DDR2: Faster memory, at a Slower Market Pace

Memory by KeithSuppe @ 2003-07-17

DDR2: Could it be used to expedite it's own Implementation?

JEDEC (Joint Electron Device Engineering Council) the major semiconductor engineering standardization body. Has approved final "testing" for the release of DDR2 SDRAM as a standard for manufacturers.

  • Go to mainpage


JEDEC (Joint Electron Device Engineering Council) the major semiconductor engineering standardization body. Has approved final "testing" for the release of DDR2 SDRAM as a standard for manufacturers. JEDEC has approved a "Clock reference board, a register validation board, and the PLL and register plug in cards." (source)

Mass production for PC's should be in full swing by 2005, in the interim DDR1-400 will fill the gap. This will not, or does not, prevent manufacturers from producing DDR2 SDRAM only that it's not "officially" supported. This of course was the case with DDR400 from its original sales to end-users as far back as March 2002. Of course most of the modules then were merely overclocked DDR333, which hadn't been "officially" (JEDEC) approved either. With DDR2, however; the scenario takes on a different twist, as DDR2 cannot simply be "overclocked" from another standard. Ironically Samsung, who's IC's which were found under the heat spreaders of the earliest versions of PC3200, will begin mass production of DDR2 SDRAM in 2004. One may expect history to repeat itself in this respect, as it will most likely be Samsung IC's found under the heat spreaders of initial of DDR2-500 modules when they reach resellers. Regardless DDR2 is now being manufactured for graphic card memory, using a “point to point” interface, which makes this early use possible. And it's in the graphics-card industry we'll first see its benefits (or idiosyncrasies) there.

Already we know the memory runs quite hot, meaning there's still work to be done at the transistor level. In so far as the SDRAM for PC standard, most likely OCZ, Twinmoss, Corsair and a few others, will be some of the first to release DDR2-500 modules for sale. This should occur long before JEDEC officially approves the standard. It's amazing to me that DDR1-400 is only just being fully integrated into the market, when there are actually few enthusiasts who would "prefer" PC3200 in their Canterwood/Springdale, or nForce2/KT400/A chipsets. In fact it's PC3500 which has occupied the DIMM's of the majority of enthusiasts PC's for the last 10 months. Most of the "high-speed" memory makers above are just a nanosecond away from releasing DDR500 (PC4000). Of course JEDEC's schedule and the schedule of the overclocker rarely coinside. We should be thankful for companies such as Abit, Asus, Epox, Soltec, and others whom write BIOS versions to accommodate the SPD of these faster memories. And I'd be remiss to exclude IC makers such as Samsung, Micron, Hynix, Infineon and Winbond whom supply the IC's to these makers of "High-Speed" memory. As the "Server" or business market's (always the last to integrate the standards pending their proven stability, usually some time after JEDEC approval) begin to fully incorporate the DDR400 standard, prices will drop even further.

The market integration schedule usually follows a specific priority. First Enthusiast/Overclocker sector provide a virtual "test" platform or market for manufacturers to introduce their "cutting edge" products. The next phase usually finds OEM vendors such as Alien Ware, Velocity Micro, ABS, Falcon Northwest, Vicious, and even Dell (dude) building for Gamers, and more performance oriented end-users. They may even see systems having a smidgen of "cutting edge" hardware (PC3500) under the hood, if their lucky. There's definitely a pecking order. And the final phase usually finds OEM mass sales, and Businesses filling the largest majority of sales. This has nothing to do with Businesses NOT preferring cutting edge technology, in fact the opposite is usually true, as it their resources in the end which fund cutting edge R&D. Businesses, and Server oriented systems must maintain a faultless standard of reliability in their hardware. And of course this usually requires hardware which surpasses manufacturer’s standards for testing, and Q&C criteria, but must also have passed the test of time in the market. Only in this respect can they be sure the hardware they place in their multi-million dollar servers will meet such high standards. You can't have memory SPD incompatibility at that late stage in the game. Not with billions of dollars in revenue depending on a system's stability.

There are extremes in the PC world. There's the person running a Prometia equipped with a P4 2.4c operating at 4.1GHz running at 12x344FSB, 1.575V, on a volt-modded Abit IC7 with custom cooled NB, And then there’s Big-Business who needs to ensure their 3000 employee's FICA are correct. I'd surmise the stress in both scenarios to be equal, except the Prometeia owner (with all due respect being one myself) is just a hobbyist. If we "blue-smoke" an Intel P4 2.4C, people don't loose jobs, and markets aren't affected. If I was under any illusions pumping 3.2VDIMM through Winbond BH-5ns IC's or pushing a i875 chipset until it screams wouldn't involve some risk, then it's clearly time to find a new interest.

So what does this have to do with JEDEC and DDR2? Well I find it ironic overclockers become so frustrated with the snail's pace of JEDEC's approval process. Especially when you consider the 300 or so companies who comprise JEDEC are "businesses" in every sense of the term. Remember this, we as overclockers, are very fortunate. Although it does cost a premium to have the fastest hardware, and cutting edge technologies, we're fortunate there are so many companies, who cater just to us. I think it's a fascinating time to be a "geek."

comments & questions can be asked here in this thread @ the forums (no registration)
  • Go to mainpage