Question Alder Lake - Official Thread

[Hulk]

With the release of Alder Lake less than a week away and the "Lakes" thread having turned into a nightmare to navigate I thought it might be a good time to start a discussion thread solely for Alder Lake.

 

[Exist50]

Context is indeed a thing, so please do go back and read his entire post in which he is addressing @epsilon84 's remark while making comments about his needs with DC as well as gaming and general computing.

The man literally writes about the following:

• P cores are better than Zen3 in terms of performance
• ADL has good pricing (if you want value, you can find it)
• ADL is wining in gaming, may not maintain the lead after 5800X3D
• hybrid not valuable for DC right now
• hybrid may make sense for other users & other uses
• he encourages people to try the product for themselves

Seems to me like you read the posts leading to that reply and decided to skip his conclusions.

Right before that he was insisting that Alder Lake consumed much more power to someone who said they were a gamer. And that's ignoring his history of commentary on the subject.

 

[DrMrLordX]

You just gave me a wonderful idea! Markfw needs to get a Mac Studio and run PrimeGrid on that. It will beat everything he has in power efficiency and he will be one happy dude!

Kind of off-topic, but there is an M1 thread. It would be entirely appropriate to ask someone who will be buying a Mac Studio there to run something like PrimeGrid.

 

[coercitiv]

Right before that he was insisting that Alder Lake consumed much more power to someone who said they were a gamer. And that's ignoring his history of commentary on the subject.

So grill him on that, or on his history of commentary. Why choose to accuse him of something else? @Markfw certainly has a very coarse view on modern computing hardware, definitely not my cup of tea in terms of priorities or patience with new platforms, but on the one topic @Zucker2k decided to challenge him yesterday... he had already spoken in a fairly balanced post.

 

[Markfw]

OK, naysayers... Here is something positive about ADL, that appears to be a weakness in Zen 3. The ADL has all available (since the 4 e-cores are disabled) cores/threads being used, and give or take individual units, is equal to the Zens 3 with less than all cores loaded. Once all cores are loaded, the Zen3 time going from 2:10 or 2:20 up to over 3 hours ! (but linux on one and windows on the other, so not apples to apples) Heat ? maybe. 142 watt max used up ? maybe. I don't have all the information, but again, I repeat, Alder lake P-core is a VERY strong core. Too bad there are not more of them.

Edit: and this information was in the pic I posted above (post 2371) but nobody except me noticed it. Nobody appears to want all the facts, except me.

 

[igor_kavinski]

Edit: and this information was in the pic I posted above (post 2371) but nobody except me noticed it. Nobody appears to want all the facts, except me.

Would you be willing to disable hyperthreading on both 5950X and 12700F and only 8 cores enabled on the 5950X and do the test?

 

[Markfw]

Would you be willing to disable hyperthreading on both 5950X and 12700F and only 8 cores enabled on the 5950X and do the test?

At the moment, our Team is having a primegrid event, so at this time, no. But maybe after its over.

 

[Exist50]

So grill him on that, or on his history of commentary.

That's exactly what was happening.

 

[igor_kavinski]

Kind of off-topic, but there is an M1 thread. It would be entirely appropriate to ask someone who will be buying a Mac Studio there to run something like PrimeGrid.

I once asked someone on these forums for a benchmark on their recently purchased MBP Pro. Got no reply. I think Mac users are too arrogant to fulfill plebian requests.

 

[Tech Junky]

Edit @Shmee , linux I don't think is fixed, win 10 I don't think is fixed, and windows 11 is questionable, opinions all over the place. Depends on load IMO.

Windows 11 no longer the fastest OS for Alder Lake: Linux 5.16 on Core i9-12900K emerges winner in most benchmarks

The latest Linux kernel 5.16 update brings scheduling and hybrid handling improvements that give Linux a significant advantage over Windows 11 in Alder Lake systems. The current Ubuntu 22.04 LTS uses the 5.15 kernel and suffers from performance regressions on Alder Lake as a result.

www.notebookcheck.net

https://www.tomshardware.com/news/linux-snatches-alder-lake-productivity-crown-from-windows-11

Or... contrasting view indicating 5.18 is the bar to reach on performance under Linux.

https://www.tomshardware.com/news/intel-thread-director-coming-to-linux-5-18

I just installed 5.17 and it seems to be performing well but, then again it's only been installed for 24 hours. Some of the prior versions though recently would show high CPU utilization or IOwait messages in glances for processes but, none so far on 5.17.

Index of /mainline

 

[igor_kavinski]

I just installed 5.17 and it seems to be performing well but, then again it's only been installed for 24 hours.

Is it easy to update the kernel (a command or two?) or more involved?

 

[Hitman928]

Is it easy to update the kernel (a command or two?) or more involved?

It's not that difficult but most distributions are designed around and validated against specific kernel versions and big distributions many times will have their own customized kernels that they use. So if you go update to the latest kernel manually, you could potentially break some things in the distribution or lose some features. It's usually not a big issue, but I've seen some weird stuff happen.

 

[Tech Junky]

Is it easy to update the kernel (a command or two?) or more involved?

It's easy t upgrade. DL the deb's and then "sudo dpkg -i *.deb" and it will install. Once that's complete remove the older kernel "sudo apt purge linux-*" and it should pick up 3-4 files to remove. Reboot.

If it has an issue then roll back to the prior version with a USB boot and chroot into the system and apply the older kernel version / remove the newer one and reboot.

I've seen some weird stuff happen.

The weirdest thing I've had issues with are the NIC not loading properly and killing the boot process / revert to prior kernel to resolve.

I have a RTL NIC also that doesn't load the correct firmware w/o blacklisting the version it attempts to use to get the newer version to load / bring up the port.

In general though the kernel upgrades aren't that big of an issue if you have a backup plan to revert to the prior version. When I had my first issue with the kernel process it took a little bit of time to get things working again but, now it's a couple of minutes and reboots. I keep a txt file in the root with the prior version kernel and the txt file has the commands I need to use. Copy / paste / reboot.

sudo mount /dev/nvme0n1p1 /mnt

sudo su
mount --bind /dev /mnt/dev
mount --bind /proc /mnt/proc
mount --bind /sys /mnt/sys

chroot /mnt

dpkg -i *.deb #older kernel

apt purge linux-<version>

 

[Abwx]

i9-12900 review at computerbase :

Intel Core i9-12900 im Test

Der Intel Core i9-12900 sieht den Betrieb von 16 Kernen und 24 Threads mit einer TDP von 65 bis 202 Watt vor. Die ganze Bandbreite im Test.

www.computerbase.de

 

[JoeRambo]

i9-12900 review at computerbase :

Kinda unexpected how CPU is choosing to clock when under PL1 limit of 65W: 2.9Ghz for big cores and 2.5Ghz for small cores.

 

[coercitiv]

Kinda unexpected how CPU is choosing to clock when under PL1 limit of 65W: 2.9Ghz for big cores and 2.5Ghz for small cores.

It's aiming for multipliers with roughly the same VID on the cores. For example, in the 12700K V/f curve from SkaterBencher, big cores @ 2.9Ghz and small cores @ 2.5Ghz have the same 0.86V voltage.

 

[Abwx]

Kinda unexpected how CPU is choosing to clock when under PL1 limit of 65W: 2.9Ghz for big cores and 2.5Ghz for small cores.

Frequencies are close enough such that there s about no overvoltage for neither cores, hence efficency is optimal for the design, wich is not the case with the top of the line i9s.

 

[psolord]

So...um....any cheap (or not so cheap) DDR4 mobos that can do bclk overclocking yet?

 

[uzzi38]

So...um....any cheap (or not so cheap) DDR4 mobos that can do bclk overclocking yet?

Not yet. There is a new version of a Gigabyte board spotted this morning, but no clue if it supports it (yet)

 

[psolord]

Not yet. There is a new version of a Gigabyte board spotted this morning, but no clue if it supports it (yet)

Thanks.

Something tells me, that after these new cheap Ryzens appear in a few days, there will be suddenly some DDR4 mobos with external clock generators appearing!

 

[epsilon84]

Thanks.

Something tells me, that after these new cheap Ryzens appear in a few days, there will be suddenly some DDR4 mobos with external clock generators appearing!

I dunno, I wouldn't hold my breath on that one! Even if they are DDR4 mobos, they will most likely be mid range at the very least, which probably means $150+ even for B660 mobos, realistically $200 for those with decent enough VRMs that you would actually wanna overclock with.

Don't get me wrong, if an affordable DDR4 clock gen mobo became available I would be all over it, just to play around with overclocking the cheap SKUs without having to worry about crazy wattages like I would with my 12900K!

 

[coercitiv]

just to play around with overclocking the cheap SKUs without having to worry about crazy wattages like I would with my 12900K!

Disable cores, if anything the 12900K is far safer to OC than any cheap bin ever will. The 12900K simply sips voltage when compared to 12400 with BCLK OC.

 

[Markfw]

I have a new issue, maybe someone can help explain this. I was running only the P-cores and 100% load at 162 watts (from the wall), I changed video cards and the power went up to 313 watts. So I put the old card back in, and it was still 313 watts. So I enabled the e-cores and blew away, and installed the same version of linux and kernal (20.3 and 5.13-37). and its still 313 watts ! But now the job ETA has dropped from 2 hours 20 minutes to 2 hours and 2 or 7 minutes ! Thats crazy. e-cores are supposed to only take about 30 watts for all 4 as I understand it. I will wait a while, then go back to p-cores only and see what happens, but I have nothing running on the video card, and the same card was getting me 162 watts with no e-cores.

 

[igor_kavinski]

I have a new issue, maybe someone can help explain this. I was running only the P-cores and 100% load at 162 watts (from the wall), I changed video cards and the power went up to 313 watts.

It was a Radeon 590 as you wrote in your other post. An RX 580 destroyed my Intel mobo's PCIe bus by pulling too much power. Worst case scenario, your mobo's power circuitry got busted somehow. But I really hope this isn't the case. By the way, which brand is the 590?

 

[dullard]

I have a new issue, maybe someone can help explain this. I was running only the P-cores and 100% load at 162 watts (from the wall), I changed video cards and the power went up to 313 watts. So I put the old card back in, and it was still 313 watts. So I enabled the e-cores and blew away, and installed the same version of linux and kernal (20.3 and 5.13-37). and its still 313 watts ! But now the job ETA has dropped from 2 hours 20 minutes to 2 hours and 2 or 7 minutes ! Thats crazy. e-cores are supposed to only take about 30 watts for all 4 as I understand it. I will wait a while, then go back to p-cores only and see what happens, but I have nothing running on the video card, and the same card was getting me 162 watts with no e-cores.

Are those (A) 24 hour power averages or (B) instantaneous values that you just happened to look at? If it is (B), then you'll drive yourself nutty trying to pin down an ever changing result.

 

[Markfw]

Are those (A) 24 hour power averages or (B) instantaneous values that you just happened to look at? If it is (B), then you'll drive yourself nutty trying to pin down an ever changing result.

Kill-a-watt. It changes constantly, so real time. Now after changing to the equivalent of onboard video its doing 78 watts ! But 6 hour ETA. I disabled some turbo things. I am going back in to enable e-cores and turbo. This thing is not fully supported by linux, and its driving me crazy.