I don't know a lot about Twitter / X or Musk, but a quick Google search indicates that Twitter rarely turned a profit and lost money about 80%+ of the time before Musk bought it. So if it's hemorrhaging money, it isn't the first time by a long shot. The exception to the rule would be: Twitter / X isn't losing money! Revenue was declining before the buyout, so the biggest loss that could be attributed to Musk is almost certainly in market cap.
I've been trying to figure out if Intel is on track to corporate health or if they are doing the wrong thing, failing, and continuing to do the wrong thing (as almost every company does). This is just one data point and not necessarily a negative one but, as I understand, Intel will include AI accelerators in every Xeon CPU complex. There could be customers asking for this, but I'm not aware of any. I'm not even aware of Tesla wanting this product. Tesla uses a different infrastructure but perhaps they will come to use Intel CPUs in time. AMD EPYC will be the same. I don't understand what is driving this but AMD and Intel are obviously massively incentivized to make customers buy a product they are not asking for. AI virus and trojans are going to be extremely difficult to detect, at least at first, so perhaps American spy crafters are forcing this? This could be an opportunity for a company like Ampere to take a piece of the data center.
Pat Gelsinger claims their 1.8um node is on track with too-good-to-be-true yields. I'm kind of choking on Intel claims but they will rock the world if they can make good and get even remotely close to their claims. Based on: Lunar Lake = 144mm^2 @ 3um Assuming Arrow lake comes in at 120mm^2 @ 1.8um (wild guess) Pat Gelsinger's claimed yield of 0.4 D0 and he said that translates to 75% yield. They would achieve 75% yield if working with die size of 100mm^2. I believe this process is still a few quarters away so they may have ground their chiplets down to 1 cm^2 or less by then. Personally, I will not gamble $0.01 on this process and yield claim. I hope it's true but I have zero trust in Intel.
I think the death knell for Intel as it is now has begun today: https://www.cnbc.com/2024/09/16/int...s-into-subsidiary-weighs-outside-funding.html God, what a horribly mismanaged company.
I bought in (again) at just under 19...sell off is overblown. Not much needs to go right for INTC to rebound to a company worth 25-30.
I've been studying performance metrics for a few months and have come to two inescapable conclusions. 1) Tech media is as trustworthy as a two year old asked to hold your cookie. 2) Both Intel and AMD CPUs are of such extreme performance, they are hitting a performance limit of windows. If a CPU were released that were 2x faster, it would be barely faster in windows benchmarks that are graphically intensive. This provides a distinct advantage to Qualcomm and an opportunity to catch up. Meanwhile, both Intel and AMD CPUs continue to scale under Linux. Pure compute under Windows also continues to scale.
I take it, you do like Intel and reading a negative view on the company hurts your eyes? It would be great if you could present a contrasting position. Why is Intel a good buy? How are they going to succeed? Will they regain market share? Do you have any confirmation they can deliver on their wildly bold claims? Is 18a as close to production ready as Pat Gelsinger stated a month ago? My questions are sincere. Please balance this thread with contrasting information. You will not be attacked for doing so. It will broaden the view of all of us.
Windows really only has two options in the long run: 1. Ditch legacy support and build the OS from the ground up for modern hardware with none of the bloat and compromise. Legacy can air gap or upgrade. Nobody is forcing them to use modern hardware/software. 2. Stay the course and perhaps cutting edge hardware companies will make a concerted push for another OS to sell more product. Windows is a total cobbled together mess under the hood. The UI sucks ass. Truly a GUI for simpletons. Intel literally had to create a hardware scheduler in the CPU to work around Windows' crappy one. AMD is having to use software hacks to achieve the same goal. All this silliness because they do not want to break legacy support. As a shareholder, FFS make a better product, Microsoft.
Tom's Hardware has done a piece on a new Intel microcode release. According to Tom's, microcode bugs caused over voltage in a few situations. This over voltage is what can damage the ring bus, according to my Namesake's web site. A fix has been released. I'm not sure what this means, as Intel is out of 14900 inventory, according to comically incorrect web reports. It would be interesting to know if Intel can ramp up 14900 production to fulfill warranty claims. That would be preferable to product refunds, IMO. They need to cling to their customer base. At this point, we don't know if the microcode fix will work. I hope it does.
Apparently, Intel has cancelled the Arrow Lake refresh. Can anyone confirm this? I'm not sure if this is a rumor or fact. Whatever the case, it's a wild turn of events. Intel has nothing to respond to AMD, except for legions of myopic YT reviewers. They are short of, or completely out of, top bin Raptor Lake inventory. Every gamer I know runs AMD/Nvidia systems.
$INTC currently trading at $23.26 ahead of new product announcements later today. This is up a tick, suggesting the market is positive about the future of INTC.
My expectations for Arrow Lake R and Lunar Lake are low. They will be good processors, from a performance point of view, but will not dominate. Intel's AI is also not world beating. Intel's next gen APUs are unlikely to be exciting but their integrated RAM strategy could be a world beater from a price/performance standpoint.
Intel's Core Ultra CPUs are slow and power hungry. The Intel fan base has been crapping on Zen 5 but they need put more energy into Intel's survival and less energy into misrepresenting AMD's situation. I hope Intel has a trick up their sleeve that isn't visible from street level. I'm not sure why they bothered to refresh Arrow Lake before 18A is released but I'm not counting on 18A being available on Intel's current schedule, either.
On package RAM has now been removed from the Intel roadmap. Lunar Lake will be the only device to have this. I'm not sure what the announcement means for ARC. They aren't dropping it, so that is probably a good thing. They aren't even backing away from having a discrete GPU but they are "not going to be aggressive" with it. I guess that means they are going to phone it in? Opinion: I like the idea of memory on package. Integration got us here. Integration will move us forward. One day, memory will be on package. This seems like the right time for such a move. I would buy a 16GB laptop that wasn't expandable, if it was affordable. Integration should make it affordable. I would buy a 32GB laptop that wasn't expandable, if it was a tiny bit cheaper than an expandable unit of similar capability. For the vast bulk of people, 16 or 32GB are well sufficient. ARC. I love the idea of a discrete GPU in m.2 form factor. There are countless uses for this configuration in everything from laptops to portable devices. m.2 is 4x PCIe lanes so with PCIe v5, an m.2 card has similar bandwidth as an 8x GPU connected to PCIe v4. Obviously, this is ridiculous with giant GPUs that dissipate hundreds of watts. For small/light/battery, it seems like an obvious evolution. I hope they stick with this.
I would only support on package RAM if it leads to substantially faster speeds due to shorter traces. And it better have more onboard than I will need initially. I am not a fan of the path to everything being integrated, as it makes everything even more consumable and harder to troubleshoot/repair/upgrade. I am still fairly certain that's where we are headed though.
I suspect most or all gamers and geeks would have a negative view to on package RAM, just as you do. I'm the only geek I know who thinks it's a good idea. People will come around to seeing the wisdom of this configuration but it will take some time. There must be 300M laptops in North America that will go their entire service lives without being upgraded. 16GB is plenty of RAM for office productivity service. 32GB of RAM is more than sufficient for just about any IT worker. The only people who want more are lunitics, like us. Gamers. Hard core hardware guys. There will always be a need for super high end, expandable, hardware but it would be a whole lot cheaper if 95% of people didn't have to carry the burden of our lunacy. 64GB laptop? OK. When will we need that? Not now. I have 64GB of DDR4 in my desktop because I'm a lunatic. I've never seen it top 30GB of use, even when I have multiple copies of CAD open and a few virtual machines running in the background. I'm sure the swap partion has never been touched by live data. If my machine was 32GB, I still doubt the swap would have been touched. Keep in mind, I'm a linux user.
