*shakes his head at all the nonsensical comments*
These workstations aren't designed for the mythical deep pocketed enthusiasts with short attention spans. If you wouldn't buy this product, then this product isn't for you. Lack of knowledge for who it's for doesn't mean it doesn't have a target audience. Bling is irrelevant outside of getting gullible PC gamers to buy crap because it has fancy LEDs and some illusion of quality because "gamer" designation. None of you are what OEM system builders like System76 are building and selling these systems for and to. Majors & Megas are designed and being sold to labs and businesses where they need either major compute resources or major multi-threaded IO capability, but don't really need very large HPC/supercomputing setups. Who would buy these? A former employer for one. She needs in-house number crunching capability way beyond that of a typical PC, but has remote access to supercomputer clusters at the US nat'l labs. That's inconvenient if all you want to do is run some Monte Carlo simulations or crunch the previous night's sensor harness test run. Whereas those simulations once took weeks to run "back in the day" on a DEC Alpha, they'll run in an hour or so on a Mega, and 512 GB of RAM is definitely enough to load the entire data set from a test run into RAM. Likewise, an employer before her would be pleased as punch to run crystal growth statistical simulations under his desk because it's limited by both parallelism and available system RAM. He can run the simulations & visualize the results locally instead of sending it off to an academic HPC cluster he may have to wait for time on. Tax payers pay for those research programs mostly through government agencies like the NSF in the US. Yet, there's plenty of cases where private engineering & research firms would want this intermediate step between normal PC workstation & HPC farm. System76 sells package deals. Time is money and no one outside of Google or Amazon wants their engineering or IT teams fiddling around building computers when they can just buy a turn key solution, load their test & in-house software on it, verify the results, and get on with their real work. For the open source crowd, you might see a System76 Major for build farms like FreeBSD's Poudriere, OpenSUSE's Build Service, etc. These are often funded by donations. Sometimes these are hand built (with the risk of unforeseen compatibility & stability headaches), but if you want warranty service, you pretty much have to go with an OEM designed and built workstation. I'd rather have some contract hire from the OEM's business/enterprise support group come fix the problem in a couple of hours rather than me spend all that time + more on the phone or more commonly days of back and forth e'mails with enthusiast grade parts vendors who keep refusing to admit a problem. I have better things to do.
These workstations aren't designed for the mythical deep pocketed enthusiasts with short attention spans. If you wouldn't buy this product, then this product isn't for you. Lack of knowledge for who it's for doesn't mean it doesn't have a target audience. Bling is irrelevant outside of getting gullible PC gamers to buy crap because it has fancy LEDs and some illusion of quality because "gamer" designation. None of you are what OEM system builders like System76 are building and selling these systems for and to. Majors & Megas are designed and being sold to labs and businesses where they need either major compute resources or major multi-threaded IO capability, but don't really need very large HPC/supercomputing setups. Who would buy these? A former employer for one. She needs in-house number crunching capability way beyond that of a typical PC, but has remote access to supercomputer clusters at the US nat'l labs. That's inconvenient if all you want to do is run some Monte Carlo simulations or crunch the previous night's sensor harness test run. Whereas those simulations once took weeks to run "back in the day" on a DEC Alpha, they'll run in an hour or so on a Mega, and 512 GB of RAM is definitely enough to load the entire data set from a test run into RAM. Likewise, an employer before her would be pleased as punch to run crystal growth statistical simulations under his desk because it's limited by both parallelism and available system RAM. He can run the simulations & visualize the results locally instead of sending it off to an academic HPC cluster he may have to wait for time on. Tax payers pay for those research programs mostly through government agencies like the NSF in the US. Yet, there's plenty of cases where private engineering & research firms would want this intermediate step between normal PC workstation & HPC farm. System76 sells package deals. Time is money and no one outside of Google or Amazon wants their engineering or IT teams fiddling around building computers when they can just buy a turn key solution, load their test & in-house software on it, verify the results, and get on with their real work. For the open source crowd, you might see a System76 Major for build farms like FreeBSD's Poudriere, OpenSUSE's Build Service, etc. These are often funded by donations. Sometimes these are hand built (with the risk of unforeseen compatibility & stability headaches), but if you want warranty service, you pretty much have to go with an OEM designed and built workstation. I'd rather have some contract hire from the OEM's business/enterprise support group come fix the problem in a couple of hours rather than me spend all that time + more on the phone or more commonly days of back and forth e'mails with enthusiast grade parts vendors who keep refusing to admit a problem. I have better things to do.
Comment