The Beast Returns - Alienware m18 - i9 13980HK 24c/RTX 4090 Mobile

[Rengsey R. H. Jr.]

any progress on the m18 repaste ?

[ssj92]

On 3/27/2023 at 2:35 PM, win32asmguy said:

Well, I disassembled my M18 this afternoon to replace the TIM and it looks like liquid metal (element 31?) was in places it should not be.

 

So now I have to figure out how to safely clean it up before applying PTM7950.

This shows why the temps aren't the greatest. Look at that cpu die... 

 

 

Aiming for 36k:

 

[Popcan]

On 3/29/2023 at 11:29 PM, ssj92 said:

This shows why the temps aren't the greatest. Look at that cpu die... 

 

 

Aiming for 36k:

 

 

 

Nice! What did you do to get those scores?  I am only getting 30000 at 85c and 31500 at 100c      100 is too hot for me.

[ssj92]

On 4/2/2023 at 10:07 AM, Popcan said:

 

 

Nice! What did you do to get those scores?  I am only getting 30000 at 85c and 31500 at 100c      100 is too hot for me.

Undervolt, max power/current limits. , max fans.

 

~32k is normal for 100c 13900HX 

 

Here's some more scores:

 

 

[MogRules]

Really interested to see how people's repaste efforts go with this and what paste they use etc.  All my P cores are hitting 100c lol.

[MogRules]

This is about where I am stuck currently.  Undervolting actually makes things worse 😒

CPU wattage never seems to pass about 117w, and I have three p cores constantly hitting 100c under synthetic loads.  But what really gets me is I see others posting stock results much higher.  Presumably just heavy thermal throttle. 

[ssj92]

1 hour ago, MogRules said:

This is about where I am stuck currently.  Undervolting actually makes things worse 😒

CPU wattage never seems to pass about 117w, and I have three p cores constantly hitting 100c under synthetic loads.  But what really gets me is I see others posting stock results much higher.  Presumably just heavy thermal throttle. 

Now that's a familiar name/avatar. 

 

Do a -100mV undervolt. then make sure your fans are 100% (might need to do custom profile in AWCC) and try again 

[MogRules]

 

Very minimal improvement overall, probably within margin of error.  The core clocks did seem to stay a little higher though.  Worth noting I did roll back to CC version 5, 6 was driving me insane with the auto theme / profile switching and long LONG load times.  

[ssj92]

4 hours ago, MogRules said:

 

Very minimal improvement overall, probably within margin of error.  The core clocks did seem to stay a little higher though.  Worth noting I did roll back to CC version 5, 6 was driving me insane with the auto theme / profile switching and long LONG load times.  

Are you on BIOS 1.6.0? Your CPU seems to be thermal throttling quite a lot, this is why the score is low. 

[MogRules]

1 hour ago, ssj92 said:

Are you on BIOS 1.6.0? Your CPU seems to be thermal throttling quite a lot, this is why the score is low. 

Yes, 1.6.0.  

I agree, likely thermal throttling hard.  I will likely try a repaste soonish.  I have seen more then a few posts from people who repasted with normal paste and it actually improved their temps, even from Dell's LM. 

[Popcan]

5 hours ago, MogRules said:

Yes, 1.6.0.  

I agree, likely thermal throttling hard.  I will likely try a repaste soonish.  I have seen more then a few posts from people who repasted with normal paste and it actually improved their temps, even from Dell's LM. 

 

I had this exact same issue.  After undervolting, it would go crazy and everytime I would put an actual load on the cpu it would drop to 800mhz.

 

No matter what I did it was stuck like that. Awcc wouldn’t actually change anything and just say my bios had been reset to default every time I opened awcc

 

So, I reset awcc and set it to max or whatever it’s called now. Went into the bios and set it to ultra performance and then confirmed the that the settings took via cinabench.

 

 I have started to save my bios settings and then use Load default custom user setting in the bios because my settings don’t always stay when I had awcc installed. I swear it was overwriting my setting.

 

as soon as that clears up I uninstalled every Alienware piece of software on my pc and I have had no issues since. Only down side is I can’t change my custom fan settings or my rgb.  But I like the way the fan curve is on ultra in the bios anyway.

 

From other posts I have read awcc is using a half baked version of xtu which really messed the bios up in my case. I have had to reset the bios via the cmos reset procedure 3 times because awcc would freeze or whatever it does and then my computer wouldn’t boot.
 

There is also a setting in bios that allows windows to modify bios settings, as per another forum post it was recommended to be disabled. I did that as well.
 

Right now I run at 15 tcc and ultra performance with 50 offset and get 30000ish on a 10min cinabench run. 

 

while gaming or benching cpu stays at 85c mostly but jumps around from time to time.

 

I just can’t bake it at 100c like its an easy bake oven…  if I turn the tcc back to stock I pretty much am pegged at 100-102c but go up to 32000ish

 

Could also be a crap paste job…  I would be interested in seeing what people get when they repaste.

 

I am on 1.6.0 and was on the latest version of awcc.

[ssj92]

I think AW meant well by putting Liquid Metal on CPU/GPU but in this case, paste would have been a better choice or ptm7950 phase change pads due to uneven heatsinks most likely. Since @win32asmguy m18 died from trying to re-paste I'm going to wait and see how others fair before opening mine up. My PTM7958 phase change paste came in 

[MogRules]

Whelp, I am on the right track at least.  30306 on R23 now with a -100 mv undervolt.  I took the plunge...someone had to go first I guess.  The paste job was not looking all too good, IMO you can see where parts of the CPU die just had garbage coverage, like there was just nothing there to work with on a large spot.  

Repasted with Conductonaut for now, but I need to possibly add more, I don't use this stuff much and haven't for a long time.  I used every last scrap I had left.  

My CPU power draw went WAY up.  It was peaking at 170w and eventually dropped back to 140w and settled there.  At 170w it was def hitting 100c on multiple P cores but once the wattage dropped back to 140w it was sitting in the 80-90c range comfortably.  I would so far call it a success, and I will likely add a bit more when I get more Conductonaut in as well. 
 

 

[win32asmguy]

25 minutes ago, MogRules said:

Whelp, I am on the right track at least.  30306 on R23 now with a -100 mv undervolt.  I took the plunge...someone had to go first I guess.  The paste job was not looking all too good, IMO you can see where parts of the CPU die just had garbage coverage, like there was just nothing there to work with on a large spot.  

Repasted with Conductonaut for now, but I need to possibly add more, I don't use this stuff much and haven't for a long time.  I used every last scrap I had left.  

My CPU power draw went WAY up.  It was peaking at 170w and eventually dropped back to 140w and settled there.  At 170w it was def hitting 100c on multiple P cores but once the wattage dropped back to 140w it was sitting in the 80-90c range comfortably.  I would so far call it a success, and I will likely add a bit more then I get more Conductonaut in as well. 

 

Wow yours looks quite a bit better than mine did after disassembly. On mine most of the LM was no longer emulsified in the paste and had made it past both the CPU and GPU barriers. Did you use acetone to clean everything up before applying Conductonaut? Do you intend to test any other options like PTM7950?

[MogRules]

40 minutes ago, win32asmguy said:

 

Wow yours looks quite a bit better than mine did after disassembly. On mine most of the LM was no longer emulsified in the paste and had made it past both the CPU and GPU barriers. Did you use acetone to clean everything up before applying Conductonaut? Do you intend to test any other options like PTM7950?

 I did not, I just cleaned it really well with cloths and Q-tips.  The die were both shiny and clean.  The only spots I couldn't get it all out of was in the crevices around the GPU die.  I kind of want to try a traditional paste of some sort, but I want to see what this thing is capable of now that I am not throttling instantly on the CPU and losing a ton of performance.  I know trying to use alcohol of ANY sort is a def nono, so I just went off the video that DJunreal did on his X17, which was just pretty much wiping it off carefully.  

I have to say, it wiped off pretty easily, much easier then traditional pastes that dry on there and have to be dissolved in many cases. 

[MogRules]

Starting to have a love hate relationship with this thing I tell you....



No idea what's going on here, but the 18k is the latest.  So my CPU is up but my GPU is down ? even though it's running at 62c lol...I got nothing.  I am going to reset the BIOS and check a few other things as well.  I did move back to CC 6.0 as I was talking with Alienware and had some issues trying to use TCC Offset, and it was determine that version 5 may cause issues.  I am going to roll that back, rest and see if that does anything. 

EDIT: 

Back to where I was before anyways.  I am going to play some damn games on this thing haha...I have been doing nothing but tests since I got it. 

 

Edited April 7, 2023 by MogRules
Added results.

[ssj92]

9 hours ago, MogRules said:

Whelp, I am on the right track at least.  30306 on R23 now with a -100 mv undervolt.  I took the plunge...someone had to go first I guess.  The paste job was not looking all too good, IMO you can see where parts of the CPU die just had garbage coverage, like there was just nothing there to work with on a large spot.  

Repasted with Conductonaut for now, but I need to possibly add more, I don't use this stuff much and haven't for a long time.  I used every last scrap I had left.  

My CPU power draw went WAY up.  It was peaking at 170w and eventually dropped back to 140w and settled there.  At 170w it was def hitting 100c on multiple P cores but once the wattage dropped back to 140w it was sitting in the 80-90c range comfortably.  I would so far call it a success, and I will likely add a bit more when I get more Conductonaut in as well. 
 

 

How long did it take? I'm thinking of just doing this to mine too. 

 

Why can't we use IPA wipes? I've cleaned LM on older laptops using IPA wipes before. 

 

New scores btw:

 

Slowly making my way up. 8th place global 4090n FS and 28th TS 

 

 

[MogRules]

12 hours ago, ssj92 said:

How long did it take? I'm thinking of just doing this to mine too.

Realistically about an hour I would say, with the cleaning and everything.  Now that I have done it I can probably do it a lot quicker.  I was taking my time to make sure I wasn't being overly aggressive with anything and separating my parts and screws for easy replacement. I had to take it back apart a second time due to some misplaced screws, pay attention to the gold vs black screw placement.  The board marks them, but I misplaced two gold ones so ended up with black ones on the top of the daughter board, which wasn't right, and had to take it apart again to fix that.  I honestly don't think this is any more difficult to take apart them the Cassini lineup, or even the 51m, there are just as many parts / pieces IMO.  

 

 

12 hours ago, ssj92 said:

Why can't we use IPA wipes? I've cleaned LM on older laptops using IPA wipes before.

DJUnreal did a teardown and tested it, it causes the LM and the traditional paste to separate and makes a hell of a mess apparently.  It breaks whatever is holding them together down and they just make a giant mess to clean up.  I didn't want to even try it, so just a microfiber to wipe it off, and it comes off pretty easily for sure.  

 

12 hours ago, ssj92 said:

New scores btw:

 

Slowly making my way up. 8th place global 4090n FS and 28th TS 

Nice!  

I honestly have no ambition to be the best, I just want it to run well.  I would be happy with the performance I have now, but I still think I can maybe do a little better on the temps.  My core differentials are WAY better then they were, I used to have the dreaded 15c+ spread, that is luckily no more.  

Keep in mind when you turn it back on it will 100% restart a few times and then get mad and say the time is wrong when it finally comes back.  Just wait for it to do it's thing, but I am sure you have seen this in the past with previous models, so it shouldn't be a surprise for you.  

[ssj92]

9 hours ago, MogRules said:

Realistically about an hour I would say, with the cleaning and everything.  Now that I have done it I can probably do it a lot quicker.  I was taking my time to make sure I wasn't being overly aggressive with anything and separating my parts and screws for easy replacement. I had to take it back apart a second time due to some misplaced screws, pay attention to the gold vs black screw placement.  The board marks them, but I misplaced two gold ones so ended up with black ones on the top of the daughter board, which wasn't right, and had to take it apart again to fix that.  I honestly don't think this is any more difficult to take apart them the Cassini lineup, or even the 51m, there are just as many parts / pieces IMO.  

 

 

DJUnreal did a teardown and tested it, it causes the LM and the traditional paste to separate and makes a hell of a mess apparently.  It breaks whatever is holding them together down and they just make a giant mess to clean up.  I didn't want to even try it, so just a microfiber to wipe it off, and it comes off pretty easily for sure.  

 

Nice!  

I honestly have no ambition to be the best, I just want it to run well.  I would be happy with the performance I have now, but I still think I can maybe do a little better on the temps.  My core differentials are WAY better then they were, I used to have the dreaded 15c+ spread, that is luckily no more.  

Keep in mind when you turn it back on it will 100% restart a few times and then get mad and say the time is wrong when it finally comes back.  Just wait for it to do it's thing, but I am sure you have seen this in the past with previous models, so it shouldn't be a surprise for you.  

I currently have 1 P-core causing 100c issue. Most of my cores are in the low 80s. This tells me a re-paste should fix my issue. I am currently working on my review video, once I am done, I will give it a shot. Might even make a disassembly video. 

 

The evolution of 18 series: 

 

[MogRules]

13 hours ago, ssj92 said:

I am currently working on my review video

In regards to that, are you locking your CPU frequency with the OC menu? Your CPU stays so perfectly level in your testing, mine is up and down, even without any thermal throttling...it's weird.  Like the guy on Reddit, my CPU score is stuck in the 17,500 range.  

[Shark00n]

On 4/6/2023 at 9:15 PM, MogRules said:

  I know trying to use alcohol of ANY sort is a def nono

 

Why is that?

But acetone is fine?

I don't even own one, just curious. Ty

[ssj92]

11 hours ago, MogRules said:

In regards to that, are you locking your CPU frequency with the OC menu? Your CPU stays so perfectly level in your testing, mine is up and down, even without any thermal throttling...it's weird.  Like the guy on Reddit, my CPU score is stuck in the 17,500 range.  

I'm not locking it. It's doing that on its own. I just max out all the power limits and fans.

If you truly aren't thermal throttling you should be reaching my scores easily 

[MogRules]

13 hours ago, Shark00n said:

 

Why is that?

But acetone is fine?

I don't even own one, just curious. Ty

According to what I have been told the alcohol causes the components of the Element 31 to separate and it becomes a pain to clean it up and remove it when that happens.  Honestly, it wiped off so easily with just a cloth I don't know why anyone would even need alcohol.  

[MogRules]

12 hours ago, ssj92 said:

I'm not locking it. It's doing that on its own. I just max out all the power limits and fans.

If you truly aren't thermal throttling you should be reaching my scores easily 

Interesting.  I will try the power limits again, the last time I tried changing that was when I hit the black screen reboot issue and had to rest the BIOS. 

I did see someone on Reddit claiming that the stock Dell image scores better then a fresh Windows install.  Not sure why that would ever be, but that's what they are claiming.  I did blow away the stock image about 10 minutes after I opened the box, and well before I did any tests, so I can't confirm.  I don't entirely feel like re-installing the Dell OS ATM either, although I did get it onto a flash drive.  

The weird part about maxing the fans is when I do it greys out the OC menu in the BIOS, so I don't know if those settings are holding once CC seems to overwrite everything.  I will have to do some more testing. 

 

[ssj92]

3 hours ago, MogRules said:

Interesting.  I will try the power limits again, the last time I tried changing that was when I hit the black screen reboot issue and had to rest the BIOS. 

I did see someone on Reddit claiming that the stock Dell image scores better then a fresh Windows install.  Not sure why that would ever be, but that's what they are claiming.  I did blow away the stock image about 10 minutes after I opened the box, and well before I did any tests, so I can't confirm.  I don't entirely feel like re-installing the Dell OS ATM either, although I did get it onto a flash drive.  

The weird part about maxing the fans is when I do it greys out the OC menu in the BIOS, so I don't know if those settings are holding once CC seems to overwrite everything.  I will have to do some more testing. 

 

Interesting, I will have to test that out because in my testing, a clean install scored really well. M

btw do you have any pics during the disassembly? Not sure how much I need to disassemble yet to get to the heatsink 

 

Review is out...