cthalupa

5 ms·

1.
by cthalupa 21d ago
I'm going based off of set standards, not local. Plenty of international organizations track this and even before you replied to me here there were other links in the thread. But the absolute numbers aren't important either way, i
2.
by cthalupa 21d ago
Japan is literally the only country in the developed world that doesn't see their trends following the path of America. Even culturally similar countries like Korea are seeing a rise in overweight and obesity similar to the trends that
3.
by cthalupa 21d ago
I would agree that they very much had business decisions around the fork - but calling them malicious is bizarre. They used the software in a way that complied with the license. When the license changed to no longer suit them for future rel
4.
by cthalupa 1mo ago
Except this entire narrative is thrown off by the fact that obesity has been a problem before any of that and is a problem in countries without any of that. Even places like Korea where you are literally shamed constantly by friends and fam
5.
by cthalupa 1mo ago
> So has anyone managed to separate the effects of semaglutide from the effects of weight loss? In general? Yes. For this specific thing? We haven't even really shown the effects matter for humans. > If this is caused by the weig
6.
by cthalupa 1mo ago
Making this argument is a strong indicator that you are extremely under-educated on this subject or motivated to make dishonest arguments. There are a variety of commercial reactor design families that CANNOT melt down. Literally impossib
7.
by cthalupa 1mo ago
I love Japan. I love Tokyo. But go look at the conviction rate in Japan - it's not because their police are just so much better than ours that they just never get the wrong person. I would never argue for the Japanese legal and polic
8.
by cthalupa 2mo ago
Just google 'mortality data statins' and there's a mountain of it. 5 years studies show minimal or no benefit for people in low-risk groups. High-risk groups see significant benefit Follow up/longer term studies show inc
9.
by cthalupa 2mo ago
VLDL is worse than regular LDL but high LDL in general is an issue. The science here isn't "If your LDL is denser everything is just peachy and you can be fine with 100 LDL," it's "You might be anywhere between sign
10.
by cthalupa 2mo ago
Significant inflammation reduction, which is another factor in ASCVD. More plaque sticks to you the more inflammation there is. It's a feedback loop, as well. There are some that will even claim inflammation is the real villain in ASCV
11.
by cthalupa 2mo ago
It (Particularly ApoB/Combining LDL-C and Lp(a) is one of the single largest leading indicators of health risk we have. If anything, the data we have is suggesting we've gotten wrong what the upper limits should be for what is con
12.
by cthalupa 2mo ago
> Low cholesterol production is an extremely serious condition. This is correct, but not in the way you mean it. We know people can effectively produce no LDL cholesterol in their liver and have positive health impacts - these MR studies
13.
by cthalupa 2mo ago
The problem is if DDR4 was viable here it would also be absurdly expensive. There’s no world in which we survive this demand without a corresponding increase in supply. It’s not like there are zettabytes of spare cheap DDR4 sitting around.
14.
by cthalupa 2mo ago
I snagged my two 4tb acer variants for under 4k each. CDW still has them listed for close to what I paid.
15.
by cthalupa 2mo ago
I’ve got a 128gb m5 max mbp and two sparks. For my real-world use cases, a single spark running DS4 Flash will have fully responded by the time my Mac has even started generating tokens. I figured I’d have more generation heavy work when I
16.
by cthalupa 3mo ago
Plenty of demand outside the US. Why would the hyperscalers not buy the chinese RAM for all of their datacenters across the world besides the US one? Rising supply from China will impact prices even in countries where there are tariffs.
17.
by cthalupa 3mo ago
Anything where you're dealing with a large volume of records/documents. Lots of people are using these for large-scale digitization of documents - scanned stuff being OCR'ed and summarized, generating embeddings, etc. Large s
18.
by cthalupa 3mo ago
> The cuda cores can be broken up into compute complexes which larger blocks of memory directly attached to the cores. Perhaps in theory, but for the gb10 stuff the memory is all on the CPU die and connected to the GPU die via nvlink-c2c
19.
by cthalupa 3mo ago
The dgx spark is the same chip and those are in the low 3s to 5 range for most of them depending on manu, storage config, etc. The dgx sparks also have connectx 7 cards in them to support the 200gbps networking for RoCE. So I would expect t
20.
by cthalupa 3mo ago
Prefill is another advantage vs. Apple. It's way way way way faster on a spark than it is even on an m5 max. Same model, same quant, same query, as close to as matched settings as I can get from vllm, and for workloads with large promp
21.
by cthalupa 3mo ago
Not true. This is aimed squarely at the Strix Halo and Mac markets. It's basically just strictly better than the Strix, and it's not clear cut vs that Macs in any sort of blanket statement. My M5 Max 128gb MBP decodes faster than
22.
by cthalupa 3mo ago
For these in specific, they appear basically transparently to the GPU. There's a lot of software/firmware stuff for this, but also a different hardware architecture - while the RAM is on the CPU die, the nvlink-c2c gives it extrem
23.
by cthalupa 3mo ago
> If I see code like the one above posted by the OP, that the author wouldn't have written, I start to pay attention. Except the author did write it. https://github.com/RsyncProject/rsync/issues/959#is
24.
by cthalupa 4mo ago
Never had this issue with any of my MBP. I never even have magsafe available, couldn't tell you where a USB-C to magsafe cable is in my house if my life depended on it.
25.
by cthalupa 4mo ago
There are a variety of inference engines that support this, regardless of whether or not there is native FP8 in Ampere - llama.cpp will do it quite happily. VLLM you can do W8A16 quant too. There are a whole lot of ways to quantize models i
26.
by cthalupa 4mo ago
I remember getting my first CalDigit TB dock and being excited - everyone seemed to love them. I expected it to largely Just Work. That thing Didn't Work more than it Worked, but options were slim. Eventually it fully died about 14 mon
27.
by cthalupa 4mo ago
You can split tensors across an AMD GPU and Nvidia GPU - different architectures are not an issue. People run LLMs across some pretty crazy setups.
28.
by cthalupa 4mo ago
I've had multiple X520-DAs in desktop cases for years without doing anything special for cooling. Hell, one of them was in a fully watercooled system that had very little in-case airflow.
29.
by cthalupa 4mo ago
To the best of my knowledge, the Indian support teams for AWS have all the same access their US/EU/etc. support teams have when it comes to escalation and communication avenues. Obviously real time communication is harder for them
30.
by cthalupa 4mo ago
From talking to people I know on support teams at hyperscalers and other tech companies: A mind-boggling huge portion. But even as an AI believer and thinking it's great for this sort of thing, I also think if it's good enough for

More ›