Stacked Memory?

[DDR4]

I heard that micron was creating a new stacked RAM memory module. http://www.anandtech.com/show/4819/intel-and-micron-develop-hybrid-memory-cube-stacked-dram-is-coming
What does everyone think? What will the system requirements be?
Thanks

 

[exdeath]

1 Tbit/sec is great. Now make it non volatile so we can eliminate the HDD/SSD. All that bandwidth and processor speed is meaningless if the initial data source is still randomly accessed from an 8 track tape at 1.5 MB/sec.

Data storage and retrieval from non volatile storage is the most serious bottle neck in this day and age of gargantuan data centric apps. Not main memory bandwidth or CPU. Make the CPU or RAM any faster, and you're just going to notice you're computer still boots and launches apps at the same speed it has been for the last 10 years.

256 GB of 1 Tbit/sec main memory doesn't mean much when you have to wait 600 seconds for Lotus Notes to launch and load your 30 GB replica into that RAM at only 50 MB/sec... don't need to load it all at once you say? Then you're just back to square one with hitching and hourglasses and pauses hitting the disk in real time as you navigate and search... then where does this uber fast DRAM leave us? Spending 10 minutes loading stuff into RAM, crunching it in .01 seconds with our uber fast CPU and RAM, then spending 3 days writing the results back to C: where it will take another program 150 minutes to load that data for analysis, hitching and stuttering and "(Not responding...)" every time you scroll the data set and cause the HDD light to light up the night.

We need a revolutionary change in our data storage technology before anything else IMO. Yeah SSDs are better, but you're still talking a scant 100s of MB/sec compared to 10s of GB/sec of current memory bandwidths let alone the TB/sec speeds discussed in the article. Data set sizes are ridiculously out of control, and faster RAM isn't going to help us load it faster from the HDD/SSD or heaven forbid spend all day transferring it from one machine to another. Faster CPUs and RAM are just going to make the problem worse as the size of our data continues to grow ever more obese while still being saved to 1950s cassette cartridges.

They need to quit with stuff like this and put more effort into stuff like large scale high density STT-MRAM, but I guess they still don't want to upset their insane flash memory profit margins just yet...

 

[Coup27]

Honestly man, do you ever stop bitching about HDDs and storage?

Now make it non volatile so we can eliminate the HDD/SSD.

The reason DDR is so fast is because its volatile. That aint gona change anytime soon.

We need a revolutionary change in our data storage technology before anything else IMO. Yeah SSDs are better, but you're still talking a scant 100s of MB/sec compared to 10s of GB/sec of current memory bandwidths let alone the TB/sec speeds discussed in the article.

This is just ridiculous. How can you compare the speed of volatile and non volatile?

 

[groberts101]

LOL.. he just hates waiting for slow ass transfers is all. Although, I do agree that less is more sometimes. @exdeath.. do you use raided SSD? because if you match that fast SSD(raided SSD is even better) with very fast raided HDD based storage(or better yet..SSD based)?.. you won't be waiting all that time while leveraging what this current tech has to offer. I spent an arm and a leg to do it.. but my system throws GB's back and forth 5-10 times faster than most others who use single HDD based storage setups.So in my case.. yes mram or another solution like this would be great.. but the speed increase would not blow my mind over what I currently have anyways. But the price for that kind of speed and cutting edge tech surely would. Just sayin', there's always a compromise to be made.

 

[exdeath]

This is just ridiculous. How can you compare the speed of volatile and non volatile?

The fact that you still distinguish between the two and feel that it should matter or that there should even be a difference just shows how most of us are still clinging to our ancient computing fundamentals are as a species. This needs to change. Time for universal memory technology, and we need it yesterday.