I think 4DS throwing that chart in was a little... misleading...

I think 4DS throwing that chart in was a little... misleading for the average person if I'm honest. Or perhaps not because for 3 slides they're explaining the problem with energy requirements for AI?

The fact is that NAND remains king for offline storage (as 4DS clearly state in that same slide). Intel's Optane was a multi-billion dollar loss technology, benefits over SSD's for the added manufacturing costs (combined with the reasons Khoutan listed). This was the area that 4DS was originally looking at and for many were hoping would be more of a universal memory that could potentially replace both SSD's and DRAM.

Now with the continued improvements to NAND, along with reduction in manufacturing costs by increasing the number of layers and density, there's literally no reason to be targeting in this space at all - which 4DS have made perfectly clear that they aren't (so - this whole discussion point is... pointless, which I don't mind as I may learn something in the process).


As 4DS have achieved 4.7ns write speeds is it? and <10ns read speeds? The concept of SCM is silly because we would be better off replacing DRAM entirely instead of having it sit somewhere in-between as an auxillary memory. With the scaling benefits of non-filamentary RRAM (to be demonstrated with PL6), it should also allow for higher density, lower energy consumption, etc. Nothing but pure benefits over DRAM.

4DS have also stated its compatible using existing fabrication equipment, so manufacturing costs should be comparable to DRAM. For D10 and D19 the yield was 97% - can't recall what PL3 was and 4DS haven't said what PL4 was. So thats a big tick as you want to be above 85-90% yield (unless they have withheld the yield of PL4 because its bad?)
DRAM wafers take about ~12-14 weeks. 4DS RRAM, especially this PL5 are taking a lot longer as Ting mentioned "taking things slow". We can guess why they've had to slow down the process between PL4 and PL5 - and as other posters have thrown out there, could be because they're doing PL6 concurrent. Don't know the reason, but that goes into the economics of the manufacturing costs and therefore value of the IP.

Now I'm just talking about what we know, and I'm saying the technology since PL4 is better than DRAM - this is why I'm holding.
If the technology can be demonstrated to be used as an embedded memory, thats another big plus and both Khoutan and MJ have covered these discussions quite well in my opinion. The technology is solid and if they can demonstrate the scalability with PL6, then it should be money in the bank**.

Now this is where I diverge from the koolaid drinkers - because I'm willing to criticise the current 4DS BOD where I think they need to lift their game. Whether its politics, corporations, universities, manufacturers... it makes no difference, we need to stop rewarding poor performance and call it out.

The shareprice isn't sitting at (checks current price) 8.3cps because the memory IP is bad... its sitting there directly because of what this BOD is doing. And the share price doesn't matter once the IP is sold so I personally don't care about the day to day fluctuations, however if the BOD stay on the path they have been, **they risk this never being sold or being largely under valued.


Just trying to keep things balanced and in check.