Go Back   Hardware Canucks > HARDWARE > RAM

    
Reply
 
LinkBack Thread Tools Display Modes
  #11 (permalink)  
Old March 2, 2007, 08:39 AM
Gav's Avatar
Gav Gav is offline
Administrator
F@H
 
Join Date: Oct 2006
Location: Canada
Posts: 1,350

My System Specs

Default

Quote:
Originally Posted by mo' power View Post
Has anyone here thought about running 3GB of ram in their rig? 2x512 in channel 1 and 2x1GB in channel 2? I've been considering this for a few weeks.
I think you should be cautious about running that kind of configuration. Anytime you try to run 4 sticks of ram in your MB, you greatly increase the chance for incompatibilities decreased stability. You also take a small performance hit when installing the 3rd and 4th stick of ram.

I'm also not sure about running 2x512 & 2x1gb in dual channel. Do all 4 sticks have to match or just each set?
Reply With Quote
  #12 (permalink)  
Old March 2, 2007, 10:21 AM
Deathspawner's Avatar
Top Prospect
 
Join Date: Jan 2007
Location: New Brunswick
Posts: 204
Default

For the sake of latency, I don't think that makes much sense. I have never done tests like that, but I think the effects would be similar to moving to single channel. You'd still have a 3GB density, but it would be offset by the lack of bandwidth.

If I can find time, I will give that shot later today, and run a few benchmarks.
__________________
Techgage.com | Editor in Chief

Intel Core 2 Quad Q9450 @ 3.20GHz, ASUS P5K Premium WiFi-AP, OCZ 8GB PC2-6400, EVGA GTX 285 1024MB
Seagate 500GB, 750GB & 1TB, Pioneer 22x ODD, Corsair 1000HX, Thermalright Ultra-120, CM Storm Sniper
Dell 2408WFP 24", ASUS Xonar Essence STX, Ultrasone PRO 750, Gentoo Linux (KDE 4.4, 2.6.32 Kernel)

"Take care to get what you like, or you will be forced to like what you get!" - H.P. Baxxter

Profiles: Last.fm | Xbox Live | Steam
Reply With Quote
  #13 (permalink)  
Old March 2, 2007, 02:49 PM
MVP
 
Join Date: Feb 2007
Location: Calgary
Posts: 493
Default

Quote:
Originally Posted by Gav View Post
I'm also not sure about running 2x512 & 2x1gb in dual channel. Do all 4 sticks have to match or just each set?
Technically, none of the sticks really have to match, but everything must be run at the same speed as the slowest stick in the bunch. In practice, this can be tricky, timings which most people don't mess with will sometimes cause problems, as it's possible for "slower" sticks to have lower SPD values, meaning both frequenct, and all timings, must be the slowest of any stick in the bunch.

In addition, Athlon 64 processors with the onboard memory controller complicate this further, as they are often unable to run 4 sticks at the same speed as 2 sticks because of the increased load. Often, changing the command rate from 1T to 2T is enough to fix this, however, sometimes it is not. This is even more pronounced in earlier Athlon 64 chip. Before the revision "E" batch, ram speeds actually had to be dropped from 400 MHz to 333 MHz when 4 sticks were installed.
Reply With Quote
  #14 (permalink)  
Old March 2, 2007, 09:29 PM
Deathspawner's Avatar
Top Prospect
 
Join Date: Jan 2007
Location: New Brunswick
Posts: 204
Default

For those interested, I did some quick tests with different scenarios to see if one hurt more than another. I will give the results in text, but you can look at the screenshots for verification.

2 x 1GB
5657MB/s - 5608MB/s (58.3ns Latency)

3 x 1GB
5154MB/s - 5149MB/s (56.1ns Latency)

4 x 1GB (Note at -no- time would all 4GB be accessed)
5653MB/s - 5652MB/s (59.4ns Latency)

2 x 1GB + 2 x 512MB
5672MB/s - 5682MB/s (59.1ns Latency)

The thing to note though, is that with 3GB or more of ram, my particular motherboard would only allow Windows to see 2.75GB. Yours will be different (unless you have the same motherboard) which would effect the bandwidth and latency results more. Overall though, it should all scale relatively close regardless of boards. Also, my 2x1GB tests were done in 2T like the rest. If you are using an nForce board you can select 1T which would add about 1,000MB/s onto the scores you see there.



__________________
Techgage.com | Editor in Chief

Intel Core 2 Quad Q9450 @ 3.20GHz, ASUS P5K Premium WiFi-AP, OCZ 8GB PC2-6400, EVGA GTX 285 1024MB
Seagate 500GB, 750GB & 1TB, Pioneer 22x ODD, Corsair 1000HX, Thermalright Ultra-120, CM Storm Sniper
Dell 2408WFP 24", ASUS Xonar Essence STX, Ultrasone PRO 750, Gentoo Linux (KDE 4.4, 2.6.32 Kernel)

"Take care to get what you like, or you will be forced to like what you get!" - H.P. Baxxter

Profiles: Last.fm | Xbox Live | Steam
Reply With Quote
  #15 (permalink)  
Old March 2, 2007, 09:35 PM
Babrbarossa's Avatar
Hall Of Fame
F@H
 
Join Date: Dec 2006
Location: New Brunswick
Posts: 3,673

My System Specs

Default

Wow-- you whipped that together quick! Certainly not worth buying more ram, is it- Although If using Vista, I bet the extra G of ram would make a noticable difference in real world applications.

Last edited by Babrbarossa; March 2, 2007 at 09:38 PM.
Reply With Quote
  #16 (permalink)  
Old March 2, 2007, 10:27 PM
MVP
 
Join Date: Feb 2007
Location: Calgary
Posts: 493
Default

Quote:
Originally Posted by Deathspawner View Post
If you are using an nForce board you can select 1T which would add about 1,000MB/s onto the scores you see there.
Is this new for Core 2 Duo and/or DDR2? Or a differnce with what Sandra is reporting compared to what I'm used to with Everst? With nforce4 S939 platforms, changing command rate changes Everst read bandwidth by about 250 MB/s at most.
Reply With Quote
  #17 (permalink)  
Old March 2, 2007, 10:50 PM
Deathspawner's Avatar
Top Prospect
 
Join Date: Jan 2007
Location: New Brunswick
Posts: 204
Default

Quote:
Originally Posted by Mars View Post
Is this new for Core 2 Duo and/or DDR2? Or a differnce with what Sandra is reporting compared to what I'm used to with Everst? With nforce4 S939 platforms, changing command rate changes Everst read bandwidth by about 250 MB/s at most.
Well first off I should correct myself. In my testing I was doing DDR2-1000 frequencies, which is absolutely impossible to do with 1T, regardless of chipset.

That aside, for whatever reason, Intel has deemed 2T to be a better choice with DDR2... for sake of stability I'd imagine. So any boards with Intel chipsets will not allow you to change the command rate, only the nForce (and possibly Crossfire as well).

Here is another test I did... changing to DDR2-800 from DDR2-1000 while retaining the same timings and such. The only other thing that changed was moving from 2T to 1T:



2 x 1GB DDR2-1000
5657MB/s - 5608MB/s (58.3ns Latency)

2 x 1GB DDR2-800 1T
5632MB/s - 5678MB/s (55.4ns Latency)

I guess the bandwidth doesn't move much ( I was thinking in DDR1 terms ), but the latency is noticeably better.
__________________
Techgage.com | Editor in Chief

Intel Core 2 Quad Q9450 @ 3.20GHz, ASUS P5K Premium WiFi-AP, OCZ 8GB PC2-6400, EVGA GTX 285 1024MB
Seagate 500GB, 750GB & 1TB, Pioneer 22x ODD, Corsair 1000HX, Thermalright Ultra-120, CM Storm Sniper
Dell 2408WFP 24", ASUS Xonar Essence STX, Ultrasone PRO 750, Gentoo Linux (KDE 4.4, 2.6.32 Kernel)

"Take care to get what you like, or you will be forced to like what you get!" - H.P. Baxxter

Profiles: Last.fm | Xbox Live | Steam
Reply With Quote
  #18 (permalink)  
Old March 2, 2007, 11:13 PM
enaberif's Avatar
Hall Of Fame
 
Join Date: Dec 2006
Location: Calgahree, AB
Posts: 10,679
Default

I laugh *haw haw* how people think this is only for Vista when its in any 32 bit environment, its simply a limitation of the architecture.
Reply With Quote
  #19 (permalink)  
Old March 3, 2007, 05:54 AM
MVP
 
Join Date: Feb 2007
Location: Calgary
Posts: 493
Default

Quote:
Originally Posted by Deathspawner View Post
I guess the bandwidth doesn't move much ( I was thinking in DDR1 terms ), but the latency is noticeably better.
For anyone else looking at the numbers, it's worth pointing out the following... while there isn't a big latency difference, the difference in latency due to 1T/2T is likepy larger than the difference portrayed, becuase latency decreases with higher frequencies, everything else constant.

All the while keeping in mind that bandwidth/latency improvements are generally important only to the extent they actually affect real-world performance of applications.
Reply With Quote
  #20 (permalink)  
Old July 15, 2007, 07:26 PM
Trial Member
 
Join Date: Jul 2007
Posts: 1
Default 2 x 2gb RAM with Windows Vista

I have the following system:

ThinkCentre A60 8700-DKU
Athlon 64 x2 - 2.2ghz
2 x 2gb DDR2 533mhz Centon Memory Sticks
1 - 160gb SATA Hard Drive
Nvidia GeForce 7950 GX2 1gb RAM
Happauge Win-TVR 1600
OCZ 700 Watt Power Supply

I have the following problem:
Windows & Bios only recognize 2430mb of RAM, I switched to 1 - 2gb and 1 - 512mb stick, still shows the same amount, 2430mb of RAM.

Is there a fix for this?

Please E-mail me at scotrinaf@********** if you know the solution to this...
Reply With Quote
Reply


Thread Tools
Display Modes

Similar Threads
Thread Thread Starter Forum Replies Last Post
x-fi elite pro and vista 32bit, anyone using? Ukonjack Audio 2 January 11, 2008 10:40 AM
32bit Vs 64bit XP enzeo O/S's, Drivers & General Software 6 September 28, 2007 02:53 PM
A Messy Transition: Practical Problems With 32bit Addressing In Windows cujo Reviews & Articles from the Web 1 July 13, 2007 09:23 AM
Vista 64bit vs. XP 32bit 4GB RAM bench comparo Eldonko O/S's, Drivers & General Software 13 May 14, 2007 02:09 PM