Відео

@@gamehopper111 No, tREFI is derived from x9

@@dainluke ty but i cant find TREFI on my bios.

@ That’s odd. It should be called DRAM Refresh Interval.

@@Decki777 I’d say so

@@godnamedtay Ok

@@Xantiem It’s usually exactly 1:1 with BUFF/CSfloat pricing.

@@ivolong It goes both ways, yes. From my perspective, I guess I just didn’t think it was that serious because I only needed like $60 of adds from him.

i really don't see how this has anything to do with the developer. The pricing data IS accurate, the developer is taking the prices directly from priceempire's API. There is nothing they can do from there - the extension user chose that pricing method themselves.

@@tommy5965 Many users select BUFF.163 for their pricing info. There isn’t really any good reason that the extension should be accepting outdated data (if it is). BUFF no longer allowing the sale of those items isn’t something at the forefront of anyone’s mind and it’s very easy for anyone trading for those sorts of items to be mislead. If the extension isn’t going to display accurate pricing data then there’s no point for it to exist. The prices seem to be pretty outdated, so the developer has had quite a long time to rectify it. Edit: I just checked Pricempire, and the pricing for all of the Stockholm 2021 capsule types displays as exactly $0.00 for BUFF.163. The capsules shouldn’t have displayed any sort of value if the extension was updated correctly.

@@doingitsidesways Edge is so good 😭

?? it's chromium based like just about every other browser

Edge is actually pretty good now. The only reason you wouldn't use it is for privacy, then again if you aren't either using pure chromium or firefox, you may as well use edge at this point

@@blackspider4 If you aren't using Firefox you are doing something wrong

edge is so much cleaner and better looking; and its based on chromium so its no difference

@@buffbacon9663 no

@@TheMongolHordes I get that, but that’s not what the feature claims to do haha. I’m just not sure why they don’t give us a proper description of how it functions.

@@YGadgETech Unfortunately 0x12B does force it. It shouldn’t matter much for gaming and heavier apps though. It forces the CPU to downclock during light loads.

@ but it seems only for ASUS it’s happening for Msi and gigabyte you can still tweak the setting and disable it for what I see

@ I don’t think there’s any real benefit in disabling it for practical use. If you need to keep it disabled, the older BIOS should still be fine enough to use.

@ but can I roll back with new Intel ME version ?

@ which mobo vendor do you have btw?

@@Mikey-lx9xs I can’t record BIOS as I don’t have a dual PC setup with a capture card, unfortunately. The BIOSs are all so different that there isn’t really much that my ASRock BIOS will convey that’ll be of any help to ASUS, EVGA, Gigabyte and MSI users.

I have a 14900ks and been running it just like your steps.. best case scenario, 0.01 LL, 6ghz all core, ht off, only 8 e-cores. Getting atound 28k cinebench score. I have had a offset applied but soon went with no offset. There seems to be more instability. LL undervolt works better for me too. Direct die cooled

@@louishoffmann4459 Funnily enough, Best Case and AC LL 0.01 are the same setting. I believe the AC LL values overrides the SVID profile.

Oh right, I wouldn't know but just want to say this method has been working very vrey well.

@@jameyt1 I don’t believe such a thing exists. The rules that have been proposed online are not followed by certain JEDEC regulations and my BIOS at auto breaks said rules when loading default settings. I personally just have a set target depending on the IC type and scale the tRFC that way. For my H16A I use 120ns for tRFC2 and 100ns for tRFCpb. For old H16M I used to stick to 100ns for tRFC2 and 90ns for tRFCpb. For H24M, I’d probably use 160ns for tRFC2 and 120ns for tRFCpb.

@@dainluke appreciate it

My board actually follows the rule (at auto) trfc value divided by 1.23 to get the trfcpb value. So i am following that rule on my board since then and it works aight.

@@Bazo712 There’s a JEDEC table we now use.

@@sakamata7 100.24 would balance it, but it stopped working on 7.02. Veii also recommends against it.

@@frankmjr6571 With MSI it’s multiplied by 100 when you go to advanced. You can also adjust the mode (downwards) to decrease it.

@@penguinbelly 🦿

@@AiWiKing Which board?

@@dainluke Z790I Lightning WiFi so same as you and 14900kf, stick are PVV532G720C34K

@@dainluke same as you

@@AiWiKing Wait, that’s odd. What RAM kit do you own?

@@dainluke youtube blocked my other comment cuz it put the ram sticks serial number

@@jokanevad So funnily enough, one more cold boot broke it. I think this board just has a very tough time consistently training 8000+ with A-die. That’s me jumping to M-die.

@@dainluke crap, sorry to hear that. I’ve read the whole of OCN thread on the lightning and it seems this board vibes with m die. Veil on OCN helped me a lot with my viper m die kit but it seemed the vipers pcb required very specific/diffrent ODTs than other vendor kits. So returned the kit and kept going with a die. Actually managed to boot 8600 a die so i think with enough tinkering i should be able to stabilize 8400. 8000 was a breeze to stabilize, above that is more work.

@@jokanevad Just beware that subsequent boots might fail even if you get seemingly stable settings. The config was actually running incredibly well given it trained properly twice with a 9 hour window in between.

@@88prince I’d say so. I’d highly recommend trying to reduce MSI Lite Load’s mode as close to 1 as possible. Lite load controls AC_LL. You might need to disable CEP to prevent performance reduction.

@@dainluke Thanks for the advice, I’ve been looking at other comments you’ve made. I went down to Mode 5 and disabled IA/GT CEP which significantly lowered temps and voltages for both idle and stress in cinebench 2024 test (which even had a slightly higher score by 50pts). Idle is floating above 1.3V which i assume is safe and will no longer cause degradation. 1.4-1.5V is a difference. I just hope my cpu hasn’t degraded already. Also I tried mode 1 but my PC just crashed upon running a stress test and idle volts still were close 1.3 V most of the time. I’d rather keep it at 5 as the difference seems small. Also for anyone that may be interested, my PL1 and PL2 are on auto (253W) which I believe are intel default settings and the microcode is x012B of course.

@@88prince Mode 5 is an awesome result

Lol what degradation mechanism causes in silicon oxidation under load? Anyone claiming that there is a "ring bus flaw" is just parroting clueless MLID comments, someone who is notorious for having a poor technical understanding.

@@JuanKolesky Hi, those settings from that “basic demonstration” video were actually just for demonstration purposes. DDR5 settings unfortunately cannot be copied because of how much electrical variance exists between different combinations of hardware. To my knowledge, however, you should have 0 issue running XMP with the latest BIOSs. If for some reason XMP is acting up, definitely switch the sticks around and see if they begin behaving better. Otherwise, if you want to manually OC, I reckon 7400 should be fairly achievable. Feel free to join the Discord as there are many great RAM OCers in there that’d be willing to help guide you.

@@dainluke Thank you so much buddy I really appreciate it allot

@@Munky332 My friend is as well. 0x12B is actually very important.

@@dainluke I just got my 14700k replaced about two weeks too. I'm seeing weird fluctuations even after 0x125 and 0x129. MSI really dropping the ball on this right now. I haven't heard of any other z690 boards not getting the 0x12b update. I'm almost tempted to put this board in the trash and go get another one and upgrade to DDR5, but at that point I kind of just want to switch to AMD...

@@Munky332 For the time being, try disabling TVB. I’m not sure if it’ll work, but if it does it would be a lot better for your chip health.

@@TheGarnerjustin73 Disable TVB if you can, and just tune your AC_LL.

@@dainluke Installed the Ox12B microcode last night. Got a Cinebench R23 score of 37800 with Intel Default Settings. Noticed Single core throttle despite good temps. Turned off Intel Boost Tech, fixed that. Definitely needs a tad bit more tuning. I'll see about the TVB as well. But all in all not terrible on stock settings.

@@FPSDuLe It’s actually a default Windows 11 wallpaper. I use this theme because it cycles a few scenic pictures every 10 or 15 minutes which helps my OLED.

@@dainluke Ahaaa 🤦‍♂️🤣 I dont use W11 so i didn't know 🤦‍♂️

@@cemsengul16 I’ll have to check. I assume you’re also on a Z790I Lightning?

@@dainluke No. My motherboard is Z790 Dark Hero motherboard.

@@ArkenShromAbuser 1.1 mOhm AC_LL will do this. Which board are you using?

@dainluke I have a Asus Prime Z790-A have loaded defult like they recommends only Apply XMP . I can se my IA AC/DC is 0.400 / 1.100 mOhm in hw monitor . And the Svid behavior is set to intel failsafe as deafult in bios wtf ?!

@@ArkenShromAbuser If you head to VRM settings, you’re likely going to see that the option to sync is enabled. You should be fine to head to BIOS and switch to Typical Scenario SVID. Just be sure to disable CEP, and it should run fine.

@@dainluke So i change the vrm too sync and change the svid too typical . And disable CEP ? 🙂

@@ArkenShromAbuser I would keep the Intel power limits and current limits in place, change SVID to Typical Scenario, disable CEP, and I’d probably disable the synchronise load line option. If you disable synchronise, and you leave LLC on auto, you’ll have the old style Asus settings. Those settings were never an issue for the i7. Just let me know what you see.

@@JJFX- What does the term “failure prone” mean to you?

@@dainluke Prone to failure. Like the 13/14th gen chips getting replaced in data centers every other day.

@@JJFX- Would you regard data centres as max utilisation use cases or low utilisation use cases?

@@dainluke Would you defend Intel without understanding the issue you're arguing against? The answer is YES. Imagine not understanding a topic and vapidly arguing against it without even offering a counter-argument. You're frying your ICC/Ringbus with 2 p-core boosting and I'm comfy knowing that 1) you didn't even know about the issue and 2) you're so butthurt about being told about the issue that you deny it is an issue without having any knowledge of the issue You're so silly.

@@jabronilifestyle When have I ever contested you on whether the ringbus is more sensitive than the IA cores themselves? When have I ever contested you on that sufficient voltages under sufficient conditions will cause degradation? The reason I am not concerned about 2-core boosts or the ringbus is because so long as the core/cache rail is kept in check, most decent chip samples will be just fine. Aside from my own CPU samples, I’ve tested others too. Two cases where individuals came to me with faulty CPUs, their BIOSs were old and they’d been using uncapped limits for months on end. It could be that high voltage from 2-core boosts degraded their ring silicon, or it could be that they’d been running at 300W with 1.25-1.3V at load. The 5.7P all-core target was always too high and if you look at the Intel spec page, they never intended the 14900K to be a 5.7P part. The same thing goes for the 13900K - it was never meant to target 5.5P at load.

@@volodumurkalunyak4651 AC_LL 1.1 mOhm was not the default setting until they began using it as the default setting in April because of CPUs not being able to run AC_LL at values of say 0.5, which were the standard on Z690/Z790 boards up until Intel began telling vendors they need to use 1.1. After AC_LL 1.1 mOhm was enforced, Intel realised that 14th Gen chips were probably going to be screwed because board vendors were targeting 5.7P, and they enforced necessary voltage caps on the new microcode. AC_LL 1.1 mOhm could not have been the cause of degradation initially, because literally no one was using 1.1 mOhm AC_LL until the degradation issue came to light. Flash any ASUS BIOS from 2022-2023 and tell me what the “Trained” SVID selects as an AC_LL. It’s definitely not Fail Safe.

@@dainluke AC_LL is only half of the picture. VRM load line is the outher half. If VRM load line is significantly higher then AC_LL then CPU wont be stable from not having enought voltage at high load. If BOTH of those are very low and match, then CEP (clock streching) is required for stability. If AC_LL is too high then CPU will aks for too much voltage.

@@dainluke of course initial AC_LL wasn't the dafault. Moutherboard manufacturers were silently undervolting the CPU's. Then CPU got unstable so Intel forced them into way worse: permanently degrading the CPU's. Having CEP on, very low VRM load line and very low (matching to VRM load line) AC_LL - Intel still refuses to recommend. They STILL want their AC_LL 1.1mOhm fantasyland.

@@volodumurkalunyak4651 CEP wasn’t a factor on older BIOSs, which is why app errors and crashes would occur prior to any clock stretching. Every BIOS prior to March had uncapped limits which also disabled CEP. The chips were crashing because uncapped current/power combined with lack of CEP and shallow AC_LL meant that 14900K/13900K samples targeting their 5.7P/5.5P VID targets would inevitably lose stability at high temperatures running heavy loads. Intel didn’t spec the 13900K for 5.5 and they didn’t spec the 14900K for 5.7. Don’t take my word for it though - have a look at the product page. Falkentyne on OCN also brought this to our attention many times, including December of 2022 when Ichirou degraded ten of his 13900K test samples. It was then that we learnt that the 5.5P 13900K VID point, and consequently the 5.7P 14900K VID point, are actually not at all safe for daily use especially when the current and power limits are maxed. All vendors used reasonable LLC settings. ASUS auto used to be mode 3 at stock, and Gigabyte and other vendors used similar conservative automatic VRM LLC values. On older BIOSs, AC_LL was always shallow. Again, don’t take my word for it. In BZ’s video from a few months ago, he investigated Gigabyte AC_LL. Hardware Unboxed initially reported that some boards were not capable of sustaining their 14900K sample’s 5.7P target, and that was because some of their test boards had AC_LLs that were far too low.

@@volodumurkalunyak4651 I agree with you that 1.1 mOhm is a problem. Intel tried to fix the initial crashing by forcing vendors to use 1.1 (with capped power limits), but then began to realise that their 14th Gen samples - and some 13th Gen samples in some scenarios - would draw too much voltage and then began addressing that with future patches. Everyone should definitely update to 0x12B, especially on 14th Gen and especially if they are not planning to manually undervolt. That’s something I’m not arguing against at all.

@@XenonG They are offering refunds as well, even to the biggest dickheads imaginable, so I’m sure that anyone who communicates with respect will be given a refund. A large batch of 14th Gen silicon and the later batches of 13th Gen silicon, according to Intel, exhibits a stronger tolerance to voltage. The early release BIOSs of ASUS boards and those of other vendors used shallow AC_LL values which is why 14900K chips often crashed under load (insufficient load voltage). Although some chips probably do have issues degrading due to idle voltages under the right circumstances, every case of electromigration that I’ve come across stems from the individual using a load voltage that was too high alongside a high current draw load. An individual in the comments was kind enough to admit that he’d been running Blender and UE5 renders on his chip without making any adjustments to the then-stock BIOS settings (with MCE enabled).

​​​@@dainlukeno. Shallow AC_LL is fine. It just needs matching VRM load line (like level "turbo") and should get matching DC_LL (the same value) and CEP on. People shouldn't adjust everything to make Intel 13gen or 14gen CPU stop killing themselves. It is an Intel fault that those unsafe defaults are the defaults.

@@dainluke >without making any adjustments to the then-stock BIOS settings (with MCE enabled). I stated the exact opposite and you're just a liar - I've caught you lying about me multiple times now. That's so fucking cringe and pathetic. You make videos lying to people then when people correct you, you make comments calling them liars. That's manchild shit and you're a proven liar JUST to white knight for Intel. That's so pathetic it's hard to fathom.

@@volodumurkalunyak4651 The reason I’m mentioning the shallow stock AC_LLs is because that condition counters the current theory that suggests that high idle/low load voltages are solely responsible for degrading Intel chips. I’m not saying that shallow AC_LL isn’t fine and I’m not even referring to the DC_LL/LLC thing regarding accurate sensors - I know. What I’m trying to say is that the fact that there were very shallow AC_LLs makes me question the current rhetoric of how people believe 13th Gen chips degraded.

@@Takeito_ Head to the section in “Advanced CPU” labelled MSI Lite Load. Select “Advanced” and check what the AC_LL is. You’re likely going to see 110 (1.1 mOhm). You’re going to want to either change the Lite Load Mode to reduce that value, or simply type in a less value of maybe 50 or 60 (as opposed to 110).

I just went into that and CPU lite load control was in "normal" when I changed it to advance it indeed showed me 110 but there are two load lines "AC" and "DC" should I put 60 on both and could you explain again what will that do in terms of volts/temps/performance? Shouldn't I try with a less aggressive decreased value and monitor?

I have 23 different modes in lite load, 16 is chosen by default (auto) maybe I can try and choose 14 or 15 and start monitoring to see how it does in terms of volts/temps?

@@Takeito_ I’m going to give you one response that’s more concise. Setting DC_LL without matching it to the corresponding motherboard VRM LLC setting (you need to know the VRM values off hand) will compromise accuracy of the package power sensor. You don’t have to do manual AC_LL values, but I’d definitely drop down to Mode 10 at the bear minimum. I think you’re going to be fine as low as Mode 5. Mode 10 equates to 0.6 mOhm of AC_LL. 0.6 is still a substantial amount of voltage. Average quality chips operate totally fine in the 0.2-0.4 mOhm range, but I’d say go ahead and testing Mode 10 and reduce from there to see how low you can get your voltages.

Thanks! I will change to mode 10 and test with hWinfo software monitoring.

@@andytroo fair point

@@dyson9422 While using TVB, did you notice higher-than-usual voltages and higher clocks due to the chip attempting to boost further? The 13900K does experience two-cores boosts, but they don’t seem to operate during all-core workloads. The 13900K has an all-core target of 5.5. In your testing did you ever observe it attempting higher due to TVB? I’ve tested around 25 Intel 10nm chips. 2X13900Ks 1X13700K 5X12900K/KFs were my own, and the other numerous chips belonged to friends and other individuals whose PCs I was able to conduct testing on. I’m just mentioning this so you know it’s not just a single chip I’ve based my deductions on.

could also simply caused by thermal buffer eaten faster with a very high short power limit, by running at lower power and thus thermal resistance it actually perform better in the longer run

@@dyson9422 higher scores with TVB disabled - that is very simple. Disabling TVB is actually botched. It should disallow highest frequencies at any temperature not at t>70°C, but instead it allows highest frequencies at all temperatures.

@@volodumurkalunyak4651 TVB’s options don’t reflect on every board, but on ASUS TVB has modes that allow extending turbo at the same voltage or targeting an all-core (boost until target). As far as I have observed and understand, TVB does not affect the all-core of the CPU at stock. It just causes that 2-core max boost thing that we all see.

@@dainluke > It just causes that 2-core max boost thing that we all see You mean the 2 core boost to p-cores that the Intel designers warned against going back to 12th gen because it was too much for the ICC and Ringbus to handle? That one? Fucking amateur bro. I called you a midwit before. I take that back. You can't even touch triple digits, 100%.