3GB RAM in Dual-Channel Mode?

Hi - I have a fair amount of DIY desktop hardware experience, but these are my first laptops.

I just ordered an Intel powered (P8600) GM45 chipset Toshiba CTO (Satellite L300 series) laptop and got only the minimum 1 x 1GB stick of RAM, and the base 160GB 5400 HDD. I figured I'd upgrade both the RAM and HDD myself for cheaper than Tosh asked, and get branded and performance parts I'd more prefer, as well.

I also got Vista Home Prem x86 (32b).

So, even tho 32b Vista will only see a little over 3GB RAM, I'm going to purchase a 2 x 2GB (4GB Total) matched pair of DIMMs in order to run in true dual channel-mode. I believe if I only add a single 2GB stick to the stock 1GB stick in the current config, I'll either have only single-channel mode, or dual-channel for the first 2GBs only - (this is how desktop mobos work - most default to non-dual-channel mode with dissimilar memory sticks, or, a very few recent cutting edge higher end mobos will pick up the amount of RAM that can be 'seen' in both channels concurrently as dual-channel, leaving the remaining RAM as single-channel).

Anyone care to comment?

Thx!

Reference: Satellite L300-ST2501 CTO L300-ST2501 Detailed Product Specification

 

Double post?

 

Hrm...I only see one...

Maybe it was my 'quickie' edit?

 

The other one was deleted...make sure that you hit the submit/create button only once.

 

Why would you care about dual-channel on a C2D setup? Please do some general benchmarks before and after and tell us what the difference is.

 

2 x 1 gb will run in intel flex mode, which is the first 1gb of each stick will run dual channel.

2 x 2 gb will run true dual channel. Vista 32 will still only read 3gb~

Difference between flex mode and dual channel is little to none as it exceeds the CPU's communication speed (FSB). Just keep away from single channel (single stick) as it will bottleneck it.

 

Oops...sry 'bout that. Thx!

OK, fair enough.

After I get the lappy (8 JAN 09 ship date), I'll run SisoftSandra memory B/W as received, and then again after the 2 x 2GB kit upgrade.

I used to be heavily into distributed computing and ran a home pharm (7 nodes). I noticed some projects were memory bandwidth intensive, and responded well to high B/W. But no, I never did any tests of dual channel vs single channel performance. All my experiments involved running lower CPU multis while maintaining the same final clocking.

Are there any tests or benchmarks you'd like to see or can recommend? I'd be happy to run off a few when I get the lappy.

 

Ill do the peak bandwidth calculation for you. Say the CPU FSB is 1066mhz and RAM is 800mhz

-CPU FSB
266mhz x Quad pump x 8 = 8.3gb/s peak bandwidth

-Dual Channel RAM
800mhz x Dual channel x 8 = 12.8gb/s peak bandwith

-Single Channel RAM
800mhz x Single channel x 8 = 6.4gb/s peak bandwith

Notice how single channel has less bandwidth than your CPU? That will bottleneck.

Flex mode is in between single and dual channel, which is still higher than the CPUs bandwith

 

Thank you.

I wasn't aware the hybrid mode was called "Flex-mode", but yes, that's what I was referring to in my [(1 x 1GB) + (1 x 2GB)] scenario. I remember seeing anecdotal reports that dual channel mode increased B/W some x percent, but don't remember the exact number. Consequently I've always run dual-channel matched modules in my desktop based nodes as SOP. I do admit its possible whatever difference there is between flex and true dual channel modes may not make a great deal of difference in real life general purpose use.

 

OK, Thx.

That is theoretical B/W and assumes the CPU:MEM divider ratio will automatically self-adjust to run my memory at full SPD programmed speed (PC6400), but yes, it does illustrate the differences well.

And yes, you've answer my question, thx again.

I've always run matched pair memory sticks and set my own CPU:MEM ratios (in my desktop based nodes). I had no idea about the behavior and features of notebook mobos...the BIOSs I’ve so far seen have been less-than-pitiful, to put it lightly!

My main concern was whether a mixed pair would default entirely to single channel. I wasn't aware if 'Flex' was available in notebook mobos (this is a whole 'nother world from the BIOS and features of higher end overclocker's desktop mobos!)

Mark Larson brings up a good point, tho: I think I'd like to run some theoretical and real world benchies showing the differences of all three modes. Any you’d care to recommend or see?

 

If i recall correctly it was like 4% higher score on the benchmark program, which pretty much equates to no performance gain in actual use.

 

I believe your memory may serve you well. Several years ago a member of my distributed computing team reported seeing ~5% difference in a machine where he had to remove a bad memory stick, but let it continue running on the remaining one. There's a sleigh-load of variables, though!

But my curiosity has been piqued.

I'll be running some benchies in all three modes for my own curiosity, and post back here, if you guys don't mind sparing the B/W (pun intended)?

 

Read the Dual Channel RAM Guide. It includes some benchmark results. There are also some more results in the DDR2 / DDR3 RAM thread.

John

 

O.K., A quickie status follow up:

I received the new laptop in mid January, and decided just to get a (2 x 2GB) DDR2 PC6400 kit, and be done with it. Only 40 bux USD, delivered, from NewEgg (G.Skill - now at 50 bux!). I got the 4GB since I decided to do a dual boot install with the current Vista HP x86 + W7 beta x64 (which I have done).

I felt this was future proofing a bit with the current inexpensive RAM prices, and let me play around with x64 and it's larger RAM addressing abilities, while also allowing true dual-channel operation.

I didn't do any benchies. The threads John linked had some good info and discussion, including further associated links. Thank you! It became clear the differences between real world performance in flex vs DC mode were likely minimal.

+rep, gentlemen.

+++

@John - Coincidently, the OP of the first thread you linked ( powerpack) resides, literally, minutes from where I currently live, if his zip-code is accurate. Oh well, stuff happens, I suppose.