redox99

33 ms·

631.
by redox99 3y ago
vsync off has better latency than gsync https://blurbusters.com/gsync/gsync101-input-lag-tests-and-s...
632.
by redox99 3y ago
It varies so much between people (and age?). My dad has a very hard time telling 60hz and 120hz apart. Meanwhile to me 144hz and 240hz are trivial to tell apart, let alone 60 to 120. One time my drivers for some reason reset my old 144hz mo
633.
by redox99 3y ago
Gsync basically doesn't matter if you're fine with disabling vsync, and at high FPS the tearing is almost invisible.
634.
by redox99 3y ago
Yeah my multimonitor setup includes both 60hz and 240hz monitors, and it's so much easier to use the cursor with the 240hz monitor.
635.
by redox99 3y ago
There's nothing wrong. What's actually wrong is always using k8s even when it's overkill.
636.
by redox99 3y ago
The CT will likely be a bit cheaper than the Rivian or Lightning. Beyond that, probably the same reasons why more people buy Model Ys instead of Mach-Es.
637.
by redox99 3y ago
Nah, the Model S (especially the plaid) is their flagship. Even if economically it doesn't make too much sense, it elevates the brand.
638.
by redox99 3y ago
If memory and compute prices fall, people will just train bigger models.
639.
by redox99 3y ago
> You know what? The equivalent of X11 in windows may be full of cruft, technical debt, poorly thought out / "unelegant" design decisions... But it fucking works. I'm sorry, but I have other things to do than dealing
640.
by redox99 3y ago
So, basic functionality is still not there 15 years after release. And I believe gnome still has that thing where if the UI lags, the cursor lags. Windows DWM, WDDM architecture, and anything graphics related on Windows is superior in so ma
641.
by redox99 3y ago
Once google only accepts hardware attestation it's over unfortunately
642.
by redox99 3y ago
What compromise? A game with 1080p internal resolution upscaled to 4k with DLSS on a 4k monitor looks way better than 1080p native on a 1080p monitor.
643.
by redox99 3y ago
Most games these days support DLSS and FSR. You should never run games at 1080p output, but instead whatever internal resolution you can handle, upscaled to 4k output.
644.
by redox99 3y ago
Once you go with a higher resolution, you realize how blurry 1080p is. For gaming, 1440p 240hz is a pretty decent middle ground considering quality, price, and fps.
645.
by redox99 3y ago
I'm not sure what you mean, but people buying $179 GPUs are very likely to have a 1080p monitor.
646.
by redox99 3y ago
You're probably less sensitive to the compression.
647.
by redox99 3y ago
It always surprises me that laptops maintained such as large market share over desktops. I'd have thought that as PCs became more of a niche, power user tool (a lot of people only use smartphones these days), desktops would prevail.
648.
by redox99 3y ago
That's a lot of hypotheticals. What I'm trying to say is that this is unrelated to regular folks' cars. For this you need to manually authorize access to your car, and then they can do things like unlock the doors[1]. It&#x27
649.
by redox99 3y ago
> Elon's probably betting on how much other people would pay for access to APIs and thus user data, and gain income on top of simply profit margin on the vehicle itself. Tesla doesn't expect end users to use this API. This is m
650.
by redox99 3y ago
Even GPT3.5 can be much more expensive. In some specific tasks, a finetuned 7B llama can work as well as GPT3.5. You can rent a 3090 at $0.20/h on vast.ai, or $0.40/h on runpod. Using VLLM at 400t/s that's 1440000 genera
651.
by redox99 3y ago
Maybe this has improved, but a few months ago, OpenAI p99 latency was much worse than a self hosted solution, which would be a problem in certain cases.
652.
by redox99 3y ago
Idk, a lot of people (including me) pay for a drill that will only be used a couple times lol. Same with trucks.
653.
by redox99 3y ago
AI outputs are not necessarily bad for datasets, given they've been verified by a human for their quality and correctness (probably the case here, but not for SEO content farms).
654.
by redox99 3y ago
Correct in what way? Not correct if you want the best performing model. And if it doesn't outperform the current best model nobody will care about you. For a company like Mistral that could be the end for them. I'm not saying it u
655.
by redox99 3y ago
Ooba is not meant to serve multiple users (no batching). Batching gives you 5x to 10x throughput increase.
656.
by redox99 3y ago
Exactly. I don't get why people (non AI researchers) discount MoE like they are cheating or fake parameters. Even if each inference pass only runs part of the network, there's still a trillion learnable parameters there lol.
657.
by redox99 3y ago
Mistral 7B is very cool for its size. But unfortunately no open model is close to GPT4 as of right now.
658.
by redox99 3y ago
This being a 3B model isn't remotely comparable to GPT4. WizardCoder 34B and Phind 34B are the only models remotely comparable, and they are still slightly worse than GPT 3.5 (let alone GPT4).
659.
by redox99 3y ago
What are the specs? The linked web page is atrocious and can't find any kind of spec there.
660.
by redox99 3y ago
> seasonal storage mechanism Is this even needed? With solar, wind, etc storing more than a few days worth of energy seems very wasteful.

More ›