Quote from: Kinay Aneo on January 27, 2024, 07:29:19Any autopilot will not be let on public roads today, doesn't matter if autopilot is good or bad. Law is there has to be a human controller to take responsibility.

This is because not a single "autopilot" has come close to the level of understanding the traffic situation and correct reaction as a living person. Those. complete failure of research in the field of autopilots. =)

Quote from: Kinay Aneo on January 27, 2024, 07:29:19Guns not being allowed in many countries doesn't make gun research a failure.

In general, complete nonsense from the average person. The weapons are not even close in terms of complexity. Weapons do NOT require intelligence. It is the person who shoots, not the weapon. And no one has yet made autonomous cyborgs for the same reasons that an autopilot has not yet been made.

You really don't understand anything about NTP. But you make extremely primitive philistine judgments. You are apparently one of the humanists...

Quote from: Kinay Aneo on January 27, 2024, 07:29:19Dunning-Kruger effect

I do not care. It is a fact that since 2016 I initially warned all those arguing with me on a bunch of forums that all autopilots would 100% fail by this date for reasons obvious to any professional - the degree of complexity is incredible, in order to completely remove the person responsible for making key decisions circuit. And everyone who tried to convince the stupid public and politicians of the opposite was just ordinary arrogant scammers who cut grants and budgets. Many of them deserve long prison sentences, but alas, such is our dirty world, where scoundrels quite easily escape responsibility with the connivance of the stupid crowd...

Quote from: NikoB on January 26, 2024, 20:24:20It is the "legal problems" that are 100% proof of the complete fiasco with actually working full-fledged autopilots.

Any autopilot will not be let on public roads today, doesn't matter if autopilot is good or bad. Law is there has to be a human controller to take responsibility.
Guns not being allowed in many countries doesn't make gun research a failure.

Quote from: NikoB on January 26, 2024, 20:24:20I initially understood the level of development well and assessed the risks well

Dunning-Kruger effect

Quote from: JohnIL on January 26, 2024, 19:05:08Intel sort of stuck in a rut and they are not making enough changes to move the needle. It seems that Intel is just trying stuff now hoping something sticks. It does appear what Intel is doing is more sour grapes with Apple dropping them and their desperate moves to prove Apple wrong. I don't even think that's a focus Intel should be on and should focus more on AMD and Qualcomm (ARM) in the PC market because Apple will never go back to Intel.

Intel just officially announced that it will make losses again in the first quarter of 2024.

Quote from: Kinay Aneo on January 26, 2024, 19:49:59Autopilots `collapsed` because of legal issues, not because they can't be made. Tesla auto-drives quite well. But no one will legally let full autopilots on public roads today though, because there's no laws to determine who is responsible if autopilot kills a man or damages property - and it will happen of course. So far legally there always has to be a human controller taking responsibility for accidents.

You just make me laugh, either with your naivety, or with your complete misunderstanding of the real situation. It is the "legal problems" that are 100% proof of the complete fiasco with actually working full-fledged autopilots. Which ALL their developers have directly admitted. But in the lying press, but in the COURT. And I have written on various forums since 2016 that they will 100% fail, because, unlike ordinary people, I initially understood the level of development well and assessed the risks well.

Quote from: NikoB on December 12, 2023, 17:35:37market for autopilots collapsed

Autopilots `collapsed` because of legal issues, not because they can't be made. Tesla auto-drives quite well. But no one will legally let full autopilots on public roads today though, because there's no laws to determine who is responsible if autopilot kills a man or damages property - and it will happen of course. So far legally there always has to be a human controller taking responsibility for accidents.
It doesn't mean autopilot research has stopped or has to stop.

Intel sort of stuck in a rut and they are not making enough changes to move the needle. It seems that Intel is just trying stuff now hoping something sticks. It does appear what Intel is doing is more sour grapes with Apple dropping them and their desperate moves to prove Apple wrong. I don't even think that's a focus Intel should be on and should focus more on AMD and Qualcomm (ARM) in the PC market because Apple will never go back to Intel.

Quote from: Just a guess on December 12, 2023, 00:12:32It might have to do with the fact that AI Data Center / Enterprise companies are willing to pay a lot more money to HBM3 manufacturers for it than all the customers that would be willing to order such laptops?

It will be like the collapse of dot-coms and video card mining - most do not understand that all these investments are unprofitable in advance and in the end there will be a shock of paying off debts for venture investors.

But the PC/laptop market is a real long-term development. Greedy people look for quick money, smart people invest for the future.

They won't be able to make any "AI" in the next 50 years, but they will master and spend huge investment budgets.

Just like the market for autopilots collapsed, which was written about a lot this fall and which I have been warning about since 2016. More than 100 billion dollars were invested - in reality, nothing came out...

Quote from: NikoB on December 11, 2023, 20:38:55I see no reason why an additional 16-24GB HBM3 cannot be soldered under vram in expensive laptops without an external video card (or with it) just for igpu operation.

I understand when they save money on cheap laptops, but why do they save money on expensive ones on the HBM3?

It might have to do with the fact that AI Data Center / Enterprise companies are willing to pay a lot more money to HBM3 manufacturers for it than all the customers that would be willing to order such laptops? And pay far higher margins for the privilege as well. Just a guess.

Quote from: Hotz on December 07, 2023, 09:28:50Yeah, but to be fair AMD was also very quiet when the 780m iGPU came out. No benchmarks whatsoever from the company itself. Though the 780m was indeed not much of an improvement over the 680m.

It must be taken into account that all igpu are as fast as possible (and this is what is advertised in the press) only with the fastest possible memory. What is the fastest memory for 780M? LPDDR5 7500. And how many laptops are there in practice with such memory? In those with 2 slots, it is not there by default. It's good if there is a discrete card, but the built-in gpu is definitely slower.

Those. people don't get a reference solution at all in most laptop models if they disable dgpu (or don't have it at all) and use only igpu if there are 2 slots for memory modules, where the maximum memory frequency is 5600, not 7500. Only on this difference, the 780M loses about 25 % graphics performance.

Until they start using HBM3 in built-in cards there will be no sense. Those. I see no reason why an additional 16-24GB HBM3 cannot be soldered under vram in expensive laptops without an external video card (or with it) just for igpu operation. With VRAM speeds starting at 256GB/s, this will be a completely different level of integrated graphics performance.

Apple seems to have added a 512-bit memory controller to speed up igpu as much as possible - but what do we see in practice? But in practice, if you believe the tests in local reviews, even the M3 Max with 300GB/s of RAM according to Apple's declarations, in reality produces only a measly 120GB/s - where are the promised 250+? It is for this reason that the integrated graphics of Apple SoCs slow down very much.

It all comes down to the extremely slow RAM on Apple and even more monstrously slow on x86 - I have never seen more than 85GB/s in the best Intel solutions.

I understand when they save money on cheap laptops, but why do they save money on expensive ones on the HBM3? Increased consumption? Well, compared to 250W+ laptops with a discrete card, I don't see any problem in this. It is unlikely that the HBM3 controller built into the SoC will consume more than 30W on its own. But even 256GB/s+ speeds up igpu several times.

There's so much trash talk on MTL that it doesn't surprise me Intel's being quite. I'm not reading anything into it.  It could either be a sleeper and better than we expect, or it'll suck.

The embargo's up next week.  We'll see then. Until then, I'm taking anything with a bag of salt and a lot of skepticism.

I'm looking to buy a laptop early next year. AMD's 8x40HS and U procs seem to be rebadges of the 7000 series, so it looks like plenty of rehashing going around.

No major improvements should at least make previous generations a bit cheaper.

Quote from: ArsLoginName on December 07, 2023, 02:54:25the quietness (and lateness of the product to market) with the recent information stating no MTL on desktop should tell you everything you need to know: MTL is a lower power design that does not offer much in terms of improvements.

Yeah, but to be fair AMD was also very quiet when the 780m iGPU came out. No benchmarks whatsoever from the company itself. Though the 780m was indeed not much of an improvement over the 680m.

Now I don't know how it was when the Radeon 680m came out. Because that iGPU was a big improvement over the Vega 8 iGPU. Did AMD show any benchmarks, tests before release? I honestly can't remember if they did or didn't.

So in the end it's difficult to say if "silence" means something good or bad. Fortunately we don't have to wait long anymore....

Quote from: davidm on December 06, 2023, 20:06:19If it can match an M3 chip -including efficiency-, they're back in the game, otherwise, after this many years, x86 is kind of a joke.

Ha ha, that's hilarious. Maybe Luna Lake (which is a U class CPU) could match M3, but by then M4 or even M5 will be out. And of course Apple's SoC will destroy the Intel in performance while using less power. And no I don't own Apple.

Nobody has anything to add regarding upcoming performance and power figures. There should be leaks at this stage but there aren't. One would think that Intel itself would start to market with ew ads since launch in 1 week. But the quietness (and lateness of the product to market) with the recent information stating no MTL on desktop should tell you everything you need to know: MTL is a lower power design that does not offer much in terms of improvements. It is really about packaging and more EUV experience because Intel wants and cares more about ARL being competitive with MacBook at similar power consumption than this stop-gap product. Intel 4 isn't even a node that will be offered by IFS if I remember correctly. Internal only. Further, they are skipping Intel 3 (which is supposed to offer another 18% perf/W increase) for consumer products. Intel is only talking about 20mA client interest. That is where they are aiming to be.

*Yawn*, more MLID bunk. If I wanted to waste my time on fake news/made up rumors, I would visit his channel. But since I don't, seems it's time to add notebookchat to my news feed blacklist. 👋

"Superior power efficiency" is more about power, heat, battery life than outright performance. It has to be "fast enough" and have the promised AI support, but if there are significant advances in efficiency, it's a win. Anyone who really wants GPU performance should get a dedicated GPU.

If it can match an M3 chip -including efficiency-, they're back in the game, otherwise, after this many years, x86 is kind of a joke.

Frankly I hoped that you had something more substantial to present than (once again) a MLID rumor.

I admit I've watched his show sometimes as well, but I've lost interest lately as most of them are only rumors, claims, hearsay. I'm tired of this s***.

That being said, it is also tiring that we still didn't get any substantial information about the iGPU performance from Intel itself. It's only 1 week until the market start of MeteorLake and we still don't have anything. I mean: w.t.f.