sacrificing RAM speed for RAM size

April 30, 2009 at 18:11:18
Specs: Windows XP MCE
Hello! I was wondering if someone could assist me with this issue. Here are my system specs:

Sony VGC-RA710G
Microsoft Windows XP MCE
Pentium 4 HT 3.20 GHz
FSB 800 MHz
2x256MB PC3200 400MHz DDR RAM
4 RAM slots total

Now I had 4 extra 512MB PC2700 333MHz sticks laying around and decided to add them to my pc for the sake of having more RAM for better performance and because im currently broke and cant buy PC3200s. Now as it is the 256MB sticks are in different banks ( I assume this because the slots are in pairs. There is a space between the two pairs.) So what I did is put the two PC3200 sticks in one bank and the other 2 PC2700 in the other bank. When i rebooted, it recognized the full 1.50GB of RAM and performance improved. Now my question is: are the PC3200 running at the speed of the PC2700 even if they are in a separate bank? If yes, then I might as well replace the PC3200 with the other 2 PC2700 and have 2GB of RAM since the size of 2GB overcomes the slower speed for better performance right?

See More: sacrificing RAM speed for RAM size

Report •

April 30, 2009 at 19:06:37
1. when mixing RAM, all sticks will run at the speed of the slowest. In other words:

PC3200 + PC2700 = PC2700

2. because of inefficiencies in the P4, it's generally best to run the RAM frequency faster than the CPU frequency. Ideally, you should run PC4300 with an 800MHz FSB P4. Since your board doesn't support DDR2, PC3200 is the next best choice. PC2700 is 3rd in line.

3. if your board supports running the memory in dual channel mode, configure the RAM so that dual channel is enabled.

Report •

April 30, 2009 at 20:52:01
I received a gift certificate for a local computer shop in town for my birthday a few weeks ago. I decided that I wanted another gig of RAM to go with the gig I already have. I asked for PC3200, which is what my system uses, and they gave me PC2700 without telling me. I got to the car and realized it was PC2700. I went back in and asked them about it and they were like "PC2700 will be fine." I installed that stick, and although the performance doesn't seem to have dropped a whole lot, the benchmark in Aida32 went from a memory read speed of about 3052MB/s, to 2143MB/s. So yeah if you mix types of RAM, the fastest you'll get is the speed of the slowest RAM.

WinSimple Software
CompTIA A+ Certified

Report •

May 1, 2009 at 10:16:26
Thank you jam and rayburn. So it doesnt matter if theyre in different banks, it will always drop to the slowest RAM. So I guess my real concern is when do you sacrifice RAM size for speed or vice versa? In my instance I had 512MB of PC3200 and when I installed 1GB of PC2700 performance improved quite a bit. Everything seems to load faster. That extra 1GB of RAM made up for that 66MHz that suffered because of the mix match. But based on what circumstances would you prefer size over speed? or speed over size if you had to choose? I guess it will be on a case by case basis and what you use the computer for right?

Report •

Related Solutions

May 1, 2009 at 11:00:34
hey rayburn thanks for mentioning that aida32 benchmark tool. thats exactly what i needed. so after testing my system, here are the results:

1) 512MB PC3200 (2x256MB)
memory read = 4655 MB/s
memory write = 1702 MB/s
PF usage = 351MB (dont know if this is relevant)

2) 1.5GB RAM (2x512MB PC2700 and 2x256MB PC3200)
memory read = 2299MB/s
memory write = 938 MB/s
PF usage = 446MB (dont know if this is relevant)

Now obviously my standard 512MB PC3200 has a higher read and write rate. And also a lower PF usage. So shouldnt my original 512MB PC3200 be alot faster (twice as much based on the benchmark) than the 1.5GB mix match of PC3200 and PC2700? Because it seems to me based on loading programs and performance that I see on the system that the 1.5GB mix match has better performance. It seems to load programs faster. My system seems faster. Any insight on this would be appreciated.

Report •

May 1, 2009 at 11:27:37
One more thing. I went ahead and put all 4 512MB PC2700 on my system and benchmarked it:

memory read = 4336MB/s
memory write = 1570 MB/s
PF usage = 372MB

Now this setup seemed to be the slowest of all three. Based on system performance and how long it takes to load programs. But according to the benchmark it should be the second fastest right? So if I had to pick between those three configurations, should i stay with the 512MB PC3200 since it has the highest read and write?

Report •

May 1, 2009 at 12:52:30
I think bios would tell you the speed.

"Best Practices", Event viewer, host file, perfmon, antivirus, anti-spyware, Live CD's, backups, are in my top 10

Report •

May 1, 2009 at 14:23:04
To answer your first question, I would choose size over speed, though it is nice to have the right speed as well as size. Because like you said, 66MHz isn't going to make that much difference.

I think the reason why adding slower memory improved performance in your case may be based on the swap file. Maybe all of the programs you have running in the background took up all or most of 512MB, then when you started running applications, the total memory that was being used went above that and into the swap file, which is slower than RAM. By adding more RAM, you raised the threshold for using the swap and was then able to keep everything in the RAM (which is way faster than swap).

WinSimple Software
CompTIA A+ Certified

Report •

May 1, 2009 at 23:06:07
unfortunately jefro my BIOS doesnt show the speed. But aida32 is a great tool to detect memory speeds/MBs and so forth. and yeah rayburn i do have 4 programs running in the background. a firewall, an antivirus, adaware and spybot. But I just fixed a customers PC so im gonna have some cash to buy 2GB of PC3200 :] thank you for all your support.

ps im gonna go make a new post concerning how much I should charge customers for fixing their PCs. I would like your input on that rayburn if its not too much trouble. Thanks!

Report •

May 2, 2009 at 00:05:15
"a firewall, an antivirus, adaware and spybot"

Yeah that'll slow it down alright, especially if you're using Norton or Mcafee as your antivirus. Personally I prefer AVG or Avast since they don't bog down the system as much as the others do.

"im gonna go make a new post concerning how much I should charge customers for fixing their PCs. I would like your input on that rayburn if its not too much trouble."

I'd be glad to chime in!

WinSimple Software
CompTIA A+ Certified

Report •

May 2, 2009 at 17:41:05
haha thanks. I have avira antivir personal which hasnt been too bad. I might have to try AVG and Avast one of these days.

Report •

May 2, 2009 at 21:30:13
You need to run the numbers to give you a better idea of what's going on.

Since this is a P4, ideally you should be running a CPU:DRAM frequency ratio of 3:4. That would mean CPU at 200MHz (800MHz FSB), RAM at 266MHz (533MHz DDR). 200:266 = 3:4.

With DDR333 RAM, you're running the ratio at 200:166 = 6:5....with DDR400, the ratio would be 200:200 = 1:1.

The CPU bandwidth at 800MHz FSB is 6400MB/sec (800 x 8). The RAM throughput at 333MHz is only 2666MB/sec (333 x 8) so the RAM is the bottleneck. If you enable dual channel mode, the RAM throughput doubles to 5333MB/sec (333 x 8 x 2). That's better, but it's still bottlenecking.

The throughput of DDR400 in dual channel mode is 400 x 8 x 2 = 6400MB/sec. Obviously better but still not ideal.

Report •

May 3, 2009 at 02:01:26
Oh the joys with the old FSB method. Funny it took this long for intel to jump on the built in memory controller method that eliminates alot of these problems that people had to deal with using FSB and getting it configured properly to run at certain ratios back then. With the Athlon 64's you never had to worry about those sorts of problems because the memory controller was built onto the cpu itself which eliminated having to worry about FSB interuptions or anything else like that.

Iron Sharpens Iron.

Report •

Ask Question