• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

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.

funnyperson1

Senior Member
Joined
Jun 16, 2001
I tried 1.26v and it lasted longer during the stress test but it was way too hot. Next I tried going back to 1.24v and lowered p cores from 5.1 to 5.0 and made it through an intel xtu memory stress test topping out at 90c. I'm thinking overall performance will be better with 4 extra e-cores than with only p-cores at an extra 100mhz but maybe I'm wrong.
Honestly, the E-cores are only going to come into play performance-wise for tasks that are embarrassingly parallel like rendering or scientific computing. There may be some games that are threaded enough to take advantage of more than 8 cores/16 threads, but I doubt that the majority are. They're also great if you are heavily multitasking while gaming or something, you can have a lot of background tasks running without impacting your game performance, I imagine that Intel is going to make a big pitch for streamers to use Alder Lake.

It's mostly a power/efficiency play though, why have your beefy cores occupied running background tasks when these E-cores can handle them just fine. It's going to be really interesting to see how these stack up with the Apple M1 in laptops.

I'm with Woomack though, unless your CPU is crunching something all the time, there's really no point in running at 5+GHz on all cores all the time. You won't notice a difference in games between 4.7 and 5.1GHz unless your performance is right on the edge of where you need it to be. I personally am still trying to find the sweet spot for undervolting at 4.7 all core, unfortunately the -100mV was not stable overnight, so I've been stepping up from -10mV and so far am at -50mV stable.

That said, 90C during the Intel Stress Test isn't terrible, I bet that while gaming your temps will be in the 70s or something.
 

Zantal

Member
Joined
Mar 19, 2009
I just tried out a beta bios for my motherboard in the hopes if would fix some voltage regulation shenanigans I noticed and now it happily applies 1.45v VCCSA, before it wouldn't even attempt to train it and I had to manually set it to 1.225 to stabilize it. That's just too much

Also I have seen a lot of people changing the ring behaviour from auto to fixed 4.2ghz which is quite a lot higher than the minimum for when E-Cores are active (3.6ghz)
I gave it a try and it tanked my performance on every bench, cinebench for example was 29k and after 21k, y-cruncher took almost 8 seconds longer to complete for 2.5b
So the software readings don't tell the whole story behind the ring behaviour, can someone try and duplicate this?

Ideally we should have 2 settings in bios 1 for P-Core and the other for E-Core active, if you decide to keep the E-Cores on you shouldn't touch ring, you either crash or reduce performance.
 

Brando

Member
Joined
Jan 9, 2006
I got an automatic c: drive repair on restart within a day of enabling e-cores. Looks like those unusually high default load voltages weren't too far off lol. I guess I'll play around with Max clocks/no e-cores vs reasonable clocks with e-cores and see if I can feel a difference after awhile. Probably not.
 

Woomack

Benching Team Leader
Joined
Jan 2, 2005
I wonder where is the safe limit for the DDR5 temps. I see some guys running them at 1.55V and I noticed that without a direct fan blowing on the RAM, there is 85°C+ during memtest or other stability tests. At 1.35V there is still 70°C.
I was also comparing what would give me going from 1.35V to 1.55V on my Samsung based kit. I can set CL32 or CL34 instead of CL36 what in AIDA64 gives me 0.5GB/s higher bandwidth and 1ns lower latency. At least in the case of Samsung RAM, I see no point in high voltages as CL is not giving as much as sub-timings that run much tighter without the need for much higher voltages.
Stable settings (memtest+AIDA64) 6400 CL36-36-36 1N 1.35V and tighter subs (not maxed out) give me 103GB/s read, 95GB/s write, 95GB/s copy and 55-56ns latency.
With some luck, I will get 1-3 more DDR5 kits for tests soon.
 

Woomack

Benching Team Leader
Joined
Jan 2, 2005
There is one sensor so hard to say as it's described as RAM temp. For sure PMIC is causing additional heat. Even lower kits run at 55-60°C. Hwinfo64 on Kingston RAM was showing RAM temp and additional info if RAM has too high temp or not. At 65°C it was still fine. ASUS BIOS shows RAM temp too.
Well, there is a warranty (if it's not a review sample ;) ) but I assume that in time, the serial number sticker will change color or fall off, and then good luck with the warranty ;)
 

Zantal

Member
Joined
Mar 19, 2009
Sounds like you got a good one or I'm doing something wrong.
If you want to see an infuriatingly good 12900k ->
He's doing 5.5 with core at 1.35 under load, we can expect 12900ks to perform like that

you should be ok with your cpu, alder lake cpus have highly varying degrees of quality across the chip
For example I can get avx512 stable at 5.2 (with bclk oc max multi for avx is fused 51) and avx2 will just crash no matter the voltage, so I just stopped trying
some people need 1.3/1.4v vccsa to get over 3800 in ram while I lucked out and need only 1.2 for 4k
 
Joined
Dec 13, 2005
Sounds like you got a good one or I'm doing something wrong.
If you're talking about the 12700 I just dropped in to ask about them. I'm hovering right around where you are (5.1 @ 1.255v in BIOS, 1.260v reported by software).

*I* might be doing something wrong though as I just did it quick and dirty whenever I had a minute free.
 

Woomack

Benching Team Leader
Joined
Jan 2, 2005
If you're talking about the 12700 I just dropped in to ask about them. I'm hovering right around where you are (5.1 @ 1.255v in BIOS, 1.260v reported by software).

*I* might be doing something wrong though as I just did it quick and dirty whenever I had a minute free.

The average seems 5.0-5.1GHz ~1.25V so you are fine. At auto, my 12900K goes up to 5.2GHz on pretty much every motherboard that I was testing, including Biostar B660. I wasn't trying max clock on my chip because of lack of time and I try to focus on other things. On the other hand, I was checking something as small as Noctua NH-L9i and I was able to run 12600K at 4.6-4.9GHz(most P cores at 4.9GHz) in games/3D tests and 4.3-4.5GHz in tests with AVX, so I'm pretty happy with the results on these CPUs.

If you want to see an infuriatingly good 12900k ->
He's doing 5.5 with core at 1.35 under load, we can expect 12900ks to perform like that

you should be ok with your cpu, alder lake cpus have highly varying degrees of quality across the chip
For example I can get avx512 stable at 5.2 (with bclk oc max multi for avx is fused 51) and avx2 will just crash no matter the voltage, so I just stopped trying
some people need 1.3/1.4v vccsa to get over 3800 in ram while I lucked out and need only 1.2 for 4k

Nice result, but 1.35-1.4V and ~65-80°C on cores max, under high load, is not a regular AIO cooler or a standard custom loop ;) Also, it's an ES and probably cherry-picked CPU as guys who test that stuff usually get more CPUs, You have to be very lucky to get something like that. It's also tested on Z690 Dark which wasn't released yet. I guess it's optimized for higher clocks. Would be nice to test this mobo but I won't pay for that and I don't think I get a review sample.
Somehow, I think we will see next-gen CPUs soon enough so 12900KS can be a waste of money. I wouldn't even care to get 12900K if it's not for competitive benchmarking and you actually have pretty good cooling.
 

Woomack

Benching Team Leader
Joined
Jan 2, 2005
New BIOS for Strix Z690-I Gaming has been released today so you can expect similar updates for all Strix/Maximus series motherboards (at least last updates were similar):

ROG STRIX Z690-I GAMING WIFI BIOS 1003
01. Improve system performance and Window 11 OS stability.
02. Update USB PD FW to 1.0F.
03. Update Intel ME FW to 16.0.15.1620 V3
04. Add Thunderbolt FW update method for onboard Thunderbolt models and ThunderboltEX 4 Card support models
05. Update Intel microcode.
06. Improve DRAM compatibility
07. Change PCIe speed hotkey item from F9 to F6.


I hope it improves DDR5 stability above 6600 and that the next DDR5 kit will have Hynix IC ...
I will test DDR4 on Z590 as everything Z690 is somehow disappointing so far and I'm not going to spend money on DDR4 mobo right now.
 
Last edited:

Zantal

Member
Joined
Mar 19, 2009
Yes I also got the v1011 for my Prime Z690-P D4 (beta bios) just to see if I could get CR1, still no go so I guess it won't ever be achievable on my board or these sticks in general, but it guesses auto voltages and trains RTL better than previous versions (had to manually lock them in at correct values as they would fluctuate to unstable settings from time to time).

Also regarding that incredible 12900k that guy has an extreme overkill custom loop and the sample he received was binned especially for him from EVGA I believe
still it shows what's possible when you get good manufactured chips, EUV is still quite new so imagine the improvements they will make and the insane chips we'll get in the coming years =3
 

Brando

Member
Joined
Jan 9, 2006
If you're talking about the 12700 I just dropped in to ask about them. I'm hovering right around where you are (5.1 @ 1.255v in BIOS, 1.260v reported by software).

*I* might be doing something wrong though as I just did it quick and dirty whenever I had a minute free.
Mine is running without e-cores so if you have yours enabled you're doing well imo.

 

Brando

Member
Joined
Jan 9, 2006
So... you've got ecores disabled? It's really worth neutering the chip for a hundred/2hundo mhz?
I'll probably flip flop on this to be honest. At the moment it seems like the e cores drop gaming performance a tiny bit which may be fixed with software at some point. Granted it's not enough to notice but combined with a slightly higher oc and reduced need for bandwidth without e cores it may be better enough to be worth it. Or not. The only intensive thing I do is gaming since I'm not a video editor or anything like that.
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
At the moment it seems like the e cores drop gaming performance a tiny bit which may be fixed with software at some point.
Are you still using W10? Otherwise, these CPUs (in default form) are a lot faster than the previous generation out of the box. I can see disabling HT, but not the e-cores...especially in W11 which is the only OS optimized 100% to use the p/e cores.
 

Brando

Member
Joined
Jan 9, 2006
Are you still using W10? Otherwise, these CPUs (in default form) are a lot faster than the previous generation out of the box. I can see disabling HT, but not the e-cores...especially in W11 which is the only OS optimized 100% to use the p/e cores.
I'm using win11. Supposedly all p cores is slightly better for what I personally use the pc for but probably not enough to worry about too much. Maybe I'll go back to stock with all cores at some point idk. Sometimes it seems to matter and sometimes not. I'm guessing performance will favor more cores as time goes by.
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
Looked like overall it was a negligible difference... and in the end said to leave the ecores enabled (11:17+)?
 

Woomack

Benching Team Leader
Joined
Jan 2, 2005
I still see no difference in most games above 6 cores but there are some titles that use whatever CPU has (just not scaling perfectly well above 6-8 threads). One example is the Civilization series that uses 24+ threads. Even looking at the test in the link, it looks like some games react better, some worse. There is also that core park/unpark software patch that most brands released, so you can check if it helps in older titles. If I'm right then problems with E-cores can happen more often in older games.

Anyway, someone mentioned bclk overclocking some days ago and yesterday I wanted to check how it's scaling on AL. It helps a bit in some tests but some others give weird results. On my setup with IGP I could go up to 175MHz without corrected PCIe errors (can check that on the bottom of the list in hwinfo64) and the max of about 190MHz. With discrete graphics, I could go up to about 230MHz but it was acting weird above 200MHz. Corrected errors = PC won't crash but can be close. Uncorrected = bluescreen or freeze.
In benchmarks like Geekbench, there are slightly better results. AIDA64 is clearly showing some weird results and it's hard to believe they're real. One example is below. Memory bandwidth went from 90-100GB/s to 155-170GB/s and latency went down from ~60ns to 33.5ns (and it's not even the highest bclk that can be set). I bet there will be some fake scores in rankings too and another witch hunt on hwbot.

6617.jpg