Where is all this power people are talking about?
Hey, it's a technical topic. I can't resist!
:-)
It may very well be that for this iteration of technology, DDR SDRAM will be the most cost effective and fastest solution for consumer graphics boards for 2 reasons:
1. cost
2. minimized stub reflection problems on surface mounted chips make it easier to run DDR SDRAM fast. It's much harder to mount them onto DIMMs and get them to run that fast.
But DDR technology does not have legs. It is very hard to scale to higher speeds. DDR is just increasing speed of the data group signals, not the address, control, clock signals. Why not? Because it's too hard to keep all the lines in sync at high speeds! Rambus takes a completely different approach of packetizing the transfers into a protocol and sending them around in a loop (kind of like token ring) between the controller and the memory.
Rambus can go up to 800 mhz with a 16 bit wide interface: 1.6 GB peak bandwidth
DDR SDRAM (64 bit interface, 2 transfers per cycle) running at only 100 mhz: 1.6 GB peak bandwidth
But it takes 8 chips to do the DDR RAM; you get 1/8 of your bandwidth from each chip. With rambus you get full bandwidth on each chip. So if you want an 8 chip solution with rambus, you could have 1.6 GB/sec * 8 = 12.8 GB/sec, if you wanted to pay for that many RIMMS (probably not on a consumer graphics card). The point is that it's hard to compare apples to oranges. BTW, this is why N64 and PS2 love rambus; they only need small amounts of memory (32MB max I think), so they can get it all on less chips. Saves them money, and great performance too...
The basic advantages of rambus are:
1. Easier to clock to higher speeds because there's only 4 lines.
2. It's built on top of SDRAM. Rambus chips have SDRAMish logic embedded in them. So as SDRAM technology gets faster, so does rambus. If rambus used PC133 type embedded SDRAM instead of 100MHz then the 800MHz would run at 1.6 GB/sec * 1.33 = 2.1 GB/sec.
3. If you're using a DIMM type package, DDR RAM has reflection problems because of the stub length off the bus for each DIMM. Rambus minimizes these with almost no stub length and a continuous bus from DIMM to DIMM, making for less signal degridation and higher clock speeds.
4. Rambus has more banks. SDRAM has 4 banks (if you have 2 DIMMS). Every rambus module has 8 devices and each device has 16 banks. So there's 128 banks/module (but you can only have 64 banks open at a time on the whole RIMM). If a bank is already open and the row you need already precharged then you can burst data from the row without another row packet and precharge. You can just use fast column packets. So the more banks the better.
Rambus is very different animal than traditional SDRAM and is hard to compare directly with it. Also (*Disclaimer*), I'm far from an expert on rambus; I'm really a software guy who plays with hardware. I just learned the above rambus info in doing research on it before I invested in it.
Plaz |
