What is HDMI port maximum resolution ?

Shouldn't the proper comparison for performance be against the 7840HS in a machine that maxes out its official sustained power at 54W? Comparing Meteor Lake against a competitor that gets to use less than half the sustained amount of power is hard to justify, though I would be curious to hear the authors reasoning.

When you do use the 7840HS instead of the 7840U (=343027&specs[]=342689]based on notebookcheck data), that convincing W of the Meteor Lake chip turns in to a tie at best and a convincing loss in most cases. Anyone can look this up on here and a comparison should have been drawn in all the benchmarks to illustrate that Intel (once again) is brute-forcing performance-parity through massively higher power consumption. Something that can be forgiven in a Desktop-PC but is absolutely critical in battery-constrained applications like laptops and gaming handhelds.

The usual technical details are there, but I am pretty disappointed by the skew given to the performance analysis of this chip. It does not seem fair or representative of what users will experience in practice.

The article is disgusting advertising.
The article says that the Meteor Lake GPU is superior to the 780M, but in tests it cannot beat it even against a low-voltage processor.
The article advertises an NPU, but in terms of performance it is inferior to the Meteor Lake GPU (an NPU is not needed
). There is no software available to users, which was confirmed when setting up the software.
All competing H chips from both AMD and Intel have been removed from the article. To make Meteor Lake look as good as possible.
The authors customized the provided equipment to make it look better, which is prohibited by testing rules.

Quote from: Ram on December 19, 2023, 15:18:05I guess the point of the NPUs, at least from the smartphones SOCs is that they are "more efficient" for those AI tasks, impacting the battery less than the GPU? Perhaps that could be an advantage in laptop SOCs?

Yeah, it's the same in Apple M chips, "slower but more efficient", but no one bought into it and everyone is just using GPU. Apple also adapted Stable Diffusion model back in the day but it never became a thing. GPU algorithms became more efficient with time due to research while NPU was just forgotten.
The only thing using Neural Engine on Apple is OS at the moment - scans your images when asleep, figures out your battery usage patters for battery charge limiter, etc. I expect the same to happen on Windows, with more telemetry of course.

Quote from: A on December 19, 2023, 11:50:54

Quote from: RobertJasiek on December 19, 2023, 09:37:59Can you support this by citations or links?

You haven't read the article? It clearly says iGPU is more powerful in AI, even with graph. Right above the flashy Stable Diffusion images, they should at least attract your attention beyond just checking out the outdated benchmarks.

Apple's NPU is almost twice as fast and still everyone is using GPU for NNs, because it's even faster. Plus NPUs are limited in data format selection. Same thing here. And NPU is used by OS to... drumroll... ocr/object detection/tag your images on PC. Totally not for ad targeting wink wink.

I guess the point of the NPUs, at least from the smartphones SOCs is that they are "more efficient" for those AI tasks, impacting the battery less than the GPU? Perhaps that could be an advantage in laptop SOCs?


Every time I read those power management sections.... I get more and more surprised. In load the med is 64W and the max over 100W? On a laptop? It's astonishing.

I'm disappointed that MTL isn't, customer-facing, as revolutionary as Intel tried to sell. However, I acknowledge it seems it's going to be the cornerstone of its strategy that COULD allow to greatly improve in the future, with the heterogenous nodes.

It is a shame that this product did not come out with AMD ZEN 4 7040 Phoenix in its beginnings 1 year ago it would have already been a sales success, but we do not know why Intel has briefcases paying the Large OEM manufacturers so that they do not release products with AMD when the Zen 4 Phoenix has been the best processor ever built waiting for the next generation.
Many people are starting to talk about Zen 5 + RDNA 4, which they say will be twice as much as this latest Zen 4 Phoenix.
I would like to see comparisons in which it was seen that the laptops had the same hardware when performing the tests except for the processor.

Quote from: RobertJasiek on December 19, 2023, 09:37:59Can you support this by citations or links?

You haven't read the article? It clearly says iGPU is more powerful in AI, even with graph. Right above the flashy Stable Diffusion images, they should at least attract your attention beyond just checking out the outdated benchmarks.

Apple's NPU is almost twice as fast and still everyone is using GPU for NNs, because it's even faster. Plus NPUs are limited in data format selection. Same thing here. And NPU is used by OS to... drumroll... ocr/object detection/tag your images on PC. Totally not for ad targeting wink wink.

Quote from: A on December 19, 2023, 07:50:36NPU [...] It'll only be used to spy on you anyway, because for every other consumer task running neural on iGPU is faster.

Can you support this by citations or links?

Came as far as running some absolutely sh*tty Stable Diffusion model on it but "forgot" to run the latest Geekbench 6.2 and Cinebench 2024.

Not even saying NPU is a thing in Apple chips for years already yet it's marketed like it's THE NEW THING all around the article. It'll only be used to spy on you anyway, because for every other consumer task running neural on iGPU is faster.

Miracle didn't happen with Meteor Lake, I'm done with Intel bs. Waiting for Qualcomm.

It's nice to see Intel's Integrated Graphics(Tile) getting closer to AMD's Radeon 780M in gaming but the Author needs to indicate for any Blender 3D Cycles rendering tests if that's the Blender Cycles CPU rendering or the Blender 3D GPU Accelerated Cycles rendering that accelerated on the GPU instead! And that's not been made clear and The Tech Press is more fond of using Blender 3D's CPU cycles rendering option at the cost of any Integrated Graphics never getting tested for Blender 3D GPU accelerated Cycles rendering that make's much quicker work of Cycles rendering than any CPU cores and CPU cores that get tied up at 100% doing the Blender CPU Cycles rendering until the rendering run is over. 

And overall the Tech Pres has been robbing the readers of much regarding Testing Blender 3D on Integrated Graphics unless they are testing Apple's products. And so when will there be some more comprehensive AMD iGPU to Intel iGPU testing where the Blender Cycles rendering is GPU accelerated as no one really uses Blender 3D CPU cores Cycles rendering save mostly the Tech Pres for CPU core stress testing when it's the GPU cores that can accelerate Blender 3D cycles rendering and get that done much quicker.

But NPUs on x86 processors may come in handy for Blender 3D AI accelerated de-noising at some point in time if that's not already being accelerated on the GPU Shader cores that have matrix math instructions as well and with the latest GPUs having their Own Matrix Math units/Tensor Cores. But I'd like to see the dedicated NPUs on x86 processor doing some AI related filtering workloads on more applications than GIMP.

And really I'd like to see Notebookcheck get some Blender 3D focused Laptop Tasting Articles published on a regular basis where Laptops and their Integrated Graphics and even Those Mini desktop PCs that use the Mobile APUs/SOCs from AMD and Intel get tested for all around Blender 3D performance on processors and those processor's Integrated Graphics. So some Blender 3D View-Port testing and Cycles iGPU accelerated Cycles rendering is done as well. There's already Plenty of Blender CPU Cycles rendering done but little to no iGPU accelerated Blender Cycles rendering done and those Mobile/Laptop APUs/SOCs are even more popular in the Mini Desktop PC from Factors as well but Little Blender 3D iGPU accelerated Cycles rendering testing is done on Mini Desktop PCs that are using those same Laptop/Mobile APUs/SOCs that see the Mini Desktop PCs more likely to come with just the Integrated Graphics while Laptops are likely to come with discrete Nvidia laptop GPUs mostly!

The good thing with meteor lake laptops seems to be standardized 32gb ram. 16gb in above 1k price should be banned. this acer and meteor lake although not so spectacular in any way but not bad at all. efficiency is better than last-gen but still sucks in comparison to phoenix (and probably to hawk point and off course strix point). and igpu is a giant step in right direction. this should have been available a year ago to compete favorably with rembrandt and phoenix. finally buying an intel powered laptop doesn't look like an stupid idea. its just not so wise unless at a good value.