RTX 4000 mobile series officially released.

[VEGGIM]

20 hours ago, 1610ftw said:

 

The tragic thing is that we have a lot of capability these days but no ambition - not from Nvidia, not from the manufacturers and not from the customers most of all.

If you look at a few tech channel guys around youtube you will now find them mentioning 6 to 7 lbs laptops as super heavy and huge when in weight they are below former 15" DTRs and in volume they are probably smaller, too.

 

Before it was important to have a powerful laptop now it is about holding it with one hand apparently, no idea what this is all about:

 

 

So not much is coming out of the capability to build a really good 8 to 9 lbs DTR with a vapor chamber and possibly water cooling when people mainly buy 14" and therefore companies innovate in those market segments?

 

I will give it to MSI that they are the only company that still has good performance and most if not all DTR features in a less gamey chassis with the GT77 but I would much have preferred a GT88 with a socketed CPU now that socketed GPUs are obviously off the table and most of all with a bigger than 250W combined power limit - looks like this is a new invisible barrier nobody dares to cross.

 

Question is how much of a benefit would it be to even make one to justify the profit margins. Pcs aren't cars where you can make unobtainable halo cars.

 

Plus there would have to be a factor that isn't just benchmarks. Die size is also a thing. If it was still 14nm+ 12nm. It would be easier to cool with what we have today. But with larger thermal density. Uhh I don't know.

 

Plus liquid cooling won't be done because it's a huge liability. 

[1610ftw]

4 hours ago, VEGGIM said:

Question is how much of a benefit would it be to even make one to justify the profit margins. Pcs aren't cars where you can make unobtainable halo cars.

 

Question is how much more it would have to cost. Something like the MSI GT77 is not that far away and would only need a CPU socket and vapor chamber cooling for substantial improvements and a slightly larger chassis for an 18" screen. A halo product has positive effects in and of itself so there would be some trickle down effect having the greatest and the best.

 

 

4 hours ago, VEGGIM said:

Plus there would have to be a factor that isn't just benchmarks. Die size is also a thing. If it was still 14nm+ 12nm. It would be easier to cool with what we have today. But with larger thermal density. Uhh I don't know.

 

Die size works for us - a 4090 desktop die has an almost twice as big die size as a 1080 and that was cooled quite well even going up to 200W. With close to twice the die size and advancements in cooling it would most probably be easier to cool a 4090 die at 300W TGP than a 1080W running at 180W TGP due to reduced thermal density.

Nobody is really expecting to run a 4090 in a laptop at 450W but even 300W would be feasible with proper cooling design although a more realistic goal would probably be 250 to 275W with the 4090 desktop die and a total of 400W between the GPU and CPU. Even at 250W I would not be surprised if the 4090 die would give about 50% better performance than the current 4080 die used in the laptop 4090. Or the simplest approach with still very good results would be to "only" go with a socketed CPU and allow a power limit of up to 250W for the current 4090 chip - even the 4080 chip can do a lot more if we give it more power as currently it is only getting a bit more than half of its designed TGP.

 

So there are options and I do not think that they would be that expensive as a socketed CPU has ben offered by Clevo in a 15" chassis and that cannot have been too successful and MSI had those , too some years ago and neither were they extremely expensive nor sold in a way that even made the customer aware they existed. So there is some room for improvement with some marketing and costs can be exaggerated for the sake of not having to answer the real reason for these not being released which is that the industry does not believe in high performance any more but mostly in slim and light even in their top models.

 

4 hours ago, VEGGIM said:

Plus liquid cooling won't be done because it's a huge liability. 

 

Has been done by Asus and Tongfang/Uniwill and I have not heard of too many issues with the second approach, Asus I do not have much experience with and it was also impractical due to size..

Also with the Uniwill approach it is just tacked on to a normal system - all it does add if you do not use it is a little height and weight, probably 2 or 3mm and about 100g.

It is a simple and very effective solution for people who want to use it but it can also not be used/offered for people who aren't that extreme and a very powerful system could still be offered without it.

 

 

[VEGGIM]

5 hours ago, 1610ftw said:

 

Question is how much more it would have to cost. Something like the MSI GT77 is not that far away and would only need a CPU socket and vapor chamber cooling for substantial improvements and a slightly larger chassis for an 18" screen. A halo product has positive effects in and of itself so there would be some trickle down effect having the greatest and the best.

 

 

 

Die size works for us - a 4090 desktop die has an almost twice as big die size as a 1080 and that was cooled quite well even going up to 200W. With close to twice the die size and advancements in cooling it would most probably be easier to cool a 4090 die at 300W TGP than a 1080W running at 180W TGP due to reduced thermal density.

Nobody is really expecting to run a 4090 in a laptop at 450W but even 300W would be feasible with proper cooling design although a more realistic goal would probably be 250 to 275W with the 4090 desktop die and a total of 400W between the GPU and CPU. Even at 250W I would not be surprised if the 4090 die would give about 50% better performance than the current 4080 die used in the laptop 4090. Or the simplest approach with still very good results would be to "only" go with a socketed CPU and allow a power limit of up to 250W for the current 4090 chip - even the 4080 chip can do a lot more if we give it more power as currently it is only getting a bit more than half of its designed TGP.

 

So there are options and I do not think that they would be that expensive as a socketed CPU has ben offered by Clevo in a 15" chassis and that cannot have been too successful and MSI had those , too some years ago and neither were they extremely expensive nor sold in a way that even made the customer aware they existed. So there is some room for improvement with some marketing and costs can be exaggerated for the sake of not having to answer the real reason for these not being released which is that the industry does not believe in high performance any more but mostly in slim and light even in their top models.

 

 

Has been done by Asus and Tongfang/Uniwill and I have not heard of too many issues with the second approach, Asus I do not have much experience with and it was also impractical due to size..

Also with the Uniwill approach it is just tacked on to a normal system - all it does add if you do not use it is a little height and weight, probably 2 or 3mm and about 100g.

It is a simple and very effective solution for people who want to use it but it can also not be used/offered for people who aren't that extreme and a very powerful system could still be offered without it.

 

 

I really cannot see 300w without watercooling in feasable temps. especially when the cpu and gpu is combined heatsinks for spacing purposes.  With the hx series cpu being there, theres now not alot of need to have the desktop cpus in laptops considering they are exactly the same. Also add that if you have the sockets that keep changing the upgradability becomes moot. The 51m is proof of this exact problem. If your upgradability only works for a certation generation or if it requires modifications to get it to work, then for the average user, it's not feasible. 

Other thing is that laptops seem to like direct die cooling. Unless someone is willing to delid every cpu that comes in, i don't see it happening. PTM7950 for example fails on non direct die cooling. that's it's weakness. 

[1610ftw]

59 minutes ago, VEGGIM said:

I really cannot see 300w without watercooling in feasable temps. especially when the cpu and gpu is combined heatsinks for spacing purposes.  With the hx series cpu being there, theres now not alot of need to have the desktop cpus in laptops considering they are exactly the same. Also add that if you have the sockets that keep changing the upgradability becomes moot. The 51m is proof of this exact problem. If your upgradability only works for a certation generation or if it requires modifications to get it to work, then for the average user, it's not feasible. 

Other thing is that laptops seem to like direct die cooling. Unless someone is willing to delid every cpu that comes in, i don't see it happening. PTM7950 for example fails on non direct die cooling. that's it's weakness. 

 

So let's make it 250W TGP with air and 300W with watercooling - will be a lot more than we have now.

 

As for TIM many manufacturers already offer liquid metal or phase change pads, that helps, the next step would be vapor chambers and bigger ones at that. It can obviously be done as laptops from the past did better even without vpor chambers and technology for cooling has improved since then.

 

The next Intel socket that should be introduiced towards the end of this year should be good for about 2 or 3 CPU generations and AM5 for 4 or 5 - you obviously want to introduce a socketed CPU at the right time and not at the end of a cycle and like motherboards there would have to be some support.

 

Benefits for the end user would not just be to have a socketed CPU but to choose which CPU to use. Want a lesser CPU with a 4080 or 4090? Not possible at this point with most manufacturers, same for a big CPU with a lesser GPU. That could easily be addressed and would really help those of us who do not need the top CPU AND GPU as especially the big GPUs cost a lot and are somehow forced on the buyer as he has to buy them together with a high end GPU.

[jaybee83]

On 1/4/2023 at 1:00 AM, Shark00n said:

It would still be 150W max TGP

 

Nvidia only launches eunuchs in mobile

 

LOL lovely, nicely put hahaha

 

On 1/4/2023 at 1:37 AM, ssj92 said:

175w will be max and match 3090Ti levels on high end systems

 

thats the thing though, right? on one hand u get those STUPID naming schemes and Nvidia doing their best to hide any shortcoming from the consumers via their non-existent spec sheets. on the other hand u get 3090-3090Ti performance in a frigging laptop at one third the board power (which is, of course, awesome, but also still to be proven by third party reviews and hands on users). in the end it just leaves you with a sour bitter taste in your mouth with a hint of confusion and melancholic thoughts towards the "good ol' days"...sigh

[Papusan]

8 hours ago, VEGGIM said:

I really cannot see 300w without watercooling in feasable temps. especially when the cpu and gpu is combined heatsinks for spacing purposes.  With the hx series cpu being there, theres now not alot of need to have the desktop cpus in laptops considering they are exactly the same. Also add that if you have the sockets that keep changing the upgradability becomes moot.

If there was a LGA socket in the chassis they would make it thicker for the socket and IHS. This would also benefits the GPU. Bigger HS cooler fin stack = More cooling capacity. 

[VEGGIM]

10 hours ago, 1610ftw said:

 

So let's make it 250W TGP with air and 300W with watercooling - will be a lot more than we have now.

 

As for TIM many manufacturers already offer liquid metal or phase change pads, that helps, the next step would be vapor chambers and bigger ones at that. It can obviously be done as laptops from the past did better even without vpor chambers and technology for cooling has improved since then.

 

The next Intel socket that should be introduiced towards the end of this year should be good for about 2 or 3 CPU generations and AM5 for 4 or 5 - you obviously want to introduce a socketed CPU at the right time and not at the end of a cycle and like motherboards there would have to be some support.

 

Benefits for the end user would not just be to have a socketed CPU but to choose which CPU to use. Want a lesser CPU with a 4080 or 4090? Not possible at this point with most manufacturers, same for a big CPU with a lesser GPU. That could easily be addressed and would really help those of us who do not need the top CPU AND GPU as especially the big GPUs cost a lot and are somehow forced on the buyer as he has to buy them together with a high end GPU.

Ptm is a phase change material. And considering asus had quality issues with lm the first few times. Yeah I don't really trust it. Since if the manufacturer botch the lm, it can cause problems. Reason why lenovo doesn't use liquid metal. 

 

But again. Pcm likes direct die cooling more than heatplates. You see more of an effect there. Heck lenovos heatsink are designed around it to the point that if you use anything that isn't ptm. The Temps suffer horrendously. 

[VEGGIM]

3 hours ago, Papusan said:

If there was a LGA socket in the chassis they would make it thicker for the socket and IHS. This would also benefits the GPU. Bigger HS cooler fin stack = More cooling capacity. 

I mean the fact that hx and the k/x cpus are technically the same now. So going to desktop cpus isn't the only option if you wanted a 1:1 desktop cpu. The hx series is that solution. You could thicken it anyway regardless of ihs or not. This is the equivalent of asking for more manual transmissions even though they are just a dying breed due to not being marketable. 

 

Same thing for lga + mxm. Mxm has been unsupported and lga makes the laptop thick looking which is not really marketable. Especially those who want a balance since gaming laptops, especially  14" ones are being seen as AIO devices. Especially with the handheld market with the steam deck. 

 

Try and market the x170smg to the average person and you could not sell. 

 

 

[ryan]

I hate to say it but its not nessessary anymore. with cloud computing and fast internet which is more common than not people can get beastly things done at a fraction of the cost on a thin and light laptop with good portability(better) and longer battery life...this is the future as things get progressively smaller(aside from the 4090 lol). We may not be quite quite there yet but its creeping in kinda like what 4k gaming did. everyone saw it as a far off thing but knew it was the future, now 8k is creeping the same way. cloud computing isn't what I want or what others want, but the pros out weigh the cons this side of not actually owning the hardware.

 

It was worded that way because intel/nvidia/amd seem to be pushing hardware prices and performance(gimmicks) that no one asked for or want. they are deciding what is best for us.

[Bullit]

Things might be smaller but screens will not be  with 4K and 8K .

I have no enjoyment seeing TV at 15.6" even in FHD that is why i have been buying 17 and 17. 3 since 2004. I wish obviously to have an even bigger but rolling extended screen, but that will be expensive in next 10 years or more.

[ryan]

i have a 16in screen and im happy with it....typically ive gone with 15.6in but i liked my 17.3in laptops..sometimes having a bigger screen is better. 18in beasts seem excessively big. but hey if it floats your boat

[Shark00n]

11 hours ago, ryan said:

i have a 16in screen and im happy with it....typically ive gone with 15.6in but i liked my 17.3in laptops..sometimes having a bigger screen is better. 18in beasts seem excessively big. but hey if it floats your boat

 

The Blade 18 is very very slightly bigger than the old Blade Pro 17. Because it's 16:10

 

I think I really like this upgrade to 18.3" 16:10 screens.

[ryan]

yeah it could be nice, I remember a acer laptop with a 21in display. cost like 10 000 but was a great powerhouse. I don't get why we dont have rollable screens yet. look at cell phones. they are basically 1 mm thick and bendable, been that way for ages

[VEGGIM]

On 1/11/2023 at 11:43 AM, 1610ftw said:

 

So let's make it 250W TGP with air and 300W with watercooling - will be a lot more than we have now.

 

As for TIM many manufacturers already offer liquid metal or phase change pads, that helps, the next step would be vapor chambers and bigger ones at that. It can obviously be done as laptops from the past did better even without vpor chambers and technology for cooling has improved since then.

 

The next Intel socket that should be introduiced towards the end of this year should be good for about 2 or 3 CPU generations and AM5 for 4 or 5 - you obviously want to introduce a socketed CPU at the right time and not at the end of a cycle and like motherboards there would have to be some support.

 

Benefits for the end user would not just be to have a socketed CPU but to choose which CPU to use. Want a lesser CPU with a 4080 or 4090? Not possible at this point with most manufacturers, same for a big CPU with a lesser GPU. That could easily be addressed and would really help those of us who do not need the top CPU AND GPU as especially the big GPUs cost a lot and are somehow forced on the buyer as he has to buy them together with a high end GPU.

It sounds like you want all gaming laptops to be de dtr levels of thick. Including the 14 inch ones. No offense, but this sounds like the equivalent of "we need to gatekeep gaming laptops." Also desktop cpus are not made for efficiency curves. By dtr levels of thick i kinda mean between 1-2 inches. Plus are we talking about 250w-300w gpu by itself or combined with the cpu. 

[1610ftw]

10 hours ago, VEGGIM said:

It sounds like you want all gaming laptops to be de dtr levels of thick.

 

It is rather strange that you would come to that conclusion as I would be perfectly happy to be able to select between maybe three DTR models from as many manufacturers - it is nice to have some competition like we had in the past with Alienware, Asus, Clevo and MSI.

 

 

10 hours ago, VEGGIM said:

 No offense, but this sounds like the equivalent of "we need to gatekeep gaming laptops." 

 

No offense but your second sentence is extremely ironic as it seems that you can only be happy with no single proper DTR in existence, why is that and why do you even care? Isn't it really you who is gatekeeping here and hopeing / insisting / explaining that not a single proper DTR shall exist?

 

I would certainly not care much if somebody would like to see a number of thin and light 14" laptops with a 4090 as long as there are a few proper DTRs around but the mere thought of even one bigger and more powerful DTR existing seems to be a threatening thought for you.

 

In the past when we had a handful of big DTRs this did not at all affect 99.9% of laptops nor would I care about them apart from a desire to not see that much disposable crap being sold with everything soldered down including network cards, memory and even storage.  That will not make a laptop very big as you can see with that new Dell prototye or the framework laptop.

 

 

10 hours ago, VEGGIM said:

Also desktop cpus are not made for efficiency curves. By dtr levels of thick i kinda mean between 1-2 inches. Plus are we talking about 250w-300w gpu by itself or combined with the cpu. 

 

HX and regular desktop CPUs are so similar that it does not make much of a difference to efficiency if you use one or the other. As for thickness it is not a quality in and of itself but even with the current 18" screens we will probably at least need 1.5" at the thickest part of the laptop to accomodate better performance unless we go back to wider bezels but that seems to be out of the question. Again with that I am talking about a handful of models at most and not more than one chassis per manufacturer. I was talking 250W and 300W GPU in that case as obviously we used to have SLI DTRs that could dissipate up to 400W GPU power alone in the past so 250W GPU with 350W combined TDP would be something to shoot for as a compromise.

[Etern4l]

38 minutes ago, 1610ftw said:

HX and regular desktop CPUs are so similar that it does not make much of a difference to efficiency

 

What do you mean? Idle power utilisation? Or do you mean that if we power-cap a desktop CPU it will run similarly to the HX? Obviously, under full load the desktop CPU can pull 325W+ and undervolted consistently run at 270W while scoring 41-43k in CB23, so I don't imagine this is anywhere close to what the mobile CPUs can achieve, at a cost to efficiency.

[1610ftw]

3 hours ago, Etern4l said:

 

What do you mean? Idle power utilisation? Or do you mean that if we power-cap a desktop CPU it will run similarly to the HX? Obviously, under full load the desktop CPU can pull 325W+ and undervolted consistently run at 270W while scoring 41-43k in CB23, so I don't imagine this is anywhere close to what the mobile CPUs can achieve, at a cost to efficiency.

 

Efficiency in the low end indeed. I doubt that at the power levels that both achieve there will be that much of a difference between an average 13900HX laptop and 13900K desktop CPU.

And no, there is no way that any current laptop can sustain 270W and I doubt we will see much more than 200W from any of them this generation with around 34 to 37K tops CB R23 in a single run and that will be with deafening fan noise and/or added water cooling.

[Etern4l]

Yeah, I don't see a reason why significantly more power would be used by the desktop SKU. Looking at the specs, it's the same CPU, the same process, it's just necessarily clocked much lower on the laptop. Unfortunately I can't check right now, but from memory the 13900K power draw goes down to a few W on idle (with UV).

[1610ftw]

8 minutes ago, Etern4l said:

Yeah, I don't see a reason why significantly more power would be used by the desktop SKU. Looking at the specs, it's the same CPU, the same process, it's just necessarily clocked much lower on the laptop. Unfortunately I can't check right now, but from memory the 13900K power draw goes down to a few W on idle (with UV).

 

This is also my impression looking at the 12800HX, 12900HX and 12900K - very similar until the laptop CPUs top out at 150W to 200W depending on the circumstances one finds them in. Would really be cool to be able to swap CPUs in laptops again.

[Clamibot]

With tuning, I've gotten my X170's 10900K to consume less than 2 watts when idle, so a desktop CPU can be just as efficient as a mobile one. Desktop CPUs are just able to achieve higher clocks since they're meant for use in less thermally constrained environments than mobile CPUs.

[1610ftw]

1 hour ago, Clamibot said:

With tuning, I've gotten my X170's 10900K to consume less than 2 watts when idle, so a desktop CPU can be just as efficient as a mobile one. Desktop CPUs are just able to achieve higher clocks since they're meant for use in less thermally constrained environments than mobile CPUs.

 

And don't forget that a carefully selected desktop chip with more performance per Watt can make a huge difference.  If my system can only sustain 150W then I do not care what the performance would be at 250 or 300W! Not much fun being stuck with a stinker of a chip in a laptop  as that will be it. Over in the Dell thread @MyPC8MyBrain went through a bunch of workstations with the 12950HX and the difference in performance was quite significant with his last unit being kind of a golden sample but I believe he went through 5 or 6 of them!

 

[ssj92]

https://www.notebookcheck.net/NVIDIA-GeForce-RTX-4090-laptop-GPU-first-impressions-pit-flagship-graphics-card-against-GeForce-RTX-3080-Ti.683346.0.html

 

This is straight up following nVidia's guidelines on how to show performance. 

 

We don't care about dlss 2 vs 3 and the performance improvement over 3080Ti. 

 

We want to see direct raw performance between 3080Ti & 4090 

[cylix]

Yes awful comparison from Jarrod... Click bait video only.. Comparing apples to bananas.

 

[MyPC8MyBrain]

2 hours ago, 1610ftw said:

I believe he went through 5 or 6 of them

that is correct, i am on the sixth 12950HX/3080Ti combo unit,
the last unit which i decided to keep out of the box was behaving completely different to the point i had to confirm the cpu sku was correct,

every unit before was just out of control temp wise doing nothing at idle every single one of them idle at 90-100c out of the box,

this unit was idling below 70c on Dells default factory settings and before they issued the latest bios with load line calibration fix,

this cpu sample is also able to withhold over -150mV undervolt across Performance, Cache, and Efficiency cores (i run with -125mV very stable),

during idle and non heavy tasks temps are in the 30-36c range (images below as i type this), on this system i am still with Dell's original paste job,

with every prior cpu i had no choice but to replace Dell's paste (i tested stability with LM TIM on CPU and CPU/GPU),

i was not expecting the current unit to perform different then the others did (and today is my last day out of 30),

but it totally did to the point that i decided to keep it,

  

[VEGGIM]

if its anything. This is from a discord