Rambus Production Capacity Switched to Make SDRAM

Jon Rabone writes " NEC, Japan's largest chipmaker is halting Rambus production to make SDRAM. Both NEC and Samsung are to switch production over to SDRAM - sounds to me like RAMBUS could be in danger of dying the death, after Intel's latest problems with the Camino chipset. At least we might see SDRAM prices fall again. "

Comments

Law of Supply and Demand

[Jack William Bell]

Need I say more?

Jack

RAM prices fluctuate like this every year.

[Wakko Warner]

The best time of the year to buy, I've found, has always been in the middle of the year. Prices, for the past few years, seem to bottom out in June, and then slowly rise to peak in October/November. They should start dropping like a brick soon and, by the middle of next year, you'll wonder why you ever spent $80 for 128 megs of SDRAM when it's $40 now.

- A.P.
--

"One World, one Web, one Program" - Microsoft promotional ad

The reason? Volume.

[Wakko Warner]

CompUSA probably has a warehouse full of RAM they bought back when prices were low. They can afford to sell it at the prices they're selling it at now and still make a huge profit because they bought it so cheaply back then. Little RAM dealers (like the kind of people you see at computer shows and on Pricewatch) buy tiny fractions of what CompUSA buys, and they tend not to have any overstock sitting around, so they *must* charge more, because they're paying today's prices, not the price RAM was in June.

As always, it's best to look at all the options before you buy stuff -- don't always just head on over to pricewatch and think you've gotten a great deal. :)

- A.P. (speaking of great deals, checked the prices for 18 gig U2W scsi drives lately?)
--

"One World, one Web, one Program" - Microsoft promotional ad

Mandated technology

[overshoot]

• Not really surprising. Engineering by fiat fails if trumped by a Higher Authority (e.g., physics)
  
• The problem isn't the Camino chip, it's the physics. Turns out that Rambus has a major signal-integrity failure mode that sort of got swept under the rug until the systems houses got bit by it.
  
• The DRAM companies never liked Rambus, but had their arms twisted by Intel. Now they have a chance to bail and are taking it.
  
• The comment about DDR (double-data-rate SDRAM) having no standard will come as quite a surprise to the people at the memory companies and in particular JEDEC's JC-42 memory committee, which thinks that they have issued one, and AMII, which is sponsored by the memory industry (including NEC and Samsung) to promote its use.

Re:Mandated technology

[Anonymous Coward]

Not really surprising. Engineering by fiat fails if trumped by a Higher Authority (e.g., physics)

If only that was always true. Intel has been successful pretty often with inferior solutions. It's just this time they're trying to push around a few large companies rather than a larger collection of smaller companies. There are a bunch of companies that make (should I use the word assemble now?) motherboards that Intel pushed around with their (at the time) inferior chipsets, but very few companies can afford the investment to start making RAM.

fiat: This reminds me of a Pontiac commercial where they say "Wider is better." Now, Intel is promoting "Slower is better." Guess what? Wider is better, but slower isn't. Therefore, RAMBUS is out. Intel can argue that significantly slowing down RAM in order to decrease the pin count is a good thing, but most consumers don't understand that so they'll go with the faster and cheap DRAM.

Samsung already switched

[Jacco de Leeuw]

A while back, The Register reported that "Samsung shut down Rambus production and shifted to SDRAM production until Intel can come up with a new Camino launch date".

RAM prices are already falling again

[Tet]

I bought my 128MB DIMM for 54 pounds sterling about 2 months ago. A week later, they'd dropped to 49, and then prices went through the roof. Last week they were up to 198 pounds (about US$317), but have now starting falling back down again, and are at 167 today. That said, I was quoted up to 245 pounds wandering round the shops in Tottenham Court Road on Saturday. All prices exclude VAT at 17.5%.

From looking at other posts, I guess memory in the US is still dirt cheap (at least compared to our prices in the UK)

Good and bad news

[R. Anthony]

Hopefully SDRAM prices will drop as a result of this, as a lot of companies stopped fabricating DRAM back in June as a result of the overabundance of cheap Taiwanese DRAM selling at $4. I'd personally like to see SDRAM fall back to somewhat more reasonable level as $360 is not a fair price IMHO for a 128 MB DIMM when the production cost is marginally less (300%).

However. SDRAM won't suffice forever as it can't (disclaimer: as far as I know) be overclocked much higher than it already is (140 MHz is the highest I've read about before stability issues arise). RDRAM on the other hand can run up to 800 MHz. The heat sync looks rather cute, but the memory is tragically flawed by it's miniscule 16 bit bus (as opposed to the 64 bit SDRAM bus to the front side system bus).

Rambus really needs to go back to the drawing board on this before they bring it back to market, if it is ever given another chance. Intel in partnering with Rambus was seeing Large dollar signs in an unending stream of royalty payments on every future RDRAM RIMM sold for many years to come. Chalk it up to another case of greed overriding sensibility.

Re:Good and bad news

[tak amalak]

What you said about SDRAM is correct. But there are companies that will be producing DDR-SDRAM (Double Data Rate SDRAM) that can send signals up and down a cycle, effectively doubling the clock rate to 200-266MHz (depending on if they are using PC100 or PC133 chips).
Benefits:
* 64bit memory throughput to the system
* Much lower latency than RAMBUS
* Uses tested SDRAM technology
* Can be easily implemented in cheaper systems
* Has 2.1GBps max transfer rate compared to 1.6GBps on 800MHz RAMBUS
* Still has headroom to grow
--

Re:Good and bad news

[hazydave]

>RDRAM on the other hand can run up to 800 MHz. The heat sync looks rather cute, but the memory >is tragically flawed by it's miniscule 16 bit bus (as opposed to the 64 bit SDRAM bus to the front side system bus). If you take a look, you'll find that the 64-bit FSB on today's P6 machines is simply a reaction to this architectural feature of the P6 bus. The SDRAM chips themselves are usually 8 or 16-bit wide devices, ganged to deliver a 64-bit bus. You can't do this same ganging on Rambus as easily, but you can certainly support several Rambus channels for fewer controller chip pins than you need for a 64-bit SDRAM bus. Today, it's not necessary -- a 100MHz P6 bus is only going to use about 1/2 the bandwidth of one Rambus channel. So this is not the doom you think. What Rambus buys for Intel, more than anything, is an easy way to move memory off the P6 bus. This will allow more I/O efficient system chips, as PCI and AGP now can have a direct channel to memory that doesn't get directly in the way of P6 activity. So even if the PIII can only use 800-900MB/s on the 1600MB/s Rambus channel, with proper buffering in the North Bridge chip, AGP and PCI will get a piece of this without conflicting as much with the CPU. This is also what you get with the EV6 bus, though in this case, by definition. And of course, Rambus isn't the only way. You can get the same bandwith with one Direct Rambus channel, a 64-bit DDR-SDRAM bus, or a 128-bit SDRAM bus (a staple on Alpha machines for years). The problem is doubling this. Two Direct Rambus channels are practically a no-brainer. 128-bit bus SDRAM-DDR might be possible in modern BGA packaging, but it's cutting it close. A 256-bit plain old SDRAM bus is not going to happen in commodity PCs. Now double it again. This is why Intel's been interested in Rambus.

Many reasons for increase

[JediLuke]

man, there were so many factors that jacked it up to over $2 a meg from $.50 a meg

1. Micron lost a months batch
   
2. Micron's lawsuit
   
3. Intel's frickin RAMBUS delays
   

we were doing just fine before that...we should trade memory on the commodities market
JediLuke

Re:Blame it on the quake?

[R. Anthony]

The reason the prices got so high this year was because of the quake in Taiwan.

Not so. I was following the market closely before during and after the quake. I used corsair PC 133 128 MB as a marker, as it uses Micron DRAM which is not produced in Taiwan, and in no way would be effected by the quake (it also happend to be the brand I choose and now own, due to it's superior quality). Here is a timeline of the price spikes:

I chose direct.multiwave.com (wholesaler) as my test bed.

Monday (prior to quake): $297

Sept 22nd (day of the quake, wednesday): $297

Sept 24th $297

Sept 29th: $358

Oct 7th: $372

So you see, the prices were already at $300 before the quake. The subsequent rises could be attributed to Micron raising the price of DRAM to over $16 after the 22th, the day of the quake.

Re:Death of Rambus?

[Anonymous Coward]

This ought to be about more than Intel being humiliated. Intel should be stopped from forcing a more expensive, ill-planned and "sole-proprietor" technology on the PC buyers of the world; and if the revolt of .tw suppliers does not suffice DOJ should see what they can find in an investigation. (I think they should start on it now. Yes, the motherboardf makers are in revolt, but they're not in a very good position to bargain at the moment due to quake. Plus DOJ should uphold principle here and not wait until all the damage is done an irreversible)

The premise of Rambus is much stronger for ultra high end server machines than for general use PC's. But Intel has been shoving Rambus down everyone's throats to finance the costs of their push into the Enterprise. If this isn't an example of leveraging monopoly power then I don't know anything that could be fairly called a monopoly: the increased cost (damage to consumers) is inarguably real and dollar-figure obvious. As is Intel's control over the "standard". They're so obsessed with this goal that little things that affect "the little people" like the design flaw of the i820 slip by.

Re:Rambus will not die

[overshoot]

Hate to break the news, but the problem isn't really with the Camino north bridge chip. It's a signal integrity problem with the Rambus system architecture. Nasty combination of crosstalk and a resonant mode in the data lines that takes received data out of signaling spec (the line fails to cross above the logic threshold when the RAM is sending a HIGH.)
Intel is taking the fall on this one as though it's a silicon problem because it doesn't really matter where the problem is, the 820-based product isn't going to ship. Inside of Intel there's a big bloodletting going on between the engineers and the suits, because the suits are having a hard time dealing with the concept that there are some things that can't be changed by management fiat and the engineers aren't real amused by egos under the delusion that they can order back the tide.
There's been some good discussion on this over on SI-LIST

Re:AMD

[overshoot]

• Rambus isn't an Intel technology, although they are pushing it.
  
• AMD is _not_ using Rambus for the K7. All of the announced K7 north-bridge chips (most notably Via's) are strictly DDR SDRAM designs.
  
• for fun reading on the subject of Rambus performance compared to old, slow, PC100 SDRAM check out the Dell benchmarks.
  

Don't read too much into this.

[hazydave]

Actually, AMD recently took a Rambus licence. Which only made sense once they had a system chip architecture (licensed from VIA, possibly a competitor now that VIA has bough the high-end Cyrix stuff). Clearly, the Rambus retreat is based on Intel's inability to get their new i820 chipset working properly. They were counting on this to make a big splash for Christmastime, and so anyone making Direct Rambus likely has stockpiles already. Knowing that's not going to happen, it makes perfect sense they're running SDRAM now, especially with the rise in SDRAM pricing. After all, the main allure of Direct Rambus for the memory vendors was higher margins. You can get that now, at least for a little while, with SDRAM, and there's no sense in making more Rambus now. Rambus itself has been proven in the consumer maket -- the Nintendo 64 uses the original 8-bit Rambus design, at 500MHz (technically, 250MHz DDR). The interesting thing is that Intel can't use Rambus anyway, now; they're priming the pumps for some future migration. One 16-bit Direct Rambus at 800MHz (technically, 400MHz DDR) peaks out at 1600MB/s, but a 100MHz P6 bus peaks out at 800MB/s. Sure, they'll run it up to 133MHz in the i820 (VIA has a 133MHz P6 bus chipset already), but it's not something you need Rambus for. AMD, on the other hand, has this hot EV6 bus. They're running it at 200MHz, but it's already going 400MHz in Alpha machines. It's kind of silly to run this on a PC100 bus. So AMD can actually take full advantage of one Direct Rambus channel. In fact, if they want to boost I/O performance, two channels wouldn't be out of line. Of course, AMD could get to a real 200MHz data rate using 100MHz DDR SDRAM (supposedly ALi is doing an EV6 chipset to support this). So they don't really need Direct Rambus, either. Unless they're planning a 400MHz FSB migration, too...

Re:Blame it on the quake?

[Sun Tzu]

Imagine for a moment that a catastrophe destroyed *all* the memory production capacity except that from a single company. By your logic, that one company's prices shouldn't be affected by the disaster. Of course, that cannot be.

If the quake affected Micron's competition it affected Micron's market. If it affected Micron's market, it stands to affect Micron's prices.

The reality of markets is, of course, far more complicated than your argument (and mine) would have it be. Even a credible prediction of a quake in Taiwan would affect prices in the markets in which Taiwan is a major player.

Re:so, this mean there will be more R&D?

[overshoot]

An AC asked,
What is a bond-out option?

Integrated circuits are little rectangles of silicon and have to be connected to the rest of the world. On the chip itself there are exposed metal spots ("bond pads") connected to the device's I/O circuitry. Various means are used to connect the bond pads to the package, but the most common are veeeery thin gold wires attached to the chip at one end and a package conductor at the other.

Since it's inconvenient and expensive to maintain separate IC designs, it's often easier to have one design which can be used in multiple ways depending on external signals, jumpers, or whatever. Rather than bring these configuration lines out of the package, though, they can be connected internally at the time the bond wires are attached. Likewise, signals that aren't needed in a particular configuration can be left out of the package I/O set.

HTH.