electrosoft

On the older systems like x68 based Macs (pre PPC), Amiga, Atari ST you could program in x68 assembly using an assembler or inline assembly via C so you could get as "close to the metal" as possible (since speed and resources were so limited relatively speaking) and make direct calls with minimal layers versus the HAL and APIs of modern OSes like Windows which rely on modern compiler optimizations to get the code running as fast as possible and making modern languages much easier to work with like 4th generation (IE Python, SQL, Perl, etc..) for programmers. Back then we didn't have such advanced compilers that could take even VB/VBA level code and have it running nearly at if not equal to VC/VC++ code. I worked on some projects back in the early 90s in C where it just was easier to code in x86 via inline assembly using Borland's C products for sake of efficiency and speed especially when you needed to communicate with specialized hardware and/or needed to write your own drivers or heavily modify pre-existing libraries or code to get them up and running (I'm looking at you endless army of communication boards of all ilks).

If an SOC setting is the only problem that has been addressed, I can live with that. This does remind of the 5000 series and how subsequent AGESA releases neutered +mhz boost options and other things as time passed. Am I skeptical because of the 5000 series debacle? Sure. Having to neuter the SOC already due to "bulges" doesn't breed confidence but I'll give AMD (yet) another chance. Right now, I finished tightening up subs and turn arounds 6000 along with PBO solidly at -30 for a mild OC from 4824 all core stock to 4974-5000 in CB23 and I've had no problems on the test rig what so ever with the 7800X3D. The 7950X3D is still sealed sitting on the table awaiting its turn if needed. If @tps3443 gets the time maybe drop those new G.Skill sticks down to 6000, 6200, and 6400 and see if they can do CL28-32-32-30 which would make them a compelling option to replace these M dies. I just upgraded my Strix Z690 D4 to 2305 and suddenly I can now run C15 on those 4000 sticks no problem where they borked out every time before so I went back through and re-cert my timings on them too and tightened them down even further. WoW runs are looking even better than before months ago with the 12900k. If the 7800X3D brings nothing to the table, I'll return it and the 7950X3D unopened. If I like what I see, I'll crack open the 7950X3D and keep most likely the 7950X3D and return the 7800X3D.

I'm one of those, "new GPU new install" or at least a Macrium GPU less OS restore kinda enthusiasts. I like my GPUs to have a fresh and shiny OS install all to their own. 😁

I just love seeing old hardware smashed against the 13900k... 🙂 It leaves nothing on the table. 🙂 OHHHH YEAH!!!! Looking forward to the results and congrats! Best 7900XTX on the market IMHO.

Official statement on the SOC issue with Ryzen.... https://videocardz.com/newz/amd-identifies-the-root-cause-of-ryzen-7000-cpus-burnout-issue-new-agesa-released

Oh snap.... .....Taichi time? 🙂

I'm trying to figure out why it would synchronize Vdimm, VDDIO MEM in tandem with vSOC? That means with EXPO (or setting your own Vmem) you are increasing SOC beyond what would be needed as a bug. Of course if you're the type that manually sets your own timings across the board including dialing in Vdimm, Vddio mem and vSOC you would never have to worry about this problem but imagine some faster and tighter memory kits that push >1.35v and SOC scaling up in tandem? Yikes....

If they can find the transient or sustained peaks that are causing the issue this may be the new normal but if you want to FAFO it because you want to OC, I'd still stick with the most recent BIOS that isn't locked down but that is at your own risk. I went back and checked my initial run and SOC = ~1.360 at its highest. Set to 1.25v = ~1.26v peak CB23 run. Scores remained the same. 7800X3D boosts to 4824 and remains there at ~80c so it is actually a better sample versus some I've seen out there with many boosting to ~4700 range only. This is with my EVGA CLC 360 using the AM4 mounting. I'm not sure if 1usmus ever updated their programs to support AM5. MSI has left all their BIOS files up so far for the Carbon x670e. MSI (unlike Asus) once you lock in your RGB it still sticks to their MB. I'm not too happy with my DDR5 temps with these heatsinks. I might seek out a better set that can handle 1.35v a bit better as TM5 run temps were creeping up on 50c.

I was following this earlier and the initial recommendation is SOC = 1.25v but it looks like Asus is setting to 1.3v? Either way, I set mine to 1.25v earlier and re-tested with TM5 and WMD and no problems but I only run my memory at 6000. I won't upgrade or change my BIOS in the fears MSI was overzealous in their restrictions while I continue to tune my memory and then PBO.

Apple M1 $4k hardware getting smashed by a $2k buildout Hackintosh.... (if you price shop you could probably go even cheaper or squeeze in a 6900xt over the 6800xt) and/or go 13900k). Of course energy efficiency wise Apple rules the roost if you care about such things...

I remember @johnksss had posted some of the earliest pictures of the dremel work needed to be done on the heatsink to make it work with 2000 series he performed on his own unit.

Unless Framework gets something out by the time you want to upgrade (they are working on a GPU upgrade enabled laptop via snap in modules), you may only be stuck with BGA all around. Clevo's last DTR'ish laptop with swappable parts was the NH55 (12th gen).

Dunno as @jaybee83 must be in full new wife/child grind. With that type of power off and re-connect it does sound possible. What types of loads are you running and how much pull? There's links in the launch page to 280w and 330w PSUs. I've been busy with work and my desktop so I haven't had the chance to dig into mine like I want. I need to finish tweaking this AMD config before I transplant it into my desktop and then swap my Strix Z690 into my test rig to test some 12900k's I have here and swap between the laptop and desktop. We know it can take a 12900ks (confirmed by @win32asmguy). I'd like to drop a 13th gen in there just to see what it does but those aren't cheap. 🙂

Congrats! Your system is now running exactly as it should with an 11900k. ~15.4k CB23 is a non-throttling score and 86c is right in the sweet spot for a CB23 run with good cooling and/or silicon quality. The fact gaming temps of ~65c just prove your config is working excellently as those are great. You now have the ability to overclock if that is what you seek or just use your system as is knowing your stock performance is right in line with no throttling. I'm sure the fans when gaming are much less too which is always a plus. Cores sound like a C state issue as they are not in use and just go into lower powered mode. As long as they don't clock down while running CB23 you're good to go. You can always lock the cores but I've never been a fan of that on laptops. Let it run cooler when it can.

De8auer following up with AMD bulge/burn issues. Non X3D chips also hit. Asus official statement:

And.....they're out of stock again. 🙂

AMD being their best AMD while AMD'ing themselves yet again..... 🤣 (Looks over at the 7900XTX vs 4090) Motherboard makers being proactive in restricting OOB voltages while AMD sorts out what their chips can and can't do. Hey.... At least my USB ports work flawlessly this time! I kid! It isn't like AMD isn't slowly chipping away at Intel's marketshare so they must be doing something right....

This. You will know pretty quick if you got a good KM. The one I first had back in 2021 was one that would just shut down under heavy load or 3Dmark Timespy and had all types of problems. It also had some problems with different memory modules especially some Ballistix sticks that would lock up from time to time. The one I picked up in December was basically flawless. I was happy (and sad lol) that it required no tweaking or such to get it working as it just worked right out of the box. Even trying different memory types made no difference. It just worked. Gaming temps stayed in the 70s or low 80s. It used a custom GoBOXX BIOS that was incompatible with the Unlock here though so you would need to flash it with a compatible revision then apply the unlock.

Looks to be Asus only. The one MSI reference was in a YT video and it was a, "he said/she said," at best. Do you have any concrete direct links to any other motherboards having this problem except Asus because thumbing the article and reddit thread all I'm seeing concretely is Asus and a potential overvolting issue which makes sense with the sensitive nature of V-cache and reduced clocks and voltages. ------ As for the 4070 and >$500 GPU sales flopping, people worldwide have other pressing issues with inflation, recession and needing to keep a roof over their head vs luxury computer devices. Toss in a sprinkle of GPU fatigue, record sales during the Pandemic/Crypto and the used market saturated with fantastic deals and there you go. 🙂 Only one not reading the writing on the wall is Nvidia who have the ability to stick to their guns and simply opt to reduce production vs lower price. The problem with that model is you reach a point where lower production still has units on the shelves and not moving. Then what do you do? Unfortunately they are emboldened with having used this model to move tons of Ampere stock in the last 6-8 months creating artificial demand.

DR is a possibility but it can be hard on the fclk for AM5 so I want to see where my 7950X3D stacks up in a week or so when I unbox it for testing. On the 7800X3D I have no problems going from 2100 (its max) to 2000 for DR memory but if the 7950X3D is a 2200 I would most likely stick with SR unless they clear up the DR situation or I just buy a set to test for myself and see if it can handle it.

Yep, it's an AMD thing. Even AM4 had higher latency and lower overall bandwidth vs Intel 10th and 11th but the end performance was all that mattered.

All things being equal, pushing fclk gives you more bandwidth unless it compromises desired voltages or timings/performance. For example, I can push it to 2167 but performance suffered so you know you're running that red line of errorville. 2133 was the same as 2100 so again I knew I was still in that no returns / redline area. I saw gains stepping 2000->2033->2066->2100. Once primaries were dialed in they were good to go 2000-2100. Hopefully the 7950X3D can be pushed a bit more but I do know these m-die's have hit a wall but I'm not sure how much more I could extract with a newer set of A-dies. Knowing me, I'll end up picking up a set and pushing those just to see to remove DDR5 limitations as the culprit inside of the AMD ~6000 (Sweet spot) to ~6600 window (fingers crossed). I've seen some fclk's pushed to 2200 with great results. I'll definitely be sticking with SR modules to avoid and headaches. I don't agree with everything Buildzoid presented but his DDR5 tuning videos are fun to watch. I'll definitely keep your groundwork for 7950X3D in mind as I plan on benching my 12900k (which I've optimized to death over the last year+) vs 7800X3D vs 7950X3D (as AMD intended) vs 7950X3D CCD off vs Lasso to see if all roads for my use cases lead to the same valley. Once I finish up terts I'll move onto PBO. I know I'm having a blast with AMD so far. No USB issues so we're already starting out good.🙂

Tuned Primaries and fclk. Fclk caps out at 2100. Onward to terts....

I wouldn't even pay that day 1 especially being a FnF crapshoot and no idea how good the IMC is or where it craps out. That is a pure SP gamble. I'll bin some CPUs when I can usually 2-4 tops. I get what I get. I've targeted a few binned chips before but they were at or near former retail (or lower) at the time. Those "in the moment," will pay serious coin for good silicon and I get it....but for myself? No thanks.

In regards to the Taichi 7900xtx, as long as it is retail and not scalpy prices, if you want it get it. 🙂 With my 4090, I knew the only two models I wanted (Suprim X Liquid or FE in that order) and just waited till it was back to retail and picked it up. I don't regret it at all (Like I did the KPE 3090ti a smidge).