electrosoft

Yeah I watched Jay's video yesterday. Final conclusion? 240mm->420mm with everything else being equal results in a ~2w differential and slightly tighter variance at 253w limit even after normalizing temps. If you have the case space and with ~$28USD separating the top and bottom models, go big. If you have a smaller case, you will still get great performance even down to a 240mm. I'd like to see a retest at 320w though....

I like that you think a 12900KS on a Z690 Hero is a "simple gaming rig" 🙂 Simple would be a 12400 on a B660 D4 board heh. But iGPU's are pretty even between the two and if I had to choose, I would go 12900KS too especially with that pricing. I'm tempted to pick up a few and bin them at that price lol. When you get the 12900KS, slap it on the encore and see where the IMC tops out. Now that I have the Lightning and 8400+ sticks in action, I'll be retesting the the 2x 12900k's I have along with the 12400 just to see where they can go on DDR5. On DDR4, the cool hand luke topped out at 4133 on DDR4, the 12900K I delidded topped out at 4000. The 12400 topped out at 3333 (yes, 3333). This 14900KS hit 4200+ on it. Good test and more importantly, it uses UE5 engine which many games will be using going forward. 4k even dropping the settings down still brings the 4090 to its knees with 69fps native and renders everything else sub 60fps yikes. That's without RT....

Well, they have 15 of them available at that price so hopefully the one they send you is a good bin. And yeah, for that price? You're not going to go much better elsewhere. ~$268 + tax shipped for a 12900KS is pretty nice. Something to keep in mind if you're contemplating an AMD build is the 7600x which trades blows with the 12900K and can be had for a tad cheaper but only has half the threads but has a good iGPU too: Timestamped at overall average but take into consideration if you're going iGPU as a priority. On the other hand, if you want to max OC a 12900k....lol

Well yeah, then you wouldn't be the @Papusan we all know and love! (minus the Crocs of course 🙂 )

There's zero downside to a bug/defect being fixed even if the benefits aren't realized due to personal use cases. But for myself, I can now start dabbling with 8600-8800 along with tighten up at 8400 without the system crashing due to insufficient SA. I couldn't do that before 0x129. Will I run anything like this D2D in the SFF? No. It will now be a tight 8200 vs 8000 for that to keeps temps in check but always better to have it and not need it than to need it and not have it. But I get where you're coming from. 🙂 "Hey look, the new model comes with 8 cup holders!" "But there will be, at most, only two of us in the car" "Maybe we'll have 4 drinks each one day and...." "...."

It's the final vs beta one previously but yeah, that 9000 series TDP is gonna rock my world!😅 Everything updated, fresh, meticulous, clean install. All hardware tested and verified.....let's go! With all the issues and problems out there, nowhere did I even imagine the SA Bug would be one of the blessings of 0x129. Everything is looking good! Now that you've had it for a bit, what is your opinion of the DD EK AIO?

@Raiderman I always say, when I want a new BIOS to drop, all I have to do is install the most recent.... 🙂 https://download.msi.com/bos_exe/mb/7D70v1J.zip Luckily, the system recert was still taking place so I updated to this one and back to testing my dialed in timings on mem and other subsystems before a clean install tonight across the board.

You absolutely extracted every ounce of value out of your P870 in every way possible. It had a good life. 🙂 The new SFF is looking good! For my in progress portable SFF buildout, I'm thinking a 4070 Super is the absolute sweet spot (or a heavily discounted 4070).

I figured with the 9600x/9700x results the 9950x wasn't going to bring much to the table. Refinements without performance advancements is this generation. Now we're talking.... 🤣 --- Rumors are of a dual X3D CCD design for the 9950X3D, but I'm just not seeing that happening. If we get a 9950X3D like the 7950X3D, that is a massive pass. This generation is all about refinements and a new node. This is the ultimate tick using Intel methodology. Hopefully the next tock is meaningful on this new fab. Honestly, they could have just named these 8600x, 8700x, 8850x and save the odd numbering for major uplift/changes as there just isn't much of anything there outside of the new fab and a touch of uplift maybe. I'll need to see some locked clock for clock testing to draw a final conclusion. Crap on the 7900xtx all you want, but it is 30-40% faster than the 6950xt overall. But this? Meh. (Back to tuning the 7950X3D and 14900KS for WoW Xpac comparisons with my 7900xtx and 4090 while I wait for Arrow Lake)

You knew deep in your heart you wanted the 4090 from the jump.... how you convinced yourself a 4080 Super was good enough is beyond me! 🤣

Yeah, I watched this last night and just went, "huh?" Here's who I rely on right now.... The massive army of engineers at Intel (for better or worse) BZ for his oscilloscope readings That's it.

Watching BZ's video about disabled 0x129, and he clearly says it is basically Gigabyte's fault and that after accepting the warning of not using Intel's default settings (it actually gives you a warning on their board), all bets are off and you're back to Gigabyte's settings which includes horrendously set LLCs and unrealistic UV which can cause poor silicon to crash right out of the box. Just piss poor settings from Gigabyte by default even with the newest BIOS. This reminds me of Asrock and their 0x125 BIOS which set at proper defaults and enforced limits was still thermal throttling and shooting to the moon at 363w but more importantly pulling a crap ton of vcore. The newest 0x129 update works properly right out of the box with enforced limits and temps and everything were within range and 6.2 boosting went bye bye to 5.9 max. Back to BZ: He literally says "Now maybe some vendors somehow like have a way of keeping it in place but at least on gigabyte it disappears as soon as you turn off Intel default settings..." - This is what I was alluding to previously. He also notes the Silicon lottery and the fact his CPU can't run R15 with the Intel default settings on Gigabyte but if you have good Silicon it can run it. This is mainly due to Gigabyte's aggressive UV/LLC out of box on 0x129. Another thing is even after disabling Intel Enforced Limits and running multiple tests, he fails to spike to 1.6+ which he was doing pre-0x129 and caps out at 1.5792v which is oddly specific and close to 1.55.....

Ditto. On an AIO and stock lid, after a lot of testing, 5.9 / 4.5 / auto is really the sweet spot and correlates nicely to the V/F point and runs great with Intel enforced limits of 253w/400a 5.9 all core. Temps usually stay nice and low. 6ghz all core is doable, but I'm not super keen on the vcore pull sometimes and temps are good but I know this is going to eventually end up on the 240mm EK Nucleus so I need temps to be great now so they can be great/good in the sff. Next step is move the 240mm in the test case and that also opens up room to slot in the 7900xtx and start testing it and the 4090 back and forth against the 7950x3d in single binned X3D CCD mode. Watched your video. Always a fun watch bro. 🙂 Exactly and quite honestly the MC update is working as intended. Anyone who goes with Intel defaults is fully protected along with the SA Bug being fixed. The fact changing BIOS options removes the protection for me shows Intel still giving MB makers control which is exactly what you want. It is up to Gigabyte (and other MB makers if they haven't already) to leave in place the full 1.55v ceiling in all scenarios unless a user explicitly removes it for whatever reasons (LN2, suicide runs, etc...). Other MBs will have to be tested to see how they handle BIOS alterations and the 1.55v limit in different scenarios.

Will have to be confirmed on other motherboards too, but the problem appears to be it is set by default but Intel still left in place controls for MB settings to disable it for LN2 and other hardcore OC'ers which is the right move. It is up to the MB makers to keep it in place in all situations except when explicitly disabled. This may be the fault of Gigabyte and how it handles settings outside of stock Intel defaults on the 0x129 era.

Thanks for giving it a test! I guess the upside is, cost aside, it performs just as well. The downside is the crazy cost per gram comparatively.

Agreed. I've tried it 3 times actually with two different BGA laptops and it sucks OR I got a bad/fake batch ordered from Amazon here. I had way better results with Nano extreme (the original go to for laptops and poor heatsink pairings). Isn't this the same stuff 4090/7900XTX users are raving over as being the best though (outside of liquid metal)?

After a lot of testing, I'm thinking I might go with the best of both worlds and do an Intel enforced limits auto optimized fixed all core config for D2D. I've been playing WoW, FO76, Fortnite and Starfield on this at 59x all core and even 60x all core with great temps and performance. I went and did some shader compiles in Fortnite and Hogwarts and decompression tests. So in the end, you end up with a variable all core but with auto settings and dialed in LL/VCComp/ACDC all running at 253w/400a. What I like about this config is for most games it will sit at 59x and 60x no problem and game away with load requirements well under 200w (usually 95w-130w). If things get heavy and require more than 253w (which doesn't happen in game yet) it will just downclock to compensate. If I really need the extra clocks (not really in games), I can push to 320w extreme limit or if push comes to shove load up one of my fixed profiles from 56x -> 59x (along with the increased package pull, heat and load vcore). After a lot of game play and futzin' around, while it's cool to see, I see zero reason to boost to 6.2. Like....none. It is fun to see pop up from time to time and flex in CB23/GB6 but in real D2D use? I'm not seeing it as just about every game I play is multithreaded now. On the new 6.02 bios w/ 0x129, I now need 1.30v fixed / 1.184 under load for 56x all core now in CB23 vs 1.28v / 1.152v under load with SA now roaming free and wild..... same adjustment up for the other profiles too. (1.35 for 5.7, 1.40 for 5.8, 1.44 for 5.9) Ugh, that feeling when you think a CPU has died and went to the great Silicon place in the sky.....nothing sinks your heart like 00 or a stuck 15 on MSI boards. Glad everything turned out ok. In those situations, depending on silicon quality, I almost always hope its the MB and not the CPU.

Most excellent! And just like that our "still good deals" SA Bugged chips are now great deals running as intended. Absolutely...... Unfortunately you're correct. I find myself agreeing with him far more often than not......pure torture. 😞

As you said, he's pushing too hard and his problem lies elsewhere. At a certain point, SA/IMC/Memory/MB will give up the ghost. Every piece of a subsystem has its limits fixes or not. Every chip has its limits SA Bug or not. Once you are >=8000, you are fully out of Intel's sanctioned IMC zone and then some anyhow. --- As for Jufus, there is definitely a need for 0x129. Without it, my SA caps at 1.18 or it's crash city everywhere. On a whim, I set my memory to 8600 and booted with SA Auto hitting 1.35 no problem. In every other scenario WITHOUT 0x129, anything greater than 1.18 on SA and memory >=7800 is insta hard lock up SA style requiring a power cycle. 0x129 puts in place 1.55 limit to stop any overshoots or excessive voltage if you fail to cap/tune it yourself or on the off chance a request/transient gets loose and unless explicitly told not to, the buck stops there regardless of everything else as a safety measure. No more; no less. If you were capping your vcore/clocks already, there is usually no problem but 0x129 is that extra layer of protection and it really is no harm in using it. And let's not forget HWInfo and other software routinely misses >1.55v transient spikes (or flat out misreports like OCCT and amps) due to polling speeds as BZ showed repeatedly in his video pre-0x129 that 0x129 is designed to stop at the base level (or you can in theory by using fixed Vcore and/or vrout max capping). I prefer to use both 0x129 and tune the bios for a win:win. Jufus is also stuck in the mentality that 1.40v will kill your CPU which it will not. He is stuck on 1.35v or lower or you're going to degrade your CPU. If that were the case, Intel would have set the cap much lower....unless (tin foil hat on) Intel knows they would have to toss out the bulk of their chips on shelf and RMA them out there all the way back to many 13900k chips with >=1.4v 6ghz/6.2ghz VIDs on top or chop off performance and the CPUs no longer perform as advertised and suffer a massive lawsuit/return landslide. (tin foil hat off) But seriously, that last half of theorizing (and acting like he is the only one to think as such) what Intel is doing including forced idle states which I hate to tell him was talked about days ago on other forums which lends me to believe he is ghost reading. 🙂 I absolutely agree with him capping top end frequency or at least picking a sane voltage:performance ratio --- I grow tired of Jufus patting himself on the back and making it a, "me against them" mentality, acting like his discord is some magical unicorn laden virtual land where all technical matters are discussed and solved and nowhere else (you gotta pay of course) and somehow always pointing to himself as some messiah/savant when it comes to basics while always shilling his bundles and services.....but I do still watch and enjoy his videos. 🙂

How does Kryonaut Extreme stack up against KPx?

Looking forward to your findings. The Lightning is actually a fairly new board released this year, so it had launch BIOS and then when I picked it up the newest BIOS was 0x125. I'm pretty pleased even at Intel's enforced limits. I tested with WoW and FO76 and it sits at 5.9 rock solid everywhere with better temps than fixed (as expected). I might push to 320w extreme limits for testing but as is 253w/400a works more than well enough for D2D especially gaming. I could adjust down to 59x all core (via per core to keep ring boosting in place). I'll check with Starfield and Fortnite later. I'm running this just like I ran my 12900k cool hand luke (which is in my NH55 now). It was the only 12900k I had that could handle lowest LLC / AC/DC settings without crashing on an AIO. When running like that, binning and then cooling (in that order) makes a huge difference. You are really testing your chips auto bottom/bottom voltage range. I'm always reminded that binning does matter. I tried to adjust my bud's SP101 14900ks using the same settings on his Hero as my Z690 Strix and Z790i Lightning and it was crash city. I ended up going with LLC3 / .70 / 320w and 56x all core to stop it from overheating on his Corsair 360mm AIO (I also repasted w/ KPx). He had also picked up a pair of G.Skill 8200 sticks and tried running them at 8200 which did boot and seem to work to him but a quick run of TM5 and they were erroring out left and right so after some adjustments and testing, he's locked in at 7600 solid. With the SA Bug now resolved, this is a perfect chip for my needs. I mean absolutely perfect. Next up is Intel APO testing in WoW, deeper memory testing and 320w limits testing.

Dialed back in auto enforced limits and it is not only running the same as before, but ~5-6c cooler and hitting 6.2ghz again at 8200 SA Auto which is 1.31: 0x125: 0x129:

Installed BIOS update 6.02 on the Asrock Z790i + SP109 14900KS. Initial notes and observations: ----------------------------------------------------------------------------------------------------------------------- Now default load optimized settings no longer blows up to 350w+ and overheats but stays at 71c tops for CB23 run while scoring /clocking the same multi (~37k) Unfortunately the settings used are still not optimal as it is pulling 1.184v under load yet still throttling down to ~4.8ghz all core versus my settings pulling 1.152v under load and staying rock solid at 56x all core and running just as cool if not a touch cooler. Auto single core only turbos to 5.9 versus my settings which hit 6.2 at basically the same pull. Exact same settings as 0x125 so I'll have to investigate to see whats going on / load up my saved auto profile as it may be power throttling somewhere with the default settings. 0x125 was blasted from the MC options and replaced directly with 0x129. SA Bug is fixed on Asrock. Initial testing SA Auto 8200, SA was at 1.301 and running TM5 no problem. Before, it was instant "SA Crash" on auto and anything >=1.20v + >=7800mhz .....w00t. I have all my old profiles saved on a USB stick so I'll start restoring them next and see how they perform. After re-validating all my dialed in profiles (Fixed_Voltage(56x-59x, 45x, Auto Ring, 8200)), I plan on testing the memory a bit now that I am not capped by 1.18v SA.

Simple enough to test. Scale by per core in CB23 and measure Vcore under load to get a spread spectrum of required voltages from 1 to 8 P-cores while verifying there is no clock stretching or downclocking. That will give you an idea but the good thing is your cooling really helps cut down on load requests especially in a varied chilled state so I don't expect anything too high. I see no downside to running 0x129 especially in your setup as it will help any errant transients pop out even with Asus controls in place.

14600k wins but I mean I guess the argument could be platform longevity as Socket 1700 is almost at an end (sans Bartlett) so 9600x is more future proof? Maybe? So far, 9000 series is more about refinements than anything....9950X3D/9800X3D needs to bring some zip to this Miracle Whip especially with Intel looming large with 15th gen. Asrock's Z790i lightning BIOS update for 0x129 has dropped, so I'll be giving that a whirl later. @Raiderman, I installed MSI's newest BIOS for the x670e Carbon and boot times are definitely faster (or at least returned to boot times 2x BIOS drops ago 🙂 ). Everything else looks and feels about the same and all old timing profiles work just as well.