paulmd

29 ms·

241.
by paulmd 2y ago
it's also worth pointing out that you have abdicated driver support on those currently-sold Zen2/3 APUs with Vega as well ... they are essentially legacy-support/security-update-only. And again, I'm sure you see it as
242.
by paulmd 2y ago
No, they don’t, the optimal amount of units vacant is not zero to begin with, and in fact the profit-maximizing number of vacant units is likely considerably higher than the market-optimal price. Sellers like a market with constricted suppl
243.
by paulmd 2y ago
That is the point of a steel man, yes. OP was asking “how could this possibly ever work” and the parent did the math.
244.
by paulmd 2y ago
I mean, you say there was “just no money” but AMD signed a deal over three years ago to acquire Xilinx for $50b. They’ve been on an acquisition spree in fact. Just not anything related to gpgpu, because that wasn’t a priority. Yes, after yo
245.
by paulmd 2y ago
No, I’m not replying in the wrong place and i cited the portion of your comment I was responding to. I think your idea that a proprietor reducing the quality of the product because they disagree with someone’s private life is literally the
246.
by paulmd 2y ago
Genetic and memetic/intellectual immortality, of course. Biologically there can be no other answer. We are here to spread and endure, there is no “why” or end-condition. If your response to there not being a big ending cinematic to lif
247.
by paulmd 2y ago
> Only a sociopath, when confronted with a question they don't know the answer to, confidently makes something up. The human may not know the answer, but they generally know whether they know the answer or not. What? No. Literally f
248.
by paulmd 2y ago
>> Objectively, no it doesn't. Even if you think it's a bubble, then the bubble is just growing. There are zero signs of anything popping yet. > Ever seen a bubble pop? It doesn't contract and then pop, usually. A
249.
by paulmd 2y ago
At the end of the day the person coming to a public accommodation for a service or good doesn’t and shouldn’t need to know anything about the religion of the owner. Why should they? What does that have to do with me buying a muffler from t
250.
by paulmd 2y ago
none of these companies offer monthly subscriptions - they offer a 12-month subscription billed monthly or annually (with a slight discount), your choice. This is adobe's whole schtick with the cancellation fees for example. You pay 5
251.
by paulmd 2y ago
> My understanding is that the generation of responses (the inference) is what most of the compute is spent upon (as opposed to the training). no, training takes hundreds/thousands of high-end HPC gpus for days/weeks/month
252.
by paulmd 2y ago
AMD has spent years developing MCM/multi-chip packaging to solve exactly this same problem. And they have Xilinx too, and Infinity Link (which is different from infinity fabric), and all these other things that help them build systems
253.
by paulmd 2y ago
sure, it's things like oneapi that have pluggable backends. spir-v/vulkan compute too. now, for how many of those things does AMD have a pluggable backend that actually works? none of them, unfortunately. Octane implemented on
254.
by paulmd 2y ago
maybe because there's actual value in the systems being built and it's not simply blockchain 2.0 just because you keep repeating it? there's a fairly incredible amount of commercial value in a system that can provide a workab
255.
by paulmd 2y ago
> If industry decides they like the aarch64 isa more than x64 I doubt it would cause much trouble to the silicon team. the problem isn't ISA so much as AMD not having a moat around its CPU revenue anymore. Yes, AMD will have ARM c
256.
by paulmd 2y ago
> For turn key proprietary stuff where you really like the happy path foreseen by your vendor there really was no way for AMD to foresee that people might want to run GPGPU workloads on their polaris cards? isn't that a little coun
257.
by paulmd 2y ago
ISA shouldn’t be more than a 10% or 20% performance impact or so, per jim Keller
258.
by paulmd 2y ago
Yes, as the article discusses, there were also the options of just using a mini-tower or SFF desktop, or a NUC/booksize PC, and often these options were comparable or less in price. For example, I owned an ECS Liva 2/32 that I got
259.
by paulmd 2y ago
same as humidifiers and they make bacteriostatic additives that help control this (although eventually it does need to be cleaned regardless).
260.
by paulmd 2y ago
I would love it if google would implement support for RCS in google voice too…
261.
IC Design Companies Adopting TSMC's 3nm Process Trigger Cost-Driven Price Hikes(trendforce.com)
2 points by paulmd 2y ago|0 comments
262.
by paulmd 2y ago
Ollama is not as powerful as llama.cpp or raw pytorch, but it is almost zero effort to get started. brew install ollama; ollama serve; ollama pull llama3: 8b-v2.9-q5_K_M; ollama run llama3: 8b-v2.9-q5_K_M https://ollama.com/
263.
by paulmd 2y ago
the problem with the GPGPU space is that everything except CUDA is so fractally broken that everything eventually converges to the NVIDIA stuff that actually works. yes, the heterogeneous compute frameworks are largely broken, except for
264.
by paulmd 2y ago
Apple specifically avoids building solutions which put them in the position of being able to do lawful intercepts. I am convinced it is an explicit checkbox in their development process now. If you’ve built the solution for commercial purpo
265.
by paulmd 2y ago
GP here: oops, I have a couple CPUs for them and I forgot that wasn’t the newer one https://i.imgur.com/eluaofD.jpeg this one, is K10 B3 stepping
266.
by paulmd 2y ago
> For Apple to build out a dedicated server farm to accommodate even 50% of all their current users would be extremely challenging fyi there have been rumors that apple is looking to scale out Mac Pro production for their own server farm
267.
by paulmd 2y ago
Microsoft needs hardware for their “AI pc” too, this is nothing new and the only alternatives are server-side (which people will complain about even more) or extremely slow performance which will chew up battery (which, guess what, people c
268.
by paulmd 2y ago
> You gonna need things like Pentium D launched in 2005, or AMD K10 launched in 2007. finally, my moment to shine https://i.imgur.com/WxmKuG8.jpeg
269.
by paulmd 2y ago
a few years ago I read an observation that Amazon is basically a hyperwebster for physical goods. If you can name a specific goal and think some product would do it for you, you simply type it into Amazon and by the time you press 'ent
270.
by paulmd 2y ago
> Microarchitecture design skills are not theoretical numbers you manage to put on a spec sheet not only can you measure this, not only do they measure this, but it's literally the first component of the Rayleigh resolution equati

More ›