I guess you just can't help it. It will either be rude remarks, or plain irony.
There's no rudeness. There's robust engagement. If you decide to unleash frontal and brutal attack, which also contains strands of inaccurate or one-sided information in the mix, expect robust response.
You just posted an angry post, because you didn't liked mine.
There's no anger either. Liking is irrelevant. We don't need to like or dislike each other in order to engage. Focus.
The last 2 years AMD is unable to go over 6 Billions, no matter what. They hit some kind of a ceiling. Hope doing more business with Samsung can somehow help them to break it. But for now they gain in one market while losing on another.
2 years is not 5 years, but I am glad we are going in a good direction with basics. Also, if you look again into that annual revenues graph, you will see that it's not even two years, but one single year. AMD posted record revenues in 2022, so 2023 is the first and only year that they were "unable to go over" and surpass the record previous year. Hardly a ceiling, no?
Context is everything. You might remember that entire semiconductor sector experienced a large dip post-Covid and majority of companies, apart from a few, experienced losses in revenues. Entire sector dipped from ~$600 billion in 2022 to ~$535 in 2023. Intel was one of the biggest losers. They lost $25 billion in revenues in last two years. Sasmsung lost billions too. Nvidia was the biggest winner due to AI craze. A few other companies escaped the dip, but others did not. AMD escaped the dip due to rebalancing of segment revenues and signing dozens of long-term deals with hyperscalers and telecoms. They were neither 'losers' or 'winners'. For one year. There is no "ceiling" trend there. Trends may be identified after 2-3 years. They expect $4-5 billion more this year, so any idea of ceiling does not seem to have merit.
If a 250 billions company can't understand market segments that are in fact closely connected to each other, then you agree that there is a problem in the company
Q1 report shows that they doubled client CPU revenues. They are also expected to reach parity in server revenues with Intel within 18-24 months, which is huge for them considering the fact that just 6 years ago Intel commanded more than 97% of this market. Several other segments are ramping up and there will be a big opening with Zen5, the most simple drop-in upgrade across segments without additional costs, plus new opportunities for those who were hesitant with Zen4. I agree that AI will gradually increase with collaboration with Samsung after $3 billion deal was signed for 12-Hi HBM3E memory for MI350.
The only hiccup is client graphics at the moment, after Xmas saturation. I'd increase cadence for GPUs to ~18 months instead of two years, but it's not clear whether this is possible until new US fabs come online, as currect capacity at TSMC is really tight on advanced nodes. Intel and Samsung do not have at the moment as competive nodes for graphics. Everybody suckles milk from TSMC. Consoles are flat at the moment as both Sony and Microsoft have long cadences for next gen, expected in 2026/2027. PS5 Pro refresh will get some traction for a year or two and bring $2-3 billion. That's one of reasons why AMD decided to develop Strix Halo, a platform for laptops and mini-PCs that will act as alternative source of revenues to consoles. More than 150 designs are planned for this year alone with Strix, plus plenty of mini-PCs. We will need to wait and see how volume production proceeds.
Excuses are not arguments, just excuses. And if you are not part of AMD's engineering team, you are just guessing here. AMD was first in chiplets and helped with the creation of HBM. They should know better if it was a good idea to bet a whole generation of GPUs on whether they can make chiplets work with the first try.
There are no excuses, just pure reasoning that should help us explain and understand what is going on. Ultimately, I do not care which company is more successful than the other, as I have no financial interest or personal attachment to any of them. You can say that I am "Samsung fan" because I have never bought Apple phone. That's as far as it gets with my personal preferences.
Chiplets needed to be introduced in GPUs too, in one way or the other. Initial Zen was not perfect either, but one needs to persevere and bite, including bumps in the road. Maturity of Zen is now in full swing across segments. First generation of chiplet GPUs is not perfect, of course, but it's good enough and it paradoxically allowed them to claim back a bit of market share by the end of 2023, as published in market analysis by Tom's Hardware a few months ago. Bumps ahead with no halo model on RDNA4, but I prefer them to shift R&D into halo RDNA5 and focus on creating attractive enough offer for widest possible market, which is broadly understood as mid range. This should be fine.
While they can't compete with Intel in capacity, because Intel owns fabs, which is a reason why OEMs will keep building more Intel based systems no matter how much better AMD CPUs/APUs are, they should have never let Nvidia become that much bigger in the gaming market. They could be making 10 billions by now, they just fear to take the risk to become bigger.
They actually compete with Intel on capacity at TSMC, as Intel is not even able to use their own fabs to full capacity now and outsources almost entire client CPU and GPU etching process to TSMC. It's ironic situation for Intel, but they also block significant amount of capacity for AMD at TSMC. It is true that Intel still keeps majority of OEMs in a tight grip, but this is gradually changing. More designs are planned this year with AMD and Elite X chips, like never before, as several OEMs got fed-up with Meteor Lake shenanigans and plan to purchase a lot of Strix APUs, as outlined by Su during the call. As a reminder, Intel had 90% of laptop market share just 6 years ago. Now, it's gradually dipping below 70%, as Apple took ~10%, AMD ~20% and Qualcomm is poised to make further dent in it. By 2030, Intel may easily drop to 50% in laptop market.
Nvidia took the risk in the mining era and while they seemed to pay the price when mining and demand for graphics cards collapsed they moved forward and now they are a $2.5 trillions company
Nvidia is on explosive AI-rodeo, no doubt, but there's a catch. Nothing happens in a vacuum. Increasing number of institutions and governments are becoming aware of and concerned about immense pressures on regional and national power grids of large AI-systems, including the US government. This is despite the fact that the same entities crave those systems. Total installation network of Nvidia AI systems in 2024 is estimated to consume up to whopping 14 Gigawatt hours of energy, almost twice as much as the most powerful nuclear energy power plant in the world can produce annually. There will be serious challenges here at some point. We will watch and see whether Nvidia is allowed to become an equivalent of Wallace Corporation from Blade Runner 2049.
Where AMD was selling more cores, now Intel sells more cores.
They sell more cores, but not proportionally more CPUs. Intel is very slowly, gradually and consistenly losing overall CPU market share. It's so slow that it's almost unnoticeable among aggressive marketing, but numbers don't lie. Yes, they had some recovery period with hybrid CPUs, but the big picture is that this has not made a bigger positive difference for them. Remember, they lost $25 billion in global revenues in last two years and significant amount of overall market share in last 6 years. The client segment has become de facto their only Trojan horse they would need to fight very hard to keep and cling to. Also, Gaudi3 is estimated to bring no more than ~$500 million this year and server could drop from $3 billion towards $2.7-2.5 in next 12-18 months. Finally, there are currently no big contracts for Intel foundry on the horizon. Have you heard of any multi-billion deal signed in recent months? I have not.
AMD needs to take advantage all the favorable occasions when they present to them. Intel might collapse or it might come up stronger with 20A and 18A latter. If by then AMD is not established in AI and gaming markets, they will have a problem. If Intel fixes it's manufacturing OEMs will having an easier time preferring Intel over AMD hardware. ARC graphics will also start looking more mature and more competitive when Battlemage comes out. What are we going to be talking then. That AMD can't handle so many market segments? Your arguments are not even arguments.
I generally follow what's happening in the industry, as you can see with the amount of information I share with you from diverse sources. Intel showed the slide below. 20A and 18A do not look very good until 2026. They have the smallest number of EUV machines in comparison to TSMC and Samsung. This was one of bigger mistakes they made which made them more dependent now on TSMC. The first high-NA EUV was recently assembled with big media coverage, but it will take years for any commercial chips, and in volume to come to your and my laptops/desktops. Years.
The best Battlemage can do is not allow GPU prices from AMD and Nvidia go crazy bonkers in coming years. So, everyone should pray it is more mature and competitive. And not subsidised by Intel. Apologetics of "duopoly" may yet become disappointed when they realize that Intel joined them to create GPU triopoly. What then? Buy MooreThreads? On the other hand, AMD does not need to be present everywhere and at all times. Like Intel that exited SSD market, in a few years, AMD may decide to leave parts of discrete GPUs for longer if efforts put in RDNA5 do not work well. No reason to keep a segment if it does not work in long-run. And that's fine. Businesses need to adapt and let go sometimes. I am ok with that.
Looking at the tree and totally missing the point? Figures.
Which figures would you like?
AMD done nothing to educate the consumers of how much misleading Intel's marketing was and is. Even now that Intel top CPUs are unstable and degrading, do you see press releases with marketing material using those new information to convince consumers that in fact AMD's CPUs are better? Have you? I haven't. Believe me Intel wouldn't had pass that opportunity. Nvidia definitely wouldn't either.
Well, we saw how Snake Oil embarrassment backfired on Intel... I do not want to see more toxic propaganda against other company. It's brain-washing. Nvidia today released "real AI PC" absurdity, a petty slide showing that their GPUs are better than upcoming AI-laptops. It's nonsense. Noise. We saw hundreds of slides with Meteor Lake architecture. Still, increasing number of OEMs are pissed and plan to release a record number of AMD and Qualcomm designs later this year. Marketing propaganda and beautiful slides can go just as much. Recent CPU instability scandal does not need more input from AMD. It's enough embarrassing on its own. Clear evidence of that is 13 consecutive months of 7800X3D being the best selling CPU in the world.
The fact that AMD needs the AM4 platform to even have a presense at low/mid range says much about their good luck and not about their good planning. Cheap boards where non existent when AM5 came out and the lack of cheaper CPUs was aproblem from day one.
This is industry-wide problem, not specific to Intel, AMD or Nvidia. Companies mostly release top products first, or a cluster of top products. I do not like this approach, but understand why they are doing it, to increase milking and desire from buyters to buy shinny new products bringing big margins.
The X3D chips came and things started turning around again.
Well, it's complex... We need to find out how production line works. It is not clear whether X3D chips can be released earlier, as EPYC is always a priority. I'd like to know whether at TSMC they can run two production lines at the same time or they must run in single batches. Entire fab machinery experiences a significant change when different chiplets are produced. This can happen only once in a while, but not every few weeks. It does take time. As EPYC goes first, they have to produce tones of vanilla chiplets first and package those, then X3D SKUs. Not sure whether this could be mixed up in future, so I am not going to go with hastily negative judgements here. Better to have vanilla chips in the market until X3D arrive than nothing for a few months and then release all of those together. We can't have it all. There is just one TSMC in the world, with a lot of hungry customers...
Tech press promotes the narratives of Intel and Nvidia. Either they start producing ""Snake Oil" Power Points" or we will be the only one talking about their products.
I am against this toxic propaganda. It usually backfires as it is seen as petty and hypocritical. Luckily, we have enough of popular techtubers, like Gamers Nexus and Hardware Unboxed who happily debunk such nonsense for entire tech community. AMD products have probably never been as recognisable and widely discussed as nowadays thanks to electronic media that everybody is hooked to. Everybody who has ever watched a few youtube videos know that the best gaming chips in the world are X3D from AMD.
Numbers say that Intel is still selling much more CPUs thanks mostly to OEM ties, fabs and marketing. 7800X3D is one battle in the CPU war that helpes to give value to the AM5 platform. But it not enough. Either AMD wins the war, or just one battle.
Intel does sell more in volume, but proportionally less, year-on-year. Big difference. The shift is, as I said, gradual. market share is a metric that moves rather slowly in this industry. On average, people keep a PC for 5 years. Besides, this shift is faster in server space, but we don't hear much about it apart from periodical quarterly or annual analysis, as this is boring and under the radar, but it's still happening even if no one shouts about it.
All the press swallowed and promoted that "AMD bad, makes developer not include DLSS" idea, making Nvidia, a company with shaddy practices for 2 decades to look like a victim and you say that it didn't worked? Oh my..... Really?
It did not work. Sure, a bit of hysteria in media and on fora, but ultimately the guy from Bethesda clearly showed a middle finger to noisy media and said that they would roll it out a few months later when it's ready. And that's exactly what happened. Done, dusted and forgotten. It was a storm in a cup of tea, a bunch of conspiracy theories deisgned to damage AMD, but it did not work and the game sold really well plus 7800XT also sold well. So, Nvidia fans of Starfield needed to learn to wait, like every child needs to learn patience and wait for a meal. You know, even internet needs to learn that real world does not suffer from hysteria and that delayed gratification is a skills that we teach little children to master.
RT performance sells cards. If you can't see it, or don't want to see it, it's not my problem.
It does sell, but also increasing number of users are educated by techtubers that VRAM capacity is equally important for RT to work properly in many scenarios. Hardware Unboxed is especially well engaged with this content in tech community and users are gradually more knowledgeable.
Θέλεις να συνεχίσουμε την κουβέντα στα Ελληνικά και όποτε κάνεις κάποιο λάθος, να το παρουσιάζω ως επιχείρημα ότι είσαι λάθος; Διότι μέχρι τώρα, μόνο τέτοιου επιπέδου επιχειρήματα διαβάζω από εσένα. Ανούσια.
I am trained to nitpick. It's my daily job
AMD gain marketshare in laptops but it doesn't seem to keep moving up. Apple is another story. Is your excuse in 2-3 years going to also include Qualcomm?
I don't really care which company has better chips for laptops. I will simply buy one that meets my needs, whoever provides CPU. I am brand agnostic and think we should have 4-5 competitors in order for everybody to put more effort into silicon.
Your rant was just been reported on CNN. Clearly your rant is more than a blip in the digital jungle.
It's the same blip, of course, but it was necessary to provoke you and tone down highly aggressive anti-AMD narrative that was not necessary in the first place. Your long response is much, much better, despite a few bites, but those should quickly disapper too as we focis more on topics.
It's clear that at AMD they realised that without parity with Nvidia in features, they can't sell hi end GPUs at the numbers they need to justify the R&D costs.
Yes and no. It's more complicated. For example, the overall ratio of sold GPUs used to be roughly 6:1 two years ago in favour of Nvidia. Now it is 5:1 overall. Interestingly, 4080 has never sold in 5:1 ratio against 7900XTX, its nearest competitor. It is roughly 2:1 ratio, which is also visible on Steam.
Hardware Unboxed is today the biggest advertising platform for Nvidia's DLSS. What year is it in your room? 2022?
I don't find this to be correct portrayal of them because I watch them regularly. I find them to be well-balanced and neutral. They mock every nonsense from any company.
