Recent content by AcostaJA

I considered it on my last answer, please read the study PDF, I find two big wrongs: it assumes national average power sources instead datacenter's power sources (often nuclear, hydro, solar etc, the cheapest), also it uses wrongs figures on power per request usage, at least 40x higher, but upto...

a thing is water cooling power and another is water consumption.

I understand some industrial chilling system spray free water on cooling towers to improve performance when the system capacity is exceed, but the methodology on this study cant be more biased and un-fair, firts it dont consider Datacenter specific energy source, it levels to US national std...

"that facilities run by companies like Microsoft used around 6 percent of all the district’s water" I miss quoted google by MS, it doesn't change the whole article is made up and ridiculous.

All the "study" purpose is to build fear and scandal, it don't overcome basic fact checking, seeing this blatant BS here (where mean reader is over mean IQ) is what disgust me. Most data center don't pump water on cooling, actually water is an server n1 enemy, closed loop liquid cooling...

The study is so picky that Blames Google on using Dalles' 6% water ... (Dalles is an small town, at about 16000 ppl) so google is spending water equivalent to 1000 habitants, that preocupant ginven the size of this datacenter and how many people works there I assume employes barely use the WC...

Pierce figures are right and legit, while its true inference is much less energy intesive, it uses the same hardware as training, his point is that dont waste water, BTW the article focused on inferencing.

Cold air comes from the environment, the liquid running on the loops comes from factories where is processed (is very similar to car coolant but there is no concentrate to dilute, all the liquid is sourced from manufaturers to ensure purity), spraying down water on the cooling towers may actual...

As powering a rockstar plane (in a year a single private Jet expend on Fuel more than Facebook on Training LLAMA3), or a politician rally? and that are not biased figures, nither comes from renewable (carbon neutral) sources LMFAO

I suscribe other' pior opinions, the study figures are opinionated at least, or madeup seeking an political headline, I'm familiar with GPU (ai) servers farms, some of them even dont use water but direct air or submerged hardware in cooling oil (the most modern), and those on liquid cooling use...

No surprise given both consumer, miner and HPC markets cooled down Nvidia needs a warehouse cleaner product, expect similar moves from AMD with theirs gcn2 HPC products.

Ethereum will switch to Proof of Stake algorithm instead Proof of Work, this will happen according the developers in the next 9 months, proof of stake dont need heavy compute capabilities, instead it rewards you in proportion of your stake (miner address should have an stake in Eth, this stake...

I totally Agree, wrong position from FCC, hardware control is not a matter of secret, but responsibility, having OpenSource there is warranty on this software to be auditable and do whatever is supposed have to do AND NOTHING MORE [PERIOD]. FCC should then establish liability for those...

The problem with Samsung is that they believes they are selling phones to iPhone owners... Anyway, while Samsung is excessive and out of focus about battery and SD memory, the fact is common people (not Geeks and Geek-wannabe) is more often to never update nor upgrade theis PCs, some reasoning...

This is the kind of project that makes no sense and which worries me about Alphabet=Android R&D, an "upgradeable" phone has sense only if you paid the same plus an reasonable prime than the equivalent non-upgradeable, project ARA hardly would meed this goal, as niche device is only attractive...