Quick comparison 3GB -> 4GB on M1530

Hi all,

I was curious to know whether there was a performance difference between 3GB and 4GB of RAM on the M1530 in regards to dual channel mode.

With 3GB, the motherboard would be running asymmetric dual channel mode i.e. 2GB would be dual channel and the other 1GB would be treated as single channel.

So, I've tried it and run a couple of benchmarks. Note: I've chosen some commonly used benchmarks but nothing that actually needed more RAM.

All benchmarks were running on a fresh install of Vista 32bit with 55 processes running. The power profile was on High Performance.

The RAM is running at 667Mhz with 5-5-5-15. The first stick (from the factory ) is Hyundai (Part # HYMP125S64CP8-Y5). The second stick is Transcend (Part # JM800QSU-2G).

Antivirus software (Eset) was installed but disabled. The graphics drivers were 169.09 from LaptopVideo2Go.

So, here we go!

3GB
===

WEI (only included as everyone running Vista can use it)
CPU = 5.3
RAM = 4.8
Graphics = 5.9
Gaming Graphics = 5.5
HDD = 5.3

SuperPi 1.5mod
1M = 20.670
2M = 50.856

3DMark06. GPU at stock speeds (475/700) = 4310
3DMark06. GPU at OC'd speeds (600/900) = 5293

4GB (Vista Reports 3581Mb Total)
=========================

WEI (only included as everyone running Vista can use it)
CPU = 5.3
RAM = 4.8
Graphics = 5.9
Gaming Graphics = 5.5
HDD = 5.3

SuperPi 1.5mod
1M = 20.93
2M = 50.98

3DMark06. GPU at stock speeds (475/700) = 4333
3DMark06. GPU at OC'd speeds (600/900) = 5320

Conclusion
========
There were slight differences in both SuperPi and 3DMark06. I would put these differences down to a margin of error in the testing process rather then a definite performance difference.

 

Thanks for the benchmarks. Doesn't look like much of a change. I guess I'll stick with my 3GB for a while then.

 

You would most likely see a much bigger difference if you had Vista 64 installed instead of 32.

 

If it was a test error, what are the odds of all the scores show that 4GB is better? I think 4GB is slightly better. Not worth it, but better. Great comparison, thanks and rep to you!

 

By margin of error, I mean a statistically insignificient difference.

Caused by anything from a minor fluctation in supply voltage to a butterfly in New York falling out of a tree

To determine whether the difference in the 3DMark06 score under 4GB wasn't a fluke, I'll run it a couple more times today and add the results.

 

Vista 64 benchmark would be interesting. I thought there'd be more of a performance difference than this.

 

Your not really straining the ram on the system! You should do some seriously RAM limiting benchmarks to see if you could get a larger margin. I don't know of any RAM intensive benchies though....

 

Do you mean I wasn't using anything that required a lot of memory? If so, that is what I intended. I wanted to see the difference that symmetric dual channel mode made of asymmetric in some common benchmarks that people use to compare stuff.

I actually thought that all three tests would show a difference because in theory, the memory bandwidth should now be consistent across the entire memory space as opposed to dropping off 2/3rds of the way through.

I was also thinking about it because I remember going from 1.5Gig to 2Gig in my Asus A6Ja making a rather large difference to the SuperPi scores. I don't have any numbers on me at the moment unfortunately.

 

Nice analysis. Useful info. Thanks!

Rep:d.

 

You didn't really do a memory test. Try a memory benchmark! And/or a general-purpose application benchmark.

SuperPi only really tests CPU.

WEI tests RAM speed, but as you mentioned you should only have a difference in the upper area of RAM (>2gb) so the results make me think it may not be testing all areas of RAM.

I don't really know much about 3DMark other than that it mostly tests the video hardware.

If you used Vista 64-bit, you probably would notice a difference -- but due to the OS change, not any RAM change. You'd have to go back and test Vista 64 with both 3gb and 4gb if you want to test the RAM change.