Anyone thinking about nibbling at it? or too many unknowns?
Not me. As you say, too many unknowns. The macro economic outlook is anything but bright, not to mention cough Taiwan.
Once inflation tops out I will be looking at a number of stocks to invest in. I think AMD is worth a look because I believe they will rule the data center until 2025 which, in tech, is the foreseeable future. If AMD continues to make the kind of moves they are, then AMD may even rule the data center market beyond 2025.
AMD also meets my minimum 3year-hold investment window as well. My only problem is my portfolio is already too AMD heavy.
I think AMD is worth a look because I believe they will rule the data center until 2025 which, in tech, is the foreseeable future. If AMD continues to make the kind of moves they are, then AMD may even rule the data center market beyond 2025.
I’ve been a bit busy, so I have yet to post about the FAD. By the way, one of the links in Vatilla’s post: https://wccftech.com/amd-exascale-heterogeneous-processor-eh… is from 2015! So if you thought you had fallen through a time warp, you had. But it shows that the old saw about microelectronics. You put the gun to your head to play Russian roulette, and five years later you find out if you won or lost. But now it is seven or eight years. I think AMD is winning. Intel? They used up one “Get out of jail free” card with the 10 nm mess. Do they have another one? Will Intel need another one?
I think they will not because AMD is such a threat. It isn’t really. AMD is only a threat because they are showing how the game has to be played today. AMD is designing three or four chiplets per year. Intel is still spreading their effort too thin, and worse, management making the same stupid mistake of telling engineering what they want, not listening to engineering when they say what they can provide. Right now this is pretty obvious with the discrete graphics cards. I don’t expect problems with Raptor Lake other than it will require more silicon. But Meteor Lake? Look at it this way. AMD is planning Zen 4 for this year, and Zen 5 for 2024. Intel is planning Raptor Lake for later this year–it will be interesting to see who gets to market first. Then in 2023, on the Intel 4 process, Meteor Lake. The only good thing I see at this point is the number of witty phrases available if Meteor Lake crashes and burns.
Oh, and I have to wonder who will need Zen 4, Zen 4 3d, or Zen 4c on the desktop. I am finding it extremely difficult it keep my new (Zen 3) 5900x at all busy. Yes, there is method in my madness. If I get a speedy matrix multiplication algorithm, I can use Atlas to build a faster BLAS. But my real goal is a faster complex matrix multiplication algorithm to use with FFT code. Yes, I know that it seems easy to optimize this code–until you try it. In theory, AES 2 provides four 64-bit floating-point multiplies in one clock cycle. But to get close to that, you not only have to marshall your data so that the values you want to multiply are in order in two 32-byte sequences, but also that neither crosses a cache line boundary. Then you will find that L1d or L2 bandwidth is limiting your code.