6GHz boost on a 13900K and a discussion on verifying dynamic CPU overclocks.
HTML-код
- Опубликовано: 21 окт 2022
- My Patreon: / buildzoid
Teespring: teespring.com/stores/actually...
Bandcamp: machineforscreams.bandcamp.co...
Twitch: / buildzoid
Twitter: / buildzoid1
Facebook: / actuallyhardcoreovercl...
#overclocking #intel #13900K 
Наука
Buildzoid - You can tell which core is being used by which worker in P95 by looking at the log line "affinity set to cpuset XX: no error" and converting XX from hex to know which one it was set to. Love your work, thanks.
Since we know a KS is likely coming early next year with a 6 GHz boost, we'll probably see a lot of 13900K stock with 1-2 crappy cores right now that were rejected during binning.
Good Point.... but i think the overall consensus here is to just leave everything at auto, as it's just nonsense with these chips to do anything else with them unless you have way too much time on your hands and like to waste unthinkable amounts of time for very little return, if any.
@@dabneyoffermein595 No self-respecting overclocker would leave everything at auto, whether it gives you a benefit or not. You're on the wrong channel if that's your argument.
Hmmm that discussion at the end was just really interesting!!! You should do that more, especially regarding stress testing.
Fantastic video BZ! There are some people that have so much time on their hands that they can afford to run various test on the boards to find different results that others don't have time for. It's all about enjoying what you're doing!
I agree with what you said on the time spent and what you gain usually but sometimes it's just fun to get the most out of our system.
agree, but tbh we are kinda masochists for overclocking hardware and esp. ram setup
yeah but if I'm serious about maxing out the system then I'll just got HWbot rankings not daily settings.
Sleeping schedule more messed up than zen4 on win11
You have an awesome stream of consciousness. Thank you for sharing this video. I hope I had seen it before I lost my temper on 13900k
Buildzoid and Intel overclocking, makes me a happy camper. Thanks for the vid bro.
It's not really practical, yes, but it's so much fun to see chips boost to silly clocks on lightly threaded workloads.
My pc isn't at all mission critical, so I don't mind having it run a semi-unstable OC. So far my OC has been working well for over a year!
The most amazing thing is that you did this next year.
So true, stress testing is time consuming when you have bad silicon. Stress tested with the heavy loads with prime everything, works snappy but I would leave and come back to my PC off or restarted randomly thinking It was electricity issues at home. all along it was the manual tune, I leave lots of browser tabs and Large PS documents open and something in-between triggered the crash. Memory Training on the Asus x570 and having to poke the reset got tiring, but its kind of fun learning what the cores are capable of individually or not if its really wack like my 3900x
did you have a log running so that if it did crash you could tell why? i realize that logs don't always tell the story. I was thinking along the lines of HWINFO logs. Also, what program do you use to overclock a 13900K?
BZ OCCT can cycle through threads at any rate you want with any amount of threads (ie so cycle single core at a time, cycle 2 cores at a time and so forth) and also cycle between single and all core. it will tell you which core errors.
i absolutely ripped my hair out trying to get curve optimiser to stop blacksreening, testing and testing for days while out and such with corecycler blah blah.
OCCT had the system stable with 0 blackscreen within about 30 minutes (ie the settings i dialed in after 30 minutes of OCCT have been stable for months now). turned out there were issues not with any single core but with certain combinations of core(s) when loads were switching between threads. OCCT was a godsend mostly because i think of this RAPID thread cycling on multiple cores combined with switching between core X and all but core X.
I've also had good experiences using OCCT for stability testing for daily use. Though it's utility for hardcore overclocking is beyond my knowledge level.
Yeeea more videos.
Stuffs insane!
Not going to help with dynamic OC testing, but I've found that running linpack and P95 blend simultaneously (each running with half as many threads as the CPU) is actually slightly harder to pass than P95 small FFTs while being significantly easier on the cooling. Passing that without errors requires the frequency to be ~50 Mhz lower than what my 8700K can pass small FFTs with.
P95 SFFT on it's own tends to not be very transient so that's probably why you're seeing that.
Grate video as usual!
I have also been asking my self if its worth doing all the combinations and the answer for me was no XD
So yea an automated tool wold be nice.
Another program I want is a memtest program that loads up the PCI-E buss while doing memory tests.
Thats because I figured out that sure memtest86 might pass your ram OC just fine 4/4 passes over and over again BUT as soon as you had PCI-E loads, GPU/nvme I would see random errors on the PCI-E buss and GPU driver crashes.
JEDEC speeds where fine and never had a problem just to verify it was a HW issue.
It got even worse when Linux kernel 5.10 came out with more optimization for better I/O die performance on Ryzen 3000/5000 and what use to run almost stable would not even get to the desktop.
I could also see that a lot of people with Ryzen system and AMD GPU's did bug reports to the linux kernel and AMD put out patches but they would come back complaining about the same errors again that I saw in my own log.
Once I knew it was I/O die related I sad fuck memtest86 and uses some normal programs I knew would trigger it if I ran them one after the other and I eventually had to lower the memory OC and loosening the timings and now its been rock solid for 18 months.
To me the I/O die behaves a bit like a multicore CPU where trying to run all core load at max clocks just wont work, but running one part like the IMC at max will work, then PCI-E parts sees hello and it all comes crashing down.
Modern CPU's needs modern testing tools like you asked for and like I am because there are so many variables to test for now that its just tedious.
Idk what is considered the best memtest for Ryzen but memtest86 did not come any where close to actually tell me if the ram or CPU could run the speeds I asked for, I can run 333Mhz higher clocks on my DDR4 ram before memtest86 complains then I can do with an up to date OS and actual software.
18+ months stable now at least with no crashes and I run 24/7.
So in the end my OC over the XMP profile it pretty small but well at least its stable and any OC is free performance I guess.
Thanks again for a good video!
Isn't memtest86 widely considered to be very bad? Don't get me wrong, it has its place (since you should stress test using multiple different tests for daily anyway), but I would never rely on ONLY memtest86 for daily.
grate
@@maxromanowski6069 Well for daily use I use Intel Linpack and Prime95 24h for the CPU side of things. For ram I historically used memtest86+, still do for older systems but since memtest86 seems to be the one in development and used today I tested that. But yes its sucks and is totally useless in my view and that was the point.
But the thing is, is it just memtest86 fault alone or do we not need new software for testing because any memtest program will stress ram and the IMC and nothing else and modern cpu's require more testing then ever.
One could run memtest programs from the OS but that gives limited access to all the ram and has its own issues.
It would be great if you could find a simple software that checks each core with one click
This will only get really useful once "Threat Director 2" as shown by Intel in a PCWorld video will go live. Than the background tasks will really be placed on the e-cores where they don't mess up the threatcount. As you pointed out correctly currently you can forget about the lower threatconts as you usually have some things in the background like steam that don't nead mutch performance but count as a used core if they happen to land on a p-core...
Also, I have the Z790 Master with 7200cl34 Hynix, and the "low latency" and "high bandwidth" options are like magic... significant increases in performance without changing timings or frequency. Totally stable in Karhu and y-cruncher - I don't know what they're doing to achieve this or whether you've talked about it or not. The Z790 Carbon I tried couldn't even keep it stable at stock XMP.
When I get either the 13700k or 13900k I will only be running at stock.It’s plenty fast enough!
glad i could help
52:50 this is the problem I have on my z97!! I had a day where I set my 4790k to 46 core 46 cache, and it passed everything, back to back! (my usual use scenarios anyway) 8~x back to back cinebench runs, an X264 encode after it, then a bunch of 3dmark Time Spy/Fire mark runs, followed by Cyberpunk benchmarking and gameplay. Everythings good!
Then I came back the next day to my computer, tried to do the exact same benchmarks, and it was instantly failing the cinebench step!!!
Buildzoid try using a package power limit as a means to drop frequency instead of the boost algorithm.
I did this with a 9900K and was successful as it prevents the cpu from pulling to much power at high frequencys causing the instability issues.
As an example I setup 5ghz fixed all core at 1.38 volts and a power limit of 140 amps. Anything beyond around 180 watts would pull down the frequency =)
I have the EVGA CLC 280, where did you get the 1700 bracket, and how well does this cool the 13900k?
Yeah seems like good cooling is the main thing to be concerned about.
I’m more focused on getting my RAM timings tight for better frame time consistency.
I’ve found a 80mm 1800rpm Noctua fan works great for cooling RAM. I’ve dropped ~ 8c with this / 45c max ram temps (R7 7800X3D / RTX 4090)
Buildzoid: You completely forgot about Stockfish AVX2/BMI2 builds? That tests all of your threads, requires like 30mv higher load vcore to pass than Cinebench 23, and doesn't massively overheat your CPU like Prime95 small FFT FMA3 (which no one without a delid is running at 5.6 ghz anyway). I use stockfish for my daily testing as it's actually sane to run. I'd rather have 88C daily 24h stable from chess testing than 110C from small FFT FMA3.
I also did a comparison with asus OCtool -->Raw VRM.
Stockfish at 5.5 ghz, 1.163v load, was pulling 230 amps and 88C.
Prime95 FMA3 at 1.172v load was pulling 320 amps and almost instant 110C. That's risking slow degradation (you're both past tjmax and iccmax of 307A).
Also I've found that Stockfish can generate a CPU Cache L0 error or Internal Parity Error (yes, they are back with RPL) where LinX 0.9.12/Linpack or Y-cruncher would happily pass.
The ONLY time stockfish will pass when y-cruncher / linx bombs out with a red critical error is when the E cores crash due to insufficient vcore. And yes, you want E cores on for stockfish because for chess engines, higher hash rate=better, but you can disable them for P core testing.
Note: W11 22H2 is best for stockfish. W11 older builds and w10 have a scheduler problem where if you alt tab out of stockfish to a browser, it will put the p cores to sleep and fully load the e-cores until you put stockfish in the foreground again (with E cores enabled).
Stockfish works best on linux(best distros for ahoc linux are the ones without systemD runs tighter & quicker).
Launching the linux kernel at boot can slap any hard ram oc and/or cpu oc crashing, but if it works & boot to the distro(i use MX linux or antiX cuz both are debian & no systemD) then use stockfish+linpack xtreme(v1.1.5) is one of my ways to stress test before installing gentoo(got one pc with 8 months uptime, no crashes)
Wy don't you use Thermall Grizzly Kryonaut Extreme? From my testing, Der8auer, Jays2cents, etc, it beats the MX-4 from about 3 to 4c
Had close to same experience and same conclusion on my 5800x.
I tried to chase the magical 5000 on my 5800x, I succeeded but just turned my CPU into A and B team cores.
With PBO target of 4900 almost all cores hit that target.
With PBO 5000 half of the cores hit the boost target, but the other half went down to 4700-4800, so I ended up with practically no performance increase.
More testing on individual cores was simply not worth the time and effort for me.
Thanks buildzoid. From stress testing cache I remember different tools gave different results. Even tm5 helped me find unstable cache.
Can't wait to get 13900k and 8000mts hynix adie
Bit of a random question, how hot does the back of the socket get pushing chips like the 13900K and 7950X in all core loads? Perhaps an AIO with 1.5mm thermal pad mounted to the back of the socket could offer a slight reduction in CPU and VRM thermals.
Buildzoid actually did this with I think 10th gen. And ended up modding one boards power delivery because it making it justifyable for the second aio.
I tried that with 10th gen or 11th gen. The AIO just ended up cooling the VRM rather than the CPU.
6ghz. sick! What power TDP were you running with prime 95?
@Buildzoid 1.what's up with Your AMD 7950x death?! any follow up?
2.Where do You see AM5 memory clocks being at in these coming months?
3.Do you disable Page File SYS when benchmarking memory?!
@buildzoid any plans on making a 13900k video with a unify X?
Are these per core settings available on Asus strix B660 ?
3:05 to 3:20, Buildzoid, would you elaborate on what you said there, wouldn't that mean disabling the e cores, would shorten the ring, and raise clocks?
Hey bro! Something strange with overclocking my 13900kf Ring Bus: whatever value i put, it remains 4500ghz. P-cores are 6.0ghz, E-cores are 4.7ghz, voltage is 1.42 + llc mode 1. Could u please help me with any suggestions? Many thanks.
46:47 what if you leave everything on auto and also increase the bclk? I imagine it would act as a sort of offset for the stock behaviour. Probably would still be a pain to stress test though.
yeah that still suffers from the stress testing difficulty.
my cooler is a Corsair Capellix Elite 360mm and I'll use Thermal Grizzley and the Grizzley frame (if you think i need the frame) and my CPU is a 13900 and I need it overclocked as much as possible. Do you think i could hit 5.6 single core? and lower on all other cores, but hoping to crank all cores to something half way decent. I have windows 10 and I have a 4090 GPU and a 1200W Thermaltake PSU and a z690 mobo Asus Strix "E" DDR5 with G-Skill Trident 6000 32GB of RAM (16x2). Is it possible to get 5.6? or am I way out of the realm of possibility? Or perhaps the better question, is there a video you did of the 13900K of more practical gaming overclocking , like where you try to peak a single core to something really high and tune the other cores down to keep overall heat down since i will only be using the Corsair Capellix Elite 360mm AIO? Flight SImulator is VERY demanding of CPU and at this juncture in time still only uses mainly one thread (single core), but in the future Flight Sim will use DX12 better and take advantage of more cores. Also Flight SIm does (MSFS-2020) does not like WINDOWS 11 on my setup above and therefore i have to stick with Windows 10. Anyway, any help with these concepts would be appreciated!!!
Standard TVB mode should get you up to 5.8GHz on a single core on most if not all chips. As soon as you boost more P cores the power goes up rapidly, as it does with E Core speed. I'm just building an OCUK 8Pack 13900K bundle into my case. With custom loop watercooling it's running all the P Cores at 5.7GHz with ease and the E Cores at 4.4GHz. My 4090 is air cooled and no real need to add a waterblock. MSFS actually runs over 100 threads but, even in DX12 is not well optimised yet, but it runs best with all the P Cores boosted. The system is a huge upgrade on my 5.1GHz all core OC 10900k and Aorus Xtreme 3090 for MSFS. I think you'd do best not going for single core performance but optimising all the P Cores to the maximum speed your cooler will allow while staying below say 85-90C. Edit: I'm running fine on Windows 11 22H2.
Thoughts on 13700K? All I really want to do is 5.5 on the p's and 4.5 on the e's - nothing crazy. It will be cooled by an ARCTIC Liquid Freezer II 280 (top mount) in a P500A with all four fans.
i wonder if one could script core affinity to jump around between cores and test all combinations.
Anyone know if the 790 Elite Ax VRM TWIN 16 is any good for overclocking the 13900k ?
Are you guys all using fixed vcore or adaptive offset for getting the 13900K prime stable? I'm having a hard time getting it prime stable at stock with adaptive offset no matter how much voltage I add or take away? What could be my issue? I'm trying to do the blend test and it passes line pack but never fully utilities the cores and drops out frequently during prime blend. What could I do to stabilize the 13900Ks on Aorus Z790 Master for prime?
props to Intel for making cpu overclocking more tedious than ram overclocking.
Someone needs to automate this. Someone smarter than me lol
The trick to setting a daily oc is figuring out how stable you really need to be. If linpack can run through its stress test a couple times without error it's good enough for me. Cinebench is probably as stable as lots of people need. Cause ppl typically buy wayyy more cpu than they will actually use lol.
Any time I'm really getting more nitpicky than that about settings im tailoring it to perform a specific benchmark. And usually 95 percent of the time spent is getting the last 5 percent of performance. Which is not at all noticeable in daily use.
This is true, if my 4790k can pass a couple back to back cinebench runs and then an x264 encode stress test pass... and can get through 5 (my fps test samples) benchmarks of the Cyberpunk benchmark + a full day of typical fortnite/black desert online/browser usage... that's good enough for me...
I find that I must have some kind of unmonitored hotspot on my cpu because it seems no matter WHAT voltage I throw at it on say 46 or 47 multiplier, it'll just bluescreen with either WHEA or Watchdog (random) the moment it sits at 80c+ for a long time, like it spikes to 90 or 100 or something in the hotspot or whatever for a brief moment that hwinfo doesn't catch and then shuts down.
That 95% of the oc is done fast and last 5% is so true tho. I could be totally fine with 46 core 44 cache... but i could boot and play games and seemingly function - and have minor but recordable changes in Aida, cine, fps scores with it.. so my brain is like YOU GOTTA WORK WITH 46/46... when it's literally like nothing..
I built my first system this year.
Here are my specs.
13900k, Asus Z790a, 32 Kingston ddr4 ram and a EVGA 3080ti.
I used the auto overclocking AI suit 3.
It automatically over clocked my 13900 to ×62 and the softest showed it was hitting 6000-6200gh. The problem was thermal throttling and the performance dropped.
I dropped it down to ×58 and the performance jumped tremendously. Problem is I have to manually lower it to ×58 every time I turn on my PC.
I'm afraid to push the chip over 95c and for daily use Sim racing the PC runs best at ×58 on the P cores and the GPU runs significantly better also.
Being new to all this I'm terrified I'm going to destroy my 13900k if I push it to ×60-62.
Perhaps a combination of ASUS ROG REAL BENCH and a PROJECT LASSO can be the she's text your looking for, it may require a few runs and configuring in PROJECT LASSO to force things to only run on specific cores, but that could be a good place to start
i have a 5800x that would be capable of like 46 maybe 47 but theres one bad core that will cause the system to get unstable and crash 24 or 48 hours after boot
56:58 lmao that Facebook notification
I was getting really poor performance in some games and 3dmark after I swapped a 13900K into my z690 apex. For some reason manually setting the ring clock fixed it.
I know it's an old question but for you guys what you consider safe voltage for all cores in Ryzen 5000? People say 1.3v and 1.35v max and I saw skatterbencher using 1.375v. :)
The new answer is that it honestly doesn't matter, as the higher your voltage - the more the CPU throttles internally, resulting in less maximum performance.
So really, the "safe" voltage is "whatever runs best in the range of 1.2-1.35v"
CLC name vs AIO what one is well accepted? Rtx 3090 ti or tie 👔 🤔 what name is accepted? We need to get these name corrected
How about running multiple instances of Prime95, and setting their affinity to a dedicated core, instead? Seems like it might save you a bit of work.
Can you do some msi motherboard and 13900k. I'm struggling with my z690 and 13900kf
Misread title "60GHz" 😱
You showed me dirty p95 stuff, I have never seen before
i9.13900k using dry ice De8brauer o/c thumped it 6.5ghz -32cel Calvin
actually hardcore stability testing
sounds like my life with my small home pc service business with Craig's momlist(craigslist) vs frame chasers cuz we are at the same city, I price more fairy low even lower for those in need. Thank god for teacher BZ i take a lot of notes almost 3 years, also I ahoc in linux btw
There's 1 article out there saying that you can lose around 25% performance when using 13900K on a Z690 mb (they were using an ASUS ROG Z690 Hero, BIOS 2103). Hopefully that isn't the case with the Unify-X I have. Does anyone have any other results for a 13900K on a Z690 mb?
It’s due to the older bios. As soon as you update the bios the issue goes away.
Wait are you not supposed to use Win11 for Raptor lake or is the new scheduler stuff is all BS?
Win11 suck
Ideally yes, but since buildzoid is just doing overclocking, and also likes to disable E-cores for higher P-core overclocks, it doesn't apply to this. You can absolutely use w10 for daily driving too, but W11 is preferred for better scheduling
win11 is only necessary if you keep the Ecores turned on
@@Coffeephile Hell No Win11 22H2 Is Way Better Then Win10 21H2/22H2 Win10 Even Has More Bugs
Did you get a new Zen 4 chip to test yet? I'm curious about memory overclocking on X670E/X670.
Somebody get the 1usmus to recreate hydra for intel cpus! Automatic curve optimizer with multicore stress testing on my 5950x was badass... Definitely felt like it maxed out the silicon, and gave me complete control over power efficiency decisions. Oh well, one man can only do so much!
I don't know about you guys but I find the anta777 absolut for TM5 truly amazing to test ram stability.
I used that when testing mine, 10 months stable so far.
Would be easier if Intel marked which cores were "best cores"? It feels like AMD's curve optimizer is nicer for per core OC.
Of all the sad words of tongue and pen, these are the saddest: It's probably worth it to just leave it on Auto.
You need some python scripts for testing automation. There should be something in the XOC community.
i have the z790 aorus master with a 12900k I'm Using GSkill Trident Z5 RGB 32GB (2 x 16GB) DDR5-6400 PC5-51200 CL32 Dual Channel Desktop Memory Kit F5-6400J3239G16GX2-TZ5RK. When i enable xmp profile save and exit, the monitor loses signal and stays that way till press the clear cmos on the i/o panel.
12th gen's memory controller is a lot weaker than 13th gen's and I wouldn't be surprised if the Z790 Master's BIOS straight up lacked some memory opimizations that the Z690 boards had. (though even on Z690 6400 wasn't exactly reliable with some 12th gen CPUs)
Maybe enabling XMP and then dialing down memory to something like DDR5-6000 or DDR5-5800 and then save and exit could help.
@@ActuallyHardcoreOverclocking Just installed 13900k, same thing, will not accept xmp profile. Anything over 5000 m/t dosent work. The bios needs an update. I need to remember i bought the z790 on launch. I installed the board 10/21/22 4 pm pacific time, with in 24 hours 2 bios updates were released. Tells me their working on it.
@@volodumurkalunyak4651 Just installed 13900k, same thing, will not accept xmp profile. Anything over 5000 m/t dosent work. The bios needs an update. I need to remember i bought the z790 on launch.
@@birdsoup777 very sad. Bios update is definitelly needed.
Do you also have 13700k to test?
not yet
Tuff Sledding Man. Try paneled sail making if you want Tedium.
2:30am, time to do a youtube vid.
Hello Bill Zoid.
The best gaming benchmark for CPUs right now would be Microsoft Flight Simulator 😅 I got all my hopes and dreams crashed when i tried to use my 12900k/3080ti combo on it... Heavily CPU bottlenecked... I uninstalled it... Gonna wait for a 16900k or something until it's actually playable(flawless) for me 😅😅
P95 for me is weird… my 5600x hits 97w 2 higher than my set value 95w but it runs cooler than if i was benching R23 on a loop
What is using power that dont crap out a tone of heat when running P95 mixed?
I would set my limit higher but my cooler sucks so 95w works year round and i can go upto 115w during the winter if my let my room get pretty frickin cold 12-14c and not turning my PC off vs what i normally have my room at 25-27c
I just like being in a warmer room than 99.9% of people… ohh and still wearing a hoodie, yeah my PC hates me but its the only way to guarantee my PC boots otherwise im in a boot loop till my RAM heats up enough to run my overclock
Ryzen on stock settings has a realy low current limit so it can't run P95 at full tilt.
@@ActuallyHardcoreOverclocking ive got mine set to PTT 95w (cooler sucks) and the other 2 at 150A
Should i raise the current limit?
Tho that would be 0.633v to draw 150A at 95w anyway… wouldn’t it?
Fun tests, but really .. as silly a cpu both the 12900 and 13900 is, its only good for identifying bad cores, but really, if we forget the redicolous power draw on the 900 models. How does 1 or more bad cores go thru quality/binning on their halo processors is beyond me. That shouldn’t really happen in a perfect scenario. Such a chip should be binned to a cheaper model where you can disable problem cores
If you go straight on ftt128 you gonna put load on only core, the more you step away from it you put load on imc
I know that. But I have other stress tests for the IMC and RAM. Testmem5/Ycrucnher/Linpack/HCI memtest.
Buildzoid do you know why MSI stop to make MEG B550 UNIFY-X if a lot of people are buying AM4 CPUs? To me this is the best B550 motherboard to combine with Zen 3 CPUs and where I found it is so expensive that appears to me that vendors know that MSI will not make this anymore and they will sell a collectors' price... =/
idk and i guess msi is more focused on the next gen for the unify anyway, i don't blame them even if that broad is my most wanted.
@@emini6 I understand you, because MSI will focus on the new platform. If you look for another way MSI is just crazy, AM4 CPUs are selling now as hot cakes and will sell more and will continue strong because of used(second 'hand') market. B550 Unify-X is the better board to make a combo and MSI is 'hurting' itself and the possible costumers!
Thermal Grizzly Kryonaut Extreme is the dog's danglies, that and a cheap from from AliExpress and I saw a decent jump in performance, like I was able to get over 12000 in CPU-Z benchmark, and beat the 5950X for the 1st time since I got the 12900k a year ago.
"dog's danglies"
wtf?
Affinity mask kills perf. Jesus.
I know you mentioned this, but I never see these boosts in anything useful, yet everyone is pushing it these days as the way to OC. All-core OC still seems like the king to me - seems like a waste of time to optimize for anything else.
I kinda agree, Witcher 3 DX12 or Eldenring have very low core activation, my 5.1 allcore on my 10850K vs 5.1-5.4 I still see these games shift the clocks 5.1, 5.2, 5.2, 5.1, 5.2,might flicker 5.3 occasionallyif at all.
Lots of cores outside the game are still minimally activated, maybe process lasso would help but thats even more work when 5.1 fixed allcorr is no extra work.
@@SKHYJINX I'm surprised you were able to see that behavior at all - I think in every game I've tried it always downclocks to all-core.
I'm just starting to watch this video, but I have a simple question why are you only using a 280mm radiator with a chip that can easily push more than 200 watts with only P cores enabled? unless you have 140mm fans that can push 3000+ rpms is kind of pointless to use that radiator. You need something like a 480mm radiator with 3000+ rpm 120mm fans to keep in under 90C for overclocking. Using ANC headphones would be recommended. If you want a stable system you Linpack it for 24 hours to start with.
The cooler is capable enough for these tests, thermals aren't really the problem but better cooling would be ideal.
@@JJFX- Not if you are not passing enough air through the radiator. If it was "enough" he would not be thermal throttling in such short tests.
@@michaelcarson8375 I don't recall him actually throttling in any of it but I could be wrong. He did cross 100C for a bit but he has the max temp cranked up in BIOS. It's definitely not good enough for OC'ing this chip. It does hurt stability, I just don't think it was significantly impacting what he was showing.
Maybe it's a silly qustion, but I haven't fully understood why we should uncheck hyperthreading.
because the hyperthreading setting just spawns 2x the threads. So if you're trying to test 1 core at a time it's not helpful.
@@ActuallyHardcoreOverclocking Thanks for the explanation！
It took me 5 years to figure out how to talk to intel chips and overclock on the 14nm derp, now they got the P cores and the E cores and the D cores and the O cores .. man, I just don't have time anymore to start all over and figure that fucketry out.
This are e-waste cores anyway... Intel amazing innovative mindset to compete with AMD. Just turn the shit off on these e defective cores
115 °C doesnt seem save for dayli use
every other ga-
do a 790 mb video plsss love you
> I'd like to make a video every other day
Who stole our BZ and replaced him with clone?
a video every other day has been the goal for this channel for a very long time.
@@ActuallyHardcoreOverclocking it was about words "like" and "make a video" in one sentence
@@funtaril I'd like a video to spontaneously come into existence every day. I don't actually want to spend any time making them
@@ActuallyHardcoreOverclocking Well I can say we all very much appreciate you not becoming one of those 'those' channels constantly pushing watered down content. Feel free to do as many "low effort" timing reactions or showcase videos as you need to.
So... all-core OC is still the way to OC in other words...
if you're lazy. Yes
idk why people waste time doing a by core usage oc, it’s such a big waste of time, it will very very rarely be useful in a real workload like you said it yourself. Why not do a proper TVBOC and actually get 6ghz on all the p cores in workloads such as games? Been doing this since RKL and getting stupid high clocks for games. Such as 5.5 on a 11900K, 5.6 on a 12900KS, its great.
What is TVBOC?
@@stevewatson6839 thermal velocity boost overclocking
@@Yahfz Thanks!
how do you stress test TVB?
@@ActuallyHardcoreOverclocking OCCT with advanced thread settings to cycle specific cores and games. Basically you need a workload that will make the cpu change freq quickly to see if your offsets are properly set, once they're in place properly you're golden. I'll very easily get 6ghz+ on games with RPL that should arrive soon
My 7950x boosts 7.0ghz on 10 cores and i dont know how or why. I havent tried overclocking it yet, so i dont know what kind of sustained clocks it can hold. Im shocked.
Under what kind of load?
You mean 7mhz?
Really? This is most probably being misreported.
@@luis4ever887 timespy
@@makingstatements812 do you see such clocks on some monitoring tool like HWINFO64, too?
BZ looking to kill another high end CPU..
I think you lost the silicon lottery on the i9.
Ah, I see you're more of the strong silent type 🤣
So it sucks because it can't do 6.0ghz across all cores. Wow. . Wow. ww..
My x299/7980xe system is overclocked by core/vid individually!!! each core has his own multiplier and his own vid etc...
your main problem is loading @ P95 load "specially small ffts like come on man" to try to verify an overclock and thats a nono for what?? gaming load is the lightest of loads out there to begin with... realbench would suffix for that task and it will flip around cores and threads...
But here comes the user with but but i need to be solid because im running cientific data sets.. then leave the crap on stock...
for a heavy realistic load after realbench runarounds do a blender loop like the benchmark one with all the renders tests is quite good for max wattage load at current settings without hammering the cpu with useless instructions just to heat the cpu like p95 does..
also you need to TWEAK windows scheduler "power settings" so it pick the right threads and such
6ghz? you mean 13900KS ?
well ya know core 6 is a beast lmao run at 58 NO i do 6 huh ok you do 6 then
Are you having fun with these, Dinosaurus chip '-')?
Well he cant use an amd cpu if he wants to go past 5.1-5.2 all core (finally they hit 5) same goes for overclocking ram past 6000
Seems like these dinosaur chips are all over the 3dmark hall of fame top records 🤣🤣
Ikr amd cpu architecture sucks and cant get up there haha thank God for terrible lizards am I right? 🤣
Dear lord, rather than do this tedious crap... better served by perfecting your RAM timings.
You need a real water cooling system and your CPU voltage is too high.
You're telling me this does not provide measurable difference in gaming ?
Cmon man...
If there is no good stress test for you, learn programming for 8 months, write your own program.
that cinebench score is REALLY low, im getting 39090-39200 score
well the E-cores are disabled so of course it would be lower than yours.
I dOn’T wAnT a SpAce hEaTeR iN mY rOoM
DeAd PlAtFoRm
wAiTiNg foR X3d
*copes*