Finally! 4GB of True 1.8V DDR2-800 Memory!

One of the things that continually burns me up on PC Memory is the over-marketing of the synchronous clock of today's Double Data Rate II (DDR2). It was bad enough when latency was jacked up high (worse) for higher synchronous clock rates and performance was no different. I touched on how to calculate real-world "latency" on DRAM modules near the end of my Spring 2007 PC Assembler's Quick Guide.

But now the majority of vendors are claiming DDR2-800 (JEDEC PC2-6400) and higher synchronous clocks that utterly violate the JEDEC spec of 1.8V that require you to run them at 2.0-2.1V typically, 1.9V if you're lucky, and 2.2V+ are not uncommon either. It was less common at DDR2-667 (JEDEC PC2-5300) for them to do this, but now almost defacto standard with DDR2-800 (even in notebook SO-DIMMs, almost guaranteeing incompatibility). To even misinform the consumer more, some even report the DDR2-667 timings (e.g., CAS 5), and the DDR2-800 gets worse (e.g., CAS 6 is common), and that's only when you over-volt too. That's why I've just stuck with DDR2-667 for now, inexpensive 1GB HP parts when rebates are offered, along with a JEDEC compliant 1.8V DDR2-667 parts from a few vendors.

Luckily I noted last week the Transcend JM4GDDR2-8K 4GiB kit, two (2) 2GiB DIMMs that are fully JEDEC DDR2 Compliant at 1.8V for PC2-6400 (DDR2-800) rated with CAS 5 timing. It is a very inexpensive kit available for under $80 at the time of this writing. It replaced my 8-month old Patriot PSD24G667K 4GiB kit, also two (2) 2GiB DIMMs that were fully JEDEC DDR2 compliant at 1.8V for PC2-5300 (DDR2-667) rated with CAS 5 timing.

On the eVGA 112-CK-NF77-A1 Mainboard (eVGA "NF77") with the nVidia GeForce 7150 IGP (with nForce 630i MCP combination), the timings were reported by Serial Presence Detect (SPD) as follows ...
- Transcent DDR2-800: CAS 5, RAD 4, RP 4 and RAS 15 (5-4-4-15)
- Patriot DDR2-667: CAS 5, RAD 4, RP 4, RAS 15 (5-4-4-14)

Essentially the exact same timings, but at a 20% synchronous clock improvement. CAS 5 at DDR2-800 is basically like having CAS 4 at DDR2-667, a RAS of 15 at DDR2-800 closer to a RAS of 12 (than 13) on DDR2-667. Most people are reporting major boosts in performance as memory latency is always a major performance hit. I.e., RAS 15 on DDR2-800 is 37.5ns (26.6MHz "equivalent" -- yes, two-six megahertz like your old 80386 clock, that's how much DRAM technology sucks at latency, such as just getting the first page read). RAS 15 on DDR2-667 is 45ns (22.2MHz "equivalent"). And this is without over-clocking/volting, stock clock and timing at 1.8V.

The overwhelming majority of DDR2-800 modules are, again, over-clocked because they are over-volted beyond 1.8V -- 2.0-2.1V is extremely common. That's why most come with heat spreaders, which also hide the true synchronous clock (reported in latency). If you're running at JEDEC compliant 1.8V, you don't need spreaders, and only need to add spreaders if you want to over-clock/volt. Frankly, I want my memory chips to last, so I'll run them at stock, keep the ambient temperature in the case low, and possibly consider heat spreaders later.

If you just want basic performance, such as a low-end dual-core system, then DDR2-667 will do fine, although make sure you get 1.8V. If you're looking for optimal performance while being compatible, get a DDR2-800 which runs at a true, JEDEC compliant 1.8V and has a true CAS 5 and similar 5-5-5-15 timings or better at actual DDR2-800. So many out there are 6-6-6-18 or similar, and that requires an over-volt to 2.0V+ to get (and some want you to do 2.3V+ to get CAS 5 or 4 at DDR2-800). The Transcend JM4GDDR2-8K 4GiB kit makes a great selection, and just separately pay $5 more for the heat spreaders if you really want to add them (especially if you're going to over-clock/volt to DDR2-1000+).

This blog has been deprecated ...

As some of you may already know, I am no longer employed as an independent consultant. The work was great, the pay even better, but I quickly got "burnt out" working long hours with virtually no breaks between clients over the last few years, especially through the weekends (and not getting to fly home to my wife). Long story short, I've been working for major Linux vendors the last 6+ months, and this is unlikely to change anytime soon.

In a nutshell, since this move now seems to be permanent, I really don't have to hunt for my own dinner anymore. As such, I really don't bother to show myself publicly except when sanctioned or I otherwise find it beneficial for my employer or the Linux community in general. Yes, there is no more benefit from arguing with retirees and other, "less busy" consultants that have too much time on their hands. No need to show I'm making points based on experience, and not what Google turns up -- which is what used to get me work, because those peers who hired me knew what I was saying was coming from experience. And, frankly, I just liked to help people, but I purposely limited my help to areas where I had corporate experience (and not "I think").

Some of you already noted this, especially since I've pulled back even more in 2007 from the public scene than even 2005-2006 prior. Less help, less personality, less presence, period. My apologies to those who miss my assistance and the real answers that work in enterprises. My apologies to those where I used to directly help on-site when someone else's suggestions (typically not from experience or experience in your field) used to get you in hot water -- even if it was on my own time and expense (and the reason why those people were my biggest advocates). Again, I used to make my points because, even though I was in the minority on things, it got me work, good work (with other peers who had similar experiences to mine), and frankly it's the type of judgment I believe a sound consultant must show (either know it, or don't, and never say, "I think").

Again, I haven't had to do that for some time now, so I haven't shared myself nearly on-line as much, at least not outside of my formal role(s). These facts, combined with Google's "interesting" copyright policies on the Blogger/Blogspot site, is why I'm deprecating this blog, and moving over to my new profile/blog on Yahoo 360. My posts will be less personal, largely technical and all professional. I don't know if I'm going to stay with Yahoo 360 as it moves out of Beta or not, but it will be a RSS 2.0 feed for whenever I get the time to get my Fedora Blog going (just way too busy, but that's a good thing, more time with wife and my life). So until then, please check for updates on my Yahoo 360 page, and not here. I will also keep a link active to my latest professional blog/profile from my LinkedIn profile as well, so you can always check there.

Thank you again for so many of you caring and "keeping in touch" over the last few years. I've kept more of my personal life closed over the last few years, and kept my focus on helping paying customers. Some of you have been extremely thoughtful and professional, and I'll never forget you. Now, hopefully, I'll have more time to add back to the community with real code and other endeavors, with a good support avenue to boot, even if still not direct at the user group or support list level as I did in 2003 and prior. As I've always said, credentials and claims of experience mean nothing, only repeat, sound technical information presented and dissected over years are the only way to show experience others will trust.

Technical verbosity has always been an asset in my experience, and I'm glad most everyone I've worked with has appreciated that, even if retirees and select, other consultants didn't want to respond with the same. Especially when I explained this when people's jobs were on-the-line, like I assumed those same retirees and consultants would want others to consider if their own jobs were on-the-line (especially when they were spreading baseless FUD and pure gossip). It's never been about me "being right," it's only been about experience, applicable experience, combined with focusing on technologies and opportunities, not blind (and often misguided) vendor alignments and marketing.

As I've stated over and over, if I wanted blind vendor alignment and marketing, I would be advocating commercial products. And it explains why I work for my current employer as well, because I believe in the effort, not any marketing (which is often seeing elsewhere). I'm sure some of these people still scratch their head why I explain (defend?) my own competitors, not realizing it's about the collaborative effort, and not some vendor/marketing non-sense. I wish they would see the error of their ways, and the collective community that is built on our commonality, not differences.

I will always and gladly stand with any minority that does. And I guess that has always summed up my professional attitude, which my wife regularly tells me is why she loves me.