DDR versus GDDR

[screwballl]

Figured I would start this to keep other threads from derailment...

are there any industry insiders that can let us know in either laymans terms or technical terms the difference between GDDR used on video cards (and the Xbox 360 as system memory as well) and regular DDR that is used as system memory in pretty much all x86 based systems?

I saw some info elsewhere but it is without a source so cannot be fully trusted:

DDR1 and GDDR1 are pretty much the same except that system memory DDR1 has 8bits wide I/O and GDDR1 has 16bit width. Basically GDDR1 just allows slightly higher bandwidth than DDR1.

But, on the other hand:
DDR2 vs. GDDR2
DDR3 vs. GDDR3/GDDR4

The difference?
Basically they all are completely different DRAM architectures. GDDR2 has nothing to do with DDR2. DDR2 is a bit like half-breed of GDDR1 and DDR2, it has features from both. The real GDDR2 hasn't been used on video cards for several years anymore, the "GDDR2" you see these days is merely standard DDR2 we all love. GDDR3 is an sort of an enhanced DDR2 - it certainly isn't anything like DDR3. And GDDR4 is just a tweaked GDDR3.

You ask why don't we see GDDR3 used as system memory since they're clearly superior to DDR2 for what comes to bandwidth? That's because GDDR3 is quite a complex chip with a nasty I/O architecture: GDDR3/4 are 32bits wide and DDR2 is 8bits wide. This is a problem since the amount of signal traces that can sensibly be built into desktop motherboards are very limited - by cost that is. 128bits is complex (expensive) enough for desktop pricerange motherboards. So when you have a motherboard with just 128bit bus you could only access 4 GDDR3/4 chips (128bit/32bit) at a time. This would seriously limit the RAM capacity of the system as GDDR3 densities are what they are - 256MB/chip, tops. Usually it's just 128BM/chip. This could be countered with making boards with a huge array of DIMM slots, but that would be even more expensive.

I did find one older but relevant piece on this:
DDR vs. GDDR Memories | Hardware Secrets

edit: more info on wikipedia.
DDR2 and GDDR2:
The first commercial product to claim using the "DDR2" technology was the NVIDIA GeForce FX 5800 graphics card.. Shortly after, DDR2 system memory made its debut. Some budget video cards after this point did end up using DDR2 chips instead of GDDR2 chips.
GDDR3:

Unlike the DDR2 used on graphics cards, GDDR3 is unrelated to the JEDEC DDR3 specification. This memory uses internal terminators, enabling it to better handle certain graphics demands. To improve bandwidth, GDDR3 memory transfers 4 bits of data per pin in 2 clock cycles. The GDDR3 Interface transfers two 32 bit wide data words per clock cycle from the I/O pins. Corresponding to the 4n-pre fetch a single write or read access consists of a 128 bit wide, one-clock-cycle data transfer at the internal memory core and four corresponding 32 bit wide, one-half-clock-cycle data transfers at the I/O Pins. Single-ended unidirectional Read and Write Data strobes are transmitted simultaneously with Read and Write data respectively in order to capture data properly at the receivers of both the Graphics SDRAM and the controller. Data strobes are organized per byte of the 32 bit wide interface.
Despite being designed by ATI, the first card to use the technology was nVidia's GeForce FX 5700 Ultra, where it replaced the DDR2 modules used up to that time. The next card to use GDDR3 was nVidia's GeForce 6800 Ultra, where it was key in maintaining reasonable power requirements compared to the card's predecessor, the GeForce 5950 Ultra. ATI began using the memory on its Radeon X800 cards. GDDR3 was Sony's choice for the PlayStation 3 gaming console's graphics processor, although the main system memory is comprised of XDR DRAM designed by Rambus Incorporated. Microsoft's Xbox 360 is also shipped with 512 MiB of GDDR3 memory, and is helping to pioneer the use of this memory as standard system memory rather than only video memory.

Advantages of GDDR3 over GDDR2

• GDDR3s strobe signal unlike GDDR2 is Uni Directional & Single-ended (RDQS, WDQS). This means there is a separate read and write data strobe allowing for a quicker read to write ratio than GDDR2.

• GDDR3 has a hardware reset capability allowing it to flush all data from memory and then start again.

• Lower voltage requirements leads to lower power requirements.

• Higher clock frequencies.

then GDDR4:

GDDR4 SDRAM introduced DBI (Data Bus Inversion) and Multi-Preamble to reduce data transmission delay. Prefetch was increased from 4 to 8 bits. The maximum number of memory banks for GDDR4 has been increased to 8. To achieve the same bandwidth as GDDR3 SDRAM, the GDDR4 core runs at half the speed of a GDDR3 core of the same raw bandwidth. Core voltage was decreased to 1.5 V.
On the signaling front, GDDR4 expands the chip I/O buffer to 8 bits per two cycles, allowing for greater sustained bandwidth during burst transmission, but at the expense of significantly increased CAS latency (CL), determined mainly by the double reduced count of the address/command pins and half-clocked DRAM cells, compared to GDDR3. The number of addressing pins was reduced to half that of the GDDR3 core, and were used for power and ground, which also increases latency. Another advantage of GDDR4 is power efficiency: running at 2.4 Gbit/s, it uses 45% less power when compared to GDDR3 chips running at 2.0 Gbit/s.
In Samsung's GDDR4 SDRAM datasheet, it was referred as 'GDDR4 SGRAM', or 'Graphics Double Data Rate version 4 Synchronous Graphics RAM'. However, the essential block write feature is not available, so it is not classified as SGRAM.



The video memory manufacturer Qimonda (formerly Infineon Memory Products division) has stated it will "skip" the development of GDDR4, and move directly to GDDR5

and now GDDR5:

Qimonda, a spin-off of Infineon, has demonstrated and sampled GDDR5, and released a white paper about the technologies behind GDDR5. On May 10, 2008, Qimonda announced volume production of 512 MB GDDR5 modules at 3.6 GHz, 4.0 GHz, and 4.5 GHz clock speeds. Samsung has announced plans for the transition to GDDR5 by the beginning of 2008.
Hynix Semiconductor has introduced the industry's first 1 Gib GDDR5 memory. It supports a bandwidth of 20 GB/s on a 32-bit bus, which enables memory configurations of 1 GiB at 160 GB/s with only 8 circuits on a 256-bit bus.
The cooperation of Qimonda with AMD and ATI has been confirmed, and so are the rumors about ATI Radeon 4-series graphics cards and GDDR-5.

[screwballl]

so based on my original questioning....
GDDR started out pretty much the same as DDR. Once GDDR3 hit the stage, it was created specifically for graphics cards.

Of course then the question remains, if the Xbox 360 can use it for system AND video memory, why don't motherboard makers start including (built in) 1 or 2GB of GDDR memory on the board itself? GDDR3 on up is obviously better at handling memory duties so why not include them on the main boards to be used for onboard video and system memory?

[Boy'nBlack]

Xbox can get away with it as it pretty much has the same narrow demands that a graphics card does. System level uses in the console are basic functions in all reality. Like running a chat session or holding up the media client (live for instance).

And as quoted, we'd have a much more complex system memory dilemma to deal with if used on motherboards, while at the same time be hindered by memory with specifically tailored duties vise vast and variable commands that pass through system memory.