aseipp

62 ms·

601.
by aseipp 4y ago
Where does it say that? The pricing page says on "Writes" in the info tooltip: "Each write operation (INSERT, DELETE, etc) consumes write units depending on the number of rows, columns, and partitions it writes to." This
602.
by aseipp 4y ago
It doesn't explicitly support but today people work around it with one-time in-app purchases for all the functionality. Two apps I can think of doing this are Halide and Darkroom. But yes, you can't just pay $X upfront in the App
603.
by aseipp 4y ago
Honestly that's something they can probably offer separately, but I really prefer this pricing for most use cases where I want a database that is always available but has bursty request/response patterns. This means I can have an
604.
by aseipp 4y ago
You can in theory clock the Agilex/S10 parts up-to 1GHz, but something like 600-700MHz sounds more realistic, yes. (Note that this is in the same ballpark as what Virtex UltraScale+ can do too.) Also, regarding point 1 -- and I'm
605.
by aseipp 4y ago
Yes, but that's not the problem. The problem is these Agilex devices are insane overkill for that task, and you likely won't be able to buy the SKUs in a reasonable form factor or price for that use case. The Mister worked so well
606.
by aseipp 4y ago
Orin has support for FP16 of course, it would be insane not to have it when INT8 isn't viable for a lot of models anyway. There are some preliminary benchmarks that can confirm this that weren't hard to find.[1] The number juicing
607.
by aseipp 4y ago
Yes, Stable Diffusion performance will improve the more VRAM you give it. This is generally true of almost every ML model, because ML models are often limited by bandwidth. There is a constant need to shove more data into the accelerator, b
608.
by aseipp 4y ago
Yeah that's fair, VR titles are still a place where GPUs need more performance. I personally have no interest in them but yes the compute needs are in general way higher. Not to mention Flight Sim is a particularly intense case...
609.
by aseipp 4y ago
The 3090 Ti series was already pushing 450W so this isn't new.[1] And it's because they clock these things incredibly high, well beyond the efficiency curve where it makes sense. Because that's what gaming customers expect, b
610.
by aseipp 4y ago
No, go pick up a 3090 Ti from Best Buy instead. It's $999 right now -- over half off, maybe more with this announcement -- and has 24GB of VRAM, and can chew through Stable Diffusion requests in seconds versus dozens of seconds or minu
611.
by aseipp 4y ago
No, but not for lack of interest. I simply upgraded from 2k res + 2080 Super to 4k res + 3080 at one point in the last year or so, but I had already beat the game by then and haven't returned to it.
612.
by aseipp 4y ago
Eh, the current 3080 can already do 60FPS @ 4k HDR on every AAA title I've thrown at it, and that's with an mITX undervolt build (cliffing the voltage at a point). "60FPS @ 4k is readily achievable" has been sort of the
613.
by aseipp 4y ago
Run 'waypipe ssh user@machine firefox' and you are done.
614.
by aseipp 4y ago
The example code uses an atomic store instruction in order to write values from threads to a memory location, and then an atomic read to read them. The system guarantees that a read of a previously written location is consistent with a subs
615.
by aseipp 4y ago
Yeah, AFAIR the basic realization at some point was that most freely submitted user reports were either of low quality, part of something that wasn't on the near-term roadmap, or part of something to be addressed more broadly in the ne
616.
by aseipp 4y ago
I am very happy to see this. Phabricator is an incredibly well written and deeply thought-out tool (even if people may disagree with some of its philosophies regarding e.g. patches and code review) and it had a tremendous influence on my ca
617.
by aseipp 4y ago
Yes, I have used it with languages like Bluespec and Lean v4, which have very little data (FOSS code on GitHub) for the model to work with. It's nowhere near as accurate as it is when using Rust or TypeScript but it's still mostly
618.
by aseipp 4y ago
Discord was originally designed for gaming voice chat. It marketed itself this way, it was positioned this way in its public perception, and most of the early feature development reflected this. The move towards "online communities&quo
619.
by aseipp 4y ago
The best you can get right now are some NXP parts from Solid-Run (bulky form factor for mobile, no GPU), the Nvidia Xavier/Orin series mobile SoCs, or some of the newer Snapdragon devices (I'm not familiar with these.) These (all)
620.
by aseipp 4y ago
Trial and error. You need to dial in the right point on the voltage/clock frequency curve for your workloads, AKA "just play some games and look at the results." Just use whatever your overclocking software for your motherboa
621.
by aseipp 4y ago
Because fork() was very simple and conceptually "easy" to do when it first was introduced, and is now massively complex and has huge implications on every part of the system. It's not compositional, isn't thread safe, in
622.
by aseipp 4y ago
Oh wow, that's quite impressive. I assume they're just using a 128-bit datapath like NEON? Thanks for the correction.
623.
by aseipp 4y ago
There isn't any standard matrix multiplication instruction set so there's nothing to standardize over. Machine-learning driven instruction sets (of which MM instructions are motivated by, but not exclusively for) like this have be
624.
by aseipp 4y ago
ARM literally doesn't allow public instruction set extensions by compliant license holders. Apple is presumably only allowed to do this precisely because they do not sell or otherwise offer their CPUs with any other software, with any
625.
by aseipp 4y ago
No, there are no publicly available chips with SVE2 support at all; the closest you can get is SVE1 on Graviton 3 from AWS.
626.
by aseipp 4y ago
The frequency scaling problems most prevalent in Skylake are significantly better as of e.g. Ice Lake. It's complex to summarize but basically those instructions request a higher "power license" for higher power delivery when
627.
by aseipp 4y ago
No, it's close to the core internals of an OLAP database. The "operators" it can execute in question are things like filter, join, aggregate, group by, projection (select), things of that nature. It makes sure to use availabl
628.
by aseipp 4y ago
At a prior gig we were using their FPGA instances and had a way of causing hard faults on the system that required reboots, with specific FPGA bitstreams. Their systems monitored this and when we reached out about it they scheduled at least
629.
by aseipp 4y ago
People tend to call them "lakes" because, I think, they are "unfiltered" and contain raw data objects and blobs, originally from the source system, unmodified. In a normal data warehouse system you ETL things and the fin
630.
by aseipp 4y ago
Also, even if you do want to have some kind of binary-ized transport with a strict type-checked schema and compilers for multiple languages -- ultimately a noble goal, I think -- there are much better options available, design and technol

More ›