning properly.
Trying 6400mhz.
- 1.2V CPU NB/SoC
- 1.35V VDD/VDDQ
- 1.43V VDDIO
Gonna try working my way up. Cleared all the settings and working my way up from XMP.
@sudden torrent I got 6400Mhz to post!
My board and CPU seems to really want to stay at 1.3V SoC otherwise a lot of things become unstable below that.
Been this way since I got the replacement.
gonna eat then strart tuning it
for some reason I had to up the voltages to 1.43 for the VDD, VDDQ, VDDIO for it to post
K I got tREFI 65,000 to post. For some reason tREFI 65,355 has never posted.
Yeah that's a common thing
Do you know the cause?
I am using TM5 with Anta Extreme 777 profile to check if the OC is stable enough for me to move into the next settings then I will run PCBdestroyer later.
The ram is maxing out at 39c.
trefi is just one of those platform quirks
Gotcha.
I doubt 355 makes a difference when tREFI is at 65K so I am not to worried.
After I am done with tREFI should I do the tertiaries or secondaries?
.
On a side note, is it fine if I leave my VDD/VDDQ/VDDIO at 1.43v? I will try lowering them after timings are done but if I can't lower them without the system being stable is 1.43v fine for daily? I am fairly certain I have a Hynix IC I do not know if it's A or M though.
It should be fine
Am I able to go higher or stick to 1.43 being my max because I know on my previous 7950x3D bumping them to 1.55v yielded around 10-20% tighter timings across the secondaries and primaries and slightly better tertiaries at 6400mhz.
Gosh I hate autocorrect.
I wouldn't go higher than 1.45 personally for daily but I know people do higher than that
You are pretty conservative and I trust you more than others.
So make 1.45v my ceiling then?
My plan is to start at the ceiling and lower wherever possible once I have my timings set.
Up to you, there's a lot of things left to figure out about DDR5 still
Yes I am pretty conservative with my settings. I'm just running at 1.35V right now.
And there's only a sample size of 3 different memory controllers so far to figure out how it behaves
I know 1.38v across VDD/VDDQ and 1.43V didn't allow 6400mhz to post for me.
Could be down to binning, worse bins need more voltage to get the same results. If I'm not mistaken you have a 5600 kit.
Unfortunately it appears unless I wanna venture into the 1.6v region I can't get 6600mhz to post (not stable) and I am simply am not comfortable with that.
I've seen up to 1.7V for benching
6000cl32-40-40-80 @1.35v
Though I am running 2x32.
But bad for daily I assume.
We still don't know, it can take several years to show degradation
Also it seems my board/CPU doesn't like going below 1.305v SoC
Board has had it set to that since as far as I can remember and when I lower it my stability goes down exponentially.
That'll affect UCLK, and above 1.35V is what was known to cause CPU popcorn so that's a hard limit in the BIOS now.
Increasing that makes UCLK=MCLK more stable at higher clocks?
I don't think I am willing to risk it though tbh. To much risk for not nearly enough of a reward. I am already doing better than the average with getting 6400 to post synced.
.
@sudden torrent tREFI 65,000 passed TM5 Anta777. Peak temps were 42.5 and 41 Celsius.
Should I do Terts or Seconds now?
K gonna do that now. Also really, thank you for helping me and I appreciate the company.
By the way, that tip with the reset button that is so freakishly helpful
Baruch Hashem to that.
K.
I've changed SCLs to 4 and tFAW to 16. Also voltages for VDD/VDDQ/VDDIO are all at 1.45v now as the ceiling.
tRAS 30
tRC 68
tWR 48
tRTP 12
tRRD_L 8
tRRD_S 4
tWTR_L 16
tWTR_S 6
SCLs and tFaw passed the test.
Should I try to lower SCL and tFaw some more?
Cool, you can do a few of those next timings at a time or all at once, your pick
SCLs might go to 2
If memory serves me right one of these at a certain point makes no difference.
That's tFAW
is 16 that point?
Yes
Gotcha. May I ask the reason?
tFAW is defined as the window in which 4 activate commands can be given consecutively
Each command takes 4 cycles
4*4=16 min
If it were higher then it would wait out the remaining cycles before more activates can be issued
Is this a AMD thing? the 4 commands part
Odd numbers are fine or? If so I am gonna try for 2 and 3.
It's a DDR5 thing, part of the spec
Sometimes
gotcha.
I don't remember if I showed you but I finally got around to upgrading my setup after yapping about for like two months.
Figured you guys would appreciate knowing that all those hours of talk and suggestions weren't for naught.
I maintain that the contact frame is unnecessary unless you use liquid metal
Ahh gotcha. I nabbed it because TG's PTM stuff showed them using the frame with it.
The drive is admittedly a bit overkill but it's been my dream drive for a long time.
are there any formulas that should be followed with any of these timings? Also I am gonna try SCLs set to 2.
Not with those timings
That's just a set of "known to work on hynix" timings
TRRD_L might go lower but start there
Once you have those working then I'd go to tRDRDSD and tWRWRSD 6, odd numbers work on these ones if that fails.
tRDRDSC 1
tWRWRSC 1
tWRRD 2
tRDWR 16
After that is primaries
Back
@sudden torrent unfortunately I couldn't get pass bios screen with scl 2.
Is scl 3 something I can try or do those need to be even numbers?
You can try it
Btw something interesting I remembered.
My previous 7950x3D couldn't get Anything lower than SCL 5 to post stable. This replacement has it beat in that department so far.
Ah nice
My previous one did 6400cl30 at 1.55v with very tight secondaries if I recall correctly but struggled like hell with terts.
Hmm. Not a good sign.
It is Cycling the Dram.
By the way, @sudden torrent besides VDD/VDDQ/VDDIO/NB/SoC are there any voltages I can adjust to increase rdm stability?
Not that I'm aware of
Ahh okay.
How risky is it to up NB/SoC to 1.35v? I keep reading that it's safe so long as it is below 1.4V.
It's theoretically safe
Do you think it's too risky?
Does bumping up SoC increase ram stability?
It can for some timings
I wouldn't do it myself but I'm quite cautious
Eh maybe I would just to see if it helped
Oh but what might also help is settings LLC level 3 from digitall power menu
So put LLC to level 3?
CPU LLC and VDDSOC LLC
I see two separate Loadline Calibration Settings.
- CPU NB/SoC LLC
- CPU LLC
Yes those ones
Both have options MODE 1 to MODE 8
mode 3
for both?
Yes much better than my cheapo DS3H B450M.
So what does the node do? Is it making the voltages more consistent by overshooting it slightly?
It's going to undershoot by design, this controls how much
Gotcha.
mode 1 is no undershoot, and can occasionally overshoot
Sure
Hmm it's dram cycling again.
Alright I think I will leave SoC be at 1.3v
Not gonna risk it. I found a reddit full of users who have run 1.35v for months on end. The common consensus is that while the cpu and it's mc will survive, the iGPU seems to die at some point if left enabled.
Unfortunately SCL 2/3 is a no go.
Oh well, not a huge loss
So move onto those timings?
Yep move on
Just applied tRAS, tWR, tRD and changed tRDRDSC/tWRWRSC to 1.
K it posted
Hmm it appears tRAS 26 is a no go.
tRAS 28 posted
tRAS 30
tRC 68
tWR 48
tRTP 12
tRRD_L 8
tRRD_S 4
tWTR_L 16
tWTR_S 6
tRC 62 posted.
Gonna try tRC 58
You're below the register limit now, tRC=tRAS+tRP
Ahh so there is a formula.
Yes some of the timing minimums are determined by combining other timings
Pretty much, but by this point you should be pretty much locked in on those timings
Can I still adjust tRAS and tRP individually each first to post so ai knows the values each goes down to?
Yeah
K
tRP and tRCD do they need to be the same?
No
Gotcha
tRCDWR and tRCDRD can also be different
Fingers crossed tRC 34 posts.
iirc tRCDWR doesn't make any difference on AMD, maybe that was DDR4
I purchased two items of TG's 40x50mm PTM. If each item comes with 2 sheets that should leave enough for me to change the pads on my ram. Do you think that'd be a good idea?
No, it's more risky to take the current pads off and there would be little benefit to doing so
Part of why I was wondering is because it would also give me an excuse to finally know which ICs are on my ram. Is delidding ram sticks that use thermal pads hard?
The risk is twofold. You can bend the heatsink so it never sits correctly again, even after bending it back, and if it's especially sticky you can rip the IC right off the PCB.
Ahh alright.
Also tRP 32 did not post.
34 posted and that is with tRC synced
@sudden torrent are there any formulas for tWR?
Wait? Why won't it let me set tWR below 48.
Yeah that's a hard limit
Ahh okay. I don't know why but on my previous now 1 year old bios it'd let me change it in bios iirc.
Yeah sometimes it the bios will let you set it below the register limit and in that case your setting just gets ignored
Since I am in the primaries section should I just get tCL and tRCD done with before I go do the secondaries you mentioned?
No the primaries are the least valuable thing to tune
tRTP 12
tRRD_L 8
tRRD_S 4
tWTR_L 16
tWTR_S 6
I am reposting this on occasion to keep it on hand
You want to do everything else first just in case a primary timing keeps a secondary/tertiary from going lower
Gotcha
So these settings and formulas for them?
Also should they be even numbers or can I do odds.
Yes those next. Evens.
Gonna try posting
It's alive!
Going to do a TM5 Anta777 run before I try tuning the settings in more
Good idea. It should go a bit faster now too.
I hope I can get my ram tuned down to 6400cl28 again
2s, 4s, and 8s are definitely my preferred numbers to look at.
definitely not because of using Minecraft to practice math lmao
CAS is a conundrum. If you set it too low you can actually regress in performance.
Really?
Oh btw somehow my bios when I updated it was able to tell my ram has Hynix ics. It had a thing for optimizing ram settings based on ic.
There's a buildzoid rant about it but I don't feel like digging for it, so I'll leave this one instead https://youtu.be/pgb8N23tsfA?si=nnasM_lQ6Khn5PXB
DDR4 PyPrime 2B record: https://hwbot.org/submission/4758130_rauf
DDR5 PyPrime 2B record: https://hwbot.org/submission/4987588_p5ych0
My Patreon: https://www.patreon.com/buildzoid
Teespring: https://teespring.com/stores/actually-hardcore-overclocking
Bandcamp: https://machineforscreams.bandcamp.com/music
Twitch: https://www.twitch.tv/buildzoid
...
It could be mentioned in there somewhere but idr
How do I know if my performance is regressing?
Aida64 is paid only now though iirc.
There's a trial
I used it.
Lemme see how much aida64 even costs.
Because iirc they have some good benches to test CPU stability. Correct me if so am wrong.
So uninstall using revo uninstaller and that should remove the registry data with the trial
That it?
yep
That's a super useful app too, I keep it installed just in case something goes wrong
After all the settings are tuned in do you think 1 hour of OCCT and a round of the PCBdestroyer profile in TM5 ought to be enough to test for ram stability?
You can skip the OCCT if you plan to run TM5 too
Maybe a few hours of y-cruncher stress test
So TM5 (PCBdestroyer) and some Y-Cruncher
yes
My Y-Cruncher install is a year old now should I update it?
Yes there's a new algorithm to use
@sudden torrent I have noticed something interesting with my SoC voltages.
I have it set to auto and I have noticed with each time that I have tuned a timing, the voltage has gone down slightly. Started at 1.305v and now at 1.294v.
And that's why I don't count on auto
You can drop this into your y-cruncher folder and run it as a shortcut to starting the test
Thank you. Also now we shall tune thy secondaries!
tRTP 12
tRRD_L 8
tRRD_S 4
tWTR_L 16
tWTR_S 6
tRTP 6 failed.
Now trying tRTP 8.
@sudden torrent tRTP 8 actually managed to post somehow
Test that one right away, at least a quick test
Anta777 work?
yep
Is that good or?
tRTP down to 8.
It's just one of those timings that's a silent killer
You don't know it's too low until you blue screen and lose your boot config
Yep there it is
Odd numbers work for tRTP?
Are any of the other secondaries like this too?
Because I'd like to keep the process fast by tuning each down till it still posts then do a test at the end of each batch.
I don't think so
All of them need to be even numbered and no formulas right?
For the secondaries, correct
Tertiaries (turn around timings) can be odd, no formulas
Make sure to save a profile before tightening terts, they're a quick way to lock up during post
tRTP is being a real pain up my ass about posting rn.
Might as well leave that at 12
It's not a big loss in performance and you get more stability
8 posted and for some reason 10 caused me to get stuck at boot.
Gonna try clearing the cmos real quick.
Yeah if I can't get this to work I will do that for certain.
Praying that this posts
@sudden torrent IT'S ALIBE!
At some point soon you should set FCLK to 2133 so you can get a minor latency bonus having it run 3:2 with UCLK
The voltages you have now should allow that
K. Should I try for faster if I am able to?
I hate how 2x32GB sticks take so frustratingly long to test in TM5....
Nah I'd rather have that synced, unless you can get it to 2200 or higher without regression
16-18m per run
How does one test for regression?
So Aida64 for CAS and Y-Cruncher for fclk.
Oh I will just use y-cruncher then
I'm getting the itch to overclock my system again
I'm just using a basic set of timings right now
tRTP 12
tRRD_L 8
tRRD_S 4
tWTR_L 16
tWTR_S 6
tRRD_L refuses to post below 8.
So I will be moving onto tRRD_S
NVM that is already as low as it gets
Now's a good time to save a profile
I have been saving at each step that passed tm5
The terts ending in DD you can ignore since you only have one DIMM per channel
DD stands for dual dimm or double dimm?
Terts you said are odds or evens and no formulas right?
Yep
Terts:
- tRDRDSD
- tWRWRSD
- tWRRD
- tRDWR
Those are the ones left to set yes
SDs you'll have to play around with, maybe start at 6
WRRD 2
Lmao I actually just set it to 6 as you said that
RDWR 16, might go lower
Looking good. Last thing is tRFC, then you can do tCL and set tCWL to tCL-2
Go lower or higher?
low
gotcha..
And don't worry about tRFC2 or 4
I am gonna try tRDWR just so I don't half ass anything then I will immediately move onto tRFC.
@sudden torrent seems tRDWR 14 is the lowest I can go and still post. Regarding tRFC I started with 450 and it posted.
tRFC will fail with heat so you'll need a longer test for that one
Well, you have a fan so it's less of a big deal
@sudden torrent so heat is my only concern? What happens if it gets to hot? Error then shut down?
Error but that's it
okay so no permanent damage
Current RAM Overclock Settings.
@sudden torrent I need to go to sleep in a moment. Should I just run PCBdestoryer while I sleep?
Hmm, what if I ran Y--Cruncher and PCBdestoryer at the same time?
Would I be able to kill 2 birds with 1 stone?
You don't have enough memory for that lol
rip
Both of those will try to fill up your memory
Not an effective test doing that
Gosh I have been at this since like 1PM my time. It's now 12:20AM.
There's no corner cutting
You corner cut by skipping numbers not tests
Do multiple numbers at a time instead of 1 at a time
numbers?
Timings
Yeah we're on the home stretch with that and that's what he did
Ah I haven't been watching
Either way you can't shortcut tests
That's how you get bad results
I kept lowering each group of timings till they can still post then once all were done for example terts I then run a test
There's only tCL and tRCD left once tRFC is proven stable
Hynix die of some sort. I can send the picture with the code.\
Fan over the ram and room temp is 16c.
You should test in warmer temps
Only do that for benchmarking strictly
It's alaska m8
also all my fans are almost always running 100% down to 70% at the lowest in a fractal torrent
That was unexpected
It's -20c ouytside lmao.
You need to test if 7800 posts
That's the quickest way to find out
7800 auto timings
If post, A die
If not, m die
Hot af sticks also a sign of m die
Reason I say is because tRFC is directly related to temperature
I usually run wide trfc intentionally so temps don't cause issues
I've dailied B die tuned at over 80C doing this lol
Never degraded
After this test is done I can try that.
I hate that my pmic temp sensor doesn't work above 1.35v
Just post? So 7800 all auto timings, voltage, etc and unsynced
Yeah just post
One thing that Arshia actually had some idea about and it's only because they were obsessed with trfc until they learned this
Wide trfc makes temps a non issue for a small loss
@sudden torrent current settings passed Anta777. Max tempts 42.5c
Gonna run PCBdestroyer whilst I sleep and try the 7800 test in the morning.
Night folks
When your mom gets into orbit
Grrrr I'm so close to stabilizing 6400
It boots but errors
It's definitely the UCLK since it's stable with MCLK/2
@sudden torrent I am confused, how do I select this? I tried running it in administrator with it inside the Y-Cruncher folder
btw the ram passed 2 rounds of PCBdestoryer (not back to back), and Y-Cruncher's default PiBench 10B Multi-Core with no errors and temps never spiked above 43.5c
So I try running it but it feels like nothing is happening
nvm now it freaking works
today has been a confusing day
@sudden torrent I gotta go in a moment, but when I am back iirc you said to tune tCL, tCWL (tCL - 2), and work up FCLK right? Also regarding the tRFC should I just keep lowing it till 1?
Did you ever figure out if it's M or A die
haven't had time to try yet today






