The Long Term Investor

It is comical to watch. For example today. Totally getting hammered, then a FOMC guy comes out at a speech and mentions...."further adjustment to rates in the near term." At exactly that moment....boom! The market takes off.

All this tells me, is that any one of these tool's, FOMC and politicians or even someone with a "voice platform" can move this type of market almost instantly either way. It is just that type of environment.

We could be "all clear" in one moment and then straight to PERDITION the next.

 

Well, today closed out okay despite the silly little ride. Actually, this week felt worse than the overall result at least for me.

Had I not poked around at some of the coverage and noise….the actual results would not have matched up to the hoopla.

Sometimes the messaging is way worse than the reality. This proved true for me this week.

Next week will be a short one. Might be good for all of us to get a break from the circus and the market too.

 

A postcard from TomB.

It's been a little while since I last posted. I doubt anyone noticed and that's completely cool.

I found some kindred spirits on Lemmy and we talk about this stuff quite a bit. It's not ideal, as they mostly think the way I do. On the other hand, discussions here resulted in conflict as to both basic facts, as well as future possibilities. I'm not sure which is more productive. To be direct, I suspect neither is all that productive.

Meanwhile, you have Smokie. If you want to know what I think about literally any topic, read Smokie's posts. The guy's thoughts reflect mine to an uncanny degree. What's more, he is far more concise and efficient with his prose.

My portfolio is doing adequately. I'm certainly not making double digit gains.

My businesses are doing fantastic. They're small businesses so I'm not getting rich off them but they have scaled beyond what I could have imagined and are starting to move the needle.

During my deep dive into AI clusters, I picked up an interest in SIMD coding. I ended up picking up a project that involves computer vision and an LLM. Imagine an old dude like me cooking up large batches of ROCm/C code. It's some of the coolest and most fun stuff I've ever done. Who would have thought?

Please accept my best wishes for health and prosperity. You folks in this thread are a classy lot and have my utmost respect.


Your friend,

TomB

 

Oh shut up, you melodramatic goofball!

I am glad you are doing well. Please feel free to post anytime you want. As W always says, the more viewpoints the better.

And AM6 better rock my socks with PCIe lanes, or I'll be pissed

 

Finally a TB16 post. Nice to see one again and thanks for the kind words. I am glad you are doing well.

I always enjoyed the TB16 thread and your posts here. In addition, the TB16 and RTN banter about all things chips and computing was always a good read as well.

We will leave the light on for you should you decide to return or swing by. I think your contributions were important and added value to the place.

 

Not an investing topic, but some might be interested in it. A very good documentary series on PBS…The American Revolution. I believe you can stream it now at your leisure.

I found it quite interesting and well done. Others might too, so just sharing the info.

 

In case you're not just being nice, here's a bulk download...

It appears 40% of all RAM wafers are contractually earmarked for OpenAI and project Stargate. This explains RAM scarcity and exploding RAM prices. These are unfinished wafers and that brings up all kinds of thoughts. I'm pretty sure they are DDR5 and LPDDR6 wafers. These contracts are over several years so the new DRAM landscape is permanent.

Meanwhile, declarations of IGZO 3d DRAM generating infinite, nearly free, DRAM turned out to be complete crap. IGZO DRAM is scheduled to come online by 2030, perhaps in time for Zen 8. That's when I expect we will see RAM on the CPU complex. There may be no AM6.

AMD will be hit hard by the DRAM crunch. More DRAM production will be brought online but it will be less than the demand increase so I expect high DRAM prices for the next two or three years at a minimum.

Intel will be hit even harder by the DRAM crunch. It might finish putting them out of the CPU business. It wouldn't surprise me if they were to attempt to merge with any company at all with a set of DDR5 masks. That includes companies like Team Group.

AMD's next gen GPU is a world beater. It will massively leap-frog nVidia Blackwell in power and perhaps triple the efficiency lead they already have. Blackwell will be blown out of the water.

I don't like AMD giving 10% of itself to OpenAI to be an exclusive supplier. I can't declare it a bad move but I consider Sam Altman to be reptilian so he is cock blocking my crush on Lisa Su and AMD.

nVidia has been plagued with an architectural issue that is bandwidth limiting bus access of their GPUs and has for three generations. nVidia GPUs perform identically on PCIe v3, v4, v5; this, despite doubling (or more in the case of v3~v4) the bandwidth each generation. They have been working on this issue for years. Someone in my ROCm group claims to be a developer at nVidia. He tells us they have a major step forward in bus utilization. Time will tell if this is true and, if so, how much CUDA code it breaks. If true, this is a massive win for nVidia that could put AMD back in their place.

AMD is absolutely dominating in HPC and workgroup servers. This new marketshare looks to grow significantly from here. The cloud on the horizon is a couple of companies have announced and shown 1024 core ARM CPUs with ridiculous PCIe and memory channels. I would bet on AMD to win this. We don't even know if these kilocore parts will come to market. None the less, these CPU complex could be ideal for specific work loads. Given the state of IPC, I'd rather have fewer, more powerful, cores but keep in mind my AMD bias.

Now over to RTN for rebuttal.

 

@TomB16 absolute chinese language to me but thanks anyway
this place has allways been also your home so youre more than wellcomed here to post whatever you want!

 

Absolutely TB16. A very informative post. I enjoy them. I have very little knowledge when you and RTN get to breaking these things down. However, I am drawn to read them to maybe learn a bit and also to see two guys discuss a topic with a good level of knowledge about it.

Thank you for sharing and posting again. I look forward to them always. You as well RTN.

 

Talk about some shady company business leadership. The big three (Samsung, Micron, and SKHynix) knew this was going to happen a mile away. They wanted to keep their capex down while driving prices up. Memory is definitely a boom/bust business, but demand will not go away anytime soon, and when you only have three companies manufacturing RAM, it makes for a rather easy to manage market.

Someone who was dying for business.... some on-the-ropes, Optane abandoning company with fabs to burn.... could definitely tool up and shake the market here. RAM demand is not going away.

This claim has been done to death on enthusiast forums over the years. I would love for this to be true, and I hope I am wrong, but my extreme skepticism is based on past generations. AMD plays second fiddle when it comes to GPUs historically. Blackwell will be replaced by Rubin, so if AMDs target is Blackwell, they are shooting at yesterday's news.

Sam Altman is reptilian for sure. Up there with all the other dystopian tech leaders. I am banking on Lisa cozying up to OpenAI to get ROCm tuned for it's software to make AMD chips more attractive for AI loads. AMD is not showing that it will beat Nvidia with raw horsepower, but if it can get an edge in drivers, it could be good. If this is the case, it is savvy as hell. If not, I am baffled.

I have not read this one before. I am seeing that the stagnant bus width is pure monetary strategy. Wider bus means more RAM chips and more traces on the PCB. The smaller nodes are off the charts in price and RAM is way up. Nvidia wants to keep its margins high, so it has been counting on RAM frequency boosts and driver developments to maintain generational increases. I would be curious to see your source.

As they should. Their CPU department put the work in and whupped Intel's lazy ass. I love to see it.

Back to you, Tom