Tuesday, October 21st 2014
Samsung Starts Production of 8-Gigabit DDR4 Based on 20 Nanometer Technology
Samsung Electronics announced that it is mass producing the industry's most advanced 8-gigabit (Gb) DDR4 memory and 32-gigabyte (GB) module, both of which will be manufactured based on a new 20-nanometer (nm) process technology, for use in enterprise servers.
"Our new 20 nm 8 Gb DDR4 DRAM more than meets the high performance, high density and energy efficiency needs that are driving the proliferation of next-generation enterprise servers," said Jeeho Baek, Vice President of Memory Marketing at Samsung Electronics. "By expanding the production of our 20 nm DRAM line-ups, we will provide premium, high-density DRAM products, while handling increasing demand from customers in the global premium enterprise market."With its new 8 Gb DDR4, Samsung now offers a full line-up of 20 nm-based DRAM to lead a new era of 20 nm DRAM efficiency that also includes the 20 nm 4 Gb DDR3 for PCs and the 20 nm 6 Gb LPDDR3 for mobile devices.
Using the new 8 Gb DDR4 chip, Samsung began producing the 32 GB registered dual in-line memory module (RDIMM) earlier this month. The new module's data transfer rate per pin reaches up to 2,400 megabits per second (Mbps), which delivers an approximately 29 percent performance increase, compared to the 1,866 Mbps bandwidth of a DDR3 server module.
Beyond the 32 GB modules, the new 8 Gb chips will allow production of server modules with a maximum capacity of 128 GB by applying 3D through silicon via (TSV) technology, which will encourage further expansion of the high-density DRAM market.
The new high density DDR4, also boasts improved error correction features, which will increase memory reliability in the design of enterprise servers. In addition, the new DDR4 chip and module use 1.2 volt, which is currently the lowest possible voltage.
"Our new 20 nm 8 Gb DDR4 DRAM more than meets the high performance, high density and energy efficiency needs that are driving the proliferation of next-generation enterprise servers," said Jeeho Baek, Vice President of Memory Marketing at Samsung Electronics. "By expanding the production of our 20 nm DRAM line-ups, we will provide premium, high-density DRAM products, while handling increasing demand from customers in the global premium enterprise market."With its new 8 Gb DDR4, Samsung now offers a full line-up of 20 nm-based DRAM to lead a new era of 20 nm DRAM efficiency that also includes the 20 nm 4 Gb DDR3 for PCs and the 20 nm 6 Gb LPDDR3 for mobile devices.
Using the new 8 Gb DDR4 chip, Samsung began producing the 32 GB registered dual in-line memory module (RDIMM) earlier this month. The new module's data transfer rate per pin reaches up to 2,400 megabits per second (Mbps), which delivers an approximately 29 percent performance increase, compared to the 1,866 Mbps bandwidth of a DDR3 server module.
Beyond the 32 GB modules, the new 8 Gb chips will allow production of server modules with a maximum capacity of 128 GB by applying 3D through silicon via (TSV) technology, which will encourage further expansion of the high-density DRAM market.
The new high density DDR4, also boasts improved error correction features, which will increase memory reliability in the design of enterprise servers. In addition, the new DDR4 chip and module use 1.2 volt, which is currently the lowest possible voltage.
38 Comments on Samsung Starts Production of 8-Gigabit DDR4 Based on 20 Nanometer Technology
I have 8 GB and turned it off with no issue what so ever... but i doubt that technology will fade away, i guess it makes a nice fail safe.
Yes please! Now thats the reason to go for DDR4, at some point.
Also once you tell Debian that you really want no swap space, it never complains about it again unless you run out of memory.
Also, you'll find servers with a lot of memory won't always tend to have swap space:
Let me tell you something, dual 10c Xeons, 128GB of memory, and PCI-E Flash SSDs is a great reason to not use a page file. Simply put, you don't need it.
msdn.microsoft.com/en-us/library/windows/desktop/aa366551
I have no idea what are you trying to achieve by disabling one of the principal Windows mechanisms, but I assure you, this will cause quite a few very unpleasant crashes.
All in all, to say that disabling the page file will cause issues is simply not true, at least not in most cases. This was true in XP days but not so much in Vista/7 days. Even on my Phenom II 940 box, I had 8GB of memory and the only time disabling the page file was bad was when I ran out of memory.
If you want to nik pik you will get a error in the error logs when a system is not shutdown right but that's due to crash data not being able to be saved.
Anyways hopefully some more kick ass low profile ram from them again soon.
DDR4 is primarily for server use and of no real advantage over LV DDR3 for desktop/laptop use so there is no point of running out and buying a system that uses DDR4 especially since DDR3 running at 1600 MHz. on a CPU powered desktop is not a system bottleneck as extensive testing with real applications has demonstrated. 2133 MHz. DDR3 for an APU is about the limit for tangible system gains so there is no need to get duped into buying faster more expensive RAM that doesn't deliver any significant system performance gains. Do your homework.
I disagree however with the assessment that DDR4 has no tangible gains. I'm willing to bet you that people with a 5960X who are using all of those cores are grateful for the bandwidth provided over DDR3-1600. However, you're right. 95% of people have no use for strictly DDR4. There is however a use for more cores and the more cores you add to any system the more demanding the physical memory requirements will be.
I suspect slower memory will negatively impact an 8c/16t CPU much more than a 4c/8t CPU but as you eluded to, most people don't need that kind of hardware.
DDR4 spec calls for stacked memory modules which could do wonders for improving how much RAM is on mobile devices where PCB space is at a premium while simultaneously increasing bandwidth. So I don't see any benefit for mainstream machines, I do see a benefit for servers and mobile devices (don't forget lower voltage and less power consumption, also good for both platforms).
Again, can you please tell me what are you trying to achieve by disabling page file? I just don't understand, I really don't. Swapping is a completely transparent mechanism that helps to prevent your RAM from over-bloating by data that your Task Manager thinks won't be needed on a high demand. You allow this thing to eat up as much processor time as it wants, purge your drive's cache, but when it comes to storing 30 MiB of changes history from your Microsoft Word... You just don't let it to! Why?..
Also I have to produce sources but you don't provide any yourself? I would be careful with a double standard like that. Then you're using it wrong because Windows can easily move page files as needed. You're not writing an OS, that's not your job. Also if you need to store data on the disk, you can use a normal file that will be temporary, not the page file. Most budget laptops with 2GB and Windows can let windows swap pages into the page file and doesn't need the application to do it for it. The page file is not the place for transient data or caching. SSDs degrade the more you write to them and RAID-5 calculates parity, something useless (and very slow) for swap space. Also if you have enough memory there is no reason to let the OS swap pages out of memory or to swap parts of the OS out.
I say "modern software" because most software won't directly interact with the swap file and will let Windows handle swapping pages. The way you describe your usage of swap space in your application tells me that you're using it wrong and that a dedicated temp file would be more suitable for your purposes.
I suspect you found a sub-optimal solution to your specific problem, but that doesn't mean it's the only answer or the most correct one.
This discussion needs a little bit of Comp Sci 101 in it.
You want your system to MAXIMIZE the usage of RAM. So if you have 16 gigs of RAM you want your OS to use ALL OF IT! Yes, you read that correct. Using only 15 gigs of the 16 gigs of RAM you purchases is wasting RAM!!! Why? Well the more data you can store in RAM the faster the system response. It is only logical.
So take that mentality and extrapolate it out to the page file/swap space argument.
If you want your OS to MAXIMIZE the amount of RAM you have in your rig then you want to ensure that the OS has all of the necessary tools required to MAXIMIZE said RAM. The page file/swap area is a mechanism to allow your OS to offload clean pages to your HDD which ineffect MAXIMIZES the amount of RAM you have. If you disable the page file or remove swap area then you have removed a mechanicsm that your OS uses to manage memory and it will instead have to start killing processes and preventing new write to memory!
So the page file/swap are argument is actually moot if you do the proper thing in the first place and make sure you utilize all of the RAM that is in your system.
What's faster? Always accessing something from memory or occasionally having to rip something out of the page file on a spinny disk? The page file is for space efficiency when resources are limited and being fought over, not performance efficiency when space isn't an issue.
I don't want to utilize all the memory in my system because that would mean I'm hitting swap which is slow. I would rather have more than I need so Windows can cache more into memory, not take more out of it to put in a page file.
Page swapping and caching are not synonymous and the former is slower than the latter.
Paging makes up for lack of resources, caching makes up for excess resources. If you don't have a shortage of memory resources and never intend to, there is no reason to need a page file. Even more so if performance is important enough where slowdown from using swap would be unacceptable. I would rather my system throw away cache data than have to swap something out before swapping anything back in its place.
Resources are only fought over if none are to be had in the first place in which case it will be slow because you don't have enough main memory for the tasks at hand and for me that's unacceptable.
I'm just saying I purposely made it so I wouldn't need a page file, so why would I need one if everything always fits in memory?
I'm not saying that paging/swap space is useless. I'm just saying it's slow and I purposely went out of my way to ensure it would never be an issue or even used for that matter on my machine. That is all.
Part of the reason I invested in skt2011 3 years ago was for the 8 DIMM slots and DDR3. I don't intend to run out of memory like I have in the past when I had DDR2 and a Phenom II 940 where 8GB was my limit and I would hit it regularly.
This statement is simply not true: It also makes no sense because if there is free memory, it's not competing for anything. Pages are small enough where you won't run into an issue where you have huge swaths of unused memory.
You're confusing memory paging and virtual memory/swap space. Windows always divvies up memory into pages even if the page file is disabled. Having a page file only lets you swap those pages in and out out of memory to and from the disk, that is all. You don't run out of space with free memory available because pages are allocated as applications need memory and the OS uses free space in pages that already exist before allocating a new page in memory and frees them up when they're empty. The NT kernel back in the day was designed to support swapping memory pages, yes. When this was designed, memory was also much more valuable and in shorter supply than it is now so the need for it was greater. Just because something still exists in the NT kernel does not mean that it must be used or that it's as relevant as it used to be in the past. I have a NT 4.11 book that you can go through where there are plenty of things that are barely used anymore but still exist in Windows 7 and the scary thing is that it's true of the majority of the book.
Your making the argument that the swap file is required for a system to run well and efficiently and I'm saying that's only the case when free memory is gone or in short supply.
You're missing my point Rhino. Disabling the page file has the expectation that you will always be able to fit your workload which is everything running, including the OS, into memory. You're not wasting anything if that's the case because without a page file, when you run out, the application closes because it can't continue and you only do this if you know you can do everything in memory. If you don't get close to running out, it's not a waste. I acknoledge that it's keeping the entirety of Windows in memory, I'm saying that's the point.
Let me put it this way:
If the OS + anything you do with your computer at any given time doesn't exceed the amount of physical memory you have, you won't lose performance because you disabled the page file. In fact, the page file lets the machine keep going even when it runs out. So without a page file you let the machine give up and kill the application before performance goes to shit. What you're not understanding is that is by design, I don't want sluggish performance when I run out of memory because I know I won't use all of it, and if I am it's probably a memory leak, but if an application is using up 13GB of memory and I need to swap the OS out, I already know there is a problem and it's not due to lack of a swap file.
You really need to cut it out with the attitude and caps because not only are you clearly getting aggravated, you're letting it show and not in a very mature way. On top of that, you're missing my point.