Intel to introduce DDR4 with high-end server CPUs in early 2014

[Shawn Knight]

Intel will reportedly introduce DDR4 into the market starting first with high end server CPUs, according to one source as reported by VR Zone. Desktop systems likely won’t see next generation memory until 2015 with new microarchitectures.

The publication reports that mainstream Haswell 4-core systems will remain on DDR3 memory as the technology approaches 3000 speed grade. Haswell-EX 4-socket platforms will introduce DDR4 in 2014 which could very well be needed as each of the four processors on a motherboard could have up to 16 cores. 

DDR4 will bring with it a reduction in operating power, down to 1.2v, as well as improved parity protection and error recovery. These features alone would likely be enough to make some administrators want to upgrade.

Both Haswell 22nm and Broadwell 14nm silicon utilize LGA 1150 sockets and as such, DDR3 will be used across both platforms. DDR4 support will likely arrive for desktop systems in 2015 but as the article points out, it’s not likely to matter much. L4 cache in Haswell and Broadwell processors will reside on a very wide low latency internal MCM bus and could function as directly addressable scratchpad memory for GPU use. The extra bandwidth should be extremely useful for the CPU as well.

All of this, however, is still a ways out as Intel hasn’t even released Ivy Bridge yet (expected later this month). Ivy Bridge is the code name for Intel’s 22nm die shrink of the Sandy Bridge microarchitecture that will be the first to use Intel’s Tri-Gate transistors.

Permalink to story.

https://www.techspot.com/news/48066-intel-to-introduce-ddr4-with-high-end-server-cpus-in-early-2014.html

 

[Guest]

Yummy!!!!!

 

[Guest]

Too bad the current SB/IV platform hardly benefits from additional memory bandwidth. Since Intel isn't requiring DDR4 for mainstream Haswell either, it looks like that architecture is once again function at 99% speed with just basic DDR3-1600.

As always, $ is better spent on an SSD, faster GPU, better aftermarket cooler to get those overclocks higher, etc.

 

[Guest]

why not a tehnology with no increased latency?
and when ddr5?

 

[sapo joe]

My system is equipped with 8gb of DDR2, and I really think this programmed obsolescence of RAM and computer peripherals is a great sh*t. Why change agp to PCI express, why change CPU sockets? TO SELL MORE MOTHERBOARDS AND MAKE YOU BUY AN ENTIRE PC. The performance gains are small, and really, they shouldn't have to change the connections to use newer technology.

Not changing memory here until I think I must.

 

[Cycloid Torus]

I'm running 16 gb of DDR2 and it works just fine. Don't need DDR3 yet - so I yawned at DDR4. Latest crop of good games (like Skyrim) might persuade me to upgrade, but I'm still playing golden GOTYs from a year or two ago.

 

[Sunny87]

sapo joe said:
My system is equipped with 8gb of DDR2, and I really think this programmed obsolescence of RAM and computer peripherals is a great sh*t. Why change agp to PCI express, why change CPU sockets? TO SELL MORE MOTHERBOARDS AND MAKE YOU BUY AN ENTIRE PC. The performance gains are small, and really, they shouldn't have to change the connections to use newer technology.

Not changing memory here until I think I must.

AGP needed to die just saying 4x 8x speed compared to PCI-E x16 is quite a leap and with out the death of AGP we would not be where we are today.

Your argument is like saying "why did we invent the car when horse and cart where perfectly functional and cheap to run?"

Die sizes are altered on CPU's (32nm) and changing sockets is also another entirely necessary step towards better things like CPU's that are cheap to run but have high calculation speed and cache RAM act.

 

[Guest]

i still have PCI-E 8x and i play games.

 

[SNGX1275]

AGP, when it was killed off, still had sufficient bandwidth to take games developed a few years after PCIe came out. But it isn't good enough now, so yes, AGP had to die. I actually bought a board (ASRock Dual-VSTA) that accepted a C2D and an AGP card (and DDR RAM too) because I wanted a new processor but wasn't ready to kill off my $215 7800GS AGP card. So there are some niche solutions for those stuck with expensive but old technology.

On a much less technical note. I kind of wish they'd change the physical length of the RAM. Every so often you hear of someone forcing RAM of the wrong type on a motherboard and causing damage. I don't understand the necessity for keeping the length of desktop RAM the same since at least the mid 90s (I don't have much expierence with RAM in machines older than about 93 or 94).

 

[Sunny87]

sngx1275 said:
AGP, when it was killed off, still had sufficient bandwidth to take games developed a few years after PCIe came out. But it isn't good enough now, so yes, AGP had to die. I actually bought a board (ASRock Dual-VSTA) that accepted a C2D and an AGP card (and DDR RAM too) because I wanted a new processor but wasn't ready to kill off my $215 7800GS AGP card. So there are some niche solutions for those stuck with expensive but old technology.

On a much less technical note. I kind of wish they'd change the physical length of the RAM. Every so often you hear of someone forcing RAM of the wrong type on a motherboard and causing damage. I don't understand the necessity for keeping the length of desktop RAM the same since at least the mid 90s (I don't have much expierence with RAM in machines older than about 93 or 94).

Actually I came across my first RAM physical size change last year in a school where I was told they have DDR3 RAM, turns out the motherboards where proprietary and featured laptop sized DDR3 RAM, it actually made sense to me to do that because I always run into issues with heatsinks bumping into the RAM sticks.

 

[Guest]

Let's start with the change from AGP to PCIe. Perhaps this move doesn't affect you, but it does affect a significant portion of the PC market. Which would you rather have, a board with dedicated sockets for each peripheral device, or one with an ubiquitous interface that can be used by any add-on card? What if you wanted to have dual graphics cards before PCIe? Would you expect the motherboard manufacturers to offer 4 different motherboards, with 1-4 AGP slots? What if you wanted to be able to do quad-SLI in the future, so you bought a board with 4 AGP ports, but then decided you wanted tri-SLI with a SATA RAID add-in card? What if you wanted to run data parallel calculations on your graphics card with AGP (where the uplink speed is significantly less than the downlink speed)? Perhaps for you, AGP was good enough, and I support you continuing to run Windows XP with outdated hardware to save a few bucks. But don't hold the other 7 billion people on this planet hostage to crappy hardware just because it is good enough for you. I choose the one expansion slot that can handle all add-in cards, which gives me more freedom in expanding my system WITHOUT purchasing an extraneous, new motherboard.

As for the socket changes, I see both sides of the argument. For some things, such as an integrated graphics processor, it seems like a socket change is required. But, there are some changes that I feel Intel & AMD can predict. Why not add M extra, currently unused, memory channel pins, and the same for PCIe lanes, and QPI (Intel)/HT (AMD)? I can see why these companies (especially Intel) don't want to hamstring their next processors by promising the same socket support, but planning ahead by adding pins that you know will be useful in the next gen seems like it would be good for the do-it-yourself crew. The problem, of course, is that the do-it-yourself crew is a vast minority, and that extra cent per socket adds up for the masses that recycle an entire computer and buy everything new every 5 years.

 

[Guest]

Theres no mention anywhere about Mobile platforms, but thats exactly what we need DDR4 for because of the power savings. Higher clocked DDR3 is ineffecient for laptops. Its absurd that there's no mention of this in the article. Where's the initiative amongst journalists these days? Just another copy-pasted press release under guise of "Article" moniker.