Intel Meteor Lake Analysis - Core Ultra 7 155H only convinces with GPU performance

[bennyg1]

There are allegations that Intel are seeding their review samples with faster LP-7467 "MHz" RAM than the actual retail units are getting (6400 or 5600).  Which, with iGPU performance being very sensitive to RAM subsystem performance, would potentially render reviews like these overstating the actual performance the buyer will receive ......
The day-1 review would be unable to mention this, but I hope, if true and legit retail LP-7467 versions of this SKU do not exist, this kind of unrepresentative review unit shenanigans would be added as a footnote to counter the potential for misleading buyers. And to hold deceptive sales practices to account.

[Somebody]

It looks like Intel lied us with its efficiency shenanigans. Look how impressive was this article/presentation:
www.notebookcheck.net/Intel-reveals-Meteor-Lake-Core-Ultra-bets-on-efficiency-AI-a-new-iGPU-Intel-4-and-TSMC.752748.0.html

Now look how it really performs. The CPU performs almost the same as or even worse than last years Raptor Lake (OK). But what about efficiency promises?

Negligible efficiency difference on CPU and GPU side (above mentioned test also proves it). It is still power hungry and hot as a hell chip. Where all efficiency promises went away, Intel?

Now I understand why OEMs are angry about Intel:
www.notebookcheck.net/Laptop-OEMs-to-allegedly-shift-away-from-Intel-s-underwhelming-Meteor-Lake-and-offer-more-AMD-Ryzen-8000U-H-models.770866.0.html

The only hope is from AMD now. Let's hope it will bring something more powerful and efficient on the table with its Strix Point

[Dan6]

What a disappointment.. Better GPU is good, same CPU performance is ok, but I expected better battery life with their new "efficiency" cores.. But looks like they are still added just for numbers in benchmarks. New AI engine? Go put it in your *!@, I mean datacenters and do what you want with them instead of making new marketing marvel out of it.

[NikoB]

Take the hint, seriously. If AMD cared about laptop users they'd not be selling all their chips to handhelds. If OEMs care more, well, they'd be releasing better devices with good pricing, availability and supply. But nobody gives a sheet, so why should you?

I already wrote in the first comment of the topic that there is and will not be any hope for AMD, because... the ratio of sold Intel vs AMD processors is 5:1 (in laptops it can be 4:1). This leaves no chance for AMD, which is dependent on supplies from TSMC/Samsung. In addition, they are forced to give a significant share of the surplus value to these companies, and Intel, as soon as it pays for the plant, takes everything for itself. The very scheme with an intermediary for SoC production excludes sufficient profit. And if you increase the price of SoC for consumers, then interest in AMD will disappear. They are interesting as long as their laptops cost the same as with Intel or are cheaper - more expensive, there is no interest.

AMD is 100% not interested in increasing market share through dumping, because it has no cache reserve and Intel diligently keeps it on a hungry leash, just so that they serve as a buffer with the antitrust authorities. No more. There is no real serious competition between Intel and AMD with the former's share being 5:1. This is all an artificial game for antimonopoly officials.

Apple could buy Intel and get all the x86 patents at once, but apparently they no longer give it power, then we would get an even bigger monster on the market vs small AMD.

The situation is simply that there is only one company on the market with its own factories and a gigantic production volume many times greater than TSMC's volume in the x86 market - Intel. It has no real competitors precisely because of its monstrous market share. All this is the fault of the US antimonopoly authorities, who deliberately allowed this situation to happen more than 10 years ago, for geopolitical purposes.

Obviously, any sensible buyer who understands the laptop market will not refuse a high-quality laptop with a 7840U at the moment, but there are simply no sane models on the market anywhere, which is what I described in the first comment. We are only discussing theoretical calculations - but in practice it is impossible to buy an interesting (adequate to progress at the end of 2023) model with Zen4 Phoenix in the 16-18" class.

Intel literally imposes its useless options on buyers, because...the ratio is 5:1 in their favor for SoC production. Well, much like a dictator creates the appearance of legitimate "democratic" elections. This is such a permanent game with the stupid, majority of the population...

[billou]

in real life gpu and cpu work together , even in windows , a simple

[RobertJasiek]

NikoB, AMD cache -> cash?

[NikoB]

cash

Are you kidding me Robert? From the context it is obvious that we are talking about money in AMD accounts. It simply does not have them in such quantities as Apple to carry out deliberate dumping against Intel products in order to squeeze out at least 50% of the market from it, after which it will be able to consistently take much larger volumes from TSMC/Samsung, in priority mode, compared to other customers to their production lines. The higher the order volume due to a larger market share, the more important such a customer is for TSMC/Samsung and, moreover, the lower the cost of batches for AMD, which means the higher the profit per SoC copy. AMD can break this vicious circle either with a super breakthrough compared to the performance of Intel processors or by methodically squeezing market share from it, which requires a lot of money, which, alas, it does not have.

[whoami?]

You might think that the removal of the P-series makes it easier to find a suitable processor, but this is not the case. As mentioned before, the H-series chips now cover an even wider TDP range and customers just don't know what performance they will get.

That was the whole idea. To mislead consumers. Consumers will be reading a review of a high cost laptop of high TDP and think that the cheaper laptop having an H series chip will be offering close or even the same performance.

[NikoB]

You might think that the removal of the P-series makes it easier to find a suitable processor, but this is not the case. As mentioned before, the H-series chips now cover an even wider TDP range and customers just don't know what performance they will get.

That was the whole idea. To mislead consumers. Consumers will be reading a review of a high cost laptop of high TDP and think that the cheaper laptop having an H series chip will be offering close or even the same performance.

This is not a problem of fraudulent business in general, but a problem of a society that is inadequate and does not want to do everything so that business and trade behave in a civilized and adequate manner. And the information about the product was complete.

Specifically for the laptop market, this means that society, in each individual country, must demand from regulators that they require, at the trade level, to report, for those same laptops, the actual consumption levels in each selected factory profile and the expected performance level based on a set of publicly known tests (for example, developed by government agencies with source codes). As an example. But in general, an accurate description of the product, which excludes fraud with the characteristics of a specific model of any product on the market. So that it is impossible to sell under one product article a completely different product in terms of characteristics and quality - everything must be clear within the framework of the mandatory tolerances that manufacturers are required to publish for each new product.

The buyer has the right to know what he is buying. Any attempts to deliberately confuse the buyer or hide important information from the product (and this is exactly what happens all the time in the market) must be severely dealt with by state regulatory authorities. This is precisely the goal of such bodies - to prevent lawlessness on the part of uncontrolled and unprincipled businesses. And constantly develop and maintain minimum requirements (and what information must be disclosed to the buyer during the sale) for different groups of goods

Try to obtain from manufacturers a mass of technical data on a specific product on the market - you will quickly discover that they hide almost everything that can provide a meticulous consumer with the opportunity to sue them for violating banal product tolerances. In fact, there are no tolerances for almost anything in ordinary goods precisely because the indication of a mass of technical tolerances immediately creates the possibility of lawsuits against them.

Why then are people surprised that Intel is doing something like this? After all, the population allows them. As do other business representatives.

Therefore, dissatisfaction with the current situation should be addressed not to Intel or some other business representatives, whose goal is always to screw you with the maximum possible profit, at the lowest possible cost, but to your local society (state) - why does it allow such product descriptions that allow type of the same product, sell completely different in quality/characteristics.

Doesn't similar things happen in other spheres of human activity? The same food, clothing, shoes, medicines, etc. This is a problem of a society that is inadequate to the challenges of the times.

And this problem can only be solved by constantly increasing the level of general education of the majority of the population, so that this majority creates sufficiently competent pressure on politicians and regulatory bodies, which the powers that be and the business layer diligently avoid, on the contrary, worsening education in public schools and universities. This is what the recent PISA 2022 results have proven.

[Sergey]

As a person who has been related to AMD for a long time, I can say that AMD is still interested in VLIW5 only servers, supercomputers, number crushers, etc.
There is no attention to consumers. The number of cores has not increased since Zen2 in either PCs or mobile computers. The APUs are practically abandoned. Video cards are a disaster due to poor architecture, lack of chipsets, too much cache and a small bus. There are no problems with production or money for development, there is no innovation because ordinary AMD consumers are not interested. If you are an AMD fan, then just accept it or leave them like me in the stagnation they are in. As for Meteor Lake, these chips disappoint with a lower frequency, almost the same energy efficiency and lack of L4 cache.

[NikoB]

and lack of L4 cache.

They do not need L4 cache and L3 is not needed, with RAM speeds of 300GB/s. These are all crutches of the x86 platform, which are designed to hide from buyers the discrepancy between the performance of slow RAM cores. It is already 4 times behind in its development, at least compared to the level that we have in PCs and laptops. Apple is a little better, but in fact they are lying about the fast 512-bit memory controller in the M3 Max - tests on this site indicate the extremely low efficiency of the memory controller of these chips - only 120-130Gb/s, instead of the promised 300-400.

With a balanced architecture where the cores are not suffocated by monstrously slow memory, no cache other than L1/L2 is needed.

Look at the insanity in the AMD Zen4 laptop series - it is stated that there are 28 pci-e 5.0 lines (that's why they are needed there and why was the transistor budget spent on them, instead of igpu, which is several times weaker there than in Zen4 Phoenix?), and in order to service them at the same time you need at least 120GB/s (like Apple) - and this is only for devices on the pci-e bus, but they should not occupy the entire memory bandwidth! But in reality, Zen4 has a shameful 60-70GB/s. Those. There is clearly a lack of memory bandwidth, which, like 15 years ago in x86, has a 128-bit bus.

What's the point of attaching high-speed pcie-e 5.0 devices to such a slow memory if they quickly eat up all the bandwidth and there is simply nothing left to work on the processor itself and the OS/software for other tasks?

[NikoB]

which are designed to hide from buyers the discrepancy between the performance of slow RAM cores.

f*cked google tr - "which are designed to hide from buyers the discrepancy between the performance of slow RAM and fast cores."

only 120-130Gb/s, instead of the promised 300-400.

120-130GB/s

[johny1]

hey, please tell us whether you can't run efficiency test on efficiency cores, or just paid not to do so?

[Hotz]

There are allegations that Intel are seeding their review samples with faster LP-7467 "MHz" RAM than the actual retail units are getting (6400 or 5600).  Which, with iGPU performance being very sensitive to RAM subsystem performance, would potentially render reviews like these overstating the actual performance the buyer will receive

True, but recently another test from bilibili was posted on WCCFTECH, with a ranking...
wccftech.com/intel-arc-igpu-on-meteor-lake-cpus-competitive-gaming-performance-amd-rdna-3/

... and you can see there he ranks the iGPU of the Intel Core Ultra 155H with 5600-RAM over the AMD 7840HS with 6400-LPDDR5 RAM. So it's not like the new Intel iGPU only wins with the fastest possible RAM.

[Hotz]

btw. I'm not saying Intel is "the" winner, but meanwhile consider it as equal in performance to the 780m.

When I first looked at the Notebookcheck results the iGPU seemed considerably worse, but various techreviewers on youtube youtube have shown it performs better (and they also compared it with non-U chips). It seems to depend on heavily on the games.

In any case this is good for us buyers as we can choose from 2 options soon.