jabbr

With the ZFS filesystem, you can maintain hot remote mirrors (vdev) and then ala @R1200CL I use jottacloud as a backup.

WiFi can be weird. I tend to turn off most optional QoS stuff because a good switch should give you 1Gbe from NAS to server and 1Gbe from server to NAA. That's to say that I have managed switches which I leave in unmanaged mode for the most part. Then there's broadcasts ... we have discussed using IPV6 as a way to reduce network broadcasts.

I know that Intel implements Data Direct IO (DDIO) in Xeons and their NICs to improve power performance and reduce latency on small packet tasks.

In summary: the entire TCP/IP stack can be implemented on the NIC: https://shelbyt.github.io/rdma-explained-1.html https://community.mellanox.com/s/article/what-is-rdma-x in any case my point is that the CPU “wake up” issue is not inherent to TCP/IP and the idea that an arbitrary protocol is “better” than TCP/IP is not in following with modern network implementations.

Err ... sorry this isn't how modern NICs work. My Mellanox ConnectX-4 and 5 NICs ***most certainly*** don't generate a CPU interrupt for each packet. In ancient days, perhaps, but modern networks just don't do this and the topic is far too complex for a robust discussion here ... see RDMA, ROCE Also look at how an FPGA processes Ethernet. A lot of these things is frankly why I'm hesitant to look at Diretta without access to a highly technical whitepaper. It's just handwaving to me otherwise.

Its never that black and white. There are most certainly very high quality SMPS which have very low common mode noise aka "leakage currents" ... of course these circuits might have on board or on chip linear regulators. An example would be a 100Gbe NIC that works in a bog standard PC with a noisy power supply. The NIC of course has its own onboard power supply that filters the junk away ... otherwise it just doesn't work!!! *** for some reason this is posted in the networking section so I am assuming power supplies for high quality network gear must be the interest??? M

Absolutely, the "hybrid" model of SMPS to act as a pre-regulator followed by a low drop out linear as Analog Devices discusses in the use of its LT3045 https://www.analog.com/en/technical-articles/low-noise-power-supplies-come-in-many-flavors-part-1-linear-regulators.html

yes, it must be the shield. It’s a test amount to technology that Cat8 exists and is able to carry 40Gbe over copper. That said the necessary shield is just begging for ground loop problems. I use Blue Jeans unshielded cables for copper Ethernet, and single mode fiber otherwise. The fiber works so well for me that I haven’t bothered with listening to the SQ of Ethernet cables in several years.

This is a well known "trick" when giving a lecture at a conference, always write up the talk as a paper because when it is published, it will get accepted without much if any real peer review. Don't believe every scientific paper as the gospel truth. If 10% of the papers published in Science and Nature every week were breakthroughs then humanity would be a lot collectively smarter than we are today. The bottom line is that you need to know the biases of the author when reading anything including a published scientific paper. We need independent journalists to keep those biases in c

apropos leakage currents ... I mean leaking discussions, the comments make their way into my feed despite blocking ... the AS blocking needs to be higher impedance!

Points all well taken, but furthermore consider the other benefits of a good isolating transformer ... and I use the beautiful Equi-Tech Q balanced isolating transformer... the readers are left to consider what the ground impedance to the main power box is, in fact. Furthermore given your balanced audio circuit, what do we expect the ground current to be exactly? ZERO!!!

That's the best way to do it ... and works! Consider not hardlinking to librocfft.so ( if no AMD GPU is present)

I saw that 460 was out for Windows and posted this before I tried. We had discussed the preferred methods of installing CUDA previously. It looks like, at the present time, the driver for Linux can be installed by installing CUDA 11.2. @Miska is there a current alternative?

It seems like the Linux version of 460 is not (easily) available yet -- I see its available for Windows

Aha, The highest driver level reported in $ ubuntu-drivers devices is nvidia-driver-455 Its possible to add sudo add-apt-repository ppa:graphics-drivers/ppa and get the experimental/beta drivers alternatively the driver can be downloaded from the nvidia site and installed. (just so folks know)