xadhominemx

6 ms·

1.
by xadhominemx 14d ago
The critical back end step for HBM is thermal compression bonding - where they actually compress 13 very thin die separated by solder balls. Cracking is a major issue.
2.
by xadhominemx 22d ago
Easy to find yourself in literally 15 seconds.
3.
by xadhominemx 3mo ago
Is that sort of research unavoidably impaired by a more crowded night sky? Or do we just have to spend more to collect the same quality of data from more or better terrestrial observatories?
4.
by xadhominemx 3mo ago
I am a science and astronomy fan, but I am sorry, in this case progress is more important. If we regret our decision, the LEOs will fall out of the sky by themselves in a few years and it will be ok.
5.
by xadhominemx 4mo ago
This comment may have been true 6-9 months ago, but not today
6.
by xadhominemx 4mo ago
To me it is crazy that you are being downvoted. My experience in academia was that an incredible amount of time was devoted to data cleansing analysis, coding, etc., which were completely non-core to the actual underlying academic pursuit.
7.
by xadhominemx 4mo ago
Orders of magnitude less compute for pre- and post-training than the the frontier labs.
8.
by xadhominemx 4mo ago
Intel made bad engineering decisions. There was not under-investment in fabs or R&D.
9.
by xadhominemx 4mo ago
This model is going to start miles behind the frontier and the gap will only grow.
10.
by xadhominemx 4mo ago
That is not even close to correct.
11.
by xadhominemx 4mo ago
It's very easy to find out when the new fabs come online. Try asking Claude or ChatGPT.
12.
by xadhominemx 7mo ago
I spend much less than that at a chain theater in Manhattan.
13.
by xadhominemx 8mo ago
AI slop article
14.
by xadhominemx 8mo ago
Yes of course. Looking at the share prices of their suppliers— ASML, Lam Research, Applied Materials, etc.
15.
by xadhominemx 8mo ago
Yes of course their messaging to customers and the investment community is that they will be rational and measured in their investments. In reality, they are adding capacity as quickly as possible as margins are too high. However, capacity
16.
by xadhominemx 8mo ago
Everyone is completely sold out and adding capacity as quickly as possible.
17.
by xadhominemx 8mo ago
CXMT sells the vast majority of their bits at the prevailing market rate, just like everyone else. They are adding capacity as quickly as they can, with a 5-10 year planning horizon, just like everyone else. It’s really not that deep!
18.
by xadhominemx 8mo ago
No one sold their capacity to OpenAI. The vast majority of DRAM is transacted in what is essentially a quarterly auction.
19.
by xadhominemx 9mo ago
Nonsense. Do you think these landlords intentionally overpaid for the properties? They are sophisticated entities who like to pay at or below fair value.
20.
by xadhominemx 9mo ago
That 1/3rd is almost all small time flippers who renovate properties before resale.
21.
by xadhominemx 9mo ago
You are wrong. Memory production is being expanded in 2026 and will expand further in 2027 and 2028 as the memory suppliers catch up on fab shell capacity.
22.
by xadhominemx 9mo ago
AI consumes about 30% of DRAM wafers. PE owns about .5% of single family homes.
23.
by xadhominemx 10mo ago
China has promised to wage war and forcibly subjugate Taiwan, a democratic ally and critical trade partner. If China backed off Taiwan for a few decades, I think the US would drop export controls.
24.
by xadhominemx 10mo ago
You should really seek to actually understand an issue before you comment so arrogantly. US authorities are ok with Chinese companies accessing GPUs in overseas DCs because those DCs will still be subject any US export controls. Right now,
25.
by xadhominemx 10mo ago
Feels at least an order of magnitude too high!
26.
by xadhominemx 11mo ago
Not really a relevant issue or concern for a nation state backed hack…
27.
by xadhominemx 11mo ago
$600 per year is a trivial cost for a professional tool
28.
by xadhominemx 11mo ago
It’s because the model weights and KV cache are stored in SRAM. It’s extremely expensive per token.
29.
by xadhominemx 1y ago
The Chinese definitely want their own frontier model. There is an enormous national effort behind building up the semiconductor manufacturing, data center infrastructure, and networking technologies required to compete with US frontier mode
30.
by xadhominemx 1y ago
No, we have certainly slowed down their progress in AI with export controls on GPUs and upstream. China would have better AI models but for those export controls.

More ›