Alder Lake CPUs: Overclocking and general ADL memory/motherboard discussion

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Woomack

Benching Team Leader
 
 
 
 
Joined
Jan 2, 2005
Yep, looks like Micron B and other IC wouldn't run at these settings. If there is no problem with the mobo then it should pass 5000 at something like CL18-24-23 or 19-25-24. As I said in the earlier post, it will scale up to ~1.65V and to pass 4800, usually needs at least 1.55V. Since it's Micron then can't do much with the sub-timings.
I feel like Z690 mobos still need BIOS tuning. I was able to run Micron or Hynix up to 5400 on Z590I Unify and I see there are problems with 5000+ on most Z690. The same with dual-rank Samsung B. I could run my kit at 4400 on Z590 or B550/X570 and I can't even boot at 4200+ or XMP (4400) on Z690 mobos.
 

wtsds

New Member
Joined
Dec 21, 2021
I wonder what IC is in this ADATA kit. Samsung would have a hard time even boot at 4600. Hynix would probably need more relaxed timings. Micron, depending on IC quality, is supposed to run higher without big problems so can be limited by the motherboard (or BIOS).

As the friend above said, I have Micron B-die 1R! Just like the picture shown!

I'm thinking you might be right about bios. It must have some configuration that when it goes beyond 4600mhz, it cannot boot. It doesn't matter what latency I put. If I put 20-30-30-60 for 4800mhz. 1.6v Ram, 1.3v SA. 1.5v Vddq. No boot.

I was watching a video that shows a guy increasing the Dram training voltage to help with overclocking as well. He constantly criticizes the Bios of an Aorus Pro from Gigabyte.

I tried for 4700mhz, but it goes back to 4600mhz. If I put 4666mhz, it goes back to 4533mhz. I'll try 5000mhz and above to see what happens.

I wanted to reach 4800mhz with these timings of 19-23-23-45 Or at worst 19-24-23-45
Post magically merged:

Yep, looks like Micron B and other IC wouldn't run at these settings. If there is no problem with the mobo then it should pass 5000 at something like CL18-24-23 or 19-25-24. As I said in the earlier post, it will scale up to ~1.65V and to pass 4800, usually needs at least 1.55V. Since it's Micron then can't do much with the sub-timings.
I feel like Z690 mobos still need BIOS tuning. I was able to run Micron or Hynix up to 5400 on Z590I Unify and I see there are problems with 5000+ on most Z690. The same with dual-rank Samsung B. I could run my kit at 4400 on Z590 or B550/X570 and I can't even boot at 4200+ or XMP (4400) on Z690 mobos.

It could be that as yet not recovered bios mature enough. All that's left is to wait and see with new bios releases, if I can get past 4600mhz. As I said, I can't use the 4000mhz profile that comes from the factory. Arrives in windows and blue screen!
 

Woomack

Benching Team Leader
 
 
 
 
Joined
Jan 2, 2005
For my Z690 Master, there were 3 BIOS updates in the last week. I guess that for DDR4 boards will be something soon too.
I'm not sure how Gigabyte boards with DDR4 are acting. I generally pick ASUS or MSI for RAM OC but now I have only ASRock and Biostar.
Re memory clocks/ratios. Some ratios don't work on DDR4 and DDR5 mobos. I think I described that based on Gigabyte DDR5 mobo. It looks like you said so 4600 works but everything up to 4800 won't work and runs at a lower clock. I assume the same is with 4933 or 5066/5100 while 5200 works.


Yesterday I got the G.Skill TridentZ 5 32GB DDR5-6000 CL36 kit which is based on Samsung IC. I didn't have much time so only checked how high it will go at mostly auto settings and I could reach 7000 CL40 1.4V. 7200+ couldn't boot but there is a new BIOS released yesterday which I didn't try.
Out of the box, at XMP settings it can make about 80GB/s mem copy and 66ns latency. I look mainly at memory copy bandwidth as it affects the performance the most. With this scaling, I should be able to reach 100GB/s and go below 60ns latency.
Samsung-based DDR5 will be still rare but should be wider available in about 1-2 months. Hynix won't be worse in this case. It's just that Micron isn't running at higher frequencies (at least not the current IC).

I made the result with a 12600K CPU, which somehow gives a bit worse bandwidth/latency than the 12900K.

7kgs.jpg
 
Last edited:

Brando

Member
 
Joined
Jan 9, 2006
I'm thinking about grabbing an asrock z690 extreme and a 12700k but my memory isn't on the list yet though I do see other b-die 32gb kits. it's probably fine right?

EDIT: Well it took an hour with best buy chat but they matched microcenter for $329.99 for the 12700k. The guy said I was super lucky. Nice guy (y)
 
Last edited:

PolRoger

Member
 
Joined
Jul 31, 2005
I came across this YT video by SkatterBencher about ADL V/F curves. It may be too detailed for some casual users but there is some good overall info provided.
 

Woomack

Benching Team Leader
 
 
 
 
Joined
Jan 2, 2005
I'm thinking about grabbing an asrock z690 extreme and a 12700k but my memory isn't on the list yet though I do see other b-die 32gb kits. it's probably fine right?

EDIT: Well it took an hour with best buy chat but they matched microcenter for $329.99 for the 12700k. The guy said I was super lucky. Nice guy (y)
It should work fine. Actually, ASRock Z690 Extreme works pretty well with various RAM, even on the first BIOS. I started tests of the Z690 Riptide and it's already worse. I mean I had to fight for some time to even boot at 4000 XMP on my RAM when on the Z690 Extreme it works without issues. On the other hand, my 4400 TridentZ doesn't work at XMP on both.

One little update on the TridentZ 5 DDR5-6000. I can make it boot or pass benchmarks at 7000 CL40 or 6800 CL38 but it's not stable. depends on settings, there are more or less errors in memtest. I also noticed that anything above 1.39V VDIMM is causing additional errors in memtest. The highest stable clock is DDR5-6600 CL38-38-38 1.39V (maybe will work at a lower voltage) and this is also the highest clock at which works command rate 1N/T. I haven't seen any difference when I set CR1, even in AIDA64. After some tweaking, AIDA64 shows 102GB/s read, 96GB/s write, 94GB/s copy, and 58ns latency. I guess it can be better, I just had no time to adjust more timings and check stability.
I'm not sure if it's the mobo or RAM. Officially, my motherboards are not supporting more than DDR5-6400 so I can't complain but I was counting on a stable 7000.
The latest BIOS for the Strix Z690I Gaming (8.11) is worse than the previous one. I mean overall, stability is worse, memory training and sometimes I get a black screen in BIOS or artifacts. I still have to check Gigabyte as they had 3 BIOS releases in the last week.
 
Last edited:

Zantal

Member
 
Joined
Mar 19, 2009
How is it possible that AIDA64 gives me 67.000 MB/s as read bandwidth when I am running 4000MT/s ? shouldn't the theoretical limit be 64.000?
Benching with intel's MLC i am getting a more plausible 63.300 MB/s
 

opt33

Registered
Joined
Aug 12, 2017
Seems like Samsung DDR5 for hci memtest stable topping out 6200-6400 across few forums with decent timings. Hynix sticks better but no where to be found.

On my samsung 5600c36 sticks screening settings with hci memtest ~ 1 hour on z690 hero (prior to leaning my hero affected with reversed polarity cap, and my 5 volt reading was all over the place as was ddr5vdd, pretty sure why had some difficulty despite running memtest overnight and prime for 2 hours on final settings)
6000c34 2T 1.35vdd/vddq 1T no boot on any settings on 4 dimm hero. s/a stock volts (0.9) or 1.0 on all my runs. mc stock 1.1 or 1.16 max.
6000c32 2T 1.4 vdd/vddq
6200c32 2T 1.5 vdd/vddq (pictured below)

6200c32memtest1hr.jpg


Same samsung 5600c36 on new MSI unify-X 2 dimm board (after realizing my Hero affected with reverse polarity cap). unifyx has been rock stable once I run memtest stable overnight.
6000c32 1T 1.4 vdd/vddq (currently using this for 24/7 settings after memtest overnight though with non aida benching trefi/twr/trrd with 56.7 latency, similar bandwidth)
6200C32 1T 1.51 vdd/vddq, to high volts for me for 24/7 especially given replacement ddr5 is challenging.

6000c32c1.jpg 6200c32t1.jpg
 

Brando

Member
 
Joined
Jan 9, 2006
Ok the 12700k is sitting here. Waiting for the msi tomahawk ddr4, nh-d15s chromax, and wd sn850 to show up in a few days. Re-using the gskill b-die 3200 cas 14 ram and hoping to get it to 3600mhz with lowest possible latency without going over 1.4v. The plan is to disable e-cores and run it as an 8 core to save heat for maximum oc on air. Probably sticking with windows 10 for now. If temps are ok I may use e-cores at some point after win11 gets a few updates but I really don't do anything that needs them.
 

mackerel

Member
 
 
Joined
Mar 7, 2008
Just saw on the rumour mill that Intel will be killing off the unofficial AVX-512 support in future microcode updates. That is another reason putting me off from buying this gen.

So I'd like to ask if anyone is up for running a few custom benchmarks for me using Prime95? I'll give the exact settings if there are any volunteers. The goals are multiple: How does unlimited AVX-512 performance of P-core compare to Rocket Lake or Skylake-X? How does AVX2 performance of both P and E cores compare to Rocket Lake/Comet Lake/Zen 2/Zen 3.

Requirements: Alder Lake system with a mobo that allows enabling of AVX-512. Ram wont matter for this test as it will run in cache.

Run 3 test cases:
1, P-cores only, E-cores disabled. AVX-512 enabled.
2, P-cores only, E-cores disabled. AVX-512 disabled.
3, P-cores disabled. E-cores only enabled. Is this even possible?

Each benchmark run time would be of the order of 10 seconds, best run a few times each, plus a little time for up front settings. In case of any limiting causing CPU frequency to vary during each run, then I'd ask it be fixed. Doesn't matter what the clock is, as long as it is known and constant. Easier to set a lower safe value than try to push it up. I'm looking at IPC so it will be normalised out anyway.
 

Voodoo Rufus

Powder Junkie Moderator
 
 
 
Joined
Sep 20, 2001
Why do you guys think the memory performance is slightly less with the 12600K vs. the 12900K?
 

EarthDog

Gulper Nozzle Co-Owner
 
 
 
Joined
Dec 15, 2008
Location
Buckeyes!
Because aida is a multi threaded test and it has less cores/threads is one reason.
 

Woomack

Benching Team Leader
 
 
 
 
Joined
Jan 2, 2005
Why do you guys think the memory performance is slightly less with the 12600K vs. the 12900K?

I have about 2-3GB/s and 2ns worse results in the AIDA64 benchmark on the 12600K compared to the 12900K. This is why I said some posts ago that the performance is slightly worse. It can be related to the multithreading nature of the AIDA64 benchmark as the same is on some AMD CPUs or earlier Intels.

Seems like Samsung DDR5 for hci memtest stable topping out 6200-6400 across few forums with decent timings. Hynix sticks better but no where to be found.

On my samsung 5600c36 sticks screening settings with hci memtest ~ 1 hour on z690 hero (prior to leaning my hero affected with reversed polarity cap, and my 5 volt reading was all over the place as was ddr5vdd, pretty sure why had some difficulty despite running memtest overnight and prime for 2 hours on final settings)
6000c34 2T 1.35vdd/vddq 1T no boot on any settings on 4 dimm hero. s/a stock volts (0.9) or 1.0 on all my runs. mc stock 1.1 or 1.16 max.
6000c32 2T 1.4 vdd/vddq
6200c32 2T 1.5 vdd/vddq (pictured below)

View attachment 215242


Same samsung 5600c36 on new MSI unify-X 2 dimm board (after realizing my Hero affected with reverse polarity cap). unifyx has been rock stable once I run memtest stable overnight.
6000c32 1T 1.4 vdd/vddq (currently using this for 24/7 settings after memtest overnight though with non aida benching trefi/twr/trrd with 56.7 latency, similar bandwidth)
6200C32 1T 1.51 vdd/vddq, to high volts for me for 24/7 especially given replacement ddr5 is challenging.

View attachment 215244View attachment 215243

On my TridentZ 5 6000 C36, I can set up to 6400 CL34-34-34 1.37V CR1 or 6600 CL38-38-38 1.37V at CR1 and max for tests 7000 CL40-40-40 1.40V CR2. 6600 CL38 CR1 passed the stability test. The higher clock, the better bandwidth, and a bit better latency, but at 6800+ it's not fully stable. There are always single errors in memtest. Max that I had with the 12600K was something like 104GB read, 100GB/s write, 99GB/s copy and 54ns latency (not fully tuned). I'm not sure if the stability issue is related to RAM or the motherboard. I feel there are still problems with BIOS so I don't want to waste time on tuning everything step by step. I still have to back to Gigabyte as they released 4 BIOS updates in the last 2 weeks and 3 are related to memory or stability with G.Skill.
Right now I'm using Strix Z690I-Gaming but the last BIOS is somehow acting weird. Sometimes it boots with a black screen and I can't see anything in BIOS. Sometimes it's not training memory in the right way and instantly throws OC error.
I haven't seen any difference going from CR2 to CR1. At least AIDA64 was showing the same bandwidth and latency.
I also noticed that the stability is worse at 1.4V+ or there are more errors in stability tests at 6600+. 1.37-1.38V seems optimal. At lower clocks, 1.42-1.45V is still fine.


I had in plans to test it some more but I was busy with other things.
One update. ASRock Z690 Riptide runs at 4000 CL14 Gear 1 without issues. Somehow it doesn't boot at 4200+, even Gear 2 and relaxed timings.
 

Zantal

Member
 
Joined
Mar 19, 2009
One update. ASRock Z690 Riptide runs at 4000 CL14 Gear 1 without issues. Somehow it doesn't boot at 4200+, even Gear 2 and relaxed timings.
What voltage are you pumping to get CL14? my G.Skill kit is clearly B-Die but it needs 1.6v+ (it will error out after a while on 1.6 even with a fan directly on top) so I settled for 4k-15-15-15 @1.49v , the performance improvements are negligible anyway.

Also strangely even on my end it won't do over 4100/4200 no matter the voltage timings or gear mode, I really hoped I could somehow get this to 4600+ in G2 but I am lucky enough to run it at 4k G1.

At this point I don't know if it's the bios, ram or cpu holding me back.

Also

I am starting to suspect that the E-Cores mess with memory access somehow, I can run y-cruncher fine at 3900 MT/s but will throw errors or crash at 4K
Since on my motherboard the only related IMC voltage setting is called vddq tx I tried bumping that and it greatly offsets the time it takes for a crash or an error to appear, yet it still doesn't guarantee a success.

If I disable the E-Cores it runs fine at 4000MT/s

this is with memory tested as stable, so how do I selectively test IMC for errors?
 

Woomack

Benching Team Leader
 
 
 
 
Joined
Jan 2, 2005
4000 C14 G1 at 1.55V on 3 different Patriot and Team Group kits. The same was on Z690 Extreme which is better in general, but both are not so great for RAM OC.
 

Zantal

Member
 
Joined
Mar 19, 2009
A word of advice for anyone trying to overclock,

Don't do like me

I am trusting these OC programs waaay too much (Looking at you AISuite and Intel XTU :mad:)

YCruncher is stable if I set VCCSA to 1.225, when I was setting it from the program it had adverse effects (system lockup or increased instability)
Time to uninstall and never go back
 

tiggerlator

New Member
Joined
Jan 9, 2022
Not long built my ADL system so haven't tried any OC yet. specs are-
Alder lake I7 12700k 12/20 cpu
Asus Rog Strix z690-A WiFi D4
TEAM GROUP 32GB (2X16GB) DDR4 3600C16
MSI GTX 980ti Gaming 6g
Silicon Power PCIe M.2 NVMe SSD 256GB Gen3x4
Crucial 512gb 2.5" SSD
WD blue 4TB HDD
Corsair TX750M 80% gold
Lian Li o11D XL
EK 360mm PE
EK 280mm CE
EK supremacy classic (CPU)
Generic (GPU)
EK FLT 240 D5 pump/res
EK Duraclear 16/10
3x Corsair LL120mm
2xCorsair LL140mm
2x Noctua 120mm
32" MSI 1080P 165hz
Anne Pro II keyboard
Razer Viper

The ram is teamgroup 8pack edition-
https://www.overclockers.co.uk/team...00c16-3600mhz-dual-channel-kit-my-002-8p.html

I am just running it stock, so don't know how well it will OC. I have some experience of overclocking but my previous setup was a ryzen 2600x so this is very much faster. There seems to be a fair few more settings in this bios than my previous one. don't really want to use auto OC though.

Do any of you guys know this ram? i guess it wasn't cheap either at £210 but i wanted samsung B die.

Any tips would be appreciated.
 

PolRoger

Member
 
Joined
Jul 31, 2005
The ram is teamgroup 8pack edition-


I am just running it stock, so don't know how well it will OC. I have some experience of overclocking but my previous setup was a ryzen 2600x so this is very much faster. There seems to be a fair few more settings in this bios than my previous one. don't really want to use auto OC though.

Do any of you guys know this ram? i guess it wasn't cheap either at £210 but i wanted samsung B die.

Any tips would be appreciated.


Optimal memory performance on ADL DDR4 for B-die is ~ 4000m/t Gear 1. I would test first with your kit's default XMP settings (16-16-16-36) at higher speeds 3733/3800/3866/3900/4000 and see what the motherboard BIOS auto settings give for VCCSA Gear ? As you increase memory speeds, you will need to find your particular combo's stable voltage setting for DRAM and VCCSA. DRAM and VCCSA voltage requirements will increase as your memory speed goes up. 4000m/t G1 will take more DRAM/VCCSA voltage than 3800m/t G1. Once you get and idea of what it take to run at 4000G1 C16 you can try to tighten up your primary/secondary timings. Most of the B-die kits that were offered by manufacturers at 4000m/t were spec'd with at least ~1.40v/1.45v DRAM. Also note that your CPU may or may not be able to run as high as 4000 m/t G1? My 12600K can boot to Windows 4000 G1 but the highest that my combo will run stable is ~3900m/t G1. I'm now running Micron E-die at 4800C18 G2.
 

Brando

Member
 
Joined
Jan 9, 2006
I'm pretty sure this is about what I can get out of my system but there may be tweaks I don't know about. I have E-cores disabled but heat still keeps me from getting 5.2ghz without throttling. 5.1ghz @ about 1.25v with ram at 3600mhz cas15 1.4v and cpu SA at 1.2v but I can't do 1T for the life of me. I can live with it. I think somebody told me once that 32gb dual ram kits have a hard time with 1T and I couldn't do it on my last system either.

work in progress.JPG