HNHacker News
TopNewBestAskShowJobs

angoragoats

890 karma · joined January 17, 2023

YC is managed by a bunch of fascist-supporting jerks.
submissionscomments
angoragoats··on Sam Altman's response to Molotov cocktail incident
You say “the post itself is authentic” and then go on to give a great explanation of exactly why I think it’s inauthentic. I think we just have different definitions of the word “authentic.”
angoragoats··on Sam Altman's response to Molotov cocktail incident
DHH has expressed clear public support for white nationalist causes and figures, on multiple occasions. What else should we call him?
angoragoats··on Sam Altman's response to Molotov cocktail incident
> Altman didn't create AI.

No one said he did.

> That disruption is already coming no matter what.

[citation needed]. Depending on what you mean by "that disruption," I might even be willing to bet against it coming at all.

> He's a fine enough steward of the tech.

He's a manipulative con-man who is mediocre at everything except convincing investors to give him money. If the tech is truly as revolutionary as it's purported to be, he absolutely should not be a "steward of the tech."

angoragoats··on Sam Altman's response to Molotov cocktail incident
> Do you frequently get Molotov cocktails thrown at your house?

Luckily, no. Do you frequently wade into comment threads shitting on others’ statements of their lived experiences?

angoragoats··on Sam Altman's response to Molotov cocktail incident
I wonder if this is the first time in recent history (or ever?) that he has felt this way. Must be nice.
angoragoats··on Sam Altman's response to Molotov cocktail incident
To be clear, I don’t want anyone’s house to get firebombed by any means. But the “I’m just a humble guy making mistakes and trying the best I can” attitude of this article strikes me as extremely inauthentic based on everything I know about the guy.
angoragoats··on EFF is leaving X
> For something to be an ad hominem you simply need to address the speaker rather than what was said

No, this is a common misconception. Addressing the speaker is part of it but is not sufficient by itself.

People who are quick to claim “ad hominem!” have been getting this wrong basically forever, so please feel free to educate yourself by reading this excellent post: https://laurencetennant.com/bonds/adhominem.html

“It is not a logical fallacy to attack someone; the fallacy comes from assuming that a personal attack is also necessarily an attack on that person's arguments.

Therefore, if you can't demonstrate that your opponent is trying to counter your argument by attacking you, you can't demonstrate that he is resorting to ad hominem.”

angoragoats··on EFF is leaving X
I agree.
angoragoats··on EFF is leaving X
For something to be an ad hominem, one needs to be 1) addressing or responding to an argument 2) by attacking the character of the person making the argument rather than the substance of the argument.

Even though OP didn’t provide them, I can think of many supporting examples for their assertion that Bari Weiss and Matt Taibbi are either intentionally operating in bad faith, or stupid, or both. So this does not at all meet the definition of ad hominem.

Put another way: “you’re wrong because you’re stupid” is an ad hominem. “You’re wrong, and I think you’re stupid because [reason]” is not. This holds even if the person making the argument does not explicitly give the reason.

angoragoats··on Qwen-3.6-Plus is the first model to break 1T tokens processed in a day
I prefer not to give eyeballs/traffic to billionaire nazis, but you do you buddy!
angoragoats··on Qwen-3.6-Plus is the first model to break 1T tokens processed in a day
I already do this, but I don't think HN should be implicitly condoning a Nazi website by linking to it.
angoragoats··on Ollama is now powered by MLX on Apple Silicon in preview
Sure, but that’s somewhat orthogonal to the point I was making, which is that LLMs are huge in size. Even in the case of a custom “LLM chip,” you’ll need huge amounts of very fast storage of some sort (likely DRAM), which places constraints on the size, power consumption, and cost of such a device. This device, if it existed, would not in any way resemble the Coral TPU product that the GP was referencing; I think in fact it would be closer in size, price, and form factor to a GPU.
angoragoats··on Ollama is now powered by MLX on Apple Silicon in preview
I think there are drastic differences between computer vision models and LLMs that you’re not considering. LLMs are huge relative to vision models, and require gobs of fast memory. For this reason a little USB dongle isn’t going to cut it.

Put another way, there already exist add-in boards like this, and they’re called GPUs.

angoragoats··on A Message from the Ruby Central Board
Me too, and because of that I feel it's even more important to use language like racist, white nationalist, and fascist when describing him and his ilk, because that's what they are. Softening the language only leads to those beliefs becoming more normalized than they already are.
angoragoats··on A Message from the Ruby Central Board
If you’d like to read, in his own words, his “coming out” as an ultra right wing racist piece of shit, feel free to look on his blog for the post titled “As I Remember London.”
angoragoats··on ChatGPT won't let you type until Cloudflare reads your React state
Nothing you do can fully prevent automation. Someone who wants to automate requests badly enough will be able to do it, especially when the “protections” are as easy to decrypt and analyze as the OP proved.

Meanwhile, the rest of us (well, not me, because I don’t use your garbage product, but lots of others do) have to suffer and have our compute resources used up in the name of “protection.”

angoragoats··on Apple discontinues the Mac Pro
1800W is the max on a 15A circuit, but yes, it’s usually under 1600W. For LLM inference, limiting the TDP to 225W or so per card saves a lot of power, for a 5% drop in performance.
angoragoats··on Apple discontinues the Mac Pro
GPU performance can be perceptibly slower with TB5 compared to Oculink. Here’s one article showing a 14% difference in one test setup:

https://www.tomshardware.com/pc-components/gpus/oculink-outp...

angoragoats··on Apple discontinues the Mac Pro
FWIW I have never used NVLink, and I’m not sure why people are bringing up “daisy chaining” because as far as I’m aware that is not a thing with modern GPUs at all.
angoragoats··on Apple discontinues the Mac Pro
> I think it's bad form to say "citation needed" when your original claim didn't include citations.

I apologize, but using multiple GPUs for inference (without any sort of “daisy chaining”) is something that’s been supported in most LLM tooling for a long time.

> Regardless - there's a difference between training and inference.

No one brought up training vs. inference to my knowledge, besides you — I was assuming the machine was for inference, because my experience building a machine like the one I described was in order to do inference. If you want to train models, I know less about that, but I’m pretty sure the tooling does easily support multiple GPUs.

> And pytorch doesn't magically make 5 gpus behave like 1 gpu.

I never said it was magic, I just said it was supported, which it is.

angoragoats··on Apple discontinues the Mac Pro
The machine I’m describing works just fine on a dedicated 15A 120V circuit.
angoragoats··on Apple discontinues the Mac Pro
> The mac will just work for models as large as 100B, can go higher with quantized models. And power draw will be 1/5th as much as the 3090 setup.

This setup will work for 100B models as well. And yes, the Mac will draw less power, but the Nvidia machine will be many times faster. So depending on your specific Mac and your specific Nvidia setup, the performance per watt will be in the same ballpark. And higher absolute performance is certainly a nice perk.

> You can certainly daisy chain several 3090's together but it doesn't work seamlessly.

Citation needed; there's no "daisy chaining" in the setup I describe, and low level libraries like pytorch as well as higher level tools like Ollama all seamlessly support multiple GPUs.

angoragoats··on Apple discontinues the Mac Pro
You can still buy used 3090 cards on ebay. 5 of them will give you 120GB of memory and will blow away any mac in terms of performance on LLM workloads. They have gone up in price lately and are now about $1100 each, but at one point they were $700-800 each.
angoragoats··on Apple discontinues the Mac Pro
This has changed since Sam Altman started buying up all the chip supply, raising prices on memory, storage, and GPUs for everyone, but it used to be the case that you could build a PC that was both cheaper and faster than a Mac for LLM inference, with roughly equal performance per watt.

You would use multiple *90-series GPUs, throttled down in terms of power. Depending on the GPU, the sweet spot is between 225-350W, where for LLM workloads you only lose 5-10% of performance for a ~50% drop in power consumption.

Combined with a workstation (Xeon/Epyc) CPU with lots of PCIe, you can support 6-7 such GPUs (or more, depending on available power). This will blow away the fastest Mac studio, at a comparable performance per watt.

Again, a lot of this has changed, since GPUs and memory are so much more expensive now.

Macs are great for a simpler all in one box with high memory bandwidth and middling-to-decent GPU performance, but they are (or were) absolutely not "untouchable."

angoragoats··on Apple discontinues the Mac Pro
This is correct; Apple has refused to implement socketed memory on any M-series machine.
angoragoats··on Apple discontinues the Mac Pro
Maybe; I'm unable to find any benchmarks that specifically compare PCs with TB to Macs to test this. But there is certainly still overhead with TB no matter what, and therefore it'll never be as fast as Oculink.
angoragoats··on Apple discontinues the Mac Pro
And a 9800X3D is not even the fastest CPU out there, nor even the fastest CPU you could use with your specific motherboard. A 9950X3D is essentially two of the 9800X3Ds combined, and would be a drop-in replacement.
angoragoats··on Apple discontinues the Mac Pro
Oculink is generally faster than TB5 despite them both using PCIe 4.0, because Oculink provides direct PCIe access whereas Thunderbolt has to route all PCIe traffic through its controller. The benchmarks show that the overhead introduced by the TB5 controller slows down GPU performance.
angoragoats··on Apple discontinues the Mac Pro
Yes, I know; this is part of what I was implying when I said "Just about every single consumer computer shipped today uses PCIe."

I don't understand how this is a response to anything I said.

angoragoats··on Apple discontinues the Mac Pro
A Ryzen 9800X3D is about 40% faster in single-core tests and the same speed to slightly faster at multi-core tasks, as compared to the M2 Ultra in the Mac Pro. In addition, the Ryzen computer would presumably be modular and allow for the user to choose their preferred configuration of memory, storage, GPU, etc, with options far exceeding those offered by Apple in its limited and non-user-upgradable machine. In addition, configuring the Ryzen machine with comparable specs to the base model Mac Pro (64GB of ram, 1TB of storage, and a low-end to midrange discrete GPU) would put you at a total system cost of something like 20-25% of the $6999 that the Mac Pro cost, even with today's inflated memory prices.

I'm not sure if this is what the parent meant by "a real modern PC," but it would certainly be 1) faster and 2) vastly cheaper than the Mac. So at minimum, your assertion that it'd be slower is wrong.

Depending on your configuration, you could likely also match the overall power consumption of the Mac as well, though yes, it is easily possible to exceed it. But the most likely way you'd exceed it is with a high-end GPU, which would vastly outperform the (fixed, non-upgradeable) GPU in the Mac.

← PreviousPage 11 of 30Next →