The hardware of photography processing

You can still edit raw files, without needing the latest and greatest, biggest and best in computer hardware. My little mini computer is ample proof of that, without stressing the bank account in the slightest.
And there's nothing stopping you from editing your JPEGs should you so desire, either.

The image quality available today, for very little money, is astounding I feel. Although the subject matter may not be to your taste, there's not much wrong with the quality of the out of camera JPEGs or of my $500 Au G100 camera, and second hand 14-42mm kit lens. See for yourself

I spent six months of my life evaluating the brand new Gray I my agency had delivered to Princeton in the early 1980's. The Cray I was large and it produced tremendous amounts of heat as it use Fairchild F100 K - fast but power hungry. The only high speed logic at the time was ECL (emitter coupled logic). Eventually CMOS caught up.

Every card had a massive heat sink which plugged into a socket surrounded by copper piping that carries super chilled water to carry off heat. The refrigeration unit that supported the computer and the computer doubled the power foot print of the building it was housed in. However, the Cray I was a massive upgrade over what was available at the time. My iPhone 14 Pro has much more computer horsepower. My MBP 14 with the M1 Pro blows it away. Gordon Moore, founder of Fairchild Semiconductor and later Intel made a prediction which came to be known as Moore's law. One corollary is the followed. For the same number of dollars the computer horse power goes up by an order of magnitude every five years. Interestingly enough that trend has held amazing accurate over time. So 1980 to 2020 is 40 years. That is a factor of 10^8 in computational horsepower for the a dollar in 2020 vs 1980. So yep - it's about right.

So one does not need a super computer to process images. However, one does need to use the current state of the personal computing art with computers that efficiently execute AI applications based on Tensor Flow and sufficient vector processors, a.k.a., GPUs. Those are the tools of the trade for imaging processing and the code is written optimized based on these specialized processing. Super computers today uses superconducting quantum based technologies, such as Josephson Junctions (which have to be super cooled to function). There are such computers alive and operating as we speak. These computers are just the tip of the iceberg of what lies ahead in super computing. We've come a long way since the IBM PC of the mid 1980s.

The trouble is that my 20-yr old Sigma SD9 DSLR only outputs raw files - no SOOC JPEGs.

But my other cameras (Panasonics) can output JPEGs complete with lens correction. However I'm stubbornly sticking to the Sigma, being old.

When I'm been trading my computers I have looked at PugetSystems' articles to see what is the current tech level in computers.
They are making customized computers to different jobs. They don't ship to Europe, but they publish interesting articles and comparisons.
www.pugetsystems.com/all-articles/?filter

When trading computers last time two years ago I spent ~3K€ with monitor. So I'm afraid you'll need a little more than £2,000.
But you might know that.😉

In grad school - this dates to the early 1970's ( Yes I knew Moses ) I supported the Center for the Social Organization of Schools which was a research facility at Hopkins' psychology department to augment my other financial aid. My job was to provide computer support. At the time the super computer at the time was the IBM 7094 which was in the basement of the administration building on campus. The center had a IBM 1401 on site that we used for a lot of task. The 1401 was a "decimal" computer operating on the basis of a variable word length. That is unlike a fixed word length, memory was a big continuous string of core memory and the programmer was responsible in setting word boundaries for his particular application. Memory was small so one had to not waste any. Memory at that time was also magnetic core - not semiconductor. One also could use ( actually had to) one of 4 switches on the computer itself in the programing. While it had a Fortran complier which was stored on tape and to compile a program one had to mount the compiler tape and a "scratch" tape to be used in the process. In general to get the machine to purr you had to write in its basic assembly language known as Autocoder. en.wikipedia.org/wiki/Autocoder For its time Autocoder was quite a nice pice of computer science and it allowed one to unlock the full capabilities of the hardware and make the machine more productive than it had any right to be. Autocoder was the basis for the assembler for several generations of IBM machines. For my wife's PhD thesis (computer science) she developed a highly efficient assembler that was an enhancement IBM BAL, a.k.a. Compass (the family of assemblers for the IBM 360 ) for scientific computing. She said that a of the basic foundations of the BAL was based on Autocoder. For the 1401.Program input was through a punch card reader. Data was stored on tape or the ultra high capacity 20 MB hard disk drive 🤪. The drive was as large as a washing machine and shook the floor when the heads were moving. It had removable disk and often one the user had to swap out disk during a run.

We've come a very long way. As a side note. I am using the original MBP 13 inch M1 as a disk server for my back up system for our multiple computers. My Intel based Intel MBP (about 5 years old) had a swollen battery so I had it replaced and gave to my granddaughter. I used the 13 inch MBP for about a year and then picked up the MBP 14 inch for my main computer and, got read of my NAS disk as it was a royal pain in the butt and the 13 inch MBP became my backup disk server. It was an easy configuration after all deep down in its pea picking heart all Mac's are Unix boxes. My wife has a 4 year old high end iMac (Intel). When she is heavily using Capture One - it sounds like an airplane taking off. On my MBP cranking away on C1 - I've yet to hear the fan.

DXO Photolab generally does a good raw processing job for me: its noise removal algorithms are excellent. With the Canon R6 20MB sensor it is not slow, but the Leica Q3 is another story -- well, it is 3x larger!

Silkypix seems to specialise in Fuji and Panasonic.

I use Photoshop CS5 for changing file size, final sharpening, and saving to jpeg. It also does a good job for me in stitching panoramas. This quick and dirty composite of four shots took a while to achieve and used a lot of memory, but the size (originally 27578 x 5436) was exceptional!

David