MetaRAM aims to bump RAM capacity by 4x overnight

We're not sure what whacky voodoo snakeoil these MetaRAM people are peddling, but the company's got some high profile behind it (like Intel, for one), is being led by former AMD CTO Fred Weber, has appears to have some potentially revolutionary RAM quadrupling technology. Claiming to have leapfrogged current RAM technology by 2-4 years, MetaRAM uses a specialized "MetaSDRAM" chipset that effectively bonds and addresses four cheap 1Gb DRAM chips as one, tricking any machine's memory controller into using it as a 4x capacity DIMM. Since a 1Gb chip is apparently far less expensive than a 2Gb chip, MetaRAM devices can multiply capacity at prices far lower than their competition; the company claims it'll be shipping in machines in the first quarter of this year, and Hynix has already announced their own 2-rank 8GB DDR2 RDIMMs for the second half.

[phanbouy @ Feb 25th 2008 12:58AM]

Intel always has the bestest partnerships eva with RAM companies.

[ben @ Feb 25th 2008 1:09AM]

RDRAM was a _total_ success.

[phanbouy @ Feb 25th 2008 1:19AM]

why, yes it was ben. say, anyone wanna buy my Rambus stock?

[ventivent @ Feb 25th 2008 1:07AM]

After reading the headline, I thought this was going to be another fly-by-night gimmick like "SoftRAM."

[phanbouy @ Feb 25th 2008 1:18AM]

but you were convinced otherwise somehow?

[Kazriko @ Feb 25th 2008 2:16AM]

Looking at the article, it's essentially a raid controller for memory chips. I can't see how it's feasible in the long term though, it can really only compete with the very top of the line memory chip because it's using 2.5x as much silicon over what you would just by using the top chip. The price premium on the top of the line chip is the only thing making this technology feasible in the market.

[mushrooshi @ Feb 25th 2008 1:09AM]

I hope its cheap!

[soviet_vexxer @ Feb 25th 2008 1:17AM]

LOL!!!!!

[phanbouy @ Feb 25th 2008 1:20AM]

in Soviet Russia, soviet_vexxer laughs at you

[Das @ Feb 25th 2008 2:01AM]

And they don't call him soviet_vexxer either
they just call him vexxer

[phanbouy @ Feb 25th 2008 4:03AM]

that reminds me of an old SNL news update w/ Norm macdonald:

"Coming in at number 1 this week is Better than Ezra. Coming in at number 2: Ezra".

[Martin @ Feb 25th 2008 1:19AM]

Capacity is great, but what about timings, latency, and ahh, oh yeah, speed... I see this technology seeping it's way into flash disk drives and possibly SSDs.

[Kamokazi @ Feb 25th 2008 1:35AM]

Yeah that'll be the big question I think. The fact that Hynix is planning on producing sticks with it adds a lot of credibility to it, IMHO.

[morcheeba @ Feb 25th 2008 1:22AM]

It looks really simple - it's basically a buffer that allows you to put multiple chips in one slot. The reason you normally can't do this is that the signals on a DIMM are set up as a stripline & must be wired point-to-point (only a connection between 2 things, not 1 controller+4 drams). The penalty looks like extra latency - this is needed to select which DRAM to talk to & rebuffer the signal. Looks sound & obvious.

Another tradeoff is, obviously, the need for extra chips. Chips are cheap, but having lots of pins (like this design requires) isn't. A denser circuit board could also lead to more layers = cost. The total cost is not going to be calculated as simply 4x the cost of the older chip.

I take issue with some of their marketing. They repeatedly claim that processors double in power every 18 months, and that dram capability requires 36 months to double... that I can believe. But, they claim that their technology closes this gap. This 4x improvement would help in 72 months, but doesn't address what happens in the 72 months after that... their controller goes to 16x?! It doesn't seem sustainable.

[SuperQ @ Feb 25th 2008 1:42AM]

Also power consumption.. How much do these extra chips use? Another method for increasing system ram without having electrical issues is FB-DIMMs. But the buffer chips use a lot of power.

[Miikun @ Feb 25th 2008 11:18PM]

I think perhaps the point is to allow you to quadruple the ram capacity without significantly retooling motherboard design, and making ram refresh and other DDR2 protocols function through the mux/demux without a lot of hassle. At least having the chip on the stick keeps the channels short and simple to design compared to quadrupling the capacity on the motherboard and dealing with crosstalk and trace length issues. Sure it adds latency, but then if this can be applied in a daisy chain fashion, it allows designers to balance latency versus huge capacity gains in existing system designs (or you can have some lower latency address ranges for instruction storage versus higher latency address ranges for data storage with its more sequential access patterns), this makes a lot of sense. Seeing that mr. bigshot came from the Opteron side of things, he's probably not thinking of PCs.

[Eufouria @ Feb 25th 2008 1:32AM]

I'll beleive this technology when its firmly secured in my RAM slot.

[hp540 @ Feb 25th 2008 1:36AM]

the secret is to punch a hole in the top right corner of your ram chips. guaranteed to quadruple your rams.

[murray @ Feb 25th 2008 1:54PM]

Woah.. you just shot me back to 1994. I used a drill. I'm guessing a fair chunk of Engadget readers won't get this reference. Thanks for the flashback.

[whimmel @ Feb 25th 2008 7:56PM]

Doesn't that let the magic smoke out?

[noob @ Feb 26th 2008 1:37AM]

Can anyone shed a light on this? I was young enough around '94 that it was my parent's computer, and since our home computer was a CAD workstation, we had a ton of RAM already.

[murray @ Feb 26th 2008 9:18PM]

@noob: Back when there were 720KB (standard density) and 1440KB (high density) 3.5" floppy disks, you could buy the cheaper SD disks, punch a hole in the corner, and the disk drive would see them as HD disks. Then the OS would let you format them to 1440KB. So get HD for the price of SD. I did this a lot with no ill-effects.

[David @ Feb 25th 2008 2:01AM]

I'm not quite sure how this works.. Is it just basically 4x1GB ram sticks all loaded into one circuit board with a controller?

If so I hope it doesn't affect the speed or timing of RAM. I'd take faster ram over more ram.

[Reader @ Feb 25th 2008 2:10AM]

I seriously barely understand what faster RAM does for the computer. Well I understand what it does, but I can't weight its effects like I can other things. I know what it'll be like to upgrade from a 2.2ghz processor to a 3.4 and what not, or a 256mb video card to a 512, but not a 600mhz to 1000mhz 1gb ram stick. Anyone have a good explanation as to how to weigh the difference between one bus speed and another in the perspective of performance?

[David @ Feb 25th 2008 2:16AM]

It's definitely important. Your computer has three main areas to store data, the on-die cache on your cpu, the ram, and the hard drive. The biggest slow down on a computer is the hard drive because it is so slow. The next would be the ram.
Having slow ram just chokes your cpu. If your cpu runs at 3 Ghz and your ram runs at 1Ghz, theoretically your cpu has to wait for every 3rd cycle before the ram responds. This is just an elementary example and certainly not the case with modern cpus because of the said on-die cache which runs at the same speed as the cpu and program optimizations.

[NearNokia @ Feb 25th 2008 9:07AM]

Yes faster is better, but in the end, for people who just use their computer for email and wp etc... Just aren't going see the difference. The extra ram on the other hand they will.

So this new set of memory, will do very nicely for the average user.

Big gaming rigs and 3D stuff etc, no, lots of fast stuff is what you're gonna need to show off to your friends, and shave off that extra 5secs a year!

[mike @ Feb 25th 2008 3:24AM]

@reader
I have the same problem, I have no idea what ram speed does.

@david
thanks that helps a little bit

[Wwhat @ Feb 25th 2008 11:24AM]

Do people that use computers only to do things like read email really need 8GB NearNokia?
Unless you mean people who read email in MS office I guess :]

[maveric101 @ Feb 25th 2008 9:19PM]

sorry David, you're pretty wrong. the actual answer is more complicated. It doesn't have to do with processor speed, it's the Front Side Bus speed. I'll use my current comp as an example (though it's kinda crap). I'm currently running a 3.0 GHz Pentium 4. Now the way you get this number is the bus speed X the core multiplier. In my case, I'm running a 200MHz bus X a 15x core multiplier = 3000MHz = 3.0GHz. Now what does this have to do w/ memory? Essentially, the bus speed is how many times per second your processor accesses RAM (then the multiplier is how many process cycles your processor does in between; because of this, increasing the multiplier on an unlocked processor doesn't help with speed as much as increasing bus speed because it doesn't increase memory access). Now my processor (and i think most Intels) has 4 channels where it can simultaneously transmit data to the North Bridge (which is in between your processor and RAM). So 200MHz bus X 4 channels = 800MHz FSB (which is what the manufacturers list).

still with me? Now I also happen to be running 400MHz DDR2 RAM. How you get this number is just memory bus speed X 2 (because DDR2 RAM can access on the front and back (or up and down) side of each spike of the clock signal). So 400MHz / 2 = a memory bus speed of 200MHz. This means that my processor and RAM are running at the same bus speed, but does not have to be so. Most North Bridges these days can run asynchronous bus speeds, i.e. not 1:1. But keep in mind that a ratio greater than 1:1 (like 1:2, where RAM bus is faster) does not make your comp faster, because your CPU can't access that fast (so there is no need to buy ridiculously fast RAM!). On the other hand, less than 1:1 (like 1.5:1, where CPU bus is faster) makes your computer slower. Last step: the way you get the total 800MHz to match the CPU rated FSB is because most computers have 2 memory banks of two DIMMS (RAM sticks). 400MHz DIMM X 2 banks = 800MHz (kinda, not really; but that's how you computer memory throughput). Also, you usually have 2 slots per channel for a total of four slots. So if you only have 2 sticks of RAM, you want to put them in the A slot of each bank. Adding sticks to the B slots doesn't increase RAM speed, it only increases size (it may actually slow down, I don't know). Basically, what this MetaRAM tech is doing is adding another level to this, splitting each slot into four more, probably slowing it down a little.

whew. http://xkcd.com/386/. LOL!!! I need to get back to my paper...

[David @ Feb 25th 2008 9:22PM]

I only read the last line of your comment. All I have to say is.. You are right my friend.

[maveric101 @ Feb 26th 2008 1:06AM]

sigh. so much work and i can't even get a "highly ranked." (hint hint, people)
lol.

[mike @ Feb 25th 2008 3:26AM]

serouzly? lol this makes me laugh uncontrollably for some reason I don't know why

[mike @ Feb 25th 2008 3:30AM]

WOW man if i had some money right now i would totally buy some put options on the memory companies, and make a shitload of money when their stocks drop to 25%

[mike @ Feb 25th 2008 3:30AM]

do i have to activate every comment in my email

if so fuck this comment system

[Verbatim9 @ Feb 25th 2008 7:53AM]

You don't...you could use your password instead (check one of the confirmation emails you received, it's auto-generated by default).

[kjb434 @ Feb 25th 2008 9:49AM]

you don't have too

[M @ Feb 25th 2008 3:37AM]

technically, they CAN charge lower prices for it, but WILL they? if it's theoretically 4x speed, then by charging 2.5x or 3x price they'd still make a huge profit

[Gilbert Tang @ Feb 25th 2008 4:03AM]

But will it blend?

[maveric101 @ Feb 25th 2008 9:21PM]

dude: take a hint. look at your comment rank.

[Kurian @ Feb 25th 2008 4:26AM]

I thought of this years ago. I'm pretty sure it will destroy performance for more space.

[Free iPod @ Feb 25th 2008 4:49AM]

... when they come up with 2 x 16GB sticks, that'd be awesome cool.

Not to say that 8GB sticks aren't good enough, though.

Schweet.

[Vince @ Feb 25th 2008 5:17AM]

I'm sorry, I really don't mean to be a grammar snob or anything, but your first sentence was really hard to read word-for-word.

[ether @ Feb 25th 2008 7:51AM]

Yeah, that's not only two sentences but they left out some words to boot. Unless it was done purposefully to sound like a senile 80 year old getting ready for a 20 minute rant about about "this here new-fangled technologies",then promptly fell asleep and someone else wrote the next 5 sentences. Come on Ryan, you can tell us if it's a whiskey hang over that made you do it and too much Adult Swim!

Disclaimer: These comments were merely made for comedic value and shouldn't be taken seriously.

[Peter @ Feb 25th 2008 7:23AM]

That's all well and good, but unless you're running a 64-bit OS and apps, going beyond 4 Gigs of RAM is pointless.
Maybe in the future when everything is 64-bit this will make more sense.

[Greg @ Feb 25th 2008 9:21AM]

How about a little grammar check on the old blog postings ... that'd be just great.

[Jeff Jones @ Feb 25th 2008 9:27AM]

"We're not sure what whacky voodoo snakeoil these MetaRam people are peddling, but the company's got some high profile behind it (like Intel, for one), is being led by former AMD CTO Fred Weber, has appears to have some potentially revolutionary RAM quadrupling technology."

This is, grammatically, the worst sentence in the history of the internet :)

[mhaney @ Feb 25th 2008 9:37AM]

They might want to do a trademark check on their logo. Looks like the Mensa logo to me; whether these guys are geniuses or not has yet to be seen.

http://www.mensa.org/

[Wwhat @ Feb 25th 2008 11:27AM]

Yeah I'm glad this makes RAM cheaper, that huge cost of 20 dollar for 2GB is just not doable eh...

[monkfishbandana @ Feb 25th 2008 12:15PM]

Theoretical 128GB RAM in a Mac Pro? Helloooooo Mama!

[Mike @ Feb 25th 2008 2:04PM]

Does this mean I could cram more memory into my MacBook Pro? That would be wicked slick!