Smuggler caught with 256 Intel Core processors wrapped around him in cling film The insatiable demand for microchips amid a global supply crunch of semiconductors is perhaps behind the following smuggling capers in China that just happened to catch our eye. In one case, on June 16, customs officials noticed the driver of a truck on the Hong Kong–Zhuhai–Macau Bridge was suspiciously shifty as his vehicle …
More likely, they were running overclocked in not ideal conditions. These are coming from people who had to invest a lot of cash into their mining "farms" and were probably not into making sure everything was running for duration in a properly cooled datacentre-like environment. I think I'd rather steer clear of any second hand GFX cards with the power to used for crypto-mining for the foreseeable future.
Yes, but what I'm saying is as long as the load was constant (and in a mining setup I don't see why it wouldn't be) then the fact that it ran consistently hot is going to put less of a strain on it in regards to thermal expansion. Nothing is melting even at higher operating temps; what ends up doing damage to components when placed under such a load is either not dissipating heat with crusty paste or constantly stopping and starting that loaf and causing large changes in temperature, the latter of which is not very often observed when mining 24/7. Sure, some of the components still might end up getting a good cooking if the temps get too hot and that can lower their lifespan, but at that point the card is being throttled or shut down, and any miner worth their salt will tune their performance to not allow this to maintain maximum hashing speed—even if that means lowering the OC a little.
I have a tad bit of experience in the mining arena and know some people that do it professionally, and this is the kind of thing I've seen in almost every shop I've looked at. Anecdotal, but there you are.
Genuine question because I have no idea how these things work.
In a PC they need drivers. Would the drivers be monitoring the cards and throttling if necessary? Or would that be done in the hardware?
I'm thinking that if they use rows of these in custom hardware then they would presumably manage card I/O directly and focus more on the code they need to thrash the card and produce output, possibly ignoring features designed to stop the card catching fire. Or maybe their rigs are actually PCs with drivers etc and I am talking rubbish.
Like I said, I have no clue.
They are driven by something with a CPU, and that could easily manage clocking (in fact, it almost certainly does). However, there's no way of knowing exactly how that was configured. Running the management system to avoid failures would make sense for a place that wanted to preserve their investment, but if they could obtain devices for relatively cheap, then it's also possible that they wrote the management software to run them as fast as possible to speed up their mining. Without knowing who was operating them, you don't know what settings they used.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
