#overclocking
1 messages · Page 10 of 1
Yeah, that's the max
AMD had to keep some things locked to keep the 5600 from beating the 5950X 🙃
I am using a 5700 though?
You can technically go higher but it would require base clock adjusting, which can be risky
Why is that risky?
I won't do it since I don't want to risk anything on a new pc but I am kinda curious why it's risky though.
I thought OC wasn't risky
for GPU no.
At 101 bclk your sata devices might freak out
At 103 bclk everything might freak out
what's bclk?
base clock
CLocK
Okay. So i I am understanding right, if someone were to attempt this, they should forgo the use of sata drives?
If you were to attempt this you'd be risking all of your data on all your drives, sata is just more prone to freakouts
Note to self: never attempt this.
This is the base frequency that everything in the pc runs at, and devices expect it to be exactly 100, give or take 0.2
So far cores 5,4,2 have all passed.
I have a very good sample of a 5800X and I can run +200 and bclk 101.5 stable enough to run benchmarks, but it is absolutely not gaming stable
thought games were easier to run then benchmarks?
I've seen people doing memory overclocking that passed a 12H strenuous memory test, and then the PC crashes at idle
Oh seriously?
Yeah memory is a PITA when you're pushing it
Any actual gains from memory overclocking?
Depends on the games but often yes. It's usually in the form of improved 1% lows and reduced stuttering.
gotcha. I think I will probably skip out on memory overclocking until the next pc upgrade then.
You had the 5700G right?
yeah
You can use pretty fast memory then. 4000 would be pretty easy to run, 4600-4800 is often the limit without penalties.
what do you mean by penalties?
Infinity fabric (fclk) likes to run synced to the memory clock (mclk). 1:1 is ideal so that latency is minimized.
If you take mclk above what fclk can do while stable, then it would be desynchronized creating a latency penalty
oh, so going above 4800 might result in higher latency
core 0 just passed.
core 7 just passed, one last core to go.
nvm two cores left to go.
@sudden torrent, all the cores passed and now it's on the 2nd iteration, should I let it run another iteration or should I start under-volting?
Lucky you, no positive offsets needed. Nice bin.
Yeah!
Yep you can jump straight to negative 10 on all cores and test again
After this run, threads that crashed get negative 5 and those that didn't negative 15
core 3 passed
next thing you know all cores pass with -100 undervolt.
-30 is the best I've seen any single core get
mhmm, it will be awesome if my cores can attain that.
5800X
nice.
I wonder what kind of sample I will get on the 7950x when I upgrade to that.
core 3 passed
core 2 passed
core 1 passed
core 5 passed
all cores passed
i am getting tired, gonna watch a couple episodes of a show then go to sleep. Before I go to sleep should I undervolt to -15 and run it over night or keep it at -10 and run it over night?
Though tedious, this has been a pretty fun process.
I guess I will do this, undervolt again now and watch something on the tv while it runs in the background all works well I will undervolt again and let it run for the night.
c'mon baby, let's aim high and score high!!!
01:40:46 - Completed the test on Core 6** (CPU 12)**
what does CPU 12 mean?
All cores passed -15 undervolt.
doing one last test at -20 undervolt, when it passes I will let it run over night and try -25 tomorrow.
Hot diggidy dayum!
-20 undervolt on all core worked.
Gonna let it run for the rest of the night while I sleep.
cpu 12 would be core 6 second thread
Aight it's been about 12hrs time to check if it passed.
All cores passed +200 overtide with -20 offset.
if only there was a way I could input say +400 override because I feel that this chip can do it.
besides override, and offsets are there any other ways to increase cpu performance?
12 thread i assume
what cpu u running
Yep.
you could venture into the depths of bclk oc but thats so unfun for nothing but a 3% clock bump
(as ssds tend to die above 103 bclk)
I tried bclk oc on the 5800x3d and it disabled boost 😦
bclk is something to avboid anyways, not fun and barely useful
Intel bclk on 12/13th gen is fun if you get a board with the external clock generator. Then you can tell it to only apply bclk to the CPU and memory and hit ridiculous numbers like 140 bclk.
what's that in ghz?
Depends on the processor
A Pentium will have very different clocks from a 13900k but you can figure it out by taking the max boost clock and multiplying it by 1.4
Yes that's a 40% overclock, yes I've seen it work
TeamGroup “Evolving Invincibility” Giveaway: https://bit.ly/3HlM0eq
Support us on Patreon: https://www.patreon.com/hardwareunboxed
Join us on Floatplane: https://www.floatplane.com/channel/HardwareUnboxed
Buy relevant products from Amazon, Newegg and others below:
GeForce RTX 3050 - https://geni.us/fF9YeC
GeForce RTX 3060 - https://geni.us/MQ...
@kind walrus Hit a full time spy run on 1019 mV 👏
Hurt performance more than helped lol
But still cool!
13900k arrived
make it boom
why are you ramming it with voltage already
it droops to like 1.34v
in r23
stock and auto llc was 1.45v, dunno load it throttled too hard
Geez.
But at the cost of making drives nearly unusable correct?
That's the kicker, the external clock generator is able to separate the blck on different areas of the board.
Seriously?
You can tell it to apply only to the CPU and RAM, while drives remain at 100
So your telling me if I get a board with that capability there's a potential for some serious OC gains?
Would this harm the CPU?
R.I.P
Was thinking maybe I could get like a 6ghz all core 7950x.
If AMD thought they could get 6GHz they'd be making that the default boost
They already push it almost to the redline
Actually that chip is already really close to the limit so bclk isn't going to help there
I kinda wish manufacturer's would also guarantee a certain stock all core boost.
Didn't they do that at one point or is my mind playing tricks on me?
The bclk is more beneficial on locked intel chips like the 12700 (non-k) that normally you can't OC and get a slight reduction in clocks from the factory
They might have in the past but modern boost logic is heavily based on temps
Proceeds to buy a car radiator and put it in push/pull, xD
Pretty sure Ltt did that once, good temps but extremely impractical.
It was only good temps because of the thermal mass of all the liquid
You might as well use an aquarium and a pond pump
I was looking at the radiator and is all the water being stored between those tiny fins?
No there's channels at opposing angles that the water runs through
I mean on an AIO's radiator if it makes a difference l.
Oh, I thought the water went in fins and the fans are blowing on the fins with the water in them.
I've seen a graphic that shows it really well before but I can't find it now...
Ahh okay.
What would be a decent am4 upgrade on the b550m DS3H if I can find it for a decent price?
5800X3D is the only upgrade worth getting
I am returning to China in about 3 months so I am putting the big upgrade on hold until I am back there so I was wondering if there were some smaller CPU swaps I can do in the mean time.
And even that only gets you about a 10% gain iirc
Would the motherboard be able to handle it?
Sure
Oh thought the vrms would struggle
It has lower clocks to not upset the 3D vcache so it uses less power
What about the 5950x?
Don't even try it
May I ask why?
VRM will overheat
Ahh okay.
And it's also not a good upgrade
Was thinking of it for workstation use.
But I guess not.
So a GPU upgrade would be better then?
Yeah that's not a workstation capable mobo, at least not with an all core load on a high end chip
I wonder if I could find like a RTX 3080-3080ti for like $400.
Define "broken"?
Broken fan or something, easy to replace
Also would the 3080 be a decent jump from the 3070?
It's roughly a 25-30% jump on average
Hmmm I wonder if that would let me get a full 144+ fps. I am getting on average about 100fps in my games at current settings.
It would be closer to 125 fps then
ouch, $400 for an extra 25fps feels truly like pain.
Such is the pain the high end gamer feels
Especially with Nvidia calling $600 "midrange" now
Hasn't that always kind of been the case though?
Budget: 100-200
Low end: 300-400
Mid end: 500-600
high end: 700-800
enthusiast: 900+
The 1080 ti, a flagship model, was $699
A bit
It's hard to know for sure since nvidia doesn't tell us
What we do know is they've been charging the board partners a LOT more for the GPU core
I did get some info that wafer prices have significantly increased
And is increasing a lot again for next gen, which will be on 3nm
Smaller nodes are more expensive, at least at first
New tech's always cost more unfortunately.
Big/little on RTX 5000 btw, so similar to RDNA. Maybe we'll get reduced costs there.
Is 3D v-cache useful on servers? I forgot.
Which cache is most useful for all-core/server loads?
What do you mean? Like L1/L2/L3?
Also do you guys know if gaming servers prefer different cpus compared to normal data center servers?
yeah.
i can't seem to find any info online at least not a clear answer as to what type of servers prefer what aspects in CPUs.
That depends on the CPU architecture and the tasks it's running. Some will thrash the L1, others like the extra L2 space
Game servers aren't picky about what CPU it's got as long as it's enough to run the server
so it comes down to x-cores and all core boost clocks?
And generation but that's the gist ya
ya 5000 is mcm soon
yes
Ryzen 1800 won't work as well as the Ryzen 5600 even though it has more cores, for example
I wonder how well the R7-5700g will be as a Modded MC server, thinking of reusing it to run 2-3 servers with the expections for about 50 or so active users on one of them and the other two for youtube purposes.
IPC gains I am assuming?
Should be fine, even modded
Not sure if it makes a big difference but the servers I run tend to use about 100 mods each.
That's one factor of many, albeit a major factor
Ahh okay.
I wish there was like a CPU handbook on all cpus and how each one performs in certain tasks.
Preferably that's also pretty easy to understand.
i really enjoy using and learing about tech as a hobby.
You can usually find info by looking into the task itself. Often there's video benchmarks, if it can be benched, like for blender.
You can often find info in general CPU benchmark videos too, but those are more generalized
I just remembered, pretty sure someone here said that getting better memory for my pc might get like a 10% gain in perf? You think that's true? Running 3000cl16 at the moment, would a 4000/4800cl18 make that big of a difference?
alright.
I'm pretty sure that was me that said that lol
Oh okay.
xD.
I wonder if my memory is also the bottleneck causing the decreased scores in R23.
Was part of my prebuilt when I was in need of a pc quick, xD.
My laptop suffered severe degrading over the years to the point of running like a potato.
So I should get a ram upgrade first then?
Seems to be the best price/perf option.
Do you think my board can handle 4800cl18?
Probably not
You'd need a lot of manual tuning and luck
what's the next step down below 4800 then that you think I could probably use?
If you want to be able to just enable XMP and call it good, maybe drop some subtimings like SCLs (highly recommended), 4000c16 or 4400c18 max
what's scls?
what was the formula for calculating latency on DDR4?
Memory subtiming, it's hard to explain what it does without explaining all the timings, but Ryzen loves having them at 4
(CAS x 2000) / Data rate
So 4400c18 is 7.5ns
oh that's really low isn't it?
Very
maybe I can up my KD from 5 to 7 with that lmao.
4000c16 is 8ns, fairly close
It might, if you're getting stuttering or lag. Better memory helps the averages a bit, but it really improves on the lows.
I tend to get a good amount of stuttering in Fortnite.
Same with halo especially after someone throws a bunch of bombs at me.
Yeah those particle effects will do it lol
The worse case of it was in Fortnite when people blew up a giant structure near me and my fps bounced between 100 down to 50fps for like 2m.
Ouch
Usually my stuttering is about a 30fps dip which since my fps is about 100 on avg is really noticable.
So if ram can fix that, I'd gladly invest.
It'll definitely help. There's no true "fix" for it but you can definitely make it more manageable.
I see.
Plus having better ram should also help the server I''d assume for once I do my next upgrade and reuse this as a server.
Bloody hell, you weren't kidding 3000cl16 truly is bad.... 10.7ns.
Could also be a problem with ur graphics card V memory being full make sure its not being used by anything in the background
V-ram isn't in full use from what I can tell.
in Halo, it's at 6/8gb.
It's a 3070 at 1080p in fortnite iirc, no danger of running out there
and Halo Infinite to be specific.
Ah ok i had a problem with that once a program i downloaded turned out to be a crypto minor leech and was maxing my card in the background
13900k had so much dielectric grease on it i couldnt stabalise stock clocks lmao
lotsa of iso later and it's fine now
what's dielectric grease?
Corsair sticks often have trouble hitting their rated XMP spec
I was looking to get all corsair brand for my PC
They use really bad PCBs so they could use the best b-die chips in the world, but it'll be throttled by the PCB
Dang that sucks
sr ram on dr pcb with unterminated ends go brrrr
Also are there any drawbacks of overclocking a monitors refresh rate?
do 4400cl18 sticks exist?
My monitor is always overclocked to 180hrz
It might start flickering if you go too high, sometimes other weird issues
Ah makes sense
There's sticks at that speed in 2x8GB kits but not easy to find in 2x16GB
Yeah not without a lot of manual tuning
How hard would it be to manually tune 4000cl18 down to 4000cl16?
Need to use amazon.com.mx import fees in Mexico are obnoxious.
Ah right
La memoria de alto rendimiento de la serie G.SKILL Ripjaws V está construida con el más reciente estándar DDR4 en un elegante disipador de calor rediseñado que es ideal para potentes equipos de juegos, estaciones de trabajo profesionales o tareas informáticas diarias. Probado bajo la extensa uite...
starts at $40 and goes up to like 200.
I had to pay a $70 import on my Fractal Torrent.
on top of delivery fees.
I hate hate hate amazon's search
this seems to be the same ram minus the fact that it's not SDram.
Query: 32gb ddr4 4000
Result: https://www.amazon.com.mx/Asus-A320M-K-Motherboard-Micro-ATX-Socket/dp/B06ZY2F35X
Desarrollado por los procesadores AMD Ryzen AM4, el ASUS Prime A320M-K combina el rendimiento y la confiabilidad en un factor de forma microATX. Acceda a la conectividad y velocidades de próxima generación con NVMe M.2, USB 3.1, LAN Gigabit y soporte para DDR4 de 32 GB. 5X Protection III asegura ...
(╯°□°)╯︵ ┻━┻
I recommend doing this.
DDR4 32gb 4000
gets the best results.
not sure why putting the storage first messes it up.
This is the same kit as above for 300 less pesos, after shipping https://mx.pcpartpicker.com/mr/neweggmx/dqbTwP
Okay looked it up, so Mexico has 20% import fees on packages over $50
bloody hell that's rediculous.
La memoria de alto rendimiento de la serie G.SKILL Ripjaws V está construida con el más reciente estándar DDR4 en un elegante disipador de calor rediseñado que es ideal para potentes equipos de juegos, estaciones de trabajo profesionales o tareas informáticas diarias. Probado bajo la extensa uite...
@sudden torrent, if these two kits are the same I am gonna buy the kit from amazon as it's $30 cheaper after import fees.
Import tax is based on the cost of the product and shipping fees combined.
So they charge tax on the shipping fees? wtf
yep.
Told you Mexico's imports are rediculous.
They charged me $47 on the R7-5700g on arrival for import fees.
So long as I use XMP it should hit the rated speed correct?
gotcha.
And after some manual tuning I should be able to lower it to cl16 or lower yes?
Of course not guarenteed.
Yeah
Aight, gonna save up for a couple weeks then buy this.
Thanks for the help, and I hope you will continue supporting me in the future!
No problem, you know where to find me
by any chance would you know a solution to fix why not all of my argb leds are on?
Which ones aren't on?
Connection seems to only go up to 3 devices.
Yeah a lot of headers have a fairly low current rating so anything past a few devices just doesn't turn on or flickers
ahhh okay.
Was concerned it might've been a faulty fan.
guess I just gotta wait till the next upgrade then.
Anyway I can use argb headers on a non argb one?
Was thinking of getting some led strips.
3 pin rgb must go on 3 pin, 4 pin must go on 4 pin
You risk killing the device if you put a 4 pin on a 3 pin header
ooof, okay.
Just a simple hub would be ideal
hub?
Yeah like this https://www.amazon.com.mx/Thincol-Ventilador-refrigeración-Iluminación-Control/dp/B08M644QQX
would that also fix the fan issue?
That would yes
Would the built in hub in the torrent work?
It takes extra power from a sata power cable, and lets the mobo handle the lighting signals
I had to double check but the torrent hub is only for the fans, not the lights
Yes
This one is cheaper and has more rgb slots https://www.amazon.com.mx/Splitter-Ventilador-Enfriador-extensión-Adaptador/dp/B0BVVY7G1J
Multiinterfaz, fácil instalación. Admite hasta 11 interfaces ARGB, despida de la molestia de interfaces insuficientes. Especificaciones: Categoría: Hub de ventilador RGB Número de puertos de ventilador: 10 Interfaz de energía: Interfaz SATA Interfaz: 5V/3PIN Longitud del cable de interfaz: aproxi...
so for the hub I just plug it into one argb header and one sata header then I plug the fan's argb headers into the hub and it will work just like that?
Yep
The motherboard knows how many devices are plugged in, as long as there's enough power to them
no issues with brand?
Uh, Asus? No
gotcha, hmmm, the fractal torrent's fan hub can control fan speed correct?
Yes, for 4 pin fans
Arctic p12 uses 4 pins right?
Actually not the brand but the seller is bad
If you get the PWM model
all my fans are the pwm pst argb.
Then yes
Thinking of moving the fans on my AIO over to the fan hub and setting that to turn to x-speed for the cpu and the bottom fans to turn to x-speed for gpu.
Do you think that'd be a good idea?
It's a good idea but most motherboards can't handle doing a curve with GPU temp without third party software
Featured here https://youtu.be/uDPKVKBMQU8
Learn more about the Corsair XENEON Gaming Monitor at https://www.corsair.com/lm/en/blog/xeneon-gaming-monitor
Check out FanControl for free here! If you like this software PLEASE consider donating to the author! https://github.com/rem0o/fancontrol.releases
Get your JayzTwoCents Merch Here! - https://www.jayztwocents.com
○○○○○○ Items featured...
ahh okay.
should I still go with this hub?
No I checked the seller, they're a chinese scammer
for the argb wondering cause you mentioned this./
Oh okay.
Here this one ships faster, 8 fans supported https://www.amazon.com.mx/erounder-inalámbrico-Ventilador-refrigeración-computadora/dp/B08SHHXYC1
Especificación: Condición: 100% nuevo Tipo de artículo: Case Fan Hub Material: ABS + PCB Color: como se muestra en la imagen Características del producto: para fuente de alimentación SATA + control de placa base y control remoto Longitud de la línea PWM: Aprox. 40 cm / 15,7 pulgadas Longitud del ...
what's your recommendation on setting that up, daisy chain 3 fans per header? Wondering cause I am trying to figure out if the cables will all reach.
Worse case I think I can make connecting per header per fan work.
You could do that but for better signal integrity it would be better to do each fan on each header
If you needed to you could daisy chain, is what I mean
bad signal integrity can cause flickering or dim lights correct?
No bad signal would usually mean it doesn't change colors like it's supposed to or when the others do
Like lag
oh gotcha.
Low power would be what causes the flicker
gimme a sec, gonna see if I can find one with 12 headers if possible I'd love to be able to get 9 fans, pump light, and 2 led strips all hooked up to one source.
You'd be better off getting 2 hubs at that point, it'll be cheaper lol
oh really?
El CORSAIR iCUE COMMANDER CORE XT te permite controlar de forma inteligente hasta seis ventiladores PWM y múltiples componentes RGB mientras monitoriza con precisión las temperaturas del sistema, todo desde un solo dispositivo. La detección automática de cualquier combinación de ventiladores CORS...
Oh here's a good one https://www.amazon.com.mx/DeepCool-ventilador-puertos-magnético-Gigabyte/dp/B0BCVZS4XL
Control de sincronización RGB direccionable: apto para placa base con cabecera RGB direccionable de 3 pines (+5V, DATA, N/A, GND), compatible con Asus Aura Sync, Gigabyte RGB Fusion, MSI Mystic Light Sync, ASROCK Polychrome Sync y SignalRGB? Salida estable segura y eficiente: más concisa y más ef...
why is ddr4 4k being added here
huh
the pricing on this one is much nicer
thanks falcie for helping me find that.
Gonna order it probably tomorrow.
Quick question before I run CoreCycler and go to sleep.
What happens if I Undervolt to much?
Will CoreCycler give an error if the CPU does not hit the override with the set Undervolt?
Just noticed core 7 has been at 0 undervolt this whole time.
There will be an error message when the error is found, and it will skip that core in later cycles. It says at the end of each cycle if any cores crashed.
So it will indeed check if the core hits the rated override with the offset?
Yes
Ahh good to know.
What happens is when it's not stable there is an error in the math it's doing (calculating prime numbers) and that triggers the message
Should I disable the igpu since I don't have it connected to any monitors?
Nah iGPU is fine
Gotcha.
Kay, -25 Undervolt is set to all cores gonna let it run for 46m and check it if I am still awake then adjust if not then I will check in the morning.
Man am I looking forward to seeing an -30 all core on this CPU.
Would it be a world record if it did that?
Probably not but it's rare
I honestly don't know how well the 5700g undervolts, it's a different die than the rest of the series
Ahh okay.
By the way, what exactly are the perks for undervolting?
I know it helps with temps but anything else?
Temps are just the first part. Boost algorithms decide what frequency to boost to based on those temps. Lower temps, higher boost.
My temps currently peak at 50c and have an average of 43c during full load for about an hour.
This is in OCCT.
Is that considered very good?
Oh really?
So for example I could get like an extra 100mhz all core if it goes down by another 10c?
Maybe not that much but yeah
The lower the temp gets the higher it can clock, that's why you can get ridiculous results with liquid nitrogen or helium
Helium?
Yeah liquid helium
How does that work?
Oh okay.
I thought you meant helium like the gas used in balloons.
🎈
So far so good on the Undervolt.
-25 Undervolt worked on all the CPU cores.
Gonna let it run for the rest of the night to test stability.
Couldn't sleep so I started wondering, would a custom bios allow me set overrides beyond 200 for the CPU?
Because I know that for example with a custom bios a person can make a GPU consume far more wattage and voltage then the GPU would allow normally.
Sure, how good are you at hex editing and repacking a capsule?
You've got flashback so there's little risk involved if you make a bad bios at least
@modern walrus speedrunning the oc on this nanya f-die
never heard of this IC before, but it definitely beat expectations
this is the second time a corsair kit has blown past expectations
💀
this is 1 days worth of progress, still working on finalizing primaries, tRC, tRFC, tCWL/tRDWR. everything else is already as low as it can go
if this test passes then this is as low as the primaries will go, and I think also tCWL/RDWR are done
so then it would just be tRC and tRFC left
oh I forgot dropping tRCDWR, surprisingly that wouldn't post at 8 so I have to investigate it later
wow SCLs at 2 is like bdie's selling point for AM4
tRC won't drop? tRP + tRAS usually works but I don't think that's an actual formula
but very very nice even where it's at
testing tRC 64 rn
can maybe go lower
who knows, not done yet
i'm honestly not sure if it's worth raising tRDWR to 15 to run tCWL 12
but it's unstable at 14
i think it's worth? just not 100% sure
this IC honestly might be better than rev. e
the 160ns tRFC makes it feel like budget b-die lol
my rev. e couldn't run tRCD 17 either
Meter tried hex editing before and same goes for the ladder.
trying to oc kingston fury 3600 mhz cl16 how is it rn (hynix d) 2x16
its not final because I haven't seen any errors yet
youve got a while to go
which ones should I try with lower vals
i think my tuned 3200 kit beats your 3600 kit in every way
hm which subtimings are not good enough
ok
little tip though, max out trefi, should bestable on most kits but it may vary, if it doesnt like 65534/65535/65536 try 40000
was it how long the ram holds charge before emptying the capacitors and I thought its intel only I couldnt find this in bios
btw this is 2Rx8 will that make any difference with 1rx16 kits which has the same chip
also its 1.4v should I go higher
...... All 8 cores passed -25 undervolt with +200 override!!!
12hrs in CoreCycler no issues.
My man is gambling
Mohahhahahahh!
nice, probably done with this nanya f-die now
make sure to do idle test as well
What do you mean by idle?
some of my cores could pass corecycler at -19 but needed to go up to -16 to not error at idle
just leave your pc on overnight doing nothing
Oh, will do that again but it passed it last time.
whenever you push further with corecycler you should also test it at idle
I am pretty sure 30 is the maximum on zen 3?
maybe cezanne is different but I don’t think so
there's a max?
my 5800x3d maxes out at 30
I thought the 5800x3D couldn't be tampered with?
I’m 100% sure the max is 30 on vermeer
cezanne should be the same
I believe raphael allows up to 50
Raphael is Zen 4 right?
yes
cezanne is zen 3 APU (what you have) or some of the zen 3 APUs with disabled igpu (such as 5500, 5700)
vermeer is normal zen 3
but yeah, try setting 31 in bios, it’ll probably autocorrect to 30 or say invalid input
initially yes
but now you can run CO on the 5800X3D
it won’t change max clocks (you can’t boost override or anything)
It's just multiplier locked now, so no all core or direct voltage control
it’s basically just an undervolt
but it can result in higher sustained clocks under heavy loads because of that
Ahh okay.
my 5800X3D at -30 all cores can sustain 4.45 all core under multicore loads where it would be dropping to 4.3ish without CO
due to thermal or power constraints
That cache do be hawt
most 5800X3Ds can just toss -30 all core and call it a day since the max clock is so low
I did a 5800X3D build for a friend where I just tossed a quick -20 all core on it and didn’t even bother to stress test it, he never had an issue anyways
could my undervolt be causing my game Halo Infinite to be taking like 30 years to load up?
oh neat
unlikely but I suppose it’s not impossible
sometimes instability can manifest in weird ways
sometimes I’ve had unstable ram settings cause windows to take 2 minutes to boot (and then crash after trying to do anything) when the normal boot time is like 4 seconds
gen 4 nvme
debloated OS with practically nothing on it
I’ve also had unstable settings completely knock out wifi/ethernet/usb ports, that’s pretty common
it's saying "Syncing Data".
odd, everything else works.
It's just Halo Infinite being slow with the syncing.

looked up a fix.
Seems it's a issue that happened to someone else.
trying the fix
fixed.
Oh gosh, I haven't turned on the FPS counter since I changed the CPU in Halo Infinite and my gosh I gained a lot of fps.
I got about an extra 30 fps on average after all the OC and etc.
Running on average about 130fps in Halo Infinite now compared to the 99 fps average on the 10700f
Noice
also stutters have been reduced a bit as well.
Still noticable though.
The epic armor I use xD.
ram oc might help with stuttering
plan on saving up and buying the ram kit Falcie recommended me in a couple weeks then doing some manual oc on that.
4000cl18 gonna try to tune that to 4000cl16 or better.
ran Cinebench and it seems the extra undervolt got me about an extra 120pts.
Hmmm, I think the ram is the bottleneck here.
seems core #1 in the array is the fastest core. it was the only one active at 4842 MHz stock
what can i do to elevate these numbers? the ak620 seems to do a good job cooling and comparing to other stock numbers online, its a decent binned 5800x
For the single core test, curve optimizer. For the all core, a manual overclock gets the best numbers.
Falcie, would you call what I did a manual overclock?
Ahh okay.
The reason that's better in your case is games rely more on the single thread performance
What do you mean by manual overclock then?
You set the core multiplier and voltage, and it holds both steady. No boosting involved.
Isn't that a static OC?
So you'd set, say, 4.8GHz all core at 1.35v
Same thing yes
That'll depend on the voltage you set
is pbo 2.0 in the bios?
Yes, in an advanced CPU menu usually under AMD overclocking
ill work with it later. only on 3200mhz so i might see a 100 pt boost at most
ram
i dont think my mobo supports 32gb ram at higher than 3200mhz
If it was made in the last 4 years it will
im just mad my ram is 20$ cheaper than when i bought it
Would the static OC be better for workstation tasks?
x570 steel legend
You can hit 4800 with 32gb then if you tune it
The ideal speed will vary by processor
Wondering cause I might save two OC presets, one for gaming and another for when I use blender or something else that prefers an all core workload.
Yeah the static clocks are usually higher than an all core boost clock
What do you say?
I fall in the middle. To be safe 1.25v
Personally I do 1.3v because I'm a daredevil
0.05v yes
Interesting.
It can quite literally be a life or death decision for your CPU
Ouch noted.
By the way, off topic. I know the 5700g has the issue of not being able to set overrides above 200 so I was wondering if Intel's K chips also have the same thing making the I9 the only one without a limit?
no
also the +200 override limit is not unique to the 5700G
same limit exists on the r3 3100, the r5 3600, the r9 3950x, the r5 5600, r9 5950x, etc.
not sure about zen 4
Ahh my bad, I didn't know.
but out of the zen 2/3 CPUs the only exception is the 5800x3d (where the limit is 0)
Why would they set the limit on the flagship chip though?
Intel also has the xtu tool to set clock speeds at different load types and levels
the limit doesn’t really matter much
you’re not actually hitting that maximum clock in any meaningful workloads
an i3-8350K won’t stop you from setting 10ghz
5800X3D isn't the flagship, it's just the best for gaming
it won’t work, but you’re allowed to try
Was referring to the 5950x.
Ohh you mean why limit to +200
I hit the max clock in OCCT if that counts.
does it matter that the pbo limit is 6ghz if it won’t exceed 5.1ghz in any workload?
Well the reason is the 5950 is already the highest clocking overall and it's actually really hard to get it higher
my 5800x had a 4.85ghz limit, boost override allowed a maximum of 5.05ghz
never saw it exceed 4.85 for more than 0.1 seconds in even light games
Bclk it to 5.2 
would usually hover between 4.7-4.8
in fact going above +200 would be detrimental to performance overall
since you’d be dropping offsets to stabilize absolutely irrelevant points on the v/f curve
even +200 tends to be pointless tbh
like I said before, zen 3 chips often achieve best performance with an override lower than +200
Strange.
it’s not strange at all if you understand how CO works
Guess I should learn how CO works in the near future then.
however, your 5700G is lower clocked (4.6ghz stock boost) and is more likely to benefit from higher overrides than a 5800X that’s pushed higher out of the box and has less room to boost higher
I see.
I’ll give you a simple explanation
let’s assume a ryzen 7 5800x (stock boost 4.85ghz)
so override +200 means it can boost up to 5.05
Yes.
now let’s say that this 5800x has a core 0 stable with -23 offset at 4.9ghz
and the same core is only stable with -16 offset at 5.05ghz
if you were to push boost override to +200 and do your due diligence with stability testing
you would observe that at -23, your cpu would occasionally throw WHEA errors and lower your offset accordingly
however, those WHEAs are being thrown from when the CPU spikes above 4.9ghz at idle for milliseconds or even shorter
if you got your core 0 to -16 to be stable at 5.05ghz, you’d also be running -16 at 4.9ghz
but your core 0 doesn’t actually need -16 to be stable at 4.9ghz, it could do -23
then you play a game and you observe that the cpu never boosted past 4.9ghz to begin with due to silicon limits
so, you got a stable curve at +200 override, but that resulted in you lowering offsets more than you needed to in order to stabilize clocks that are not actually relevant to performance
if you instead used a +50 override (boost limit 4.9ghz), you’d be getting the same performance as +200 override, but at more aggressive offsets (-23 vs -16)
Think I understand you so far.
if you then ran an all-core load, you would be getting more performance, because now you’d be running 4.6ghz at -23 (the stable value for +50) instead of 4.55ghz at -16 (the stable value for +200)
and that’s basically all there is to it
So it's better to optimize boost clock you see most often rather than the highest clock attainable which the CPU rarely reaches?
TL;DR: you sacrifice offset magnitude to stabilize clocks that you are not reaching in reality
yes, pretty much
I owned that 5800X for like a year and it probably spent like 10 microseconds at 5.05ghz
gaming was basically always 4.85ghz or below, occasionally would jump to 4.9 in light games
so I would have gotten the same or better performance using a +50 curve or a +0 curve
since I could have run more aggressive offsets
The clocks I see reasonably often in OCCT most on this CPU is about 4.75ghz.
so that’s +150 above stock
5700G stock is 4.6
I think you could probably reasonably see some 4.8 in gaming
so for your chip it probably makes sense to use +200
Gotcha.
just ran OCCT and 5 cores hit 4.8/4.85ghz and the other 3 hit just shy of 4.8ghz by like 10mhz.
what software is good for testing GPU stabillity?
MSI Afterburner is the recommend GPU overclocking tool correct?
Yes afterburner is the most used/recommended
The best test for stability is gaming
gotcha.
With GPU overclocking there isn't any voltage, wattage, and etc concerns correct because it sets safey limits by default?
oh just realized I can't adjust voltage?
Oh, do I adjust voltage in the curve editor?
+200 core
+400 mem
seem like a good starting point?
then do increments of 25?
You can use curve editor for voltage but honestly not usually necessary
What GPU again?
Seems memory clocks really high based on what you said?
Yeah it tends to clock really high
And GDDR6 you don't need to worry about heat as much as 6X
also set a pretty aggressive fan curve just incase starting at 30c with multiple points leading to 90c.
Halo Infinite on the big maps should work pretty well for stressing the gpu I'd presume?
Oh, I thought it was synced to my bottom fans.
Nope the 2 fans on the GPU can be controlled independently
It only says 2, still applies though
Things to watch for, outright crashing is usually core too high, weird textures or glitches is usually memory too high
got it.
🤞 with all the OC and tuning I can hit stable 144+ fps so I can take full advantage of my monitor.
now loading into a match.
Increment by 50 or 25?
Currently incrementing by 25 on core and 50 on mem.
What type of visual glitches would indicate that it's the memory?
artifacts
somethin weird like this
oh gotcha.
Gosh looks like venom is taking over there xD.
now testing +175 core +1150 mem
afterburner also has a built in overclocker, wont work as well as manual but you can kinda get a baseline
gross
ahh will keep that in mind next time, I've already started the pace on manual.
i havent used afterburner a ton so idk, but when i did i just used that and increased it
its pretty useless, takes too long and also doesnt do memory
took a break, coming break and now gonna try:
Core: +225
Mem: +1250
That's odd, would having those set to high cause discord and my game to crash but not my computer?
yes
which one would cause that?
both, dont clock them at the same time
find max stable for core, reset core, then do the same to memory
Oh, that's what I've been doing.
this method.
clocking them at the same time leaves you with problems like you have right now
i'll put my money on core doing that though, if driver or anything didnt crash
So, find the stable point for both then try turning them both on at the same time?
yes
I updated my driver earlier today.
Undoing memory clock for now then.
Discord just crashed again?
Why is it only discord and not the other apps?
discord could be corrupted
fix it by reinstalling discord?
yeah
will do that real quick.
cause Halo, and Discord are the only things that crash.
back
okay so far so good 😄
Oh so that's what's causing it.
For some reason applying new settings closes discord.
+250 passed now doing +275 core.
What's considered a good GPU OC?
would increasing the voltage help me attain higher clocks and possibly more stable as well?
afaik, not necessarily since it'll increase temps
Currently running 3Dmark demo with the GPU on the default clocks to check what my baseline is.
Oh DAYUM, my R7-5700g hit 5ghz during the test.
actually increasing voltage, sure, but the voltage slider in afterburner is basically just a placebo. you can go ahead and max it out anyway though 
It won't let me move that slider though it does allow me to adjust the voltage curve via the curve editor by the looks of it.
Aight, it seems +230 is what's stable for me so far.
Also seems my CPU maintained a solid 4.8ghz through most of bench.
CPU hit 5.415ghz for a few seconds during the Time spy run!
Without moving the voltage slider I am limited to:
Core: +200
Mem: +870
Let's see if increasing the slider will let me get pass that hard limit.
so far gpu OC is much more of a pain then CPU.....
Currently at:
+150 core
+1100 mem
stabillity is so so, usually works but still crashes. Will do more testing tomorrow, for now I am a try to run CoreCycler with -30 undervolt see how that goes.
my entire pc isn't crashing.
Just discord, and Halo Infinite.
Chrome and everything else stays open.
how can I clear cmos without having to open the case
will some jumpers and 9V input be fine or just a button system for 2 pins
my motherboard doesnt have one of that clear cmos buttons
Not sure if you can. Why not just open it?
its unpractical for me bc pc isnt on the desk and its hard to connect them with a screwdriver head like this
should I set the tRDWR and tWRRD ?
they are 11 and 4 on auto rn
same question for tRDRDSD and tRDRDDD
2x16(2rx8) hynix d die
You can hook up the reset button to reset cmos
Orientation doesn't matter as it just needs to short
But just remember not to hit it after you're done setting it all up
I can connect the reset button again when I'm done, thats a great idea thanks
which ones should I try pushing, tRCDRD at 18 doesnt post
maybe higher voltage will do it but I'm preparing it for daily usage so I don't want to go over 1.4v a lot
also tFAW at 16 increased the latency in AIDA64 for some reasons
ok 68.6 with tFAW at 16, hypervisor is active in Windows so might be a bit more than what it should be
tWRWRSD and tWRWDD at 4 didnt post, I think I'm slowly reaching the limits of this ram
tWR is now 10, tried making tRDWR 8 but didnt post
I feel like maybe I've reached the limit of either this bin or generally hynix d at 1.4 volts
rn its fine with 1.38v as well
unlike my previous hynix c kit this one never errored in benchmarks it either doesnt post, doesnt boot or works fine
pretty sure it was a floating point rounding operation to test cpu stability and in ur case it wasn't stable and throwed an error
7/8 cores passed the -30 undervolt.
Okay it seems turning core 3 up to -28v did the trick.
What would be the downside if I set a static OC of: +200 override and -30 undervolt with the exception of core 3 being at -28?
wondering because I noticed that in benches I can hit 4.8ghz very regularly.
-28v sweats
???
i think its a little different from -28v, likely 28 voltage points, not just raw volts
if it was -28 volts that'd be impossible 😅
Yeah lmao.
gigachad cpu running at -26.7 volts
?
1.3 - 28 = -26.7
oh yeah, after all the oc and undervolting my multicore score in r23 went from 12,100 something to 14,386pts.
Would Static OC harm my cpu if I'm still running at an undervolt?
should I trust HWinfo's voltage count? I set it to 1.35v and it says my CPU is only currently pulling 1.27v.
Depends
Ended up upping the voltage.
Yeah funny how quick you can go from undervolting to over 1.3v load
Yeah lmao.
definitely not bad, seems about right for a zen3 8 core
Scored 15,209 in r23 with 1.35v @4.6ghz R7-5700G.
letting you know now, 1.35 is very high for zen
Will shorten life by about 2-3 years from 10 correct?
shorten life is not the problem lol
degradation, you wont keep stable at xx volts at xx freq over prolonged use
life span has never and will never be a worry for me while overclocking
How bad is the degradation? Also during the run although it's set to 1.35v it was mostly running at 1.27v.
not something i can give a set and stone answer to
this is a lot safer, PBO changes the voltage of the cpu literally thousands of times per second, while fixed votlage is well, fixed
Support me on Patreon:
https://www.patreon.com/der8auer
Save 10% on your iFixit purchase: DER8AUER10
https://eustore.ifixit.com/der8auer
Pro Tech Toolkit: http://bit.ly/2JOFD8f
Find my products at Caseking:
Delid Die Mate 2: htt...
this is at a higher voltage, but its something to keep in mind
for PBO?
ohh okay.
PBO, unless ryzen 1 or 2K will not hurt cpus if you dont do absolutely stupid settings obviously
xD, +200 offset yes, yes.
I currently have my PBO preset at +200 override with 6/8 cores at -30 offset.
the other two are -25 and -28.
fine if stable
stable.
As long as you don't touch LLC, or set the scalar higher than 1x the smu will keep the cpu safe
All the files used for this tutorial: https://drive.google.com/file/d/1rRkVDpruyRo21SGdLG7U8QcgShYt0p4J/view?usp=sharing
Known issues: GF114 isn't working with this mod, earlier Teslas like G92 might work with 2.0 automatically but general solutions haven't yet been explored.
0:00 About the mod in general
2:38 Saving BIOS Tesla
3:18 Saving BIO...
is this not meant to work?
i was running hd 6770 which is pcie 2.0 on my 13900k last night
wow that is a bad picture
got it
hmm, it was running pcie 1.0 4x lol
i had it in 2nd slot so maybe that
ill plug it into main slot and test rn
What's a good DDR4 Ram OC guide? Also does anyone by chance know what die is on my DDR4 3200Cl16 Oloy kit?
I am reading it's most likely Samsung C or D die.
That probably means you have a broken capacitor on the pcie slot

could this also put main slot into 8x
wat
my top pcie slot has been stuck in 8x ever since corrosion incident, i was assuming it was just more corrosion under pcie
either way, i cleaned out pcie yesterday. still assuming it was just corrosion.
\
16x pcie 1.0
no broken cap, my bottom slot just maxes out at x4
how’s the bench setup
where's the bucket of ice water
cry about it
are you running your 11900k rn
700k but yes
huh how
Tried my hand at seeing if the 12900kf I sold had degraded after some benching
Yeah it was uhh... definitely dead
💀
the chip is dead?
No it boots
so whats the problem
This is as far as I got 3 times
stock? lol
how did you degrade it so fast
oh...
Had it benching at like, 10C for 3 or 4 days running 1.45V
💀
Ya
oh thats not even bad
eh
Either way I couldn't get it to work
well you have a real cpu far
what does that mean lol
BTW far that mobo still isn't responding at all. I've done everything but pull the cooler and check the pins but I don't see why it would just refuse to try to power on.
I'll work on it when I'm not feeling so sick
You could try reseating the CPU. I figured you fixed it because you didnt respond to my DM
Nah I've just spent most of the day in the bathroom
It's entirely possible something hit the socket in shipping, the box was fairly beat up. I'll keep you updated.
Well the chip was in the board right? At worst it might've just moved the chip a bit since the keying is a little damaged.
Yeah that's along the lines of what I thought as well
Once I'm up to it and I find my extra paste (used the last of my mx-5
) I'll give it another go
alright. keep me updated. sorry about that, not sure why it's acting up
I don't blame you, I knew what I was getting into lol

Once it's done I'll have to show you the jank af setup I had to do
The only screws I had that were long enough and fit my radiator were some old AMD cooler screws so I had to double up on fans to get it to fit
Good thing there's lots of fan headers lol
@sudden torrent, are you available currently?
Between bathroom runs
okay. when this says Reset, does it mean full pc reset or just thaiphoon or restart?
It would be a restart but with a full power down between
Give it a try to see if AB will work afterwards then?
Sure, worth a try
Oddly enough I'm hoping you have c-die because I know very well how they overclock (not great lol)
If i'ts d-die might be better off lighting it on fire
is d-die that bad?
Eh depends if it's 4Gb or 8Gb really
Right, the capitalization was important there
Gb is not GB
My bad, GB is gigabyte and Gb is gigabit correct?
Yes
fixed.
1 byte = 8 bits so 1GB = 8Gb
gotcha.
So what should I do if AB refuses to read?
I already tried reinstalling it.
the ram sticks I have look like this incase you somehow know what die this particular stick uses.
SO MUCH INFO on that DDR4 guide....
Bro go to school😭
loser
Real, know me to well.
my rig was like this until 4am
(i went to bed at 4:30)
i wasnt lying bro i do this almost everyday 😭
i have d-die laptop experience

the only ddie you should be talking about
but my laptop 😦
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...
Is there a way to up the power limit on my R7-5700G and if so would it help with performance?
Yeah you can do that under the PBO settings
It's not as simple as setting it to max and calling it done, there's a sweet spot.
How does one find this sweet spot and would it help with performance?
I am currently seeing my CPU pull about 60w on average on all core workloads.
Testing
Bench and compare scores vs different limits
Yep. Run cinebench, record score and limits, adjust, repeat.
My motherboard has a "motherboard limits" setting and if I'm feeling lazy I'll just use that. It sets 275W PPT, 144A TDC, and 192A EDC.
damn you got one of these for 7700x 👀
It's the same procedure
Everyone would get slightly different results, even using the same CPU
Besides extra heat any issues I should be concerned about when increasing the power limit?
It'll only use what it needs so even if you maxed everything out you wouldn't be risking much
Just like the 600W vbios on the 4090, it's not actually using anywhere near that
Gotcha, in the event it does what would happen out of curiosity?
If a bunch of safeties failed and it actually pushed 400A through the CPU? It'll probably trip the PSU protections first but it might fry the CPU/mobo.
Gotcha.
There's at least 3 safeties that would have to fail for that to happen
I managed to defeat one with a bent pin I fixed on my old 2700, it stopped reading current but nothing else changed.
Gotcha
You can see the result of that bug here actually. It thinks it ran at 774W 
I'd imagine.
yet another example of why i never trust what the benchmate window says 
That sub got me gold for a while until some jerk slapped a LN2 pot on a 2700
No one else bothered to do LN2 on it lol
post a link lol
Blew me away https://hwbot.org/submission/5074202_jedi95_cinebench___r23_multi_core_with_benchmate_ryzen_7_2700_14272_cb/
yea pretty recent




