Discussion Alder Lake - Builders Thread

[AdamK47]

This thread is for those of us that own or are looking to get an Alder Lake CPU.

Bought mine on release day. 12900K. Memory selection was limited to only Crucial DDR5 4800 at the Micro Center I went to. Bought two 2 x 8GB kits. Motherboard is a Gigabyte Z690 Aorus Master. I did buy a 360mm Lian Li Galahad 360mm AIO. Ended up returning it since the Lian Li Socket 1700 backplate that I bought separately worked well with my NZXT Kraken X73.

 

[coercitiv]

In the BIOS on my ASUS board when you hover over the XMP settings the info pop up that appears at the bottom of the screen says that XMP I are the memory timings tuned by ASUS.

Here's the specs to your memory kit, luckily Kingston does this right and keeps proper spec sheets.

Are the profiles you see in UEFI the same as the ones bellow provided by Kingston themselves?


And here's the validation report from Asus themselves for this kit:


I would be very curious to know why Asus would include the option to run more relaxed timings after validating the vendor XMP profile.

 

[Hulk]

See the screen shots below. I didn't notice that XMP II was at 3000, not 3600. But notice the note saying "settings were optimized by ASUS" for XMP I. What do you make of that?

 

[coercitiv]

See the screen shots below. I didn't notice that XMP II was at 3000, not 3600. But notice the note saying "settings were optimized by ASUS" for XMP I. What do you make of that?

I can't make anything of that, the language ASUS uses is far too vague. "Other memory parameters" could mean anything to me, even VCCSA & VCCIO voltages (or whatever their name is with ADL-S). I'm also serioulsy wondering why ASUS mentions changing BCLK with XMP 1.

 

[Hulk]

I can't make anything of that, the language ASUS uses is far too vague. "Other memory parameters" could mean anything to me, even VCCSA & VCCIO voltages (or whatever their name is with ADL-S). I'm also serioulsy wondering why ASUS mentions changing BCLK with XMP 1.

Yes, it's kind of cryptic. I'm sure it's translated from Chinese and the translation might not be the best.

But in general RAM qualified for a particular motherboard should mean it's been tested and works at vendor specified parameter, correct?

 

[coercitiv]

But in general RAM qualified for a particular motherboard should mean it's been tested and works at vendor specified parameter, correct?

Yup, that's the entire purpose of the process.

 

[Ed1]

I put together a 12600k last week, parts list.

1)12600k, Tuf Z690 Plus wifi D4, 16g crucial Ballistix 3600@c16, RM750x PSU, Lancool II mesh performance.
Cooled with NH-U14S

No issues and all work well.

Ed

 

[JoeRambo]

One thing that impressed me big time with Alder Lake is just how power efficient it is. I am running it with fixed high perf plan, fixed volts, 5Ghz 4.7Ghz Uncore, E-Cores disabled and getting idle power consumption numbers i have never seen on desktop before.


Incredible chip, going from 0.565W to 250W on full load. Except it is idling most of the time obviously.
Happy owner of 12900K + MSI EDGE Z690 + DDR4 3600C14 handtuned cooled by X63 Kraken.

EASILY the snappiest system i have ever owned, 5950x with handtuned memory can't get anywhere near.

 

[Markfw]

One thing that impressed me big time with Alder Lake is just how power efficient it is. I am running it with fixed high perf plan, fixed volts, 5Ghz 4.7Ghz Uncore, E-Cores disabled and getting idle power consumption numbers i have never seen on desktop before.
View attachment 53462

Incredible chip, going from 0.565W to 250W on full load. Except it is idling most of the time obviously.
Happy owner of 12900K + MSI EDGE Z690 + DDR4 3600C14 handtuned cooled by X63 Kraken.

EASILY the snappiest system i have ever owned, 5950x with handtuned memory can't get anywhere near.

Have you owned a 5950x ? that statement can not be true when it wins most 5950x benchmarks, except for a few single threaded ones.

 

[JoeRambo]

Have you owned a 5950x ? that statement can not be true when it wins most 5950x benchmarks, except for a few single threaded ones.

I do, wrote about it here and praised the efficiency and performance of that chip in high MT load scenarios:

Page 45 - AMD Ryzen 5000 Builders Thread

Page 45 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

What i did not praise it, was using it as personal desktop machine - It just felt ~smooth as my static 5.1 10900K with 3900C15 memory.
12900K is on another level, frankly Anandtech and quite some other reviews undersell it due to memory configs they are using, nothing currently can touch ADL coupled with tuned DDR4.

P.S. And on topic of 5950x => here is power usage from completely idle 5950x 4.4Ghz fixed. Of course it is not stock ( but neither is 12900K, locked clocks, voltage, uncore, memory OC ):

 

[Markfw]

I do, wrote about it here and praised the efficiency and performance of that chip in high MT load scenarios:

Page 45 - AMD Ryzen 5000 Builders Thread

Page 45 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

What i did not praise it, was using it as personal desktop machine - It just felt ~smooth as my static 5.1 10900K with 3900C15 memory.
12900K is on another level, frankly Anandtech and quite some other reviews undersell it due to memory configs they are using, nothing currently can touch ADL coupled with tuned DDR4.

With all the benchmarks, saying "5950x with handtuned memory can't get anywhere near. " is a wild exaggeration. Saying its better is fine, but don't stretch the truth like that.

 

[OlyAR15]

Can't wait for my parts to arrive for my new game machine. 12700KF, Gigabyte Z690i mITX board, XPG Gammix 2TB ssd, Silverstone 800W sfx-L psu. Re-using my RTX3090 and 32GB DDR4 ram. EK waterblocks for the cpu and gpu, and building a custom case to fit it all.

 

[JoeRambo]

With all the benchmarks, saying "5950x with handtuned memory can't get anywhere near. " is a wild exaggeration. Saying its better is fine, but don't stretch the truth like that.

Can we agree to disagree Mark? The thing is, i think 99% of what gets done on desktop machine is web browsing and gaming and ADL when coupled with proper memory is very strong.

On exactly the same browser i am getting 240 vs 300 in Speedometer 2.0 web browser test, that is substantial advantage for me. And gaming advantages are all over benchmarks.

https://www.capframex.com/tests/Alder Lake with fast DDR5 against Ryzen 5000

Very substantial gap in CPU performance.

 

[Markfw]

Can we agree to disagree Mark? The thing is, i think 99% of what gets done on desktop machine is web browsing and gaming and ADL when coupled with proper memory is very strong.

On exactly the same browser i am getting 240 vs 300 in Speedometer 2.0 web browser test, that is substantial advantage for me. And gaming advantages are all over benchmarks.

https://www.capframex.com/tests/Alder Lake with fast DDR5 against Ryzen 5000

Very substantial gap in CPU performance.

In the first place thats stated much better than the first time. But you are stating the gaming and single thread are better. that we all know, just not as much as you stated the first time.

 

[JoeRambo]

Hard for me as non native english speaker to quantify "can't get anywhere near".
But i feel 20% advantage is large enough to feel everywhere. And 5950x can't close that gap in snappyness, no amount of tuning will narrow the gap.

 

[Ed1]

One thing that impressed me big time with Alder Lake is just how power efficient it is. I am running it with fixed high perf plan, fixed volts, 5Ghz 4.7Ghz Uncore, E-Cores disabled and getting idle power consumption numbers i have never seen on desktop before.
View attachment 53462

Incredible chip, going from 0.565W to 250W on full load. Except it is idling most of the time obviously.
Happy owner of 12900K + MSI EDGE Z690 + DDR4 3600C14 handtuned cooled by X63 Kraken.

EASILY the snappiest system i have ever owned, 5950x with handtuned memory can't get anywhere near.

One thing I was not expecting coming from a 3570k@4.5 was how good the temps are in Prime95 and CB xx.

My old system with the same cooler would be in mid 60's C in prime95, this system I have not passed 60c yet, maybe if I let it run prime many hr's it might pass that but if so not by much.

Another thing is power used under max load, ASUS by default has power limits set to optimized which puts it with 4k limits.
Under prime95 and CB20 I hit around 115w package, if I set it to Intel power limits (150w) there no difference.

 

[Markfw]

Hard for me as non native english speaker to quantify "can't get anywhere near".
But i feel 20% advantage is large enough to feel everywhere. And 5950x can't close that gap in snappyness, no amount of tuning will narrow the gap.

There are no single threaded benchmarks that I remember being that large of a different, but still, "can't get anywhere near" would be like 50% difference. I think you want to say "noticeably faster". That would reflect about what the benchmarks say in single thread.

 

[SlimFan]

Have you owned a 5950x ? that statement can not be true when it wins most 5950x benchmarks, except for a few single threaded ones.

Why is this response appropriate for an Alder Lake builder's thread?

 

[Markfw]

Why is this response appropriate for an Alder Lake builder's thread?

Because the poster was comparing his new Alder lake CPU to his 5950x. Can you not read ?

 

[SlimFan]

Because the poster was comparing his new Alder lake CPU to his 5950x. Can you not read ?

I believe I can read, which is why I was able to read the words you wrote as well as cobble together letters into a form that enabled me to put my scrawl onto your screen in such a way that you were able to understand my question.

Here's what he said in terms of the 5950x:

EASILY the snappiest system i have ever owned, 5950x with handtuned memory can't get anywhere near.

I did not see anything in your response (again, in the Alder Lake builder's thread) that clarified why his impression of "snappiest system" was in error based on your experience with your 5950x. You later gave your own opinion on what you might view as the definition of "snappiest system" (+50%), the poster clearly has a different opinion (as does, IMHO, the rest of the industry/users). Even your response seems to belay your own personal perspective "except for a few single threaded ones" which some might argue is a legitimate proxy for "snappiness."

 

[Markfw]

I believe I can read, which is why I was able to read the words you wrote as well as cobble together letters into a form that enabled me to put my scrawl onto your screen in such a way that you were able to understand my question.

Here's what he said in terms of the 5950x:



I did not see anything in your response (again, in the Alder Lake builder's thread) that clarified why his impression of "snappiest system" was in error based on your experience with your 5950x. You later gave your own opinion on what you might view as the definition of "snappiest system" (+50%), the poster clearly has a different opinion (as does, IMHO, the rest of the industry/users). Even your response seems to belay your own personal perspective "except for a few single threaded ones" which some might argue is a legitimate proxy for "snappiness."

My response was mostly based on reviews. The 12900k wins in gaming and single threaded apps in varying amounts usually in single digit %. That is hardly ""can't get anywhere near". That opinion(THE WORDING) borders on trolling. Then he admitted that his english was not that good, so then I suggested better wording for very small % differences.

As for the 50%, that was a suggestion for how much faster something would have to be to qualify for ""can't get anywhere near".

 

[DrMrLordX]

@JoeRambo

What do you attribute the "snappiness" of the system to on your 12900k? The memory/memory controller or something else? If Intel can replicate that experience on the 12400 then it will be a nice QoL upgrade for a lot of people on budget machines.

 

[coercitiv]

What do you attribute the "snappiness" of the system to on your 12900k? The memory/memory controller or something else?

He disabled the E-cores, the system is running excelent low latency memory and 12900K has the biggest cache in the ADL-S lineup. Everything is aimed at feeding the big cores as quick as possible.

The 12400 will offer a big part of this experience, but it will definitely not feel the same: lower ST clocks, lower uncore clocks, considerably smaller L3 cache, slower memory. Value

 

[coercitiv]

Incredible chip, going from 0.565W to 250W on full load. Except it is idling most of the time obviously.

Is the system running with a dGPU installed?

 

[JoeRambo]

What do you attribute the "snappiness" of the system to on your 12900k? The memory/memory controller or something else? If Intel can replicate that experience on the 12400 then it will be a nice QoL upgrade for a lot of people on budget machines.

Its as @coercitiv already spelled out. Disabled E-Cores to avoid scheduler bs and increase uncore clocks to shave some ms from each L3 access and memory latency. Combined with overclocked and tuned memory:


Not that "tight" but most i can currently do with dual rank memory on this MB and BIOS, others are hitting 4000ish on Gear1.

+ My standard tuning -> locking Core clocks to 5Ghz and uncore to 4.7Ghz, so they don't downclock and waste time increasing clocks when load comes.

12400F will be great value, but memory tuning will be as important. Hopefully Intel will keep memory OC on cheaper chipsets.

Is the system running with a dGPU installed?

Yeah, reused my 3090.

 

[Ed1]

++++++++++++++++++++++++++++++++++++++++++++++++++++++
Its as @coercitiv already spelled out. Disabled E-Cores to avoid scheduler bs and increase uncore clocks to shave some ms from each L3 access and memory latency. Combined with overclocked and tuned memory:
View attachment 53483

Not that "tight" but most i can currently do with dual rank memory on this MB and BIOS, others are hitting 4000ish on Gear1.

+ My standard tuning -> locking Core clocks to 5Ghz and uncore to 4.7Ghz, so they don't downclock and waste time increasing clocks when load comes.

12400F will be great value, but memory tuning will be as important. Hopefully Intel will keep memory OC on cheaper chipsets.



Yeah, reused my 3090.

when you set uncore to fixed value (in your case 4.7) do you also set ring/LLC clock the same, they both seem to have matching values stock.

PS: My ASUS MB doesn't idle no were as low as yours, on balanced power plan it about 10w package with default bios values but Intel limits (that doesn't change a thing). I see the lowest downclocking to 500mhz on P cores/400 on E cores. That's with a listed Vcore min of 0.231v which is really low.