PDA

View Full Version : ddr4 delayed until 2014


wheeljack12
06-27-11, 06:15 PM
http://www.xbitlabs.com/news/memory/display/20110627143149_DDR4_Memory_Now_Projected_to_Debut_ in_2014_But_Ramp_Up_Rapidly.html

Runningman
06-27-11, 06:34 PM
Well, the question is begged to be asked, is memory the bottleneck in a system and will be benefit from faster memory?

Dazz
06-27-11, 06:57 PM
Well i feel better about buying DDR3 now but lets face it DDR4 when it comes out will be massivly over priced anyway. The advantages of DDR4 will be higher memory density, 8 and 16GB modules at decent clock speeds.

AngelGraves13
06-28-11, 02:16 PM
DDR4 will be affordable in 2016 then at the earliest. Oh well, so much for looking forward to 2400MHz DDR4 in 2013. By the time they hit that speed, it will be 2016 and DDR5 will be in the works, unless something different comes along or Rambus makes a return.

ViN86
06-28-11, 02:43 PM
Fine by me. This way Intel will keep using DDR3 (since they always adopt the latest DDR right off the bat). With their quad-channel Sandy Bridge E coming out, 4x2GB setups will be very affordable. Hell, even 4x4GB setups will be pretty affordable. :)

nekrosoft13
06-28-11, 04:42 PM
Well, the question is begged to be asked, is memory the bottleneck in a system and will be benefit from faster memory?

biggest bottleneck still is the storage (hdd/ssd), even SSD are to slow.

Viral
06-28-11, 06:58 PM
Memory is a bottleneck for APU's, the more bandwidth that can be supplied the better GPU they can put on there.

Bman212121
06-28-11, 08:51 PM
biggest bottleneck still is the storage (hdd/ssd), even SSD are to slow.

Yes and no.... I think the biggest bottleneck now is single threaded workloads. The CPU is still slowing down a lot of tasks simply because they don't utilize the processor efficiently. There are so many caches between an SSD and the processor that most of the time your storage isn't what the pc is waiting on. If you check out Anand's latest review on the Revo Drive x2 you'll see that a drive that's effectively 4 vertex 3's in RAID 0 offers very little performance gain over a single vertex 3 simply because most programs don't queue multiple reads. Obviously for a HDD a single thread would deliver the best sequential performance, but on an SSD you need to have a lot of I/O to actually get the full benefit from it.

If programs could actually utilize 8 threads and fully load down the CPU, I think that memory would have a bigger impact than it does right now. A good example would be to look at the relationship of the GPU and it's memory subsystem. There is a lot more bandwidth on any high end GPU and you can still be bandwidth constrained. The main difference is that the GPU has always been massively parallel so it will generate the workloads CPUs cannot.

hirantha
06-29-11, 11:19 AM
Fine by me. This way Intel will keep using DDR3 (since they always adopt the latest DDR right off the bat). With their quad-channel Sandy Bridge E coming out, 4x2GB setups will be very affordable. Hell, even 4x4GB setups will be pretty affordable. :)

This is true. DDR3 prices are already low and with SB E its a nice time to upgrade to a whole lota RAM :D