DDR2 Ram and FSB, now let me see if I understand all this...

Ok, I've been doing a bunch of reading on this subject (and everyone seems to have a different opinion) but this is what I've gathered so far.

If you have a DDR2 RAM module with a speed of 400MHz (just for the sake of an example) and you use a pair in your notebook, then the "speed" of your RAM would effectively double which would give you 800MHz.

Now if you were to use 533MHz DDR2 RAM and you installed this in a computer with a 533MHz FSB, the "extra speed" of the RAM would effectively be lost since the computer could only handle 533MHz.

Now if I understand this correctly, to see any benefit from running DDR2 RAM in identical pairs would be if the FSB is greater than the speed of an individual RAM module. If the FSB is the same speed of the RAM module, than it wouldn't make any real world difference.

Do I understand this correctly or did I get the whole thing completely screwed up?

 

If you have two sticks of the same speed memory, they run in dual channel mode, which raises the amount of bandwith and speed of applications the notebook runs. If the FSB of the processor is 533, then the DDR2 RAM would run at that speed and there would be no loss in that sense. Right now the difference between 667 and 533 are really close together until they can drop the latency of the memory(they are just taking a long time .

 

NO
A dual channel memory has no performance effect on notebooks (maybe 0.1%).It's because of it's different design with PC.The benchmarks show that (search for yourself (dual channel vs single channel) it's just the lowered price that matters for anyone especially if you have Ram Maximum (e.g. 2 GB)

 

No, your DDR2 400MHz RAM is already running at dual data rate mode. The effective clock speed of DDR2 400 memory is 200MHz and the manufacturer then rate it at 400 including the DDR factor.
For example: Say you have 400MHz old SDR(single data rate) RAM, then its data trabsfer speeds will be equivalent to that of a memory running at 200MHZ but capable of DDR.

Again, your memory is not running at 533MHz, but at 266MHz, so whatever bandwidth is available the cpu is capable of using it, but in most cases it is left unused.

With FSB based systems you can use memory upto the speeds of the FSB, use of memory anything higher it will downclock itself to that speed. DDR and dual channel are 2 different things, dul channel effectively doubles your available by widening the memory channel(equivalent to say a 2 lane road). DDR on the other hand effectively doubles the data output, by sending data both on the raising and falling edges of the frequency. The data transmission rate is determined by the clock speed of the memory and the FSB.

 

I love when people say "What I claim is true, but I can't be bothered to provide anything to back it up"
The burden of proof is on you, if you're the one making the claims. If you have links to benchmarks, just post a link please.

Also, that's a pretty broad, sweeping and inaccurate statement. Dual channel has an effect on notebooks. It just depends on the rest of the system. DO these mythical benchmarks of which you speak test with various CPU types and motherboards, and a wide range of tasks?

Anyway, normally, RAM can usually run faster than the FSB without a problem. The problem is that the FSB determines when data can be sent from RAM to memory. That happens exactly once per FSB clock cycle.. So if it's not synchronized with the memory, the memory will sometimes (not always) have to wait a bit for the FSB to start a new cycle. But you'll still get *some* benefit from the faster RAM. It's only sometimes the RAM has to wait for FSB. Not always. (At least, as far as I know. Miner seems to disagree. Unfortunately, I don't have a FSB-based system to test it on, only an Athlon 64)

And no, dual channel doesn't affect the memory "speed". It's more like the two RAM modules operate in parallel. So you can transfer twice as much data, because you can work on both modules simultaneously.

And like said above, the way to determine if you'd benefit from faster memory is simply to look at the current speed. If you've got 533 mhz, don't bother, because 667mhz is hardly faster at all with current hardware.

Well, that's where it gets tricky. The memory is running at 266mhz, but it transfers data twice per cycle, which means it fits exactly with a 533 mhz bus, which can only transfer data once per cycle.

 

This may help: RAM Explained

Google is your friend.

 

Jalf, excuse me. Do you have anything to help the original poster's question? Miner, right or wrong in your opinion, was trying to help.

Can we just work together here a bit?

 

I think Jalf's post has some nice technical information regarding this issue.

Just a minor point - DDRII does four bits of data per clock cycle because the core runs at half the frequency of the I/O buffers. The data buffers are running 2x as fast as the core speed. Now slap on DDR protocols, and you have four bits of data per cycle.
DDR memory does two bits per cycle because the data buffers run at the same frequency as the core.

 

I wasn't criticising Miner at all, so not sure what you mean. I said that I thought what he said was wrong about how the system handles ram speeds higher than the fsb speed, and then I explained my understanding of the issue. But I wasn't entirely sure, so I couldn't test it, and so he might actually be right. And that's what I said (or meant to say. Apologies if that wasn't clear)

If I pissed off Miner, I apologize, because that wasn't my intention. But I think I did help answer the question. And what better way to work together than to help correct each others mistakes?

Yes, four transfers per cycle, but it only transfers data twice per cycle, just like DDR. The improvement is that it makes two transfers each time, where DDR only made one, so you get more transfers in the same time, but they occur only twice per cycle.
At least, as far as I can remember. Been a while since I really read up on RAM technologies.

 

The data buffers in DDRII RAM can work on four bits per clock because the data buffers run at twice the core speed. Now, I don't know if that is theoretical or not, and whether they -always- work on that much, but it is possible.

 

I can help


As you can see quite clearly, the memory is quite capable of running faster, yet is running at 1:1 with the bus. It automatically downclocks to that speed.

 

Well, aaaanyway, the point was that what normally happens when RAM and FSB runs at different speeds, is just that sometimes, a request from the FSB will arrive on RAM a bit "out of sync", so it'll have to wait for the RAM to begin on the next cycle. And sometimes, the RAM will finish reading a bit before the FSB is ready, so it has to wait a bit there. That's why it's usually best to keep RAM and FSB at synced ratios (1:1, 2:1 or 1:2 or something that doesn't give you too many "odd" numbers when counting cycles.)

At least, as far as I know, that's how it works. According to Miner, what happens is simply that the RAM is downclocked to fit the FSB speed. That might be true too. I can't test it at the moment, because Athlon 64 doesn't have a regular FSB.

But anyway, technobabble ahead: (Feel free to not read that. Just want to clarify some things with Chaz. He looks like he knows his stuff... )

Ok, I looked it up. Couldn't remember the details either.
DDR transfers data on the rising and falling edges of the (internal) clock, which allows the (external) clock to run at twice the speed. It fetches two bits in each such read.
DDR2 simply doubles the prefetch buffer size, so it grabs 4 bits on each (external) clock cycle, that is, on falling and rising edges of the (internal) clock, but it doesn't read data more often (That is, still only twice per core clock cycle)

 

Aah, thanks Miner. Nice to get that sorted out.
And you can't override that in bios then, I guess?

 

I would have taken a screenshot of the BIOS had it not been so crappy. Notebook BIOS(I took the CPU-Z scrshots from an Acer aspire 3623WXci) have no options what so ever. So if you want a BIOS confirmation, then you will have to look into a retail desktop motherboard, one which is not tinkered by any OEM.

I still disagree with you on the advantages of dual channel though.

 

Notebook BIOS' are ususally very primitive indeed. Even on the most advanced ones, you still have virtually no options besides changing the boot order and maybe one or two other things.

I'd say that you know quite a bit more than I do. You're really helpful around here.

 

To Jalf, my apologizes for being a bit short back there.

The thread had a happy ending and I had a good read.

Cheers