ABSOLUTELY EVERYTHING ABOUT LOAD LINES ON LGA1700 // why intel needed to add a 1.55V VID limit
HTML-код
- Опубликовано: 2 окт 2024
- My Patreon: / buildzoid
Teespring: teespring.com/...
Bandcamp: machineforscre...
My game: buildzoid.itch...
Twitch: / buildzoid
Twitter: / buildzoid1
Facebook: / actuallyhardcoreovercl...
#intel #overclocking #lga1700 #14900k
Ah, the short version
Failed to mention temp/resistance v current but it's there
😁🔥
Yeah definitely interesting but definitely too long...
@@CappellaTheCat Why its relaxing
Me: in the BIOS, configuring individual fan curves for my 10 case fans for an hour
My 14900K: **sitting there at 1.6V** "...any moment now!"
Since it is all about transient load, let's update the microcode so the CPU always crunches AVX512 in idle state.
this is type of the content that makes youtube worth to watch .
Got that right
More captivating than the Borderlands movie
Almost two hours, FEATURE LENGTH!
director's commentary version
The uncut it 5+ hours lol
@@AndroidBeacshire I only got half way through the first one, came to finish it in case there was more after the chat interaction and found this instead.
@@Raxiel497People were rotating in and out and asking different questions, so it basically became a few practice runs of this video
Me: The Prime 95 is not installed on this computer, it cannot hurt you
13900K: **slowly Thanos-snapping itself out of existence**
A couple of questions that I might have missed the answers to:
1) When you see the VID voltage for each core, (for example in HWinfo) is that before loadline compensation, or is that the actual voltage requested. For example with the 14900KS which might be prepared for 400A, assuming 1 milliohm LL, the max core VID we would expect to see would be 1550 - 400 = 1.15 V with all cores loaded? If you change the LLC to level-6 (Asus 0.49mOhms?) and set AC&DC to match would we then get higher fully loaded core voltages of 1550 - 196 = 1.35V? Or is the VID shown the actual voltage requested from the motherboard after compensation, so 1.550V in both cases? Does the flatter load-line allow higher boost states within the core whilst obeying the 1550mV request limit?
2) Does the VR Voltage limit override the intel default limit of 1550 if you increase it (to say 1650)?
3) I have an Asus W680 motherboard (for the ECC RAM) this does not have adaptive mode for SVID voltage, only auto and manual, I can only offset the VRM output voltage. How would you optimise this board? I was thinking to set LLC level 6, set AC and DC load-lines to 0.49, set the VRM output voltage offset to -100mV, disable CEP, and (maybe) increase the VR Voltage limit to 1650 (to allow the higher VID requests because we then deliver 100mV less than this)? Is this safe? Any idea what the max overshoot/undershoot voltages would be on a motherboard like the W680?
at the time scales we're talking about you can almost think of charge pumps. each cycle a bit of energy is pushed into the capacitors increasing its voltage, power pulled out decreases the voltage. the controllers job is to pump the correct amount of energy, as each power stage is roughly a constant current source, so needs to be on a certain amount of time to balance everything.
"Inherently unsafe power delivery."
That made my morning.
Bro literally taught me more about how power delivery works on modern-day ICs (from Intel, anyway) in just one video than my Electrical Engineering college professor did in an entire semester.
Edit: Tbf, I am a CSE undergrad, but we did study Electrical Engineering in our course for 2 semesters.
Honestly, this is so in the weeds that nobody could teach this except Intel themselves talking to board vendors, and apparently they didn't even do that successfully.
(As an aside, it's not bad to get a broad view, so you can actually survive job interviews at more than one company. A lot of bad schools just treat themselves as job training for a single company, so the students are stuck at the mercy of that company's openings for low pay.)
@@AySz88 yeah I don't trust the curriculum of my college to teach me much of anything actually relevant in the industry today anyway. It's crazy how 99℅ of colleges (in my country, anyway) barely teaches you anything relevant about the industry (even the ones that do, don't do so through the means of their curriculum but rather due to a good peer group that is made somewhat possible by the obscenely difficult entrance examinations) but still charges absurd amounts for it and we're just expected to pay it up because almost any and every relevant companies expect you to have an engineering college degree or your application won't even get accepted (or rather, most of the time you can't even fill up the application without mentioning your engineering degree in detail). The Education industry is such a scam, man.
@@AySz88 IDK, it seemed like an application of the basics to a very specific scenario with specific descriptions of the components and their roles in it. The knowledge of the specific components, their behaviors and layouts is the part you don't find anywhere else.
Yesterdays live was AWESOME. This one definitely clarifies MORE. Thank you for putting out these awesome informative and educational videos.
ALSO I can confirm that High Performance mode (On my machine) doesn't cause the system to run at full speed however ULTIMATE PERFORMANCE mode absolutley does run the system at full boost FOREVER and it's the most annoying thing in the world. ESPECIALLY when windows randomly decises to put the power plan into Ultimate Performance because it thinks it's smart. I strictly use Balanced mode and it works great.
C-states (core parking) and TVB voltage optimizations were “saving” them from their loadline issues. This is what happens when you rely on the “safeties” and “workarounds”, then actually build the tech on top of it.
But only if the Cores are really set to sleep. I guess having only stopped (C1 or something) but active Cores might be bad if they are considered in the prediction. But I don't know if this is a thing.
@@bauerns5er not C1, has to be C1E or deeper states (C3, C6, etc.). Those are not factored into the SVID request at that point. But MB vendors had C1E disabled by default for years; latest updates reenabled them by default.
But still dumb for Intel to rely on a sleep state for voltage regulation for several reasons.
Want to watch a movie? Nah, let's watch a BZ video on load lines
Extremely grateful for your videos!
Thank you for making this. I've been following your videos on this subject, including the live stream version of this yesterday. I thought I mostly did understand load lines but I think this one finally fills in those last gaps with the distinction between the voltage drop due to the impedance of the components, and Vdroop. I'm sure some will complain about the length but I don't see how it could have been much shorter. There wasn't any waffle or non-sequiturs, It just takes this long to explain in terms that someone who doesn't already have more than a basic understanding of the relationship between potential, current, and resistance. I've seen a lot of what was covered here before, but only with some of the more foundational details given in context did it finally click.
And the rubber meets the road 1:20:00
That moment when realization hits how crazy the simple Vdroop compensation hits... Good explanation!
The only thing I want to know now... is what settings I should use for each of the LLC settings o.o for the AC and DC loadline...
Wish there was a handy table somewhere.
I was told to use LLC 4 with 0.98 iirc... but the source was a random youtube comment. (It worked... but cant verify....)
BZ covers this somewhere but AC/DC load line shoudl basically the same because one is for the CPU and the other is for the VRM, you tell different humbers for the same thing. Not sure why you want this.
If yo ucan type in mohms than it's pretty easy. I ~110 should be for a standard motherbaord so I did 55 since I have a hgiher end boards less resistane between vvrm and CPU. Works fine. However the key for me was to set the voltage cap in CPU powermangge of my Asus MB to 1300 or 1.3 voltes and I set maxium current at 375. It only boosts to 5.2 and it only uses 180 watts to hit 33000 in R23 but it cool, it's low voltage and I'm happy.
@@allanwilmath8226 You want these settings right to get CEP working correctly.
According to intel spec AC/DC loadline should be the same. CEP will then reduce overshoot and understood, but will do so by clock stretching. If you undervolt with the wrong settings, it really eats into performance.
Got mine at LLC 4 with 0.98/0.98 and a -0.10 undervolt. CEP is on and working properly. I get 35k with a CBr23 on my 14700k.
I capped voltage at 1.5V, but managed an overclock of 5.9GHz altho the undervolt is reduced to -0.02V with the VF curve.
The problem is... bios doesnt stick to intel specs, using 0.4/0.9 at stock... causing it to crash cinebench r15.
So its really something users should adjust but have no good knowledge of...
Unfortunately the AC/DC values that match the various LLC levels differ by vendor and sometimes by board.
For example my MSI Z790 Carbon Wifi has a 1.10 mOhm impedance for LLC Auto (which equals level 8 there, the level with the highest amount of Vdroop) and 0.36 mOhm for LLC 5, but according to an overclock. net thread, for Z690 there are different values used (0.8 mOhm).
So you either have to be lucky find the info for your board, or measure the values yourself. Or you have an Asus board, where there's an option to automatically link the AC/DC setting to the selected LLC setting.
Buildzoid is the GOAT
We will understand Load Lines, after this... right? ;)
I sure hope so.
@@ActuallyHardcoreOverclockingstill no Comprendo, can we get another feature length film? 😅
Long play director cut version
@@ActuallyHardcoreOverclocking I'm just finishing it - and it was all understandable. There were like 2-3 places where you misspoke, but nothing big. You really did a very good job.
@@Mglunafh Yeah - and it's fun ;)
Thanks, Buildzoid, for all the time and effort you’ve put into investigating the Raptor Lake degradation issue. Your deep dives, even when they’re 2 hours long, are packed with so much valuable knowledge that’s helped us understand the problem better. We really appreciate your dedication and all the hard work you do to keep the community informed. Keep it up!
Watched this whole video hoping to understand better but it was so long I feel like I was forgetting half of what was said by the end. I've been running my basic undervolt on my 13900ks for 20 months now without issue. I just set a negative offset voltage for the vcore and svid to -0.160v for both and locked the cores to 5.6 GHz max. My vcore will stay between 0.768v to 1.280v with the svid showing between 0.777v to 1.285v. No crashes, maxes out around the high 60 to low 70c range under full load while running R23 and using a max of 235w scoring 40,500 points. I've been extremely happy with it but now I'm wondering if I need to tinker with load lines instead of using the auto setting. BZ making me question reality lol
Thank you so Much for explaining this. very informative
Keep up the good work!
🙂
My 9600k requires 1.7. Gotta pump those numbers up !
the cooler you keep it the less it requires. get the best cooler you can, but also make sure it is compatible with whichever platform you are considering next.
my xeon x5470 needed 1,45v for 4.5ghz whet it approached 60.
kept strictly under 50c, it was fine with 1.38v
"you still can't run more than 4.8ghz, so why even bother"
Brilliant😂
Note to self. Stop buying MSI motherboards.
eh I wouldn't go that far they make some good boards. Also some bad boards just like any other manufacturer. They just never have a proper Vcore reading from the Super IO chip.
@@ActuallyHardcoreOverclocking This is one thing that I missed in the video. What is the VR VOUT sensor in HWiNFO actually reporting then? Is it the "copied" value that you explained? If so, MSI also does have that VR VOUT sensor, but it's not the same as for other boards?
Thank you for this video. This video is literal gold, I feel like I need to watch it 2 more times to fully soak in all this information. So glad this dropped, I've been messing with my new z790 Dark Hero and the bios and power delivery options are more intense than I've ever dealt with. I know Intel sucks for all this, but it's been a heck of a learning experience.
Also I'd not entirely take away blame from motherboard makers - I very much doubt that most motherboards are bad enough to need the top of the spec for resistance, 1.1 ohms. It doesn't look like anyone is actually measuring resistance to set the LLC to the actual physical resistance (which Intel says you're supposed to do and nobody actually is doing). Maybe I'm wrong but it seems they're either sitting LLC too low so we get undervolting or too high so we get overvolting. So I'd blame Intel for a bad design, and motherboard makers for compounding this with further bad default configuration settings. Sure we'd have voltage spikes even if motherboard makers set LCC conservatively, but many made Intel's problem even worse than it is by blindly following the spec instead of doing resistance measurements and tuning.
well with how fast these chips went from design to release, I believe I read ~10 months, I'd assume manufacturers of motherboards didn't really get a whole lot of time to test these things.
it literally doesn't matter what the impedance of the board is. 1.1 ACLL will wreck chips on 4 layer crap boards just as hard as it does on boards like the Apex. The whole point of this video was to show how at steady state the voltage regulator literally doesn't care at all about the board impedance and if the CPU isn't doing anything it's at steady state.
@ActuallyHardcoreOverclocking I understand that at steady state a 1.1 LLC will always involve a high voltage because as you pointed out the VRM measures the voltage on the silicon itself . The point is that I highly doubt that most boards actually need it set as high as 1.1. And if they were measured and set accordingly we'd probably have fewer issues.
Great explanations.Quick question for a GB Z790 Master X mobo:
I've set LLC to Turbo and AC/DC 28 both + an adaptive offset -110mV.
This way VID matches Vcore/VR VOUT and I have low voltages and good temps.
Are these settings safe or do they encounter high under/overshoot ?
Should be safe.
Has anyone been able to definitively tell what AsRock LLC values are per setting?
I can also tell you my PG L D4 will not have settings 1 and 2 available for 13700k. Only 3, 3.5, 4 and 5 with auto being 3
With LLC at lv4 and a DC LL of .53 I can get my average VID to be within ~1mV of Vcore, does that mean anything useful?
28:25 Can anyone cite any material that explains this clock stretching mechanism as a safe protection in case of a voltage drop?
Hey man, after everything that you've said in this video. Do you have any theory on why 12th is seemingly not affected by this even though it's on the same socket and very similar design?
I only know they work just fine no issues on the Encore ;)
My brain hurts...
So I didn't open a browser and come here just to like the video for nothing because it's actually at 420 likes. lol
Can you give us a bios settings recommendation for the 14900KS 0x129 microcode, to get the most performance out of the CPU without causing degradation?! At least in the short term
Thanks for the rambling videos... 😇
Regards from Portugal 😎
When you stop to think that we are talking of hundreds of amps crossing a thin layer of silicon with tiny wires of microns in size. How in the world doesn't the electron migration make this a fancy heater? 😅
Why did you do this to yourself, BZ? I hope it was wor...actually nevermind.
The VRM knows what the Voltage is, because it knows what the voltage isnt...
Damn RUclips not recommending my subscriptions...
Maybe I'd like some overclocking jargon after listening to the greatest technician that's ever lived.
It's interesting that due to thermal throttling or current/power limits, the CPU can handle a bigger under-volt at high current, than it can at low current.
For example my 14900ks can handle -0.140V @ 400A with all p-cores being used, because cores don't boost above 5.8GHz
However when running both hyperthreads in a single p-core boosting above 6GHz the CPU can only manage -0.050V
If we assume 50A per p-core, then we get the following simultaneous equations:
140mV = 400A * R + V
50mV = 50A * R + V
Solving we get
R=0.26 mOhm
V=37 mV
If I set LLC=6 on my ASUS 680W motherboard I think that's 0.49 mOhm (validated by checking power reading at limit).
I can't set a voltage offset of exactly -37mV so I can go with either:
R=0.24 mOhm V=-40mV
R=0.23 mOhm V=-35mV
So if I set DC_LL = 0.49 (to match LLC6) and AC_LL to 0.49 - 0.24 = 0.25 and voltage offset to -40mV I can optimise the under-volt for both single core and multi-core loads.
Since updating 0x129 microcode on asus z690, hwinfo shows short 800mhz dips/minimums on my 13900k. I was super nervous then noticed BZ's does the same in asus z790 video, and didnt mention it.
Just curious if anyone knows what's going on? I'm familiar with, and have increased polling rates/refresh/frequency, whatever, in the past, watched "minimums" before but never seen this until now.
Not mega concerned since Bz didn't bring it up but still curious!
hey will you be testing the new bios for intel on gigabyte z790 aorus elite ax? please do
Could you please do a shorter version about what we can do on our boards(MSI, Asus, Gigabyte) when using 13th and 14th gen processors.
I have patiently watched all your videos but am still confused about what to do, for simply running my 13600K in an underclocked, undervolted state with low-temps, without any risks whatsoever.
In one video you mention we should not use the AC_LL and in the other you are saying that's the only way to do it .. :(
Basically, it would be good if you could simplify it for the lesser mortals like us, with steps on what to do with Lite Load Control(DC_LL,AC_LL), LLC(Load Line Calibration) and Vcore Offset/Adaptive modes.
So, am I correct in assuming that:
as long as your VID and Vcore are within safe ranges, then there's no such thing as an unsafe load line/ACLL/DCLL etc. ?
Or maybe I'm missing something regarding the transients?
Hi, as a FPS gamer I got some issues setting up AC/DC loadlines for my CPU 14900k. Every time I change AC/DC or RAM frequency I get different mouse response, either slower or faster feel. Everything that has to do with the voltage affects it, not sure if something is clashing in my system or this is normal and everyone is experiencing this.
Lenovo is shipping laptops with AC/DC LL of 1.7mOhms... What do you think about it?
Thanks for this extreme explanation xD And also for proving how lame Intel cpus are LOL
Now i'm curious how all that works on AMD cpus plus their PBO2 and CO especially, so can you as well make a short video about that?!
Something of note. You were talking about dell/hp/lenovo/etc degrading faster than normal because their boards are trash, and you're both right and wrong there. They are so undercooled and underpowered, that during an all-core load they normally run at like ~0.8v because of thermal constraints. Idle windows is way more dangerous to a dell desktop than cinebench or prime95 would be.
dejavu
they made that whole setup so dumb i could understand it
are you going to cover ryzen 9000 overclocking and ram overclocking?
"Life is very hard for CPU power delivery" - Buildzoid
Normies think they are electrical engineers. LMAO.
I want to love you Intel, and justify my 14600K buy... but you guys go out of your way to make it hard to love and justify
this is such a valuable video to me. like the explanation of ddr4 timings and other videos, actually understanding how this all works means far more to me than just being told the 'best settings'. your content is fantastic and important, and i absolutely love learning the details. throwing money.
If you had a perfect vdroop setting that completely matched the worst downward spike in voltage the CPU sees when a heavy Amp load is applied and then went back the height of the worst upward voltage spike when that heavy Amp load was released, wouldn't that just be spamming the worst upward spike volts to the CPU whenever a single core load was applied, compared to a vdroop free setting that was undervolted to the same load voltage that the perfect vdroop setting had?
Since the 13+gen chips seem stable with the brief loading downward spike, would setting the load line settings to droop be little more than wasted power and elevated volts for no reason other than tradition?
And if the load line settings were set to droop more than the ideal value, wouldn't they also make the CPU run at low load at higher volts than the instant you stopped P95? And have to run at higher power in every situation but the worst case scenario, so long as the CPU was given enough volts at that worst case scenario?
I think Intel left the door open to this and motherboard manufacturers drove a big slimy turdmobile through it trying to cheat the no voltage tuning at stock rule to make their brand look better.
do you guys think having 1 or 2 cpu connectors plugged into the motherboard is better?
I wish my engineering lectures could have been this interesting
Amazing video mate 😍
Hey BZ, if they could get rid of all degradation it really would be nice!😅😊
Can you explain what CEP actually does? Not why it triggers, but what it actually does
Jesus Christ
so basically do not buy used 13th 14th i7 i9 lol
You're the best!
Quick question. Is it ACLL that effects VID request as it seems you are suggesting? So VID adds voltage depending on the current draw when it requests voltages from the VRM controller. And does HWINFOs VID reflect this addition?
the Hardware info VID has DCLL subtracted from the SVID request. If you want to see the raw SVID request you need to set DCLL to 0.
Hi friend, your channel is really cool! But the problem is that I'm very new to the subject. I only know how to play, but I've had a problem for a long time, and I'm almost convinced (90%) that the problem is with the motherboard and the ram. Because my RAM makes a strange noise, I don't know if it's the RAM or the motherboard, I just know it's close to the RAM. It sounds like a whine in the coil, this happens when I enter a game where the fps is unlocked and consequently uses more processor, RAM, etc. The coil whine noise is not from the video card. It's like there's a bottleneck in the RAM, I don't know how to explain it, but that's what I feel. and I'm almost sure that it might be possible to fix it by changing the voltages, etc., something must be wrong. and I don't know who to ask other than you. Could I add you on discord or something to help me? I can pay if you can help me, please friend!
Thanks.
I get that RAM whine when I torture the poor things with the Aida64 memory benchmark.
gpu loadline radeon geforce intel arc
Ah so
Crikey
I think I've learned something.
Kudos bz
Frame chaser sweating bullets now 😂😂
Jufes most definitely is!
All day everyday since the 13/14th gen problems started imao
I've posted this yesterday, but for some reason it didn't show.....whatever.
I've wanted to ask about old school overclocking: fixed voltage & frequency. I have an 14700kf and I'm keeping him at a fixed 1.285 V (in BIOS, 1.292 in monitoring software) and 5.6 GHz all P-cores / 4.4 GHz all E-Cores / 4.5 GHz ring. Also using high performance power plan in Windows, so my CPU is always running at these settings. No voltage & frequency fluctuation, just flat 1.285 V @ 5.6 GHz. No problems whatsoever (no freezing, BSOD, crashes.....). I did had to set a 350 Amp current in order to keep that frequency, though. In Cinebench 23, the power consumption goes up to about 340 W, a difficult task for my cooling loop, but it does manage to keep my CPU at @ 95 deg. (room temperature @ 28 deg.) However, that was only for testing. In games, it rarely pass 50 deg., except for the compiling shaders part, when it reaches about 70 deg. I think it's better than having 2 cores asking for more than 1.5 V.......
best movie of the year
Hello ,my question is how much peak spike of voltage can be , i have 13700 non k ,ASUS ROG STRIX Z690-A GAMING WIFI ,package limit 111 watt, c-states off,279 a ,intel profile performance(4.9ghz p core 3.7 e core) llc lvl 6 ac load line 0.210 mohm and dc load line 0.490 .In idle mode , in windows and in bios i have 1.19 v on cpu, and in r23 only 1 volt.ty for video.
I have the i7 13700kf with gigabyte z790 ud ax,after the bios update cpu never goes past 4.9 ghz all cores in gaming,5.3ghz only in light load.Gigabyte messed up the load line,should i wait for the next bios,temps and power consumption are very low
IDK, but too me this LoadLine stuff seems to be a quite dumb solution for the undershoot problem in the first place.
Because while this does avoid the use of unnecessary high voltages (and therefore power consumption) under load,
it does not at all avoid unnecessary high voltages during idle!
And even if these high idle voltages are not high enough yet to kill the CPU, they still lead to increased idle power consumption (think esp. of leakage power, which gets worse and worse with newer processes).
IMHO it would make more sense to set the VRM to the voltage that is actually required to sustain a certain clock frequency, and let the CPU handle the undershoots by some other means.
I mean think about it: They spend the effort of adding the Vcore sense lines to measures at/in the CPU directly, and then artificially re-created the VDroop anyway. In that case you could probably just have measured at the VRM output on the MB in the first place and let the CPU estimate the required Voltage offsets for SVID to compensate for Vdroop.
Only advantage of using Sense Lines + LoadLine seems to be a very precise, predictable and "only as high as needed" voltage when under full load. In all other scenarios the voltage is higher than needed.
Which maybe was not considered to be an issue, when this was invented, but nowadays it should be, just for idle/partial-load efficiency alone.
Ok so, I guess I should lower the ACLL to around 0.5ish to help prevent the insane voltages,
But should I also match the DCLL to the AC as Intel specified ? Or will that just mess with the power readouts if it's not set to 1.1 ?
Watched the whole thing. Lots of good info here. Personally ive had issues with crashes with 12th gen intel laptop cpus compiling UE shaders.
Thank you, that was fantastically cool. Is there a chance to keep the 14700k processor alive on the GB Tachyon or will it still turn into beach sand in a couple of years?
btw., a bios actively learning the load line by applications running, could be an improvement.
Looking at the current consumption of a single pcore of ~50amps, how much sense does it to increase the number of cores at all, reaching a limit of 6 or 8?
What have you done?😢 I was planning to watch the entire 5h live 😢 Please bring it back up... or send the link if it's set to unlisted :(
hello sir, ive been following all your videos, im running a 13700kf on air *thermalright peerless assassin" im having issues with temps on prime95 ofc. I updated the microcode to 129 on my MSI Tomahawk, the default AC DC for it was 1.100/1.100 mOhm, level 12 load lite. Im having issues to make it very stable and cool...
Currently i have set intel defaults, 307A ICCMax, 253w PL1, load line calibration AUTO, AC DC LOAD manually set to 0.600/0.600 mOhm, voltage adaptive + offset -0.070 (My cpu is trash, anything past 0.080 makes it have WHEA errors in HWINFO64 during prime95)
since my air cooling is not that powerful i settled putting pcores down to 51x and ecores to 40x, ring runs steady at 45x.
Theres something im very confused about.. Is it better a high or low AC DC? Default is 1.100/1.100 and lowered to 0.600/0.600 also tried 0.55/0.55. This combined with undervolt (adaptive -0.070) this with core ratio limits at 51 and 40 respectively. Any suggestions? or advice? 1.100 vs 0.600 what should i set it to... i tried doing UV past 0.080v but no, my cpu doesnt like it.
With these settings, prime95 does pull the entire 253w power limits, once reached, pcores go as low as 50 and goes back to 51 when drops below 253w as expected. But no CPU errors, max core voltage reported 1.326v. Max temp 95C
Hey guys. Are we always using balanced power plan in windows for 13700k i7? Is there any advantage using high performance or e.g. Bitsum Highest performance? I am only gaming on my pc. I already did all the stuff (A/C load line settings, newest bios) to adress the degrading issue. With balanced plan all these settings I am getting 30500 cinebench 23 score. And is there an advantage using parkcontrol to unpark cores? Or just enable c-states and let windows and cpu handle it
No help?
I personally prefer and have a better experience using balanced. Just me. I also have no issues using C-states.
44:05
Show us the voltage regulation Nyquist plots, you know you want to :)
so a higher load line setting on Asus, like 6, vs 3, means less voltage to the CPU, but a lot of overshoot?
Who else watched the 5h version first?
Do one for AMD please
After I updated to the latest update 7D91vHD(micro code 0x129) and saw AC/DC LL(1.1/1.1 mOhms) and before update it was (0.4/1.1) on 7D91VHB(micro code 0x123). Now in BIOS CPU light load is set to Mode 16 but on the previous it was set to Mode 9 by default. What should I do to lower power consumption for cooler operation.
MSI went from 0,5 ACLL pre - 0x125 to 1,7 ACLL after 0x125
1:26:37 "like i actually just cannot" 🤣😂🤣
On my MSI z690i board I have compared voltage VROUT and vcore reported to the VID requested voltage and they align when the ACLL and DCLL is set to 0.8 when VRMLL is at Mode 7, so I think that means their VRMLL is 0.8. Which is not so great when the new Intel default sets it to 1.1 as that makes the voltage get rather excessively high.
On my MSI Z790 Carbon Wifi the AC/DC LL actually seems to equal 1.10 mOhm when LLC is set to Auto resp. level 8.
So they seemed to have changed this across generations or even across product lines.
Wow another masterclass from BZ 😶
for the algo
Thanks Buldzoid😀👍❤️
So glad I bought alderlake 😮💨
I was working on finalizing my undervolt and after watching this video decided to try to run P95 Small FFT with only 8 cores and no HT to really push the P-cores and check for instablity. I received the out of memory error and noticed that MSI AB crashed. This does not happen with 24-cores.
Current settings when MSI AB crashed:
Vcore to 1.4v (IA VR LIMIT Bios setting), set 253w/400A limits, use LLC#6 with ACLL@0.49 and DCLL@0.54(calibrated), enable CEP, use V/F curve 6,7,8,10 undervolt -0.13.
I reduced my undervolt in increments of -0.05 and each step MSI AB would still crash within Test 2 of the Small FFT run. When i finally hit -0.11v MSI AB no longer crashes and I am able to get through many more runs of P95(3 cycles of 6 tests each). Does the out of memory error only happen when the chip has degraded or can it also happen with a too severe undervolt? In other words, is the out of memory error simplly an i/o issue with too little Vcore, or does this only happen once the chip has degraded and more voltage allows you to hide it for a while? Prior to this I was R23/R15 stable and have not received any any WHEA errors (weeks), but had not yet tried P95. I am curious if i need to be concerned.
If you've always been undervolting, chances that your CPU has degraded are near zero. I have two systems with MSI boards, one 13900KS and one 14700KF, and settings *very* similar to yours for load lines, except the 13900KS can do -0.19 undervolt as its binned silicon really makes a difference, the 14700KF does -0.12 on the pcores and -0.08 on the ecores. In my experience P95 freezes or WHEA errors are too much undervolt on the ecores, and P95 crashes are too much undervolt on the pcores.
@@thomasg5554 thank you sir!
ASUS boards have an SVID behavior opton. Choose the best VID table to match your silicon. Then start messing with voltages and loadline.
the SVID behavior option literally just changes the AC/DC LLs and shouldn't be used. You should use the sync AC/DC to VRM LL option in the Digi+ menu. Honestly I'm surprised intel hasn't forced ASUS to remove the SVID behaviour setting.
@@ActuallyHardcoreOverclocking haven't had voltage go over 1.25v. I'll give it a try.
So TLDR, is it safe to buy a new i7 or i9 for 4-5 years or dont bother? (yes i have a 7800X3D gaming rig already but i could use a 13/14900K for productivity)
personally I'd say an i7 14700k would be a better buy than an i9, you don't really get a lot extra for the 4 extra threads for the i9, while it can be far more power hungry than the i7, and the voltages it runs at isn't as insane as the i9 voltages. (Using an i7 14700k in my own system, haven't really had any issues with the thing yet.)
Just get 7950x3d for both
3rd
I would think if you set your max core speed to 5.6 or 5.7 on an i9 so it wont boost to those dumb high clock rates for a millisecond doubt cpu would request any high voltage at all.
I tested that idea on a gigabyte board with 0x125(and intel default settings). It still peaks at 1.6V.
@@ActuallyHardcoreOverclocking hmm interesting. Even with TVB on still did it?
If i recall from your one video about gigabyte board setting voltage limit of 1400 seemed to work with old microcode correct?
Thank you good sir