There is talk about quantum computing and limitations of current computer architecture/ hardware. HPE is working on prototype of memory driven computing prototype where you can hopefully address tens of TB even hundreds of TB memory, GPUs used in some setups for extra processing power. Much work is obviously to be done which ever solution is the next viable step or even ground breaking. Not saying it is one of the mentioned ones and don't really understand well the quantum computing to start with.

However my question is not specifically about the next technology but limitations programmers here experience when building solutions from HW/architecture perspective or ideas what they would like to do if able to think big? For some it may not matter so much whereas others may be pushing the resources to the limits.

From my little knowledge it seems like AI is the next big thing so stuff like Google's hardware that assists AI data processing will be the next thing before quantum computing.

https://www.wired.com/2017/05/google...d-new-ai-chip/

Along with AI I would imagine better and better sensors and data collectors for AI to collect and process data from the real world will grow along with it.

Edit: Also more and more robotics performing everyday tasks.

RAM & Disk Space will not be separate units in the future, SSDs will be fast enough to be used as both. I wonder if there are more components that can be merged that way.

The pursuit of quantum computers is due to the limitations of continued semiconductor transistor shrinking. This isn't really competing in the same realm of AI hardware architectures. You could have a quantum computing AI chip though architecture will be different to take care of the multiple quantum states.

Gullanian, I am highly skeptical of RAM/disk space merging. Do you have some articles? It is not just SSD access time that is the bottle neck. It is addressibility. There is a reason you have multiple cache levels.

https://arstechnica.co.uk/informatio...details-price/

Yeah a 64 bit address space is huge.

My thought about the architecture of the future is which architecture are we talking about? Supercomputers, microprocessors, DSPs, microcontrollers, GPUs?

I'm interested as well if anyone here is working on projects which push the current resources to limits.

Say,we would need to be able to have more RAM and multiple cpus processing the data.
For example We need to load 10TB of data to memory to work with large datasets but now our servers normally have only 500GB or 1TB of memory and multiple cpus cant work effectively on that data.
I gave an example what hpe is trying to do with model with hundred TB of memory and data scientist be able to work with massive datasets. Below link with picture of the idea
https://news.hpe.com/memory-driven-computing-explained/


Interested of any real world examples what people/developers may have experienced or feel are limiting factors rather than only merely looking at current development in industry.

AI is the future, and also robots, but mostly AI.

There needs to be processes to determine how to more efficiently utilize the processing power today. A quantum processor will improve things, but won't address the real bottlenecks in technology, IMO.

Agree. My understanding was that quantum computing doesn't necessarily solve the physical limitation issue of traditional transistor based computers, it's just suited to better model and compute computational problems that are otherwise impossible to model on current architecture.

Sent from my SM-G900R4 using Tapatalk

I was looking at refurbished servers, solid state options, etc. Seems like data is cheap. Interpreting data on the other hand...

Deep Blue in some respects showed the power of brute force. The AI scientists since have been thinking how to use other methods, in particular, machine learning comes up a lot.

AI was claimed to be the future when I was an undergrad in the 80s and Minsky was king. Other things that were predicted to be the future were 4th Generation Languages, Bubble Memory, Optical Transistors and the 3 day working week.

I think you are right in that AI will certainly be a significant growth area now, but I also suspect it won't be in the form anyone is currently predicting.

Media has always gotten AI wrong. They were so far off in the 80s, it was ridiculous. The mind is still a mystery. It'll be automation in the short term. Even today, the media gravitates towards these friendly do nothing robots who come in peace. It's insane. But we can already imagine automation, such as drone delivery as not far off.

Automation, AI, then the super-duper processors, imo. In the interem there's going to be all kinds of new softer tools to make the process of making the processes cheaper and easier.

We do bits and bobs with bonkers amounts of data. One of our clients has in the region of 100T of data.

We tend to use AVX for data parallelism. It works reasonably well. GPU would be nice, but memory limitations tend to make it not viable.