Asus ROG Strix Scar 17 SE 2022 Discussion Thread

[win32asmguy]

This is Asus's flagship gaming laptop for 2022 using Intel 12th Gen HX BGA CPU's.

 

https://rog.asus.com/us/laptops/rog-strix/rog-strix-scar-17-se-2022-series/

 

Notable features:

• Full sized vapor chamber for CPU and GPU
• 12950HX with 175W PL2 and 170W PL1
• Nvidia RTX 3080 Ti 150W+25W Dynamic Boost
• Combined 230W CPU+GPU TDP
• Reduced size 330W power brick (maybe GAN?) with barrel connector
• 240hz QHD 99% DCIP3 17 inch display
• Dual DDR5-4800 sodimm slots
• Dual M.2 PCIe Gen4 slots
• M.2 2230 slot for Wifi
• AMI BIOS Features:
  • Disable E-cores
  • P-Core undervolting
  • Set P-Core multipliers
  • Enable overclocking

 

More information to come!

[win32asmguy]

Review (in progress):

 

Chassis:

 

The chassis is a metal lid, plastic palm rest and plastic bottom panel. No creaking and the palm rest remains relatively cool.

 

The membrane keyboard is full sized including a number pad. The keys have good travel and are quiet. The layout is not cramped nor are any keys abnormally small.

 

The trackpad feels like a glass surface and has support for the usual gestures which seem to work fine. No strange rattles when tapping / clicking.

 

Display:

 

It is a 240hz QHD panel identified as BOE NE173QHM-NZ2, 8 bit color 99% DCI-P3, and supports Adaptive Sync in hybrid mode but not GSync in dedicated mode. It also supports operating at 60hz to conserve battery.

 

CPU:

 

12950HX with 8 P-cores and 8 E-cores. It has 175W PL2 and 170W PL1 and scores 23,400 out of the box in Cinebench R23 Multi. There is thermal throttling at those power limits.

 

GPU:

 

Pretty fast. Time Spy score is 13584. I am trying to find a game that stresses CPU and GPU to confirm the max combined load limits.

 

SSD:

 

By default it shipped in a RAID configuration. My hope is that one of the drive slots is Gen 5 connected directly to CPU PCIe lanes, but I need to reinstall to determine this. The stock drives were Micron 3400 1TB Gen4.

 

Video Output Support:

 

Thunderbolt 4 connected to Integrated GPU

USB-C connected to Nvidia GPU (GSync supported)

HDMI connected to Nvidia GPU

 

Sleep Modes:

 

It appears to support both S0 Modern Standby and S3 Standby according to Windows 11 powercfg. I will need to test if it can properly resume in S3 mode.

 

Noise:

 

Very quiet in all modes. I need to test more though.

[-batab-]

Any heavy scientific benchmark (most likely avx) to check temperatures/noise on sustained loads?

 

What about throttling / temp / noise if you run such bench on 8 or 16 threads thread only?

[win32asmguy]

3 hours ago, -batab- said:

Any heavy scientific benchmark (most likely avx) to check temperatures/noise on sustained loads?

 

What about throttling / temp / noise if you run such bench on 8 or 16 threads thread only?

 

Sure, let me know a specific benchmark and I will try and run it.

 

This morning I tried booting up a Fedora 36 Live USB but ran into an error:

 

error: ../../grub-core/kern/mm.c:376:out of memory.

 

Best I can tell it happens on certain laptops and is related to either the TPM (which is bad because there is no bios option to disable that), or Intel VMD being enabled (the factory config is Raid 0 SSD's). So now I am looking at making an Asus restore image so I can reinstall without VMD/Raid0 configured.

[electrosoft]

Looking forward to a full head to head Scar vs Raptor....

 

 

[-batab-]

Hard to say. I guess anything you can put your fingers on, be it a real scientific software (numerical simulations or whatever that allows you to throw some multi threaded load at it) or some more classic stuff like prime 95 or even something else that uses AVX.

 

 

[jaybee83]

4 hours ago, electrosoft said:

Looking forward to a full head to head Scar vs Raptor....

 

 

no competition on the GPU and RAM side, but CPU should be interesting 🙂 although im a bit surprised it only allows 230W total TDP for CPU+GPU, even though it provides a 330W psu? whats that all about?

[electrosoft]

4 hours ago, -batab- said:

Hard to say. I guess anything you can put your fingers on, be it a real scientific software (numerical simulations or whatever that allows you to throw some multi threaded load at it) or some more classic stuff like prime 95 or even something else that uses AVX.

 

 

 

Prime95 is a near waste on desktops and a super waste on laptops unless the idea is to smash it into the ground with unrealistic loads. Ranks right up there with running Furmark to stress test your GPU.

 

 

[Clamibot]

The fact that this laptop has good BIOS options (ex: being able to disable the e-cores) is promising.

 

How high can the CPU's P-cores be overclocked with the e-cores disabled? I'm glad there's an option to disable the e-cores since the P-cores are what matter in games anyway. I think I saw someone before do 5.5 GHz all P-core on a 12900K while pulling 187 watts. They disabled the e-cores to achieve that. Assuming that the i9 12950HX is a binned 12900K, 5.5 GHz all P-core may be achievable within the 175 watt PL2.

 

Assuming this laptop gets a refresh with an i9 13950HX when Raptor Lake releases, we may be able to achieve 6 GHz clock speeds in this laptop when the CPU is overclocked. Now that would be an amazing feat, even if it wasn't across all the P-cores. Even having that on one or two cores in a laptop would be an amazing feat. Maybe I'm being overly optimistic, but hey, one can dream right? The power allowance should allow for such a thing to happen if the thermals can be kept under control.

[-batab-]

9 hours ago, electrosoft said:

 

Prime95 is a near waste on desktops and a super waste on laptops unless the idea is to smash it into the ground with unrealistic loads. Ranks right up there with running Furmark to stress test your GPU.

 

 

How is it that unrealistic? My laptop is usually hotter during finite element simulation than under prime95 to be honest. And I've had my current Clevo run hundreds of 4-12h simulations on all cores during the years.

 

It's also very different than gaming because you basically are on 100% cpu load and 0% gpu load.

 

Unless he has some real scientific software to use for testing, that's the first decent thing I've thought about. Any other suggestion then?

 

 

[electrosoft]

8 hours ago, -batab- said:

How is it that unrealistic? My laptop is usually hotter during finite element simulation than under prime95 to be honest. And I've had my current Clevo run hundreds of 4-12h simulations on all cores during the years.

 

It's also very different than gaming because you basically are on 100% cpu load and 0% gpu load.

 

Unless he has some real scientific software to use for testing, that's the first decent thing I've thought about. Any other suggestion then?

 

 

 

I'm lean more towards real application testing like the software you're using now. Have you tested your software on a well cooled desktop vs Prime95? On a laptop it could be a case of Prime95 forcing a more brutal sustained throttle scenario via a combination of heat or pull (or both) vs your software giving the impression your software is hitting it harder and hotter when in actuality it is not. I've seen this happen before as Prime95 overwhelms a laptop. What is the name of your software? Which Clevo do you use?

 

 

 

 

 

 

 

 

[win32asmguy]

10 hours ago, Clamibot said:

The fact that this laptop has good BIOS options (ex: being able to disable the e-cores) is promising.

 

How high can the CPU's P-cores be overclocked with the e-cores disabled? I'm glad there's an option to disable the e-cores since the P-cores are what matter in games anyway. I think I saw someone before do 5.5 GHz all P-core on a 12900K while pulling 187 watts. They disabled the e-cores to achieve that. Assuming that the i9 12950HX is a binned 12900K, 5.5 GHz all P-core may be achievable within the 175 watt PL2.

 

Assuming this laptop gets a refresh with an i9 13950HX when Raptor Lake releases, we may be able to achieve 6 GHz clock speeds in this laptop when the CPU is overclocked. Now that would be an amazing feat, even if it wasn't across all the P-cores. Even having that on one or two cores in a laptop would be an amazing feat. Maybe I'm being overly optimistic, but hey, one can dream right? The power allowance should allow for such a thing to happen if the thermals can be kept under control.

It is still missing options that would make me consider it on par with their Strix desktop boards. No PL1/PL2 config, No fan table config, No RGB config or atleast a disable override, No MUX switch config. So basically you still are tied to Armory Crate in Windows. The RGB gets reset to the default rainbow preset with a bios update, so you probably want to always keep it around even with services disabled, unless there are Linux tools that have reverse engineered Asus Strix RGB protocols so it can configured that way. There is a bios option to enable/disable the boot up sound at least.

 

With e-cores disabled the P-cores can sustain 4.2ghz. There is thermal throttling and core differences between 87C-96C so it settles on pulling 142W. So it may not have a great LM job from the factory. I saw up to 17400 CBR23.

[-batab-]

30 minutes ago, electrosoft said:

 

I'm lean more towards real application testing like the software you're using now. Have you tested your software on a well cooled desktop vs Prime95? On a laptop it could be a case of Prime95 forcing a more brutal sustained throttle scenario via a combination of heat or pull (or both) vs your software giving the impression your software is hitting it harder and hotter when in actuality it is not. I've seen this happen before as Prime95 overwhelms a laptop. What is the name of your software? Which Clevo do you use?

 

Generally, I can generate a lot of heat with any scientific software. The most sustained loads however were via Matlab, Abaqus, Solidworks and few others. Abaqus and Matlab were definitely the heaviest. However those are not freeware.

 

I guess that similar workloads could be generated with the free softwares like Calculix and Octave since they are (respectively) very similar to Abaqus and Matlab.

Actually, I think that Anandtech uses Calculix as one of their scientific benchmarks. 

 

My Clevo is a P650HP6 with a 7820HK and right now I keep HT off to gain a few °C since I love quiet systems and HT on doesn't make a big difference on such scientific software. If I'm in no hurry I keep it clocked at 3-3.1 GHz all core and I'm around 70-75°C at worse (also using a laptop stand). If i get the clock to 3.8 - 4.0 GHz, turn on HT and throw 8 threads of such software I usually get easy throttle even if this CPU is quite lucky and I can keep it at -120mV undervolt.

 

[jaybee83]

4 hours ago, win32asmguy said:

It is still missing options that would make me consider it on par with their Strix desktop boards. No PL1/PL2 config, No fan table config, No RGB config or atleast a disable override, No MUX switch config. So basically you still are tied to Armory Crate in Windows. The RGB gets reset to the default rainbow preset with a bios update, so you probably want to always keep it around even with services disabled, unless there are Linux tools that have reverse engineered Asus Strix RGB protocols so it can configured that way. There is a bios option to enable/disable the boot up sound at least.

 

With e-cores disabled the P-cores can sustain 4.2ghz. There is thermal throttling and core differences between 87C-96C so it settles on pulling 142W. So it may not have a great LM job from the factory. I saw up to 17400 CBR23.

hm that CB23 score is kinda disappointing, i actually expected more...or was that with only P cores enabled? how do the clocks compare to the NH55?

[electrosoft]

22 minutes ago, jaybee83 said:

hm that CB23 score is kinda disappointing, i actually expected more...or was that with only P cores enabled? how do the clocks compare to the NH55?

 

17400 is definitely with e-cores off as in an earlier post he said he got 23400 at full tilt.

 

It definitely is a touch better than the NH55 atm but it is using slightly higher power limits and on paper and visual inspection has a much better cooling system so we will see what happens as tweaks are made. There is also the random luck of silicon too which can't be user addressable on BGA laptops.

 

 

[Clamibot]

2 hours ago, electrosoft said:

 

17400 is definitely with e-cores off as in an earlier post he said he got 23400 at full tilt.

 

It definitely is a touch better than the NH55 atm but it is using slightly higher power limits and on paper and visual inspection has a much better cooling system so we will see what happens as tweaks are made. There is also the random luck of silicon too which can't be user addressable on BGA laptops.

 

 

 

Unless you're someone like Khenglish and have some mad soldering skills. That dude soldered a freaking desktop GTX 980 GPU core to an MXM card that originally had a 980m GPU core on it, and then he made it work perfectly!

 

He could probably do something similar with this laptop, that is, he can bin some 12900Ks, desolder the CPU off the laptop, solder the best 12900K in its place, and then bam! You now have a better binned CPU in the Scar Strix 17 SE.

 

It's very impractical, but still doable, just not for the average user. It's not even doable for a lot of us on this forum, myself included 🤣.

[win32asmguy]

4 hours ago, Clamibot said:

 

Unless you're someone like Khenglish and have some mad soldering skills. That dude soldered a freaking desktop GTX 980 GPU core to an MXM card that originally had a 980m GPU core on it, and then he made it work perfectly!

 

He could probably do something similar with this laptop, that is, he can bin some 12900Ks, desolder the CPU off the laptop, solder the best 12900K in its place, and then bam! You now have a better binned CPU in the Scar Strix 17 SE.

 

It's very impractical, but still doable, just not for the average user. It's not even doable for a lot of us on this forum, myself included 🤣.

Maybe ZtecPC could be convinced to bin the CPU on the Strix Scar 17 SE if they sell them. I am sure there are customers who would pay extra money for a known good sample, especially if they are repasting them or doing other mods already.

[jaybee83]

16 hours ago, win32asmguy said:

Maybe ZtecPC could be convinced to bin the CPU on the Strix Scar 17 SE if they sell them. I am sure there are customers who would pay extra money for a known good sample, especially if they are repasting them or doing other mods already.

dude that would be next level binning, to do that with whole laptops lol

[-batab-]

How low can you get your temps if you tone down the TDP in order to get roughly a 15000 score in cinebench R23? 

I'm curious as this seems the score of a 6900hx.

 

I know the 12950hx should be less power efficient than the 6900hx but it seems like there is some huge margin in terms of raw power.

[win32asmguy]

7 hours ago, -batab- said:

How low can you get your temps if you tone down the TDP in order to get roughly a 15000 score in cinebench R23? 

I'm curious as this seems the score of a 6900hx.

 

I know the 12950hx should be less power efficient than the 6900hx but it seems like there is some huge margin in terms of raw power.

E-cores enabled: 55W PL1/PL2 15064 Core temps: 65C fans 2000rpm Ambient temps 76F

E-cores disabled: 90W PL1/PL2 14748 Core temps: 65C fans 3000rpm Ambient temps 76F

 

[-batab-]

8 hours ago, win32asmguy said:

E-cores enabled: 55W PL1/PL2 15064 Core temps: 65C fans 2000rpm Ambient temps 76F

E-cores disabled: 90W PL1/PL2 14748 Core temps: 65C fans 3000rpm Ambient temps 76F

 

 

This seems really good wow. I guess the system is also very quiet at 2000 rpm right? Can you still tweak everything via throttlestop in 12th gen cpus?

 

I hope someone will actually review 12950hx vs 6900hx at different tdp levels. Feels like Intel almost caught up in efficiency too.

 

Also, can you reach similar scores at similar temps with the clevo?

[electrosoft]

16 hours ago, win32asmguy said:

E-cores enabled: 55W PL1/PL2 15064 Core temps: 65C fans 2000rpm Ambient temps 76F

E-cores disabled: 90W PL1/PL2 14748 Core temps: 65C fans 3000rpm Ambient temps 76F

 

 

So how goes the comparison with the NH55 vs the Scar? Leaning one way or the other yet?

 

I suspect the Scar cost much more but the one stand out feature I do like is the 17.3" display. I wish they offered a version with the full fat HX but 3060 at a much reduced cost. Any side by side pics?

 

 

[win32asmguy]

9 hours ago, -batab- said:

 

This seems really good wow. I guess the system is also very quiet at 2000 rpm right? Can you still tweak everything via throttlestop in 12th gen cpus?

 

I hope someone will actually review 12950hx vs 6900hx at different tdp levels. Feels like Intel almost caught up in efficiency too.

 

Also, can you reach similar scores at similar temps with the clevo?

Yes it is pretty quiet in general. Even with the 3080Ti at 175W and 12950HX at 50W running a game. Throttlestop has everything available. XTU also supports manipulation of almost everything except memory related stuff. XTU even lets you customize the VF Curve voltage offsets.

 

I have heard that AMD is more efficient at lower wattages.

 

The Clevo can do the same but you need to be willing to spend the time to find a good CPU and ensure it has a good paste application and potentially delid as well.

 

7 minutes ago, electrosoft said:

 

So how goes the comparison with the NH55 vs the Scar? Leaning one way or the other yet?

 

I suspect the Scar cost much more but the one stand out feature I do like is the 17.3" display. I wish they offered a version with the full fat HX but 3060 at a much reduced cost. Any side by side pics?

Nothing formal so far, but I will say that neither machine has full Windows 10 support. The Clevo could really be a winner if it had a MUX and we had access to a shop like silicon lottery where you can buy better bin CPUs easily. I do prefer 17 inch screens but prefer FHD which the Scar does not have.

 

The scar also has some pretty crappy bios bugs out of the box. In order to boot linux, it has to have both USB-C ports unplugged and the thumbdrive in the bottom USB-A port or else there is not enough free memory in the first 4GB address range for the kernel to live in. So basically it is very much a "designed for Windows 11" machine.

[electrosoft]

MSI's version of the 12950HX hitting 24140 @ ~160w / 95c -50uv in MSI BIOS out of the box in CB23:

 

 

[win32asmguy]

33 minutes ago, electrosoft said:

MSI's version of the 12950HX hitting 24140 @ ~160w / 95c -50uv in MSI BIOS out of the box in CB23:

He got pretty lucky with the paste job if he can sustain 160W on the 12900HX. The phase change pad on the GE76 12900HK had very high core temp differentials.