DDR2 - I'm confused

[otheracco]

What kinds of DDR2 RAM can I run on this motherboard at 1066MHz?

http://www.asus.com/products.aspx?modelmenu=2&model=1870&l1=3&l2=101&l3=345&l4=0

I've read another post that attempted to describe, in detail, how DDR2 DRAM works, but so far I don't understand anything past DDR1. If someone could just tell me what RAM and clock settings I would have to use it would be a big help.

 

[cdogg]

There have been a lot of confusing threads about memory, but all you need to know is that your motherboard supports 1066MHz DDR2 (also called PC2-8500).

So you should have no problems buying memory like the ones you see listed on this page (check to make sure you select a reputable vendor before buying):

http://www.pricewatch.com/system_memory/ddr2-1066_pc2-8500_2gb_kit.htm

I pointed you to 2GB kits assuming that you would like to run the memory in dual-channel. Keep in mind that dual-channel requires at least two sticks of identical size and speed for it to work. Also realize that depending on the application, dual-channel may or may not make a difference in overall performance.

Hope that helps...

~cdogg
"Insanity: doing the same thing over and over again and expecting different results." - Einstein
[tab][navy]For posting policies, click [/navy]here.

 

[otheracco]

Well, I already knew I could run the 1066 (8500) RAM, but I would really like to spend half the price for the same performance.

I've also read that all PC2-8500 RAM needs to be manually set in the BIOS as 1066 to achieve it's rated performance. Although, I read that from an ebay seller who 'might' be lying to sell his product.

I want 2x2GB = 4GB to use as a test server at home. That's why it would be nice to buy earlier PC2 RAM, to save lots of money.

Can I use 800MHz PC2 6400 RAM (for half the price) and OC it to 1066MHz?

Is it true that PC2 8500 is really just OC'd 6400?

 

[BadBigBen]

Can I use 800MHz PC2 6400 RAM (for half the price) and OC it to 1066MHz?

probably NOT... There are RAM sticks out there that are overclockable but they costs almost or more than the native rated RAM... and I would never overclock RAM if I was to use it in a SERVER environment, as you want STABILITY, and OC RAM gives NO stability at all...

Is it true that PC2 8500 is really just OC'd 6400?

NO...

btw. DDR2-800 RAM is plenty fast already, and the speed difference is only noticeable in memory intensive applications...

Ben

"If it works don't fix it! If it doesn't use a sledgehammer..."

 

[kjv1611]

==>and the speed difference is only noticeable in memory intensive applications...

As Ben is saying, you probably won't know the difference anyway with most everything. And even when you do notice a difference, it probably isn't worth paying for - unless it'll somehow make a difference in your income.

If it's that much price difference, I'd definitely go with the cheaper RAM, no question.

--

"If to err is human, then I must be some kind of human!" -Me

 

[cdogg]

That's why it would be nice to buy earlier PC2 RAM, to save lots of money

Yes, this is true. PC2-6400 is almost half the price of PC2-8500 when you buy retail. Not much I can say there...

Can I use 800MHz PC2 6400 RAM...

Yes, you can use PC2-6400 RAM as Ben points out. High-end PC2-6400 RAM with low latency timings can be overclocked to 1066MHz to match the latency of low-end PC2-8500. You must realize that although the timings of your PC2-6400 RAM might look superior on paper, that advantage goes away when you overclock it.

I understand now that your main objective is to save money, so in that case, I would say that overclocking PC2-6400 is a desirable option. In general, Ben is right about stability when overclocking memory. But again, if you start with high-quality PC2-6400, then it is designed to be overclocked and stability won't be as much of an issue. You just have to hope your motherboard is able to correctly adjust to the modified timings as a result of the increased speed.

Would I overclock memory to save a few bucks? Not likely.
Would I do it to save like $75? I'd definitely consider it.

~cdogg
"Insanity: doing the same thing over and over again and expecting different results." - Einstein
[tab][navy]For posting policies, click [/navy]here.

 

[cdogg]

otheracco,
I also forgot to address your last question about whether or not PC2-8500 is just overclocked PC2-6400. The short answer is, YES!

The same is mostly true when you compare any DDR2 module with a slower or faster one. The technology used is nearly identical across the different speeds. Improvements in the manufacturing process such as die shrinks allow for higher voltages (or more efficient use of the voltage). This in turn gives different speed options when the technology is essentially the same.

So the point you're making isn't as alarming as it may seem.

 

[kmcferrin]

Cdogg,

I disagree on the technical point that it's not actually overclocked. Overclocking something is running it at a clock speed that is in excess of the speed for which it is rated. DDR2-1066 is rated for 1066 MHz, so running it at that speed is not overclocked.

It is true that the memory modules themselves may have come off of the same lines that produces DDR2-800 or DDR2-667 and is identical in design and manufacture, but that doesn't make them overclocked. The memory manufacturer (much like CPU manufacturers) builds the chips the same way across the board. Afterwards they test them for errors and validate that they will function at their intended speed. Modules that pass validation at higher speeds usually get sold as the higher speed modules (for higher prices). Modules that don't pass validation at higher speeds get labeled and sold as slower modules.

________________________________________
CompTIA A+, Network+, Server+, Security+
MCSE:Security 2003
MCITP:Enterprise Administrator

 

[cdogg]

kmcferrin,
Yes, it is essentially the same verification process that CPU wafers are tested by. I would not have used the term "overclock" to describe the difference in speed, but without going into too much detail, I gave the short answer 'Yes'.

Though back to a technical point for a moment, overclocking is part of a loose terminology. It can apply to range of methods like increasing the clock multiplier for a CPU or changing the voltage. If you look at some of the core differences between slower and faster DDR2 modules, you will see a difference in the voltage used. It can be argued to some extent that this reflects the thinking behind overclocking, except that the manufacturer's guarantee through testing is the only difference that keeps it from being labeled as such.

~cdogg
"Insanity: doing the same thing over and over again and expecting different results." - Einstein
[tab][navy]For posting policies, click [/navy]here.

 

[wahnula]

This is why I like TT. Not only answers, but explanations and deeper information about a problem or question. Sometimes, like in this case, it's an interpretive thing, with no clear black-and-white answer, just information for us to reach our own conclusions.

I don't consider the higher-spec'd RAM (or CPU's off the same wafer) as overclocked, just better-made parts capable of better performance. The "overclocking" is done to parts from the same batch but from different bins, and is done by the customer without warranty.

Tony

Users helping Users...

 

[cdogg]

Tony,
It's always great to have you weigh in!

In this situation as you've pointed out, there's really not a clear-cut answer. We have to look at it on a case-by-case basis. Although kmcferrin and yourself point out that the best performing modules (or CPU's) are singled out from the rest of the bunch and labeled accordingly, this is not always what determines the speed at which they are sold. I'll explain.

Take the old Athlon XP Thoroughbred 'B' as an example. The first iterations of this model ran at 1400MHz (1700+). The latest ran as high as 2.2GHz. It can be argued that early on, most if not all cores yielded from each wafer had the potential to perform faster than the 1400MHz spec. However, AMD (as does Intel) has a general timeline they follow in releasing faster cores to the public. Unless the competition steps up the pressure, they like to stick to that timeline as much as possible. This buys them time to develop next generation models while milking as much as they can out of the current one. As a result, it is fair to say that many of the cores were labeled well below their potential output.

If you then study the difference between slower and faster Thoroughbred 'B' CPUs, you would discover that the only real difference here is the CPU clock multiplier and in the final three iterations, the speed of the FSB. Sure, as AMD approached the upper limit of the Thoroughbred's potential, they were likely yielding a lot less CPUs per wafer that were capable of this speed. That's when the verification process kicks in and the slower ones are labeled as such.

The point I keep coming back to here is that it is not always the case either way. You have to look at each case individually. I don't know as much about memory, but I suspect there is a similar comparison between manufacturing and performance.

_________________________________________________________________________________

And since we brought up wafers, I came across a page that had a picture I thought you'd find interesting:

http://www.dailytech.com/article.aspx?newsid=4336

~cdogg
"Insanity: doing the same thing over and over again and expecting different results." - Einstein
[tab][navy]For posting policies, click [/navy]here.

 

[kmcferrin]

You are absolutely correct. AMD, Intel, and just about everyone else in the semiconductor industry practice "binning", where faster/better performing parts are sold as lower performing parts in order to fill demand or adhere to roadmaps.

________________________________________
CompTIA A+, Network+, Server+, Security+
MCSE:Security 2003
MCITP:Enterprise Administrator