The first electronic speed computer memory (as opposed to rotational speed of a drum memory or the re-circulating time of an acoustic delay line memory) was the "Williams (Kilburn) Tube Memory" developed in England.
see Wikipedia's Williams tube.
A testbed for this technology was the "Manchester Baby" or more formally "Small-Scale Experimental Machine (SSEM)". This ran its first program on 21 June 1948

This memory was used in all "high speed" computers until the invention of coincident current magnetic core memory. (First installation on the Whirlwind machine in the summer of 1953.)
This new magnetic core memory was much more reliable and robust than the Williams Tube Memory and quickly replaced Williams Tube Memories in many existing machines and all new machines. "Core is King" lasted two plus decades until replaced by the silicon based large scale integration we are still using today. Chip memory is so cheap that basically "memory is free", and "lightning fast".

The Williams Tube Memory technology pushed the state of circuits art in its era. Microvolt transient signals in an environment of much larger noisy signals. Thousands of volts to drive the CRT tubes, hundreds of volts of deflection volts to point the electron beam. A real test of circuit design and shielding.

Oddly, memory parity was not used until the very much more reliable core memories arrived, where it was hardly needed.

IBM's "first" commercial scientific computer was the IBM 701. The first two pages of this oral history of J. C. Logue mention some of the memory development problems involved with this machine.

To develop aircraft and nuclear devices, high speed computers are extremely useful. Los Alamos National Lab (nuclear) secured an IBM 701, which required quite a maintenance staff to keep running usefully. Bob Erickson was one of the IBM maintenance staff.

from Robert Garner - Sept 21, 2021

The serendipitous acquirement of the IBM Williams tube module started in January 2012 when Judith Haemmerle spotted an article on Cnet’s website by its chief editor, Lindsey Turrentine, on the discovery of a 1950s IBM storage module in her grandfather’s “pole barn" near Tracy: "Antique IBM memory box + math = mind-blowing: CNET Reviews Editor in Chief Lindsey Turrentine finds a 1950s machine component in her grandpa's shed--and discovers anew just how blazingly fast technology has progressed." http://news.cnet.com/8301-17938_105-57364633-1/antique-ibm-memory-box-math-mind-blowing/ After emailing and talking with Lindsey, Bob Erickson and I were able to speak with her grandfather, Bill “WW" Moore, 89, in October. Bob and Bill had actually overlapped in a several areas at IBM in the 1950s — 701 CEs and RAMAC work in San Jose — but didn't know each other. When I spoke with Bill on March 28, 2013, he indicated that he would like to donate the dual Williams Tube module to Bob’s Williams tube reenactment project at the CHM. On Easter Sunday, 2013, Stan Paddock and Judith picked it up from Lindsey at a home in San Jose. > - what he intended to do with the demo rack Bob's plan for the Williams tube rack was toggle switches where one could manually read/write any bit location in a 8x8 array. The control circuits were all tube-based. (Bob didn’t have confidence in transistors... :) > - current schematics Bob’s initial schematics are here: http://ed-thelen.org/comp-hist/EricksonB2011-6.jpg. More schematics and design notes are stored together with the rack. In Jan, 2015, Bill Newman had the following thoughts about reinvigorating the project: "Bob did approve of me using an Arduino. Discussions with Marc Verdiell have led me to believe that Arduino can deliver the needed timing resolution one way or another. And, with enough HV MOSFETs, we could make all its I/O pins look like 6SN7's (~ equiv 12AX7 which I think is what was used originally). I'd follow Bob's specs (e.g. 64 bits, not 1024, and his timing notes). I think all the tech is there and doable. I'd like to see the drawer as functional as it could be. At 16 vacuum tubes it isn't a huge reliability issue. About like a 60's black and white tv. I do put it at 500-1000 hours to first glimmer of working. I think the UI would follow what Bob did originally, but just in software. We might be able to improve a little by a demo that marches different bits through it, for more visual impact (so it would be a bit more than Bob's switch interface)." ... One thing that still needs to be done is to get the Williams tube face re-metalized (for capacitive detection of the cathode ray beam), which perhaps could be done by his son, Dave Erickson, who has a career in coating optics: https://www.linkedin.com/in/dave-erickson-8b701693/ Dave retired in May, 2020: "I am now developing instrumentation and sensors for astronomical, optical, atmospheric, and geophysical interests in my home shop.” Previously he did "Remote Controlled Observatory Construction and Operation" and has an large observatory in Kennedy Meadows, southern Sierra Nevada: http://hbastro.com/index.html Before that he had a long career in optics, photonics, and lasers.** "The team successfully developed remarkably low absorption films with single digit part per million performance, suitable for the Space Based Laser Project. This accomplishment led to a contract to build the Large Optic Coating Facility located in our Temecula operation. A parallel project led to a significant product break through with commercial applications to the industrial CO2 laser market with the announcement and production of ultra low absorption MP5 antireflection coatings." I spoke with Dave back in January, 2020, right before the pandemic ramped up. He said he would be happy to metalize the face of the tube. I need to get back in touch with him and his sister, Robin, regarding the much postponed memorial 1401 luncheon for Bob and his wife Eileen. Take care, - Robert p.s. I had lent to Bob a small Russian textbook on Williams Tube memory systems, but we’re not sure where it is now…

from Robert Garner - Sept 24, 2021 - from "Proceedings of the I.R.E." October 1953
IBM 701 Electrostatic Storage

This was in response to questions, including was the contents of second tube a duplicate of the first? The author was Frizzell, as in Clarence E. Frizzell. There is a Youtube series (could use a lot of editing) on the "Development of the IBM 701" including Clarence - - segment 1 slow start, - segment 2 components, electrostatic storage - always close to disaster, drums, (the IBM 650 was 2 years later), - segment 3 Field Engineering, - segment 4 NoName machine, - segment 5 more - a little tale - in 1980, while feeling prosperous, our family moved to near the top of "Canyon View Drive" in Saratoga, CA. The neighbor at the top was a friendly old couple, the Frizzells. As I was employed by Measurex, not IBM, we did not talk business.