Nvidia Ada RTX 40 Laptop GPU analysis: 175 W RTX 4090 Laptop GPU surpasses a 350 W RTX 3090 in performance and power efficiency

[NikoB]

4090 mobile is actually the 4080 desktop while the 4080 mobile

No. As the first tests showed, even in MSI Titan - mobile 4090 loses to the desktop version exactly 1.5 times. Therefore, the mobile 4090 is equal to the desktop 4070 and slightly faster than the desktop 3080Ti.

And not only that - as tests have already shown, the mobile 4090 still does not pull current projects at 60fps+ in 4k, so it is garbage against the background of the desktop one, which pulls everything.

Playing on a laptop is stupid. Noise, slow, dangerous for hardware and quickly becomes obsolete without the possibility of an upgrade. And at the same time more expensive than the desktop build 1.5 times faster...

[RobertJasiek]

NikoB, please explain "dangerous for hardware"!

[LL]

Couple days ago i got scores in Blender benchmark:



NVIDIA GeForce RTX 4090 Laptop GPU 8256.73 4

NVIDIA GeForce RTX 4080 Laptop GPU 5992.52 1

For comparison the top desktop 3000 series card:

NVIDIA GeForce RTX 3090 Ti 6163.43

Currently the database do not have any 4000 Laptop card in database, those results were erased at least for the moment.

[Vaidyanathan]

At the moment, there aren't too many 4090/4080 samples floating around.

Do you know why? I thought there would also be some Aliens and Legions. So far only Titan and Razer 16 plus a few reviewers each got one other different notebook. But some got several presamples, too, which are hardly reviewed yet. I expected hundreds of reviews on Feburary 7 but only a dozen appeared. Do some manufacturers only send tiny numbers of review samples to carefully selected reviewers, which would not reveal the real weaknesses? It was completely different for desktop graphics cards: countless "reviews".

Not sure exactly why in this particular case, but in general a lot goes behind the scenes. It all depends on how individual OEM product marketing/PR wishes to seed samples, Nvidia's own plans (or even AMD's/Intel's, for that matter), segregation of samples to different media for an equitable distribution, and so forth. Standalone desktop GPUs are relatively less-complex pieces of hardware compared to a full laptop computer with its own display, audio, input, security, etc.

Ideally, manufacturers would want their products to be shown in the best light possible. But no product is perfect. Independent testing such as ours will uncover most weaknesses — a simple stress test alone can tell a lot lol.

Even though you may find a particular product on the shelves on launch date, the reviews can still take some time to come out. That's because not all companies take samples from retail and use for media sampling or vice versa. Sales, marketing, PR, are different verticals with their own defined budgets with little scope for cross-connect unless it's a major marketing exercise.

[Vaidyanathan]

4090 mobile is actually the 4080 desktop while the 4080 mobile
No. As the first tests showed, even in MSI Titan - mobile 4090 loses to the desktop version exactly 1.5 times. Therefore, the mobile 4090 is equal to the desktop 4070 and slightly faster than the desktop 3080Ti.

Yes. I mention in the same comment that memory and power differences account for performance differences between the same GPU on the desktop and laptop. I was only referring to the actual GPU die and not the performance equivalence.

[Vaidyanathan]

Couple days ago i got scores in Blender benchmark:



NVIDIA GeForce RTX 4090 Laptop GPU 8256.73 4

NVIDIA GeForce RTX 4080 Laptop GPU 5992.52 1

For comparison the top desktop 3000 series card:

NVIDIA GeForce RTX 3090 Ti 6163.43

Currently the database do not have any 4000 Laptop card in database, those results were erased at least for the moment.

Which particular Blender test is this? We are also in the process of sorting out Blender 3.3 CUDA anomalies with RTX 40 series. Apparently, the test needs to be done twice in quick succession due to the manner in which the render kernels are built. But this is only with RTX 40 cards, though for some strange reason.

[RobertJasiek]

We are also in the process of sorting out Blender 3.3 CUDA anomalies with RTX 40 series. Apparently, the test needs to be done twice in quick succession due to the manner in which the render kernels are built. But this is only with RTX 40 cards, though for some strange reason.

If you still have the already tested RTX 4000 dGPUs, can you please update with the correct Blender tests?

[NikoB]

NikoB, please explain "dangerous for hardware"!

As an analysis of the failure statistics on the forums showed, for those who use "gaming" laptops constantly in games (for hours every day) and without any optimizations, as is from the manufacturer, iron dumps most often begin after 1.5 years, at best after 2.5 of the year. This means. that you will simply lose iron outside the warranty period with a high probability. On a desktop, this is impossible, if only because motherboards and video cards have a warranty period of at least 3 years, and not 1-2 years (more often 1 for gaming hardware), as in laptops. Therefore, playing on laptops is a priori stupid for all of the above reasons. Noisy, hot (dangerous for iron), slow and expensive. It is a priori impossible to play multiplayer on the couch - there is no such comfort of using the keyboard as on the table and in the chair - but if on the table - why do you need a laptop? Buy a lightweight surfboard and put an eGPU on the table via TB4/USB40 eventually, or just have both a gaming desktop and a lightweight (hardware) 15.6-17" laptop for surfing on sofas and beds. It's obvious that 2 devices are more convenient, but for serious games, a desktop is always better, cheaper (at the same speed), quieter and safer for hardware, because it has a much better cooling system.

[RobertJasiek]

When you mention iron, do you mean the chemical element or do you use the word figuratively?

[NikoB]

I deliberately demonstrated, once again, the level of "AI" from google translator. What even a child of 3 years old understands - it does not understand the context. Like all these "ChatGPTs" the same nonsense.
;)

[Vaidyanathan]

If you still have the already tested RTX 4000 dGPUs, can you please update with the correct Blender tests?

Will surely look into this.

[Durzo123]

Which desktop GPU does the RTX 4080 laptop compare to? Is it faster than a RX 6800XT in 1440p gaming? Thanks.

[RobertJasiek]

RTX 4080 Laptop @ 175W ~ RTX 3080 10GB Desktop

[Durzo123]

I see, thanks @RobertJasiek