Sony's ReRAM plans - what can and can't ReRAM bring to a console? *spawn


Ok.

I can't talk about what I don't know. It's a good position to come in with some skepticism because it takes a while for things to change.

I'm no developer and I'm maybe completely wrong on this one. Please don't insult me if I am.

How about AI and Machine Learning? I've read some articles detailing how Machine Learning will be a new thing in video games enabling more complex AI, dynamic and realistic worlds and another article that says persistent memory is an important tool and will accelerate Machine Learning and AI. One of the focuses of Sony's new AI division is gaming.

ML_Game_Development.png



Artificial Intelligence (AI) and Machine Learning (ML) applications are being developed for the enterprise and consumer markets at an exponential rate, but few developers are aware that persistent memory can play a critical role in optimising access to large data sets.

AI and ML technologies create highly demanding IO (input and output) and computational performance for GPU accelerated Extract, Transform, Load (ETL). The key challenge developers must overcome is to reduce the overall time to discovery and insight within data-intensive applications. Varying IO and computational performance is driven by bandwidth and latency. Therefore, the high-performance data analytics needed by AI and ML applications can be addressed by persistent memory solutions that offer the highest bandwidth and lowest latency.
 
Last edited by a moderator:
I think realistically in the next 5 years ML would only be used during training which happens during development and not during actual playing of the game.

I hope sony's new AI division whose focus includes gaming will accelerate things faster though. (with or without reram in the ps5)
 
How about AI and Machine Learning? I've read some articles detailing how Machine Learning will be a new thing in video games enabling more complex AI, dynamic and realistic worlds and another article that says persistent memory is an important tool and will accelerate Machine Learning and AI. One of the focuses of Sony's new AI division is gaming.
Most of our challenges are on the training side when concerning the larger datasets.

With completed models; which is what you’d use to do the predictions; the amount of space isn’t that large. Though as some savvy people will point out that some of the models can get very large in size; I don’t know if that means keeping AI models sitting in ReRAM as being necessarily efficient use of that type of expensive memory.

My general consensus with this is that I’d steer away with trying to use information to explain your theory. As someone who works in data science; we can make any data point tell any story we want. A good analyst will corroborate other data to prove it is probabilistically real.
Take for instance someone who tortures POW for information. They torture them for information but anyone will give up any information under torture. You don’t take any lead and run with it. The person doing the torturing must have other data points that they can trust to verify or see how this information aligns with past information. If it does not; it’s likely wrong. So they continue to torture.

data science is actually like this. We find all sorts of correlations but none of it may make sense and we go back until we find a theory that can hold a lot better; then we start machine learning to predict on it.

so part of the process of theory making is having people crush your theories. Thats a good thing. If no one can crush it that
Doesn’t mean you are right yet. You have one step to go. Because you’re only at the stage of the Mentalist. A bit of a “reader”. So perhaps you got some things right.

so if your theory is ReRAM will be in the next PS5; then as much as you are using data to explain it; your theory should also be able to predict what items will happen next. this is difference between the theory of gravity vs the theory of attraction.

Gravity predicts when objects will fall and at speed consistently. Any object with any setup. We can predict precisely with a high degree of accuracy what will happen.

Where the theory of attraction.shit that they try to sell you on midnight late TV is some bullshit to get you to buy their dvds and learn about how wishing about something will make it come true. There’s enough data points in the theory of attraction in which we draw parallels between wishing it happens and it actually happens. But it cannot predict it. Therefore; those correlations are about as equal as saying the MIcrosfit stock price is directly correlated to the number of Kylie Jenner Instagram followers.
 
That's deep. I understood some of the words you said. lol I'm looking for info to confirm my bias, I understand that. I'm doing an speculation and whatever I say will never amount to a confirmation. I'd be happy to have the preponderance of evidence favor my theory which, so far, lean against it. So I'm on the hunt looking for more info that would corroborate my theory.
 
So I'm on the hunt looking for more info that would corroborate my theory.
The problem there though is you're kinda reaching, and the discussion is being watered down from quality technical considerations to ReRAM speculation in pursuit of (a somewhat strange) wish fulfilment.

Hmmm, I can solve that spinning of this past few pages...
 
This is better I think. The last few posts I did, I did them because the practicality and necessity of inclusion of ReRam to PS5 was challenged. As a non-developer I cannot offer counter-arguments. So I posited some leading questions to some members with few researched data so I know I'm not too off-the-mark.

I hope you debate this well. I'll just be reading on your insights about this. I cannot participate here as I have no personal technical knowledge of what can or can't reram bring to a console.

I'll just be on the hunt to look for executive statements or burried news regarding reram's readiness to be manufactured in quantity as to increase the odds of having it included in the ps5.
 
Last edited by a moderator:
I'll just be on the hunt to look for executive statements or burried news regarding reram's readiness to be manufactured in quantity as to increase the odds of having it included in the ps5.

Why dedicate time to this then? ReRam vs whatever they have won't give you as a end-user much of a visible improvement. I'd rather wish for a more powerfull gpu, solid and fast RT hardware, and a near silent box, then ReRam.
 
That's deep. I understood some of the words you said. lol I'm looking for info to confirm my bias, I understand that. I'm doing an speculation and whatever I say will never amount to a confirmation. I'd be happy to have the preponderance of evidence favor my theory which, so far, lean against it. So I'm on the hunt looking for more info that would corroborate my theory.
Sure and that's fine. I think if we only waited for the absolute amount of evidence of everything, this forum would have 0 posts and would be dead. Part of discussion is discussion, and discussion and active participation leads to learning. The most important thing I guess, if you have a theory of things, try predicting what should come next before that information actually arrives. That would show you that you are on the right track. Doing it the other way, say waiting for information to confirm your bias; well that can actually turn out wrong right up to the end.


ie; if I asked you to make predictions of what the artist is drawing while he is drawing it. Those are like data points and you're making predictions. Ask yourself if you got it right; if you did that's great, because if you got it right away you could predict which parts he would draw next.

But if we let our biases override our thinking entirely;
well then we can get into this situation; even when the solution is so obvious our brains can anchor so deeply into one area and not realize that another obvious answer is staring at you in the face. A scenario where he put in every single vowel there is into that space he would have broken the possibilities into at least 50% (naive probability [I, A, O, E] is actually 25%, but only 2 are contextually correct [E and A], but E was already on the board leaving A), and then from there go with the correct answer because the other one is inappropriately wrong and contextually wrong.


So yea, try to predict what will happen next in the string of events; not just the end state. I think you'll know yourself if you are on the right track without having us to validate your theory.
 
Last edited:
I think you'll know yourself if you are on the right track without having us to validate your theory.

The necessity/benefit of the inclusion of ReRAM in ps5 is being questioned/challenged. What can I offer as counter-argument when I'm not a dev? I posited some leading question to you and others here to extract an answer that I cannot give. I gave a link saying how persistent memory with high bandwidth and low latency benefits machine learning and an image saying how machine learning can benefit gaming. But I didn't want to be presumptive to understand what I have just quoted hence I involved you to validate or invalidate THAT CLAIM, not my theory.

Anyway, I've received too much lectures today I'll just let it go in this forum. I will stop arguing my theory, let alone how it can benefit the ps5 as I don't have anything valuable to contribute there. I just hope you'll find it interesting enough to continue to debate this thread as I'm interested to here your thoughts what this new technology can or cannot bring to next-gen gaming. Thanks! ;-)
 
The necessity/benefit of the inclusion of ReRAM in ps5 is being questioned/challenged. What can I offer as counter-argument when I'm not a dev? I posited some leading question to you and others here to extract an answer that I cannot give. I gave a link saying how persistent memory with high bandwidth and low latency benefits machine learning and an image saying how machine learning can benefit gaming. But I didn't want to be presumptive to understand what I have just quoted hence I involved you to validate or invalidate THAT CLAIM, not my theory.

Anyway, I've received too much lectures today I'll just let it go in this forum. I will stop arguing my theory, let alone how it can benefit the ps5 as I don't have anything valuable to contribute there. I just hope you'll find it interesting enough to continue to debate this thread as I'm interested to here your thoughts what this new technology can or cannot bring to next-gen gaming. Thanks! ;-)
no worries. It's been good reading about ReRAM and sharing about it. I haven't dismissed ReRAM yet, the probability is very slim though. Greater than 0 but... probably less than 5% in my mind. Too much of the debate is around functionality, in which, even then every developer will be more or less arm chairing it because the hardware doesn't exist and no one has had a chance to use it.
Other topics like Risk Analysis; time to market etc; are probably much more critical in terms of whether or not it will arrive. A good example was the growing rumours that Scorpio would contain Zen 1 CPU. That certainly had better risk analysis than ReRAM. AMD is going to produce it, they have finished designs, they have the foundries to do it, the designs were made and publicized well in advance.

There might be other ways to showcase that Sony is on this path. But honestly I feel you've done a good job exhausting those possibilities already. I think it's probably a good time for us to just sit back and wait for more information.
 
NVME4.0 would be fast enough for what they did with the spiderman demo, and maybe users can change the storage also then, without having to buy Sony products to replace, cause then it's going to be expensive.
 
Even PCIExpress 3.0 has enough bandwidth for the memory they loaded during the loadtime comparison demo. The raw numbers are in the other thread. The limiting factor is how slow a PS4 can actually load data assuming no other time had to be spent elsewhere.
 
Back
Top