News Intel and Lenovo Develop Future of PCs in Shanghai

[bit_user]

The article from the European research center IMEC (which is likely the world’s most important research center for developing semiconductor manufacturing tools and processes) seems to indicate that VG-SOT-MRAM has all the requirements to replace (at least some level) cache (at least L3 and maybe further down).

The closer you get to the CPU cores, the more stringent the bandwidth and latency requirements get.

The main thing I'm saying is that if you want to solve a problem, try to learn everything you can about it. If you're pitching it to me as a solution for cache memory, I shouldn't have to remind you about tag RAM. You should already know all about it, and how your MRAM-based solution can most-effectively be adapted to it.

For all I know, there might be interesting ways to use some property of MRAM to make tag lookups more efficient. That's on you to figure out, since you're advocating for it.

It is likely that before end of 2025, there could be some prototype chip demonstrating the viability of this (it seems they are already working on that…).

Prototypes are great, but you need sound theory for even believing an idea is taking that far. Clearly, someone did that homework.

For AI applications, it seems that VG-SOT-MRAM could be used differently and seems an interesting candidate for implementing multi-level deep-neural network weights for analog in-memory computing, and likely much, much better suited than highly energy inefficient DRAM.

Could be, but you ought to know the argument for doing so, before trying to make it. To convince someone, you should at least be able to quote the most relevant bits of the documents you cite. If you believe in the idea as much as you say you do, that shouldn't be too much to ask.

I'm not opposed to anything you're saying - just skeptical. I could be convinced, but you're currently a ways away from that.

Good luck on your journey.

 

[Diogene7]

The closer you get to the CPU cores, the more stringent the bandwidth and latency requirements get.

The main thing I'm saying is that if you want to solve a problem, try to learn everything you can about it. If you're pitching it to me as a solution for cache memory, I shouldn't have to remind you about tag RAM. You should already know all about it, and how your MRAM-based solution can most-effectively be adapted to it.

For all I know, there might be interesting ways to use some property of MRAM to make tag lookups more efficient. That's on you to figure out, since you're advocating for it.

Unfortunately, I am clueless about tag RAM you are talking about but I am confident that if IMEC has already progressed VG-SOT-MRAM to building an array (I would assume a cache memory array), they are quite confident it could replace some cache levels.

So wait 1 to 3 years that IMEC publish research articles that demonstrate that VG-SOT-MRAM used as cache is at least as fast while also more power efficient : I would think it is a reasonable enough proof.

Prototypes are great, but you need sound theory for even believing an idea is taking that far. Clearly, someone did that homework.

Yes, IMEC did that homework in the articles I attached previously. I know it takes a bit of time to read them, but if you do, you will see their reasoning and the technical challenges they had to solve and successfully solved.

Could be, but you ought to know the argument for doing so, before trying to make it. To convince someone, you should at least be able to quote the most relevant bits of the documents you cite. If you believe in the idea as much as you say you do, that shouldn't be too much to ask.

I'm not opposed to anything you're saying - just skeptical. I could be convinced, but you're currently a ways away from that.

I understand but unfortunately my technical understanding of the inner workings of electronic components is limited (let say I understand the concepts but not necessarily the deep technical details).

I would think that if we had again this conversation towards 2024/2026 after IMEC publishes research articles that demonstrate that VG-SOT-MRAM used as cache is at least as fast while also more power efficient, it would be easier to provide you with the proof you are looking for.

If you could take some of your valuable time to read the IMEC articles, I think it would already help you have a different view on this topic.

Good luck on your journey.

Anyhow, thanks very much for your constructive feedback : very helpful, I appreciate it a lot.

 

[bit_user]

Unfortunately, I am clueless about tag RAM you are talking about but I am confident that if IMEC has already progressed VG-SOT-MRAM to building an array (I would assume a cache memory array), they are quite confident it could replace some cache levels.

I know that, but you can clearly read and are resourceful enough to find sources. I'm suggesting that if you're planning on advocating for such a solution, then you should educate yourself more about it. The more you know, the better you can advocate (or maybe you find something that changes your mind about whether you should).

Yes, IMEC did that homework in the articles I attached previously. I know it takes a bit of time to read them, but if you do, you will see their reasoning and the technical challenges they had to solve and successfully solved.

You can never expect people to read your sources. I'm not invested in this outcome, you are. So, you should be quoting the key parts from your sources that you think are most important and relevant to the discussion. That's what I do and it gives the best chance of the other person actually seeing the part you want them to.

When quoting something in a post, you can simply wrap it in a pair of [quote] ... [/quote] tags, but since those are easily confused with the quotes of the other person's post, I prefer to use [indent] ... [/indent] block and add the " delimiters, myself:

"This is a quoted passage from my source. I might underline key parts, to add emphasis."
Source: link_to_source
​

If I do use the [quote] tag, I will at least wrap it in an [indent] block to help set it apart from quotes of the other person.

I understand but unfortunately my technical understanding of the inner workings of electronic components is limited (let say I understand the concepts but not necessarily the deep technical details).

So is mine (I create software, not hardware), but there's a lot of information out there about this stuff, and you'd probably understand more of it than you think.

Anyhow, thanks very much for your constructive feedback : very helpful, I appreciate it a lot.

Thanks for saying that. I was hoping you felt you got at least something of value from it.

 

[Diogene7]

I know that, but you can clearly read and are resourceful enough to find sources. I'm suggesting that if you're planning on advocating for such a solution, then you should educate yourself more about it. The more you know, the better you can advocate (or maybe you find something that changes your mind about whether you should).


You can never expect people to read your sources. I'm not invested in this outcome, you are. So, you should be quoting the key parts from your sources that you think are most important and relevant to the discussion. That's what I do and it gives the best chance of the other person actually seeing the part you want them to.

When quoting something in a post, you can simply wrap it in a pair of [quote] ... [/quote] tags, but since those are easily confused with the quotes of the other person's post, I prefer to use [indent] ... [/indent] block and add the " delimiters, myself:

​

"This is a quoted passage from my source. I might underline key parts, to add emphasis."​

Source: link_to_source​

​

Yes, I think you are right : it would help make it easier for the other person and speed up his logical reasoning.

If I do use the [quote] tag, I will at least wrap it in an [indent] block to help set it apart from quotes of the other person.


So is mine (I create software, not hardware), but there's a lot of information out there about this stuff, and you'd probably understand more of it than you think.

Thanks for saying that. I was hoping you felt you got at least something of value from it.

I might be a bit lazy so not sure I will apply everything but I will try to work towards the goal you described.
However I confirm your feedback is very constructive, and I can feel you are willing to help : so it is helpful !!!