If I have a 64-bit wide memory bus with 100MHz SDRAM I would get:
64bits x 100 = 6400 bits per second / 8 = 800 bytes/s x 1 million (100 MEGAHERTZ, remember?) = 800MB/s = 0.8GB/s
100MHz DDR:
64bits x 100 x 2 = 12800 bits per second / 8 = 1600 bytes/s x 1 million = 1 600MB/s = 1.6GB/s
100MHz DDR2:
64bits x 100 x 4 = 25600 bits per second / 8 = 3200 bytes/s x 1 million = 3 200MB/s = 3.2GB/s
100MHz DDR3:
64bits x 100 x 8 = 51200 bits per second / 8 = 6400 bytes/s x 1 million = 6 400MB/s = 6.4GB/s
100MHz DDR4:
64bits x 100 x 16 = 102400 bits per second / 8 = 12800 bytes/s x 1 million = 12 800MB/s = 12.8GB/s
Now I can't reconcile 12.8GB/s to the 2.13GT/s and 4.26 GT/s (assuming GT/s is giga transfers per second) so I'm not sure if the trend/tradition of doubling the throughput is going to continue or not.
100MHz *G*DDR5:
Based on Hynix's memory bandwidth of "20 GB/s on a 32-bit bus":
64bits x 625 x 8 = 320000 bits per second / 8 = 40000 bytes/s x 1 million = 40 000MB/s = 40.0GB/s
They're running a clock speed of 625MHz. Normalise that to 100MHz and it brings you back to DDR3 (because GDDR5 is based on DDR3 - I didn't know that!):
64bits x 100 x 8 = 51200 bits per second / 8 = 6400 bytes/s x 1 million = 6 400MB/s = 6.4GB/s
So then what differentiates GDDR from DDR? That fact that it's SGRAM: a specialised form of SDRAM.
While checking myself and getting the Wikipedia links I found out that they are using *my* 100MHz example! Maybe I should issue a DCMA takedown notice to them...Wikipedia wrote:Synchronous graphics RAM (SGRAM)
SGRAM is a specialized form of SDRAM for graphics adaptors. It adds functions such as bit masking (writing to a specified bit plane without affecting the others) and block write (filling a block of memory with a single colour). Unlike VRAM and WRAM, SGRAM is single-ported. However, it can open two memory pages at once, which simulates the dual-port nature of other video RAM technologies.
![Smile :-)](./images/smilies/icon_smile.gif)
Regarding RAM speed vs Size:
Unless you have a very, very specific reason, rather go for more RAM than faster RAM. The same goes for filling up all the available memory slots on your motherboard, e.g. if your motherboard has two slots, buying 2x 4GB sticks of RAM filling up both slots. The difference in performance when running RAM in dual channel isn't that great unless one is using an APU.
My recommendation in the above example of 2x 4GB sticks of RAM in both available slots on the motherboard is to rather buy 1x 8GB stick and leave one slot open to help make upgrading cheaper in future as you will only need to buy 1x 8GB stick as you have one slot open instead of throwing out your 2x 4GB sticks and buying 2x 8GB sticks.
PLEASE NOTE/PET PEEVE:
It is *NOT* "duel" channel or "tripple" channel. It is DUAL channel or TRIPLE channel.
//TODO: Latency, HMB and GDDR5X. Full write-ups on dual channel and triple channel RAM.