Alder Lake CPUs common discussion

[The King]

For those who are on the fence about DDR4 vs DDR5 ADL boards. This maybe useful.

 

[RandallFlagg]

Cripple? Not by a long shot. Anand didn't see much difference between DDR4-3200 and DDR5-4800 (those are the officially supported speeds, sans overclocking), except for highly multithreaded stuff.
In other words, in the absence of very intensive memory access, DDR4 already offers all the bandwidth you need.

I will tell ya, I have never been impressed with Cutress. He is so wrapped up in his spec marks, he just makes assumptions based on his theories and runs with them, never bothering to check if he is right or not.

So one of the reasons I like TPU is because they benchmark apps that, in my experience, *most* people actually use.

AnandTech does not. They make stupid assumptions that lead to comments like the one you just made.

For the record, millions of people use spreadsheets every single day. Many of these are highly complex, with hundreds of thousands of cells, dozens of graphs, images and scripts. I've known people who basically made a career out of knowing how to make stuff like that.

The below is not uncommon use cases. It's fairly normal for power users. And there a fuck ton load of people using apps like this, as opposed to 'using' specmarks and pov-ray.

Emperical facts. Tom's used DDR4-3200 and DDR5-4800, similar to AT.

You are looking at > 10% on a spreadsheet script due to DDR4 vs DDR5. This particular spreadsheet workload likes bandwidth :

Oh but wait, it can also be latency sensitive :

 

[Deleted member 24505]

For those who are on the fence about DDR4 vs DDR5 ADL boards. This maybe useful.

So there is possibly advantages to going DDR5, though might be worth waiting till next year when higher speed low latencie kits will be available. I would deffo go Z690/DDR5 now given the cash, sit on the slow sticks till faster ones become available to pop straight in.

 

[RandallFlagg]

So there is possibly advantages to going DDR5, though might be worth waiting till next year when higher speed low latencie kits will be available. I would deffo go Z690/DDR5 now given the cash, sit on the slow sticks till faster ones become available to pop straight in.

DDR5 out of stock now though. Pretty much everything I see is 4800 and 5200, but gone.

 

[GerKNG]

i have my hands on a 12600k + Aorus Pro DDR4.

am i the only one who has insane problems with this platform?
3200mb/s RAM (2x16GB) CL16 does not even POST.
the Bios that was flashed was from August and was not able to boot into anything (had to flash it before i could even install windows)

the whole PC is extremely unstable (memtest errors at 2933 Gear 1) only 3000 Gear 2 works somewhat okay.

the E Cores are not used in Benchmarks like Cinebench, i have to reset the bios every fifth boot...

Alder Lake is barely functional in this state.

 

[RandallFlagg]

i have my hands on a 12600k + Aorus Pro DDR4.

am i the only one who has insane problems with this platform?
3200mb/s RAM (2x16GB) CL16 does not even POST.
the Bios that was flashed was from August and was not able to boot into anything (had to flash it before i could even install windows)

the whole PC is extremely unstable (memtest errors at 2933 Gear 1) only 3000 Gear 2 works somewhat okay.

the E Cores are not used in Benchmarks like Cinebench, i have to reset the bios every fifth boot...

Alder Lake is barely functional in this state.

Suggest you check here, you'll find a lot of people testing ADL hands on. Too much AMD at TPU :

Overclocking ADL - 12900k etc results, bins and discussion

www.overclock.net

 

[Psychoholic]

I Plan on keeping my 12900K for a while, if need be in a year or two i'll switch over to a DDR5 board when the FAST ddr5 is out, I assume there will be pretty huge improvements in DDR5 over the course of a year or 18 months.

 

[RandallFlagg]

?

Edit: NVM, AIDA has a bclk related bug with AL

 

[tabascosauz]

am i the only one who has insane problems with this platform?
3200mb/s RAM (2x16GB) CL16 does not even POST.
the Bios that was flashed was from August and was not able to boot into anything (had to flash it before i could even install windows)

Gigabyte got hit by ransomware in August. Basically all the BIOSes on my B550I Aorus AX went to shit after F12 or so, in that time frame. Customer support was completely offline until like October.

Gigabyte apparently got hit again by the same group in late October. I have a soft spot for Gigabyte boards, but I'd stay away until they actually emerge from this hellhole. Don't expect their boards to be ready.

 

[bug]

I will tell ya, I have never been impressed with Cutress. He is so wrapped up in his spec marks, he just makes assumptions based on his theories and runs with them, never bothering to check if he is right or not.

So one of the reasons I like TPU is because they benchmark apps that, in my experience, *most* people actually use.

AnandTech does not. They make stupid assumptions that lead to comments like the one you just made.

For the record, millions of people use spreadsheets every single day. Many of these are highly complex, with hundreds of thousands of cells, dozens of graphs, images and scripts. I've known people who basically made a career out of knowing how to make stuff like that.

The below is not uncommon use cases. It's fairly normal for power users. And there a fuck ton load of people using apps like this, as opposed to 'using' specmarks and pov-ray.

Emperical facts. Tom's used DDR4-3200 and DDR5-4800, similar to AT.

You are looking at > 10% on a spreadsheet script due to DDR4 vs DDR5. This particular spreadsheet workload likes bandwidth :

View attachment 223863

Oh but wait, it can also be latency sensitive :

View attachment 223866

I fail to see the contradiction. Some workloads are impacted, some are not. What Anand says, is that most of them fall into the latter category. I'll believe that, until more reviews come out saying otherwise.

 

[birdie]

A nice compilation from 3dcenter.org

Games (CPU limit)	11600K	11700K	11900K	5600X	5800X	5900X	5950X	12600K	12700K	12900K
Cores & Topology	6C RKL	8C RKL	8C RKL	6C Zen3	8C Zen3	12C Zen3	16C Zen3	6C + 4c ADL	8C + 4c ADL	8C + 8c ADL
AnandTech (8T, 1080p 95th)			86.2%		89.3%	88.6%	87.9%			100%
CapFrameX (10T, 720p avg)			87.3%			89.9%		88.8%		100%
ComputerBase (9T, 720p avg)	78.9%		91.6%	87.4%	90.5%	93.7%	94.7%	90.5%	94.7%	100%
Eurogamer (8T, 1080p 5%)	67.8%		75.3%	75.9%			82.0%	89.0%		100%
Gamers Nexus (7T, 1080p 1%)	87.3%		93.8%	85.8%	90.4%	91.4%	91.4%			100%
Golem (8T, 720p P1)			87.0%			82.1%	84.6%			100%
Hardware luxx (4T, 720p avg)	86.5%	88.4%	91.4%	86.2%	88.6%	88.7%	88.5%	92.2%		100%
Igor's Lab (10T, 720p 99th)	76.9%	81.3%	88.4%	81.7%	87.3%	88.4%	88.1%	90.6%	95.0%	100%
Le Comptoir (11T, 1080p 1%)	72.8%	76.4%	79.9%	80.7%	85.0%	86.8%	87.9%	93.1%	97.0%	100%
Linus Tech Tips (6T, 1080p 99th)	81.8%		86.8%	85.7%		91.7%	91.4%	96.3%		100%
Notebookcheck (9T, 720p avg)	86.7%		92.3%	95.5%	98.9%	99.6%	95.4%	89.2%		100%
PC Games HW (20T, 720p avg)	75.2%		87.1%	80.0%	82.9%	87.4%	91.1%	88.8%		100%
PC world (12T, 720p avg)	80.1%		85.9%	87.7%			91.1%	91.8%		100%
SweClockers (5T, 720p 99th)	76.6%		85.9%	81.9%		86.9%	83.6%	90.3%		100%
TechPowerUp (10T, 720p avg)	81.2%	84.5%	86.6%	85.5%	89.4%	90.4%	89.6%	93.7%		100%
TechSpot (10T, 1080p 1%)			88.5%			94.3%	94.9%			100%
Tom's Hardware (6T, 1080p 99th)	85.2%	86.4%	92.3%	82.6%	83.9%	90.8%	86.4%	92.5%		100%
Averaged game performance	78.1%	82.3%	86.6%	83.4%	87.2%	89.3%	89.4%	91.5%	95.8%	100%
List price	$ 237	$ 374	$ 519	$ 299	$ 449	$ 549	$ 749	$ 264	$ 384	$ 564

Game perf.	vs 11600K	vs 11700K	vs 11900K	vs 5600X	vs 5800X	vs 5900X
Core i5-12600K	+ 17.2%	+ 11.2%	+ 5.7%	+ 9.8%	+ 5.0%	+ 2.5%
Core i7-12700K	+ 22.7%	+ 16.5%	+ 10.7%	+ 15.0%	+ 9.9%	+ 7.3%
Core i9-12900K	+ 28.1%	+ 21.5%	+ 15.5%	+ 19.9%	+ 14.7%	+ 12.0%

 

[Operandi]

Intel has pushed their P-cores for MT scenarios to extremes to be the absolute performance king and at least be faster than 5900X and rival 5950X in many cases. This results in an insane power consumption out of the box but only for heavy MT tasks, e.g. video encoding, rendering, software compilation, math calculations - not something average people do daily.

Is it jut me or is there something about the way Intel does things that allow their CPUs to scale frequency at the extreme ends of power scale better than AMD CPUs. You can manually push all core overclocks on AMD CPUs and power goes up but you get very little from it where as Intel CPUs yeah the power consumption becomes pretty impractical and more than a bit of a problem but you get performance that seems to scale better. Not sure if this has been discussed elsewhere but its an interesting observation.

If AMD tweaked the Zen 3 core to scale to ADL levels of power they'd be pretty much at parity?

 

[RandallFlagg]

A nice compilation from 3dcenter.org

Games (CPU limit)	11600K	11700K	11900K	5600X	5800X	5900X	5950X	12600K	12700K	12900K
Cores & Topology	6C RKL	8C RKL	8C RKL	6C Zen3	8C Zen3	12C Zen3	16C Zen3	6C + 4c ADL	8C + 4c ADL	8C + 8c ADL
AnandTech (8T, 1080p 95th)			86.2%		89.3%	88.6%	87.9%			100%
CapFrameX (10T, 720p avg)			87.3%			89.9%		88.8%		100%
ComputerBase (9T, 720p avg)	78.9%		91.6%	87.4%	90.5%	93.7%	94.7%	90.5%	94.7%	100%
Eurogamer (8T, 1080p 5%)	67.8%		75.3%	75.9%			82.0%	89.0%		100%
Gamers Nexus (7T, 1080p 1%)	87.3%		93.8%	85.8%	90.4%	91.4%	91.4%			100%
Golem (8T, 720p P1)			87.0%			82.1%	84.6%			100%
Hardware luxx (4T, 720p avg)	86.5%	88.4%	91.4%	86.2%	88.6%	88.7%	88.5%	92.2%		100%
Igor's Lab (10T, 720p 99th)	76.9%	81.3%	88.4%	81.7%	87.3%	88.4%	88.1%	90.6%	95.0%	100%
Le Comptoir (11T, 1080p 1%)	72.8%	76.4%	79.9%	80.7%	85.0%	86.8%	87.9%	93.1%	97.0%	100%
Linus Tech Tips (6T, 1080p 99th)	81.8%		86.8%	85.7%		91.7%	91.4%	96.3%		100%
Notebookcheck (9T, 720p avg)	86.7%		92.3%	95.5%	98.9%	99.6%	95.4%	89.2%		100%
PC Games HW (20T, 720p avg)	75.2%		87.1%	80.0%	82.9%	87.4%	91.1%	88.8%		100%
PC world (12T, 720p avg)	80.1%		85.9%	87.7%			91.1%	91.8%		100%
SweClockers (5T, 720p 99th)	76.6%		85.9%	81.9%		86.9%	83.6%	90.3%		100%
TechPowerUp (10T, 720p avg)	81.2%	84.5%	86.6%	85.5%	89.4%	90.4%	89.6%	93.7%		100%
TechSpot (10T, 1080p 1%)			88.5%			94.3%	94.9%			100%
Tom's Hardware (6T, 1080p 99th)	85.2%	86.4%	92.3%	82.6%	83.9%	90.8%	86.4%	92.5%		100%
Averaged game performance	78.1%	82.3%	86.6%	83.4%	87.2%	89.3%	89.4%	91.5%	95.8%	100%
List price	$ 237	$ 374	$ 519	$ 299	$ 449	$ 549	$ 749	$ 264	$ 384	$ 564

Game perf.	vs 11600K	vs 11700K	vs 11900K	vs 5600X	vs 5800X	vs 5900X
Core i5-12600K	+ 17.2%	+ 11.2%	+ 5.7%	+ 9.8%	+ 5.0%	+ 2.5%
Core i7-12700K	+ 22.7%	+ 16.5%	+ 10.7%	+ 15.0%	+ 9.9%	+ 7.3%
Core i9-12900K	+ 28.1%	+ 21.5%	+ 15.5%	+ 19.9%	+ 14.7%	+ 12.0%

That 12600K is just amazing. I don't recall ever seeing a release from either AMD or Intel where their mid tier performance chip defeated the top tier performance of the last generation CPU in gaming overall.

 

[GerKNG]

Gigabyte got hit by ransomware in August. Basically all the BIOSes on my B550I Aorus AX went to shit after F12 or so, in that time frame. Customer support was completely offline until like October.

Gigabyte apparently got hit again by the same group in late October. I have a soft spot for Gigabyte boards, but I'd stay away until they actually emerge from this hellhole. Don't expect their boards to be ready.

do you really think all of these issues are because of the gigabyte board?
because it feels like i am testing alder lake the day after they made the first internal silicon

 

[bug]

Is it jut me or is there something about the way Intel does things that allow their CPUs to scale frequency at the extreme ends of power scale better than AMD CPUs. You can manually push all core overclocks on AMD CPUs and power goes up but you get very little from it where as Intel CPUs yeah the power consumption becomes pretty impractical and more than a bit of a problem but you get performance that seems to scale better. Not sure if this has been discussed elsewhere but its an interesting observation.

If AMD tweaked the Zen 3 core to scale to ADL levels of power they'd be pretty much at parity?

Intel lets these unlocked CPUs to use max power all the time (as opposed to say, ~1 minute, which was the norm so far). It basically allows the CPU to overclock itself indefinitely, as long the cooling can keep up. It's an easy way to extract every little bit of juice out of the box, with the downside that effective performance will now vary based on the cooling solution.

 

[PolRoger]

i have my hands on a 12600k + Aorus Pro DDR4.

am i the only one who has insane problems with this platform?
3200mb/s RAM (2x16GB) CL16 does not even POST.
the Bios that was flashed was from August and was not able to boot into anything (had to flash it before i could even install windows)

the whole PC is extremely unstable (memtest errors at 2933 Gear 1) only 3000 Gear 2 works somewhat okay.

the E Cores are not used in Benchmarks like Cinebench, i have to reset the bios every fifth boot...

Alder Lake is barely functional in this state.

I picked up a Z690i Aorus Ultra (DDR4) and i5-12600K combo yesterday and I'm not having the type of stability issues that you are describing.

My board also came with an August F1 BIOS which I did start out with but I've now updated to the launch release F3 BIOS.

I'm testing with W10. I had some quirks/issues (my LAN cut out) but getting the correct drivers via the Gigabyte Launch App helped to clear that up. I'm currently just testing with default auto settings and XMP enabled (using a 2x8GB 1R 3200C14 kit).

I have found one particular issue while running World Community Grid... Where the program only seems to run/load the E-cores while the P-cores all remain idle??

However Asus RealBench and Prime95 will both run/stress with all cores fully loaded. I haven't tested/run Cinebench or AIDA64 yet.

 

[R-T-B]

Gigabyte apparently got hit again by the same group in late October.

OT a bit, but do you have any supporting evidence for that part of your claim? Feel free to PM.

 

[Operandi]

Intel lets these unlocked CPUs to use max power all the time (as opposed to say, ~1 minute, which was the norm so far). It basically allows the CPU to overclock itself indefinitely, as long the cooling can keep up. It's an easy way to extract every little bit of juice out of the box, with the downside that effective performance will now vary based on the cooling solution.

Yes I am aware of that. My point is there seems to be something different about how Intel CPUs scale with performance relative to power. AMD CPUs don't seem to have that top end ability to push that little bit of extra performance at the cost of extreme power usage. You can do it, cooling isn't reallly a problem but you get almost nothing in return, with Intel's designs of late you get something for that insane power consumption.

 

[birdie]

Who are those GODS?

The best 5950X result under Windows: 2104/23600. 12900K is 33% faster in ST. That's impossible.

12900K results sorted by ST and MT.
5950X results sorted by ST and MT. (MacOS results are higher and probably shouldn't be taken into consideration as this OS has a different scheduler and the app is compiled slightly differently).

 

[GerKNG]

I picked up a Z690i Aorus Ultra (DDR4) and i5-12600K combo yesterday and I'm not having the type of stability issues that you are describing.

My board also came with an August F1 BIOS which I did start out with but I've now updated to the launch release F3 BIOS.

I'm testing with W10. I had some quirks/issues (my LAN cut out) but getting the correct drivers via the Gigabyte Launch App helped to clear that up. I'm currently just testing with default auto settings and XMP enabled (using a 2x8GB 1R 3200C14 kit).

I have found one particular issue while running World Community Grid... Where the program only seems to run/load the E-cores while the P-cores all remain idle??

However Asus RealBench and Prime95 will both run/stress with all cores fully loaded. I haven't tested/run Cinebench or AIDA64 yet.

other benchmarks and applications run fine.
just cinebench R15 and R20 (sometimes!) drops out the E Cores.

btw. did you checked your Core Temperatures in HWInfo?
i have two cores constantly sitting way below ambient and dropping to 0°C (maybe it's just a HWInfo bug)

 

[tabascosauz]

do you really think all of these issues are because of the gigabyte board?
because it feels like i am testing alder lake the day after they made the first internal silicon

There's a million things it could be, and ADL owners' sample size isn't nearly big enough, but most things come back to the board. Remember how Rocket Lake had a launch day microcode patch to improve performance?

Are you still on F1 BIOS or something? Whichever it shipped with is not available on the website, the website only lists F4 from November 1st as the initial release.

The 0 degree reading is weird though. Is your HWInfo up to date? Couple of versions of HWInfo now, all with improvements for Alder Lake reporting. Possibly bad CPU (very slim chance though), but firmware also has a hand in just about everything these days.

 

[PolRoger]

other benchmarks and applications run fine.
just cinebench R15 and R20 (sometimes!) drops out the E Cores.

btw. did you checked your Core Temperatures in HWInfo?
i have two cores constantly sitting way below ambient and dropping to 0°C (maybe it's just a HWInfo bug)

I have noticed some low temps being reported sometimes? But usually only when a core is in an idle state...

With the new F3 BIOS I've been able to disable the E-Cores so now WCG will run fully loaded @12 threads. My system Kill-A Watt is showing a reading of ~191/192 watts for this 6C/!2T load.

Cooling is open air with an old (4 heat pipe) Noctua NH-U12P with Scythe S-Flex fans in push/pull config.

 

[GerKNG]

Are you still on F1 BIOS or something?

i was not even able to boot to anything (NVME drives were not even supported) with the bios F2.

i flashed F4 (the only available one) and only after that i was able to install windows.

I have noticed some low temps being reported sometimes? But usually only when a core is in an idle state...

exactly this.
at idle i see below ambient on 2-3 cores and they occasionally drop to 0°C.

 

[Deleted member 202104]

I had a 12700k and DDR4 board in my cart at Newegg this morning to try out. It was going to be a few weeks to get the LGA1700 mounting adapters to fit any of my coolers. Gave up and bought a set of fast RAM instead for the main rig. I usually want it now or not at all when it comes to that type of purchase.

I read Asus mobos have mounting holes that line up with previous gen coolers, I hope I can use my current cooler.

I need an Adult.

 

[Deleted member 215115]

That 12600K is just amazing. I don't recall ever seeing a release from either AMD or Intel where their mid tier performance chip defeated the top tier performance of the last generation CPU in gaming overall.

5600X vs 3950X

3600 vs 2700X

Maybe 9700K vs 8700K & 8600K vs 7700K