There are lot of confused comments in this thread.
TFA is not about ordinary bromine used in the chemical industry, which is extremely cheap and easily available everywhere.
TFA is about semiconductor-grade pure bromine, which, like all chemical substances used in the semiconductor industry is very expensive and it is not produced by most bromine producers.
Nobody in this thread has pointed to any evidence that USA produces semiconductor-grade pure bromine. The fact that it produces ordinary bromine is irrelevant.
You're all over this comment thread spreading your speculation as fact. Semiconductor-manufacturing grade bromine (specifically anhydrous hydrogen bromide) is, in fact, made in the US.
You may be right but it would have been nice to name the company that produces it.
The USGS article pointed by others contains absolutely no information about a US producer.
The parent article lists at least 3 US producers of ordinary bromine, but it claims that none of them produces semiconductor-grade pure HBr and the conclusion of the parent article is that someone should build in USA a purification plant, to avoid risks.
Perhaps TFA is wrong and there exists a US supplier, though even if that were true it is unknown whether it could increase its production if a shortage happened.
If you know information that contradicts the parent article, you should provide it, i.e. by naming the company.
In general any company benefits when more people know what it produces, so I cannot believe that you are bound by any confidentiality constraint to not name the producer, when you know it.
I have replied in many places in this thread because all the comments were wrong. This is no speculation. I have worked in semiconductor manufacturing and I know perfectly well the difference between semiconductor-grade pure substances and the ordinary chemical substances, while all those comments are wrong, by confusing these things.
Nothing that I said is speculation or wrong, because I have just pointed that those comments state conclusions that do not follow from their premises. The fact that USA produces ordinary bromine is no evidence that it produces semiconductor-grade pure bromine.
If you are right, that does not invalidate anything that I have said. It invalidates only the parent article, which concludes that a purification plant should be built in USA, while you say that such a plant already exists.
The linked article from USGS says nothing about semiconductor-grade purity bromine, but only about ordinary bromine that is used in the chemical industry.
Semiconductor-grade purity bromine is orders of magnitude more expensive than ordinary bromine and the vast majority of bromine producers do not make it.
The USGS article provides no evidence that such bromine is made in USA. I would rather expect Japan to be a producer, not USA, because for many semiconductor-grade purity chemical substances there are major producers in Japan.
Korea does not like to depend on imports from Japan, so I would not be surprised if there was a Japanese source of pure bromine but Korea prefers to import it from Israel. If this were true, they could still switch suppliers in case of a shortage.
That makes just bromine suitable for ordinary chemical processes.
Bromine with a semiconductor-grade purity, like any other chemical substance that may be used in semiconductor device manufacturing, must pass through a very long and energy-consuming purification process, which can be done in few places besides that from Israel that is mentioned in TFA.
At the moment. We could purify bromine gas anywhere and extraction and purification don’t need to be co-located. But at the moment, the purification and extraction in Israel are co-located, which is why this is more of an immediate risk than a long term one. However, it does take time to get new production online and no one will spend the capital to build a new purification facility that will go unused after the conflict is over.
Right. Commercial "high purity" bromine is 99.5% purity, and easily available at reasonable cost. Semiconductor processes want 99.9999%. Merck offers that. [1] €2500/liter from a lab supply house. Very few companies offer the post-purification.
That is exactly the conclusion of TFA, that in order to avoid such risks purification plants should also be built in other places, including in USA where the local producers of bromine can provide the raw material.
However, the construction of such a purification plant can take years, so TFA argues that it should be done ASAP, instead of waiting for some catastrophe that would destroy the existing plants, when this would be too late.
That’s not the issue. Someone could build the plant domestically (US), sure. It would take time, but getting a plant build isn’t hard. But who would do that? If the plant isn’t financially stable during non-crisis situations, then it won’t get built without subsidies. A plant that can’t offer a competitive price will never be used during “normal” times, so it becomes financially untenable.
We had the same issue with PPE manufacturing during Covid. We lack production capacity locally (US) because it’s normally cheaper to source from outside suppliers. When we try to build that capacity locally, it fails in the marketplace as soon as the crisis is over and the company is left with an expensive unused factory.
The hard part isn’t building a new plant. It is the commitment to ongoing support for maintaining a diverse supply chain that is more robust to geopolitical disruptions. We are not good at factoring risk into pricing, so we only accept the cheapest prices, to the detriment of a robust supply chain.
TFA is about high-purity bromine, not about ordinary bromine.
The purification processes for any of the substances used in the semiconductor industry are quite complicated and they are done in few places around the world. For many pure substances, major suppliers are located in Germany or Japan.
The substances with a semiconductor-grade purity are much more expensive than the ordinary substances. Being one thousand times more expensive is not unusual, which demonstrates the difficulty of the purification processes.
Your link does not provide any evidence that USA produces any bromine for the semiconductor industry.
Bromine itself is extremely cheap and easy to produce, like silicon.
Nevertheless pure bromine and pure silicon are very expensive and they are produced in few places around the world.
So you may have millions of tons of bromine, but if none of it has the required purity grade you must stop semiconductor device production until you build a purification facility, which requires money, time and know-how.
You can run huge local models slowly with the weights stored on SSDs.
Nowadays there are many computers that can have e.g. 2 PCIe 5.0 SSDs, which allow a reading throughput of 20 to 30 gigabyte per second, depending on the SSDs (or 1 PCIe 5.0 + 1 PCIe 4.0, for a throughput in the range 15-20 GB/s).
There are still a lot of improvements that can be done to inference back-ends like llama.cpp to reach the inference speed limit determined by the SSD throughput.
It seems that it is possible to reach inference speed in the range from a few seconds per token to a few tokens per second.
That may be too slow for a chat, but it should be good enough for an AI coding assistant, especially if many tasks are batched, so that they can progress simultaneously during a single read pass over the SSD data.
You can do that, but you're going to have rather low throughput unless you have lots of PCIe lanes to attach storage to. That's going to require either a HEDT or some kind of compute cluster.
Batching inferences doesn't necessarily help that much since as models get sparser the individual inferences are going to share fewer experts. It does always help wrt. shared routing layers, of course.
After NVIDIA essentially removed FP64 from consumer GPUs (their 1:64 performance ratio is worse than what you can obtain by software emulation, so it is useless, except for testing programs intended to run on datacenter GPUs), AMD persisted for a few years, but then they also followed NVIDIA.
AMD Hawaii GPUs still had 1:2 FP64:FP32, while the consumer variant of Radeon VII dropped to 1:4. The following AMD consumer GPUs dropped the FP64 performance to levels that are not competitive with CPUs.
Nowadays the only consumer GPUs with decent FP64 performance are the Intel Battlemage GPUs, which have a 1:8 performance ratio, which provides very good performance per dollar.
5090 is an overpriced outlier. A typical consumer GPU, like RTX 5070, has a 3-times lower memory throughput.
Even a RTX 5080 has a lower memory throughput than a Radeon VII from 2019, 7 years ago, while being much more expensive.
The memory throughput of GPUs per dollar has regressed greatly during the last 5 years, despite the fact that the widths of the GPU memory interfaces have been reduced, in order to decrease the production costs.
RTX 5080 has a 256-bit memory interface, while the much cheaper Radeon VII had an 1024-bit memory interface. RTX 5080 has almost 4-times faster memories than Radeon VII, but it has not used this to increase the memory throughput, but only to reduce the production costs, while simultaneously increasing the product price.
Modern GPUs like RTX 5080 are much faster for the applications that are limited by computational capabilities, mainly because they have more execution units, whose clock frequencies have also increased.
I suppose that most games are limited by computation, so they are indeed much faster on modern GPUs.
However, there are applications that are limited by memory throughput, not by computation, including AI inference and many scientific/technical computing applications.
For such applications, old GPUs with higher memory throughput are still faster.
This is why I am still using an old Radeon VII and a couple of other ancient AMD GPUs with high memory throughput.
Last year I have bought an Intel GPU, which is still slower than my old GPUs, but it at least had very good performance per dollar, competitive with that of the old GPUs, because it was very cheap, while the current AMD and especially NVIDIA GPUs have poor performance per dollar.
TFA is not about ordinary bromine used in the chemical industry, which is extremely cheap and easily available everywhere.
TFA is about semiconductor-grade pure bromine, which, like all chemical substances used in the semiconductor industry is very expensive and it is not produced by most bromine producers.
Nobody in this thread has pointed to any evidence that USA produces semiconductor-grade pure bromine. The fact that it produces ordinary bromine is irrelevant.
reply