Slashdot Mirror
Intel To Support 128GB of DDR4 on Core 9th Gen Desktop Processors (anandtech.com)
Ian Cutress, writing for AnandTech: One of today's announcements threw up an interesting footnote worthy of further investigation. With its latest products, HP announced that their mainstream desktop platforms would be shipped with up to 32GB of memory, which was further expandable up to 128GB. Intel has confirmed to us, based on new memory entering the market, that there will be an adjustment to the memory support of the latest processors.
Normally mainstream processors only support 64GB, by virtue of two memory channels, two DIMMs per memory channel (2DPC), and the maximum size of a standard consumer UDIMM being 16GB of DDR4, meaning 4x16GB = 64GB. However the launch of two different technologies, both double height double capacity 32GB DDR4 modules from Zadak and G.Skill, as well as new 16Gb DDR4 chips coming from Samsung, means that technically in a consumer system with four memory slots, up to 128GB might be possible.
Normally mainstream processors only support 64GB, by virtue of two memory channels, two DIMMs per memory channel (2DPC), and the maximum size of a standard consumer UDIMM being 16GB of DDR4, meaning 4x16GB = 64GB. However the launch of two different technologies, both double height double capacity 32GB DDR4 modules from Zadak and G.Skill, as well as new 16Gb DDR4 chips coming from Samsung, means that technically in a consumer system with four memory slots, up to 128GB might be possible.
VMs obviously. Adobe as well.
Virtual machines ... CAD software ... databases ... rendering software ... huge data sets.
There's a lot of things for which "too much RAM" can never be true.
On a desktop I can burn through 16GB without even trying, and 32GB I can fill without trying that hard.
I can guarantee you, someone somewhere can chew through 128GB of RAM for their specific problem.
I'm dealing with 100GB-1TB databases at work on a regular basis — I'm sure, others have encountered even bigger ones. Fitting all — or most — of the dataset into RAM is greatly speeding things up. Indeed, there are database-software packages already (such as, ugh, "memcache"), that must load it all into RAM, offering dramatically-improved speeds in exchange for this requirement.
On the OS level, swap — and the associated complexity of the kernels — is becoming unnecessary in more and more cases. On some of my FreeBSD machines, for example, I'm already compiling the kernel with options NO_SWAPPING.
On the filesystem-level, ZFS — the revolutionary filesystem — can offer much better speed with more RAM. The abundance of RAM is also making its advanced features (like deduplication) practical.
And for a layman's personal computer, editing a 4K video becomes much snappier too, if the the entire (uncompressed) clip fits into RAM.
And then come things like "machine learning" — I'm waiting for a Thunderbird add-on, for example, to automatically sort my incoming e-mail. Not just "spam/not-spam", but all of it, based on the ongoing analysis of how I've been sorting it through the years... For those things to be effective, they need both CPU and memory — continuously...
Other examples — legitimate and otherwise (like Chrome) — abound...
In Soviet Washington the swamp drains you.