EVGATech_LeeM

Kathum

joneffingvo

Kathum
No one here noticed unstable adie xmp / mem oc in combo with an 13900K rpl cpu?

I'm getting unstable a-die xmp. Can't run my GSkill 7200 at XMP, can only get stable at 6200 mhz

Thats the point , i have multiple adie kits here . All unstable . Even old mdie OC‘s wont go as high as they did before .

Please provide the model numbers for each kit so I can have our team look into it.  What are you using to test stability?

Oloy MPOWER DDR5-6800 C38-42-42-109 1.45V 32G(2x16G) - M-Power6838AW16
G.Skill Trident Z5 RGB schwarz/black DIMM Kit 32GB, DDR5-6800, CL34-45-45-108 - F5-6800J3445G16GX2-TZ5RK
G.Skill Trident Z5 RGB schwarz/black DIMM Kit 32GB, DDR5-7200, CL34-45-45-115 - F5-7200J3445G16GX2-TZ5RK
 
I use multiple Tests like: 20 Cycles of Testmem5 1usmus_v3 config, atleast 2000% on HCI RunMemTestPro and 3x y-cruncher Pi-2.5b runs.
Coldboot stability is weird, probably some weird Training issues/behaviour. After i got something "stable", same setting wont pass my stability testing after 3rd or 4th coldboot.
(Coldboot=shutdown & psu switch off & waiting a few mins) And my Board still wont properly shutdown, it just restarts.

I noticed on the Classified with 13900k and 2.03 bios, if I left it on default settings it never boosts to 5.8GHz, even with temps are below 60°C. At default I was running CB-R23 with a package temp of 65°C and hitting 40.1k points, but running single-thread it never went over 5.5GHz. I remember people talking about this in 12900k. Was it C-states that had to be disabled to make that work right?
 
I did a ratio-limit OC and that made the CPU actually boost for lighter work-loads, but at default it would not. Just thought I'd report that.
 
OH...and I did find one old bug. If I have HWiNFO64 open and try to run CPU-z, it crashes on the SPD read. I had gotten to the point where this no longer happened with my 12900k with BIOS 1.10. I did have C-states disabled though (on 12900k), so I'm not sure if that made a difference with this or not.

I'll have to confirm this for RPL, but Intel has sort of a hidden "all-core max speed" spec on processors, which limits the top frequency based on the number of cores you have under load.  You can probably hit 5.8 on all cores if you don't put any load on the processor and set Windows power options to Maximum (or Ultra) Performance and leave it idling there.  As you start putting a load on a certain number of cores, that top core speed will start dropping until you hit the all-core max speed.
 
Typically C-States needs to be enabled for Turbo/speedstep to function properly.  Turning it off would do the opposite.

Update/Continuation of the story from https://forums.evga.com/FindPost/3583997
 
Changed the GPU: from 3090Ti Kingping -> 4090 AMP Extreme got the 53 54 7F loop even with the Corsair kit of DDR5 (this was working fine before)
Replaced Corsair KIT with the the 2 GSkill DDR5 kits mentioned in the forum post above (still looping) 
Replaced the G.skill kits back with the Corsair and Replaced the 4090 with 3090Ti Kingping .. all good
Replaced the 3090 Ti Kingpin with a 3090 Ti EVGA FTW3 Ultra .. all good (with Corsair DDR5 kit) 

I hope something can be done - is kind of silly just to replace the GPU and board started to act funny, also the incompatibility with gskill memory module is kind of bad also. 

Dschless
Currently getting boot loop with two DIMMs:
* Loop 00,12,28,7F on 2x16GB
* Loop 00,12,28,54,7F on 2x32GB
 
With one DIMM in DIMM 2 slot, boots OK.
 
Specs:
* 13900k
* G.SKILL 2x16GB F5-6400J3239G16GX2-TZ5RK (Hynix M-die)
* G.SKILL 2x32GB F5-6000J3040G32GX2-TZ5RS (Hynix M-die)
 * BIOS 2.03
* K|NGP|N Hydro Copper 3090
 
Tired clearing CMOS, pressing power button to allow residual power to drain, lots of BIOS flashes (2.02 and 2.03), tried different memory OCs as well as stock settings. Tried mixing RAM kits, also with a single DDR5 4800 stick in the mix. Currently running BIOS 2.03.
 
Just bought a Hynix A-die kit that'll arrive on Friday. I wonder if it's an M-die issue.

shouldn't be m die. what is the "diff memory OCs" in this case? can you try XMP with like:
 
1.55 vdd
Auto vddq
Auto gear mode
Auto pmic turbo mode
1.3 cpu vddq
Auto SA
 
and report back? 1.55 might be way too high (and by way too high, in no way is this dangerously high - thought you will need to enable extreme voltage mode if you haven't yet - 1.55 is not extreme, promise) for m die on the newer bios too, i can test later when i'm home but your debug codes are def mem. goal is to get the voltages right for a successful post, then you can go back in and tackle raising frequency and finally tightening if you're comfortable with that. but XMP and voltages should at least get you booting in to windows, m or a die. GL

Dschless
Thanks zippytek! That worked! I put in one DIMM, applied settings, then put in the second DIMM.

Not your fault: memory OC is currently unstable (screen artifacting), and I need to reinstall Windows since it got corrupt during troubleshooting.

To answer question about OC’s tried previosly: XMP profiles, daily OCs found in extras, and bumping mem VDD a bit.

Hopefully with a future BIOS I can just use XMP and boot with two DIMMs on stock BIOS. Or maybe everything will just work with Hynix A die. I’ll report back on Friday. Newegg packaging new memory up today.

glad to hear you managed to get in! some notes about windows behaviors to help you along...
 
-if you can post and boot in to windows at 1.55 vdd/1.3 cpu vddq, try lowering each (one at a time) by 10-20 mV and testing again. ideally you want to run the lowest *stable* voltages you can manage for the given frequency. scaling is such that if you increase frequency and/or tighten timings, your voltages will have an impact and may need adjusting. this is why i recommend finding the target frequency + voltages first, then work on tightening timings and fine tuning voltages for stability again at the end.
 
-screen artifacting is unusual, honestly. that's maybe a sign of hardware failure or bad connections. i'd power off and make sure both dimms are well seated before doing more testing. you're well on your way now, though. 
 
-congrats on the incoming A die! it's VERY FUN and really flies on raptor lake. keep in mind that a die is much different in how it behaves both frequency wise and voltages. so you'll want to follow the pattern i outlined here, start higher than you will likely need for daily stable at the target frequency and work down from there.

Kathum

EVGATech_LeeM

Kathum

joneffingvo

Kathum
No one here noticed unstable adie xmp / mem oc in combo with an 13900K rpl cpu?

I'm getting unstable a-die xmp. Can't run my GSkill 7200 at XMP, can only get stable at 6200 mhz

Thats the point , i have multiple adie kits here . All unstable . Even old mdie OC‘s wont go as high as they did before .

Please provide the model numbers for each kit so I can have our team look into it.  What are you using to test stability?

Oloy MPOWER DDR5-6800 C38-42-42-109 1.45V 32G(2x16G) - M-Power6838AW16
G.Skill Trident Z5 RGB schwarz/black DIMM Kit 32GB, DDR5-6800, CL34-45-45-108 - F5-6800J3445G16GX2-TZ5RK
G.Skill Trident Z5 RGB schwarz/black DIMM Kit 32GB, DDR5-7200, CL34-45-45-115 - F5-7200J3445G16GX2-TZ5RK
 
I use multiple Tests like: 20 Cycles of Testmem5 1usmus_v3 config, atleast 2000% on HCI RunMemTestPro and 3x y-cruncher Pi-2.5b runs.
Coldboot stability is weird, probably some weird Training issues/behaviour. After i got something "stable", same setting wont pass my stability testing after 3rd or 4th coldboot.
(Coldboot=shutdown & psu switch off & waiting a few mins) And my Board still wont properly shutdown, it just restarts.

CMOS then load profile, thats how u check training stability. also look at the post code sequence if u see 54, that means smth is wrong, cause it means board is unable to train properly, it needs to retry. 

Thanks for help! Just wanted to report back that I’m back to memory post code boot loops with the 7600 MT CL36 Hynix a die: F5-7600J3646G16GX2-TZ5RS. I wasn’t able to work around and provide manual OCs to make this work with two DIMM. One DIMM worked fine though.

I’m going to remove the motherboard this weekend, look for bent pins, and install a contact frame. I’m also ordering some no name A die from china. Something has got to work, haha. 🤞

wbarber07458
Any idea when this will go up as a final version?  Seems like the best shot at stability with a 13900k.

Main reason why my 13900k is still in the box. Waiting for stable.

I didn't get a dark kingpin to run ddr5 at JEDEC speeds. I'll wait.

Sajin

rgarodnick
I didn't get a dark kingpin to run ddr5 at JEDEC speeds. I'll wait.

Have fun waiting overclocks are never guaranteed to be stable.

Of course they're not guaranteed. But when major issues pop up on a beta bios, it's reasonable to wait for a stable release. I'm running 6400 stable on ADL with 1.x bios. You do you. I'm waiting for a stable release

rgarodnick
I didn't get a dark kingpin to run ddr5 at JEDEC speeds. I'll wait.

just get the z790 apex dude. its up on newegg for preorder. gskill did 8k on their new 8k c40 xmp kit.so yea 8k on the apex will def work.. 

Dschless
Thanks for help! Just wanted to report back that I’m back to memory post code boot loops with the 7600 MT CL36 Hynix a die: F5-7600J3646G16GX2-TZ5RS. I wasn’t able to work around and provide manual OCs to make this work with two DIMM. One DIMM worked fine though.

I’m going to remove the motherboard this weekend, look for bent pins, and install a contact frame. I’m also ordering some no name A die from china. Something has got to work, haha. 🤞

which bios are u on?

babayega, I’m on 2.03. rgarodnick, +1. EVGA will sort it out. Right now Z690 apex is running same memory settings as Z790 apex. EVGA Z690 will be the same way. Only diff between the apexes is that the new apex comes with a DDR5 fan with factory mount. I’ve got no shame slapping a $5 fan on my RAM sticks. Also on the dark we don’t have that weird M.2 slot next to the RAM, and we can properly run 3 M.2 between the two PCI slots.

TheAffxct

babayega

rgarodnick
I didn't get a dark kingpin to run ddr5 at JEDEC speeds. I'll wait.

just get the z790 apex dude. its up on newegg for preorder. gskill did 8k on their new 8k c40 xmp kit.so yea 8k on the apex will def work.. 

I’m not asking for a fight, but I’m just going to say, G.SKILL putting up some arbitrary result of them doing 8000MT/s on an Apex does not indicate stability/doability. It really depends on their cooling solution, HCI coverage, and if they used a binned MB/CPU (which they almost certainly did).

If you’re being sarcastic then I apologise for being a dufus.

well. they said they will launch 7.8k c38 and 8K c40 xmp kit. so i would assume .. well XMP should work on the boards that are on the QVL for that kit.