Rambus introduces 4.3Gbps mobile memory
By Nilay Patel 
posted Feb 2nd 2009 1:32PM
posted Feb 2nd 2009 1:32PM
It's not all just lawsuits and overinflated posturing for Rambus -- the prickly memory firm will later today demo what its calling the Mobile Memory Initiative, a 4.3Gbps phone-oriented variant of the XDR RAM used in the PS3 that requires just 100mV of power. That means companies can either build super-powerful phones that take advantage of a complete MMI system's estimated 17GBps throughput, or build cheaper devices that offer performance equivalent to today's memory with fewer MMI parts. Rambus won't say who it's working with to bring out MMI, but we're going to have to wait a bit before we all score cheap HD videophones -- the first products aren't scheduled to hit until early 2010.[Via Electronista]
"100mV of power"
V = IR (volts)
P = VI = V^2/R = etc etc (watts)
So how many amps does this stuff draw?
test
Currents are small. So less than 100mW.
Nice.
It draws 100mV of power? Nice one, Engadget... Back to Electricity 101 for you guys ;-)
This probably won't be cheap but oh well.
"a 4.3Gbps phone-oriented variant of the XDR RAM used in the PS3 that draws just 100mV of power."
Thats what she said!
Nilay isn't a girl silly!
niiice
to hell with Rambus, I hated buying RDRAM paired modules for 5x the price of everything else.
I completely agree, the fiasco that was RDRAM ruined upgrading making it more expensive than buying a new computer.
Aye, if ever given the choice I will avoid Rambus. That exclusive they had on Intel chipsets way back burned me good.
I agree that Rambus sucks. They are patent trolls that tried to screw the consumer over with their trash. These guys do not deserve any support.
RAMBUS makes a come back...
So they develop a new ram tech thats probably going to be much more expensive and way overkill for a phone. Sounds like a brilliant plan. Then they will try to sue everybody that puts RAM in a phone.
their business model works well in the realm of devices because consumers don't pay for upgrades on the "open" market like with PC RAM. Where Rambus failed was in way overestimating their value to the process. Could you imagine paying $1 per chip royalty on DDR2 modules? (that's $16 off the top... more take than the retail store, or the OEM that actually MADE the chip gets after materials and JDEC didn't bite) But in mobile devices where chips are sold directly to OEMS and use more frugally, they clean up nicely and do some neat things. In PCs they were a bust because they tried to corner the market, and they made complex technical choices for "user upgraded" memory.
Rambus was more expensive because they tried to sue OEMS into making their product... so OEMs stopped making their product. (imagine that!) Also squarely blame Intel that basically told the Ram OEMs to talk to Rambus because they had a financial stake in their business. basically trying to leverage their Pentium near-monopoly (remember P3 820 chipsets) into getting monopoly licenses on RAM too.
There are programs that would make use of that speed, ones that before now wouldn't be on a phone. Why so ready to be negative?
Oh Rambus, we are not falling for you again.
Go away patent trolls.
a method to force patent trolls to leave, hmm...
OFF TO THE PATENT OFFICE!
Guess they're putting all that money grabbed from law suits to some use.
Oh really?
How would i utilize that sort of memory on a phone, my phone has twice the ram of my first computer and im happy with that.
Well, 4.3 Gbps per lane... likely talking four lanes (one per bank)... so for example USB3.0 SuperSpeed gives 4.6 Gbps, you you could use this to basically dump your payload directly into RAM at speed, which would be nice. Look at the bandwidth requirements for high performance 3D rendering, and you can see why this might be attractive to GPU or AP manufacturers.
Then you have to look at other factors, like board routing and package size. LPDDR2 at 400MHz will give you 3.2GBps of throughput, but you need 32b of data, some 24 odd b of c/a plus eight control lines and clocks - some 64 odd bits total for the interface, matched lengths, on-chip termination for each, etc etc... massive package at a minimum (just look at some of the JEDEC footprints). For this, you're talking maybe 8-9 pins per bank, total of 32-36, 40 tops for 17GBps... consider this is likely going package-on-package, and you can see why this would be useful.
Or you can't because you're not in the industry and have no clue what I just wrote.
It's too bad RAMBUS has such D-bags working for them because it appears they have some excellent IP and engineering staff that always seem to make high performance components but always have screwy business practices..
Let me get this straight: Rambus technology will enable better performance in a smaller package with longer battery life and they have the temerity to expect to get paid for this? I thought the best things in life were free?