electrosoft

Good news: ------------------------------------------------------------------ Been testing the TG vs Patriot sticks and the TG sticks are vastly superior. Better XMP timings and lower voltage (1.4 vs 1.45) at the same speed. TG passed 8000 90 min TM5 XMP right out of the box even hitting close to 70c no problem. Patriot erred out at 8200, 8000 and on a long run of 7800 (before 90 min). As a test, I set the Patriot sticks to TG timings and voltage and they erred out immediately at 8200, 8000 and within 10 minutes at 7800 with temps either only hitting 55c or in the case og 8200, 43c. They're garbage and going back. Crazy how much beefier the TG heatsinks are. I weighed each stick and the Patriots weigh 47g/ea and the TG sticks are 84g/ea. I'm most likely going to hold onto the TG sticks but..... Bad news: ------------------------------------------------------------------- After more granular testing, I'm pretty sure this chip has the SA Bug. The TG sticks let me find it as the Patriot sticks would error out on TM5 before triggering it. TG sticks at 8000 and 8200 just sucked it up till the bug triggered (complete system lock up, black screen, requiring hard power off/on). No problems with DDR4 at 1.35v and no problems at lower frequencies at 1.35SA which is must likely an issue of setting vs requested voltages. As I scale up, requested voltage triggers it and boom. Once you're at 8000 and 8200 auto SA (or hard set SA) and you go to run TM5, instant (8200) or within 5 minutes (8000) lock up, black screen and power reset required. Testing SA shows 1.18 is rock solid. 1.20 it will eventually SA Bug out. The higher you scale the quicker it triggers. IMC on this chip is indicating 8200 @ 1.18v it will eventually hit a wall and need more SA and bug out. Hard set to 1.18v at 8200 and the system will just reboot 20-30 min into testing. I'll run a few more tests with a touch more Vmem just to completely rule out the sticks vs SA wall but either way, it looks like overall this chip is SA Bugged. --------------------------------------------------------------------- Not sure if I'm going to keep it but everything else about it seems pretty solid. I just have to decide if I'm good with the SA Bug as 8000 tuned was my D2D goal.

Intel: I would prefer to see actual Vcore/ out idle and under load as that is the most accurate read under load not VID requests.... (IE, what is requested vs what is actually given). With every 13th and 14th gen so far tuned, VID pretty much does it's own thing but scroll down further under motherboard and read Vcore / out and you will see the actual fixed Vcore idle and the droop under load. Now if under load Vcore/out is hitting those levels that's another story..... AMD: Remember, I received a brand new SEALED 7950X3D from best buy with a 7950X in it.....QC run amok. Still hilarious though. 🙂 Speaking of Intel and limits: This post over on the OCN forums makes much sense: Stepping through my amps, ~325a set stopped any type of downclocking at 56x / 45x / 50x which coincidentally using the calculation showed 307a bios set wasn't enough (307a set = 204a actual) but Intel allows 400a under their new guidlines set = 576a which is overkill if you're looking to dial in your ICC to that particular setting. ~341a-342a bios set is the inflection point it seems if you're looking for equilibrium between set and actual. Or we can take the @Mr. Fox approach and open up everything to the wall! 🙂

What are you using to cool them? I picked up the TG because of the superior HS and to compare/contrast with the Patriot sticks but I sit there and watch the Patriot temps edge upward and know when they get to ~ 53-55, an error is incoming. I even closed the case completely up and let it get toasty in there (there are no fans except the AIO intake in the front so that is hurting not helping lol) I can get them to crash in ~10 minutes with the case closed. 15 minutes case open and 25 minutes with a fan right on them at 8200. I haven't even checked to see how high they will post on either set. I'll check that later. I did check the newest 0X125 vs 0X11F Microcodes versus each other just to see if there was a difference and there wasn't for my needs which was expected since I tuned everything manually but I DO know even at stock default everything with 0X125, that thing shot to 363w and overheated almost immediately when I first tested it last week before "The Tuning-ing" Going to sort out / have fun with the memory over the next several days then dial in 5.8 all core just to see where it lands with this SP109. ETA for the SFF transplant is probably a month at the rate I'm sauntering along in the testing/dialing in plus I still plan on slapping the 7900xtx in there and then 4090 to collect all new sets of data since the WoW Pre-patch Xpac has dropped (final upgrades/settings for the new Xpac GPU engine).

I suspect many many 14900K and especially 14900KS chips will degrade over the next year or so. I also suspect many users are going to be irritated when the new MC patch neuters a lot of those boost speeds as many of us are seeing testing even current limits and actually running Intel new limits which as found doesn't stop lesser core boosting to 1.5v+. Helping out a few friends and a few guildies with their systems = Two 14900KS chips and two 14900K chips and three were "flame on!" out of the box at stock settings tanking 3x AIOs and one CL without tuning. Even the SP115 13900KS was approaching the sun out of the box before tuning. One 14900KS was an SP98 on a Z790 Strix and it capped out at 1.40 BIOS setting and pulled close to 1.3v to be stable under CB23 load testing. I ended up setting it to 53x / 43x / 45x to keep it stable and cool on his AIO but it still games like a monster. The universal answer of enabling CEP, cutting PL to 253w and ICC to 400a just results in Silicon Lottery grossly coming into play and the better your silicon, the more performance you will get when capped across the board but either way it is nowhere near what it should be as advertised and still doesn't stop 1-4 core loads from spiking over 1.5v. I wouldn''t be surprised if Intel doesn't cap it to 1.5v or even 1.45v across the board regardless of situation with the MC update. I still firmly believe that capping your PL is not the solution nor the problem. It will always be dialing in and controlling your Vcore and as a safety measure just cap your amps to 400a which in just about all normal like scenarios won't even be approached. Just using CEP, PL and ICC is an indirect way of limiting your Vcore but not in all scenarios if unchecked.

State of the SP109 14900KS: For my most likely SFF settings of 56x/45x/50x 8000 for CB23 loop testing, Vcore is BIOS set 1.28v fixed , LLC 3 and ~ 1.152v under load resulting in 72c on the Lian Li Galahad II Performance 360mm AIO w/ KPx. PL are unlimited and ICC is capped to 400a. This results in pulling ~323w in HWInfo verified with ye ole Kill-o-Watt meter. This chip is hungry but runs relatively cool. I worked with AC/DC LL and it really didn't make much of a difference and added in a little bit of instability, so near stock (.74) is good enough. I was able to shave off ~9-10w dialing it in and it wasn't worth it overall. SP115 13900ks required fixed 1.33 for the same level of performance and that was with E-cores @ 4.3 vs 4.5 on the 14900KS. SP109 14900KS can do 5.7 same settings as above and top out around 77-80c on the AIO at 1.33v fixed CB23 pulling ~356w. Ambient temp in my computer room is ~70-72f year around. I haven't really worked with 14900KS chips, but i'm kinda amazed how cool then CAN run even on an AIO dialed in with no delid. I haven't really dug into 5.8 too deeply as I moved onto memory a bit when I had time but I want to see where that tunes into too. -------------------------- Right now, I'm fighting with the memory so as a control I ordered a set of Teamgroup 8200 sticks that will be here tomorrow. Problem it looks like is the heat sinks (aka heat retainers) of the Patriot sticks hit ~53-55ish and they error out at 8200 in about 15 minutes. If I put a 140mm fan directly over them, they can last about 25 minutes on TM5 before hitting the same temp. This is before dialing them in along with Vmem but the writing is on the wall for temps being an issue especially when I move it to the SFF case. XMP on the Patriots 8200 requires 1.45 vs 1.4 for the Teamgroup 8200 (not that it will matter in the end as I scale and tune manually but I still want to compare their full timings for comparison just to see). Teamgroup has actually somewhat usable heatsinks but in the end, I feel like i'm going to order a set of these to try and minimize heat as much as possible. If there are better heatsinks out there, please let me know. I'll work with both sets over the next few days and return the weaker set (or both if needed) which I feel is going to be the Patriots but you never know. Maybe a good, tested 8400+ set will pop up on the forums for sale in the next week or so. 🙂

You coulda put the Crocs BEHIND the box ya know? 🙂 Nice setup there @Papusan!

Whew, that looks beautiful! I've thought about picking up an EVGA 3090ti for the wife's system just to round out that final homage to EVGA with the DG-77 case.

As suspected for quite some time in the OCN forums and elsewhere and as Jufus had said for quite some time (Over a year ago he said boost Vcore in non all core was insane). All his kits he sold were locked down to all core for stability and fixed Vcore tuning which is....shockingly....what many of us have done over the years to reign in Intel's insane limits along with MB makers pushing as hard as possible so their MB would look better than their competitors.....

Testing different settings and profiles on the SP109 14900KS and it is definitely superior to my SP115 13900KS using Falkentyne's baseline testing here which I used on the SP115 13900KS but more importantly in real world testing it can do at 5.8 and 5.7 for SFF parameters what the SP115 could do at 5.6 all core. Running a 30min CB23 at 5.6 all core even on an AIO and temps never break 70c and score ~42k which is good enough for me but I like having a realistic option of 5.7 and 5.8 settings even on an AIO which will push into the 80s. Onward to memory tuning.... Yeah this pricing is very nice and bodes well for their X3D models.

Very reasonable! And it looks like AMD learned their lesson from the 7950x is that is the pricing. They are really putting their boot on Intel's neck big time and I'm loving it! Glad they did and very happy to see your work in the thread bro.

That's some damn good work Chew! I wonder if anyone else in the thread has hit 8600 with that level of stability along with some benchable 8800 runs, nice bro!. When all is said and done, I doubt I'll be hitting those levels, but it shows the potential of the Asrock Lightning board.

I never took the time to test cache scaling so a video like this shows minimal gains but they're still there in pockets here and there for gaming. I think I did do a few tests in WoW back with my 12900k but it didn't show much gains at all. I may retest. As we suffered through the slow death of truly upgrade enabled laptops...... 😞 https://videocardz.com/newz/socketed-cpu-laptops-and-high-end-mxm-graphics-are-fading-into-history https://www.notebookcheck.net/Socketed-mobile-CPUs-the-modular-laptop-potential-that-lived-in-the-worst-time-possible.862212.0.html

I like the diversity of laptops and desktops along with everything associated with both platforms. Even with the death basically of *new* mxm/slotted CPU laptops, every laptop has potential for hardware and software enhancements and modifications. I've ripped apart just about every BGA laptop that's crossed my path to enhance its cooling while also doing everything possible to dial in the CPU/GPUs. In summary, as long as the forums retain the heart and soul of NBR and remains the go to place for NBR refugees still wandering in and the great ever growing resources for laptops of all ilk, whatever grows and prospers up and around it works for me and then some. I'm extremely flexible and even minded (in case that isn't readily apparent). 🙂 More exposure and growth is always a welcomed addition.

The Bartlett 12 P-core model really works for me. I'd maybe settle for a 10 core. 8? I'll just turn off my e-cores and keep it moving....

Both runs with this sp109 14900KS on the Z690 Strix D4 and Asrock Z790i netted about ~40k and monstrous pull and overheating. This was pure stock, defaults out of the box settings nothing adjusted (not even memory). Both motherboards were running the newest BIOS as of July 2024. The 41k+ from my last few posts are starting the fun dial in / tuning locked to 56x / 45x while I continue dial that in as tight as possible, then move to 57x, 58x and 59x rinse/repeat to establish 4 optimal all core profiles to adjust accordingly for the SFF build with it. The asrock lets you pick from 3 microcode options in the bios which is nice. I haven't explored these yet and am running 125 atm. I don't know what MB makers are doing with the newest BIOS settings, but seeing what the D4 and Lightning did out of the box is not fixing the extreme overheating or pull. Probably a good idea. What MB and memory are you using again?

A bone, stock 14900k with no BIOS adjustments on defaults? 40k is about right....

Agreed, might as well shoot for the big dog and get a 14900k variant especially if your Alienware has a 13900HX which is basically a desktop chip slapped in there like the AMD X3D laptops (MSI and Asus). Why step backwards? It's going to vary with so many things in play, but out of box should be ~39k; and unlike our laptops, it can do it indefinitely with zero cooling issues (or it should). 🙂 Since they're all basically the same thing (13900k -> 14900KS), it will come down to cooling, tuning and silicon quality.

Um, I'm going to assume you meant to save breathe. 🙂 Fully no longer touching the sun now (max temp = 74c on AIO) but going to need to tighten it up more to qualify it for the 12.4L SFF build while finding that perfect all core vs temp range.

Sure! I always wanted to walk on the surface of the sun.... 🤣

Walking into hades vs starting to dial it in a bit (PL = 4096, ICC = 512a, LLC auto, 56x / 45x): Hades: Dialed in a bit (much more to come) to finally get at least 41k without exploding 🤣 56x / 45x:

Using the Z790i Lightning so far.... I did re-verify with DDR5 (already had checked with DDR4) to check for SA Bug (set SA to 1.35)on the SP109 and nope, no bug. I let TM5 run for awhile. Speaking of which, the new, cleaner look is very nice. Asrock read the V/F 6.2 point at 1.484 (vs 1.488 on Asus). While the Lightning has a better VRM design (14+1+1 vs 10+1 on the Edge and Asus Z790i boards) this SP109 is a furnace with these early runs. This little board unleashed I was pulling 400w and 320a on this 14900KS holy moly before throttling with this Lian-Li Galahad performance AIO which is a beast. Problem is CB23 keeps capping out around 40k. Looks like the board tops out at a -100UV which can't be right. That's laptop shenanigans. I must be missing something but I'm beat so I'll come back at it tomorrow. Playing with LLC vs PL vs UV. I might have to dial in fixed to calm this thing down. 🤣 I may have finally met my match the first time with a chip that requires a delid and/or better cooling (CL, DD, Chiller), but we'll see. 🤗

7950x will absolutely face smash the 7800X3D for encoding. I went with the best of both worlds and just grabbed the 7950X3D. Compared to the 7800X3D, you are looking at a slight performance uplift for WoW with the Non-X3D CCD turned off. I just hop into the BIOS and turn it off for gaming but you could also try to get Gamebar or PL to play nicely too. I find just turning it off works the best overall. Then I turn it back on after. I sometimes just leave it off though or forget to turn it back on. 🙂 Agreed. With that being said, a 7600x can beat a 5800X3D in gaming and it just gets better as you scale upward. 7950x is a very competent gaming chip. If you don't want to deal with PL, GB or BIOS disabling, you can't go wrong with a 7950x. @win32asmguy, I would go with a 7950X3D and get the best of both worlds. If multithreaded performance is important to you significantly, I would skip the 7800X3D especially with 7950X3D pricing as it is atm. $459.98 @ Amazon is a damn good deal. This. Holding back and slowly doling out chips to the consumers for maximum profit is old hat for Intel. Switching chipsets and even blocking older chipsets that are socket compatible from running newer chips is also classic Intel. They could learn a thing or two from AMD and not only extend the socket life (while also introducing newer chipsets for new customers or current customers if they want with QOL/upgrades) but also showing their true chip roadmaps to give buyers future assurance of chip purchases. That is one thing I always liked about AMD.

@Raiderman New BIOS 18.7.2024 dropped today: https://www.msi.com/Motherboard/MPG-X670E-CARBON-WIFI/support Good thing I did one last check before updating and nuking my system after retesting all my components....

Definitely an AMD focused tool that has some minimal Intel functionality. I'm more curious to try it on some of my laptops (both AMD and Intel) to see if it helps dial in some cooling profiles.

I can't imagine a 12 core no HT part but with the state of Intel these days. I'm still hoping AL surprises us all but while technically offering advanced architectures in the past, the 1st gen early versions were slower than their predecessors: P4 regressed in performance vs P3 back in the day till it got its clocks going and surpassed it... 11900k had worse MT performance than 10900k but laid the groundwork for 12th gen... We shall see. With Asus, you have to wade through their overpriced chafe to get the worthy wheat.....ugh.