Unsurprisingly, you cannot assign a single intent to 166k+ people.
Just like Microsoft there are parts of the company who are hostile to open source, and there parts of the company whose success is attributable to open source.
True, but you can compare them to, say, Google, which maintains thriving OSS projects like Chromium and AOSP and generally does a way better job at publishing code and research.
I wouldn’t mention as positive example. I wouldn’t even mention them as example.
Apple cooperates within WebKit well with WebKitGtk. They supported LLVM when it is in their interest.
Chrome is used as proprietary web-engine to vendor lock-in the web. While often used by others, I’m not aware of a broad cooperation. Android is a shadow of Linux, merely using the Linux-Kernel, not GNU. Plus a lot of closed-source code (PlayServices, App Signatures, Google Cloud, Google Apps).
Googles open-source projects seem often exclusive Google only projects? Google works together with others! But especially Chrome and AOSP are…causing worries.
AOSP is the foundation of GrapheneOS, LineageOS and dozens of other patently non-Google systems. Chromium is the foundation for Edge, Brave, Opera, and every single AI browser being churned out by the dozen. Many of the Chromium forks are specifically designed to block Google ads.
There's a reason most of these projects picked AOSP over iOS, or even Chromium over WebKit. Google just engages with the community better than Apple. It's silly to pretend like they're on the same level.
AMD have had people contribute optimised ROCm kernels in the past. They closed the PR without merge. ROCm are not interested in this. Baffling behaviour.
LLM performance depends on doing a lot of math on a lot of different numbers. For example, if your model has 8 billion parameters, and each parameter is one byte, then for 256gb/s you can't do better than 32 tokens per second. So if you try to load a model that's 80 gigs, you only get 3.2 tokens per second, which is kinda bad for something that costs 3-4k.
There's newer models called "Mixture of Experts" that are, say, 120b parameters, but only use 5b parameters per token (the specific parameters are chosen via a much smaller routing model). That is the kind of model that excels on this machine. Unfortunately again, those models work really well when doing hybrid inference, because the GPU can handle the small-but-computationally-complex fully connected layers while the CPU can handle the large-but-computationally-easy expert layers.
This product doesn't really have a niche for inference. For training and prototyping is another story, but I'm a noob on those topics.
Running LLMs will be slow and training them is basically out of the question. You can get a Framework Desktop with similar bandwidth for less than a third of the price of this thing (though that isn't NVIDIA).
You're right about "reasoning". It's just trying to steer the conversation in a more relevant direction in vector space, hopefully to generate more relevant output tokens. I find it easier to conceptualize this in three dimensions. 3blue1brown has a good video series which covers the overall concept of LLM vectors in machine learning: https://youtube.com/playlist?list=PLZHQObOWTQDNU6R1_67000Dx_...
To give a concrete example, say we're generating the next token from the word "queen". Is this the monarch, the bee, the playing card, the drag entertainer? By adding more relevant tokens (honey, worker, hive, beeswax) we steer the token generation to the place in the "word cloud" where our next token is more likely to exist.
I don't see LLMs as "lossy compression" of text. To me that implies retrieval, and Transformers are a prediction device, not a retrieval device. If one needs retrieval then use a database.
> You're right about "reasoning". It's just trying to steer the conversation in a more relevant direction in vector space, hopefully to generate more relevant output tokens.
I like to frame it as a theater-script cycling through the LLM. The "reasoning" difference is just changing the style so that each character has film noir monologues. The underlying process hasn't really changes, and the monologues text isn't fundamentally different from dialogue or stage-direction... but more data still means more guidance for each improv-cycle.
> say we're generating the next token from the word "queen". Is this the monarch, the bee, the playing card, the drag entertainer?
I'd like to point out that this scheme can result in things that look better to humans in the end... even when the "clarifying" choice is entirely arbitrary and irrational.
In other words, we should be alert to the difference between "explaining what you were thinking" versus "picking a firm direction so future improv makes nicer rationalizations."
It makes sense if you think of the LLM as building a data-aware model that compresses the noisy data by parsimony (the principle that the simplest explanation that fits is best). Typical text compression algorithms are not data-aware and not robust to noise.
In lossy compression the compression itself is the goal. In prediction, compression is the road that leads to parsimonious models.
Some US washers don't but many do. However, US washers tend to not heat water as quickly or to as high of a temp. The video cites two reasons: 1. US power being 110V vs 220v. 2. US dishwasher heating elements being limited to 800 or 1000 watts because many are designed to potentially share one 20A residential circuit with an oven and/or fridge due to possibly being retrofitted into a kitchen built before built-in dishwashers were standard and manufacturers not wanting to create different models for retrofit vs new installs.
This plus the comment about sharing a circuit with an oven. If the oven is electric, even in the US it is 220v. If it is gas only, then it could be 120v as it only needs to run the igniter and other circuitry without running any heating elements.
I think he said sharing a circuit with a fridge, which are generally 110 in the US -- i think this is how my apartment is wired (2-phase 30A to oven dedicated, one 20A for the whole rest of kitchen)
Trying to run a resistive heater on the same circuit as a fridge compressor without tripping leans towards very conservative wattage
That's funny. Code in Ontario Canada is that the fridge needs to be on its own circuit. It's funny because we have an extra-big-ass inverter drive fridge that never draws more than an amp or two, even at startup because it's inherently soft-start.
But also helps avoid the case where your coffee maker trips the breaker shared with your refrigerator and you don't notice until the food in the refrigerator is warm. (which was a risk in my previous apartment - the counter circuits were shared with the refrigerator). I think it makes sense to have it as a separate circuit.
My thought was to share it with the lights, so you get an early indication if/when there is a fault than just your fridge going out.
> But also helps avoid the case where your coffee maker trips the breaker shared with your refrigerator and you don't notice until the food in the refrigerator is warm.
Didn’t notice the coffee was cold?
Overall, given the massive fears of a fridge failure, which can happen beyond just electrical failures, very very very few people have any kind of monitoring/alarming for this event. You’d think that would be the first requirement.
> counter circuits were shared with the refrigerator
Ouch. Code here (Ontario) is that not only does the fridge need a separate circuit, but counter outlets need two separate circuits: each socket on the duplex outlet is required to be on a separate circuit (although multiple outlets can all share the same two circuits, but you're supposed to alternate top and bottom).
Of course, if your home is older than I am or it's a handyman special, all bets are off. If I run the microwave while someone is vacuuming in another part of the house it'll trip the breaker.
Good point. I haven’t tripped a GFCI in a long while but I don’t actually know if my fridge will lose power when I do trip the GFCI. My guess is that it will since it does have a water line and ice dispenser so probably requires being wired into the same circuit.
Electric ovens in the US have required dedicated 40 or 50 amp circuits for decades per the NEC. Dishwashers, as well, have required dedicated circuits for a while but the 20 amp requirement is a more recent development (although it's probably been at least a couple decades).
Kitchens in general have required 20 amp general purpose circuits since at least the early 80s. However the NEC (but not the Canadian equivalent) allows for 15 amp duplex receptacles on 20 amp circuits so home builders looking to save a few pennies often use those. Besides, there are few if any, residential appliances out there that have NEMA 5-20 plugs. Then again hardwiring dishwashers was pretty common up until recently.
in traditional times it was customary to buy a few outfits high quality clothing that would last, and wear the same clothes for a week at a time, and then really boil them clean. This is the European market.
post world war 2 consumer choice culture in the US led to people buying cheaper clothing but varying their outfits every day and cleaning them (with copious availability of water) with less intensity.
once these patterns are established in the market, they become more like customary and it's what consumers expect of their appliances, detergents, etc.
American dishwashers are typically hooked up to hot water. Some will have heaters but they're not that powerful and they may only run for the main wash cycle
I can’t speak to Australian dishwashers, but trying to skip the video by catching a summary has failed you. Heating is discussed extensively in the video
The ones that do vary in ability by overall dishwasher quality.
The ones that don't are hooked up to the kitchen's hot water line.
This is considered more energy efficient because a home's hot water heater (whether electric, gas, or another fuel) is better at heating the water in a bulk capacity than a tiny heater in the dishwasher.
The downside is that the cold water between the big water heater and the dishwasher has to be purged first for it to be really effective. If your hot water heater is in the other side of the wall, no problem. If it's six rooms away, problem.
Hot water from the house supply isn't that hot though? My dishwasher gets MUCH hotter than the hot water supply... and I don't think the heater is "tiny" I think it's a rather substantial element. The dishwasher also doesn't need to heat up a "bulk" amount of water, just the amount of water used for washing the current load of dishes.
Hot water from the house supply isn't that hot though?
Depends on how you have it set. My current and previous hot water heaters had thermostats which permitted adjusting the temperature.
They also had warning labels on them about scalding water. If it's hot enough to scald, it's hot enough.
The dishwasher also doesn't need to heat up a "bulk" amount of water, just the amount of water used for washing the current load of dishes.
If you're washing dishes and someone is, or has recently, taken a shower; or someone is, or has recently, done laundry; or someone is, or has recently shaved or done any of the other dozen things that draw from the hot water heater, then the water is already hot and available and doesn't need to be heated all the way from cold by the dishwasher. A properly insulated hot water heater can retain heat for quite some time.
My cheap GE dishwasher uses a hot water line, but also has an internal heating element to "boost" it, and help dry. My electric bill definitely suffers if cold water is used.
Watch the video; it makes a huge difference even though the hot water input is not as hot as the water can get when the dishwasher runs its heating element.
Also the size of the heating element is irrelevant. What matters is the power dissipated. Most dishwashers in the US will use only about 900 watts of power even when plugged into a circuit that supports 1500 watts. In the EU they often hit 3000 watts. Even when just heating up a gallon or two of water that makes a huge difference.
Bosch dishwashers are NOT heat pump dishwashers! They have a pump that also heats the water. They call it a “heat pump” but I find that terminology a bit strange, as its clearly not a heat pump (which is a term of art in appliances and HVAC). It’s just a resistive heater that is no more efficient than any other dishwasher heating element.
Crystal dry uses a heat exchanger for drying, so I guess that is what they are referring to. I guess you could try to claim that only the way HVAC's move heat around qualifies for being called a heat pump, not any other way of moving it should count right?
Here is what I got from Gemini:
Bosch does not refer to their dishwashers as having a "heat pump" in the same way the term is used for HVAC systems; rather, the part is often called a circulation motor and heating assembly or a combined "heater/pump" unit by users and repair sites. Bosch dishwashers use a flow-through water heater (a type of resistive element) to heat the water and a different, non-refrigerant-based system for drying.
How the Bosch System Works
Water Heating: All Bosch dishwashers use a flow-through water heater, which is a heating element integrated with the circulation pump. This system rapidly increases the water temperature to the required level. It uses electrical resistance, not the reverse-refrigeration cycle of a true heat pump.
Drying: Bosch dishwashers (especially the higher-end models with features like CrystalDry) typically do not use a separate heating element for drying. Instead, they rely on a process involving a stainless steel tub and a heat exchanger or a mineral-based drying technology (like Zeolith for CrystalDry) to condense moisture and wick it away from the dishes. This is an energy-efficient method of moisture removal, not active heat generation for drying.
> I guess you could try to claim that only the way HVAC's move heat around qualifies for being called a heat pump, not any other way of moving it should count right?
Anything that moves heat from one area to another, not just evening them out but actually forcing the heat to move, counts as a heat pump. HVAC style, or peltier style, or other methods with tension or chemical reactions all count.
A combination heater and circulator does not do that. It is not a heat pump.
CrystalDry is an “upsell” feature on the high end versions of their dishwashers for more efficient drying. It’s not the thing that they’re calling a heat pump, that’s the water pump that also does resistive heating. Bottom line is, Bosch (nor anyone else afaik) does not make dishwashers that are more efficient due to using “heat pump” technology.
Sure, but most people don’t have a modern dishwasher. It’s an appliance that lasts 20 to 30 years ergo most people have old dishwashers that were manufactured decades ago.
Ya, but eventually the new tech will take over and we will be talking about very different things. Even if you just consider new builds, or hundreds of millions of Chinese buying their first dishwasher ever. Also, lots of utility districts offer incentives to upgrade to more efficient appliances.
Also, I’m way too lazy to look it up right now, but I’m quite certain I’ve heard of dishwashers that run the hot water for a little bit before letting it fill the basin. Like, I’m pretty sure this sort of thing is commonplace.
It’s not like the engineers for heaterless dishwashers are just too stupid to realize there’s an obvious workaround for having to purge the line before filling the basin. Especially when the performance is so much measurably better when you do it.
Like I said though, it’s a guess. It’s also possible efficiency certifications ding you for the excess water use.
Most of the new ones (at least higher end ones?) have heat pumps that heat water and handle drying. They are efficient enough to work on 110V, and the trade off is longer cycle times. Bonus: no more plastic utensils melting because they fell to the bottom resistive heating elements.
A dishwasher cycle is usually only going to run for a specific period of time. Its more effective it if starts that time closer to the proper temperature rather than relying on waiting for the heater to get the temperature up to that time. Especially on the pre-rinse cycle, where the heater may or (probably) many not engage.
The video explains that dishwashers sold in 110V countries often has a hot water connection as it's too slow heating water off a 110V/10A circuit so it is more efficent to utilise the hot water pipes. However we live in NZ, a 230V country so we get dishwashers that can heat water from cold fine off a 230V/10A circuit so no need for a hot water connection.
Modern heat pump dishwashers will heat water on 110V just fine, but you are looking at 3 hour wash/dry times anyways. My Bosch isn’t connected to hot water and even has a sanitize mode.
Bosch dishwashers have something they call a “heat pump” that is a water pump that also heats the water with resistance heat. It’s not any more efficient than the normal heating coils in every other dishwasher, and it isn’t a “heat pump” that uses a refrigeration cycle (as one would use to e.g. heat their house).
We’re talking about heating the water, not drying. The discussion started with regards to energy used to heat water vs. using water from the hot water tap.
I feel like it's probably pointless. The dishwasher will be full of water before the hot water starts coming out the pipe. Depending on how far the dishwasher is from the water heater I guess.
In most kitchens I've seen, the dishwasher is pretty close to the sink. In fact the sink and the dishwasher often share a shut-off valve. So if you run the water at the sink until it's hot, then start the dishwasher, it will get hot water.
Problem is, that most dishwashers have a prewash and a main wash. By the time the prewash is finished and the main wash starts, the water in the supply line will have cooled off quite a bit.
Is that the point of the air gap? I can't even get a straight answer from plumbers on what it's for. I don't see how that could possibly help with a clogged drain, just seems like a secondary point for the drain water to come out.
I'm fairly sure the point of air gaps on drainage is to prevent sewerage water from backing up in to appliances if the sewerage line is blocked. It will instead spill on the floor where it will be more easily noticed and cleaned.
That’s exactly what it’s for. If you block the sink drain and fill it with water, you can have water flow down the dishwasher drain hose and into the sump in the dishwasher. If that happens during the rinse cycle you’re rinsing with grey water.
Pumped out water has to go somewhere . With the airgap, it will either back out your garbage disposal or pour out your airgap into the sink basin, depending on the location of the blockage.
The airgap causes the pump to be physically incapable of backfeeding the drinking water supply with dishwasher waste
iirc its less about contaminating drinking water (there is a valve and pump to get through. rather tricky) and more about waste getting into dishwasher during cycle and you getting contaminated dishes.
my wife once decided to dump into garbage disposal a bunch of uncooked broccoli at once. it clogged garbage disposal and drain. when i tried to unclog it with plunger it backed into dishwasher (was hooked directly to garbage disposal bypassing airgap). took me hour to get everything out of dishwasher.
Thus the video's advice (also in my dishwasher's manual) is to run the water from a nearby sink until it's hot before starting the dishwasher. Because it helps significantly to get hot water at the input when US dishwashers are limited to 1200W of heating.
When I do the dishes I hand wash those that can't be put in the dishwasher before I start the dishwasher. This ensures that the water that goes into the dishwasher is already hot.
I don't think the dishwasher will be "full of water" as it doesn't actually fill up - rather, it only uses 2 gallons maximum per cycle, about the amount that would be the bottom of basin of the washer.
That's what I meant. The water drawn from the dishwasher is small enough to not even purge the cold water from the line in many houses. So you would just be wasting heat by filling the pipe with hot water while only taking the cold water from it.
This seems like something that only makes sense when water is scarce but electricity is cheap. You’d be constantly losing heat to the poorly insulated pipes.
They do. I didn't realize this until my natural gas supply company decided to replace my meter on a Friday. Without alerting me ahead of time so that I could, you know, plan to be gone while my house had no hot water.
Whenever natural gas supply is turned off in the US, for any reason, only the gas company can turn it back on. And they can't do so if there's a leak at all. You have to call a plumber to come out, detect the leaks, and fix them. After that, you can call the gas company to come back out (but not on a weekend) to turn it back on. And a same-day request for service requires someone to be home ALL DAY after it's called in.
And this is how I ended up showering at work for three days that week after not having had one over the weekend.
My parents used to have an old cooker which rather than having a spark button, had individual pilot lights for all of the hob burners and the grill. My mother was forever worried about whether one of the damn things had gone out (which they occasionally did). I think if you switched the supply off, switched it on again, and someone has left their house for a week, it might build up a significant amount of gas. Although they are supposed to be small enough not to. Presumably there were hardly any of those left now, but they can't assume they're all gone.
Pilot lights are often designed so that the heat from the flame holds a bimetallic switch in the open position. Should the light go out, the bimetallic switch will shut as it cools.
TBF the amount of gas used in old style pilots is really tiny. I’m sure it’s possible to accumulate dangerous quantities somehow, perhaps in a sealed subterranean basement if using propane instead of natural gas.
Natural gas is mostly methane, which is lighter than air and easily escapes most structures.
Natural gas today is mostly methane, but in the past it often had large concentrations of CO. In 1950 you can turn the gas on and stick your head in the oven as a form of suicide - won't work anymore (unless you get the house to explode).
Fascinating. I double-checked with ChatGPT (FWIW), and it confirmed. It said that currently, natural gas is extracted and shipped in its mostly pure form. In the mid-20th century, natural gas was "town gas," manufactured by heating Cole in the absence of oxygen. That produced a lot of carbon monoxide.
often around here in texas, when the gas is turned off due to an issue, the gas company disables the meter, or even removes or bypasses it. And I live in gas land, where we have natural gas piped in to the kitchen, bathrooms, laundry, outside for grills, as well as the furnace. We've seen it a lot, if you call the gas company about smelling gas, they come and remove your gas meter until you hire a plumber to go find the leak.
>Whenever natural gas supply is turned off in the US, for any reason, only the gas company can turn it back on
I had a seismic shutoff installed at my gas meter and the plumber who installed it had no problem turning off the gas and turning it back on when he was done. (and then turning it off again to demonstrate to me how it worked).
He re-lit the water heater pilot light before he left. The gas company was not involved at all.
Every country I have ever discussed with its residents has something that, on its face, is a reasonable safety precaution (I definitely don’t want to blow up my house), but in practice is just a way to make your life miserable while helping the people who work there have an easier day.
This just happens to be the one that affected me. Like modern gas water heaters that have electric ignition instead of pilot lights, because the one serious reason to have gas water heaters is that they work when there is no electricity. Now it’s just a price distinction.
This has always struck me as dumb, as until recently it was far cheaper to use your existing (gas-fired) hot water than to use a resistive element. However, with gas going out of fashion (and already hugely expensive in the Eastern states), and abundant solar PV, the calculus has changed.
The problem is that the first few litres of the water coming from the hot water pipe may be cold or warm. Therefore adding a resistive element is a better solution to guarantee a specific temperature.
Gas (especially just in time) still works well for water heating even if you can use heat pumps for everything else. No sure when that will flip, I assume it will eventually.
Gas is already outlawed for new builds in Victoria, despite vast gas resources in the Bass Strait. Presumably that's the direction other states are heading too.
It was a direction some states in the USA were heading before Trump, but now… anyways there will come an economic/technological point where electricity just makes more sense like it does for almost everything else. No need to legislate a transition when one will happen naturally, but we aren’t there yet.
My fairly cheap dishwasher in the UK has its own heater, but you can attach it to a hot water supply, which may save money as gas is so much cheaper than electricity.
traditionally (in household washing machine time) US houses were large and had a lot more hot water capacity for the whole house, and putting a heater into individual appliances was not necessary/cost effective.
retrofitting old traditional houses (especially stone) with higher capacity plumbing was expensive and infeasible, so putting heaters in appliances was a cope for markets that needed it.
I somehow have the memory that there was an extremely narrow time window where the speedup was tangible and quantifiable for Gentoo, as they were the first distro to ship some very early gcc optimisation. However it's open source software so every other distro soon caught up and became just as fast as Gentoo.
Fun tinkering but I don't find it very compelling to get root access by becoming the root user.
A more exciting usage is pairing this with qemu userspace emulation, so you can uncompress a filesystem from another architecture. You can have a terminal into a Raspberry Pi OS without a Raspberry Pi.
If you have ~$25k to buy a H200 then don't buy one. Rent them out much cheaper and keep renting newer models when your H200 becomes an outdated paperweight.
Assuming you ran inference for the full working day, you'd need to run your H200 for almost 2 years to break even. Realistically you don't run inference full time so you'll never realise the value of the card before it's obsolete.
The company I work for is in the defense industry and by contract can't send any code outside their own datacenter. So cloud-rented H200's are a no-go and obviously commercial LLM's as well. so breaking even is not the goal here.
Just like Microsoft there are parts of the company who are hostile to open source, and there parts of the company whose success is attributable to open source.
reply