Sign in with
Sign up | Sign in
Your question

My A8N-SLI Premium has 4 memory slots, right now im using...

Last response: in Overclocking
Share
December 11, 2005 7:30:04 PM

My A8N-SLI Premium has 4 memory slots, right now im using... the two blue ones. I have 1GB dual channel OCZ in each one, can I add two more 1GB dual channel OCZ sticks of the same type to the remaining slots and get the same timing and performance?
December 11, 2005 7:50:36 PM

You could do that, and im not saying this cause im jellous of you having 4gb of ram but i heard 4GB slows you down more then it helps you. I wouldnt mind having 2GB though that would be nice.
Related resources
January 30, 2006 7:21:54 PM

Quote:
You could do that, and im not saying this cause im jellous of you having 4gb of ram but i heard 4GB slows you down more then it helps you. I wouldnt mind having 2GB though that would be nice.


Why would more RAM slow you down?
January 30, 2006 7:47:29 PM

Well not sure to what degree it would be noticable but, the dual channel only works on the two blue slots, so if u fill the wrong slots or in the case of 4x 1gb strips u fill all of them then the dual channel function will not work, so in theory it will slow you down, i have 2 1gb KVR strips on my A8N-SLI, and its all i need, theres no point getting more unless u do video editing and graphics work, in my opinion anyway.
January 31, 2006 12:43:26 PM

4x1gb - 4gb would slow you down for the same reason 4x512mb - 2gb would on an amd processor (or Intel for that matter...).

4 sticks of ram, will force itself down to 333 memory speed, changing the the speed of the cpu:ram to 400:333, from 400:400. Which is somewhat significant.

That said, if you're running a 32-bit OS, you can't use 4gb anyway. 32-bit OS's will only use up to 3gb of total memory (that's including the page file). In order to use 3gb of RAM you have to disable your page file and set a special boot setting in Win XP. Disabling your page file can cause issues since some programs require it. Alternately you can change the preset 3gb boot switch to a smaller amount and enable page file (but make sure the page file is very small; and make sure the memory specified in the switch and the page file don't add up to more then 3gb).
January 31, 2006 12:57:57 PM

Its A Limitation Of Current Desktop Chipsets
January 31, 2006 1:26:49 PM

On the intel side yes, on the AMD side its a CPU limitation, since the memory controller is on the CPU.
January 31, 2006 1:30:59 PM

Oh Yeah, Thats Right, But You Knew What I Meant :lol: 
January 31, 2006 1:52:53 PM

Roight Roight! =D
January 31, 2006 2:11:30 PM

You can still run 4 sticks of RAM in dual channel mode. The 333Mhz limitation is because of motherboard signal traces and the 2T command rate with 4 sticks of RAM is an AMD memory controller limitation.

So, regardless if you have 4GB using 4 modules or 2GB using 4 modules, it will be slower because of that.

Intel, but it explains the dual channel configurations:
http://www.intel.com/support/motherboards/desktop/sb/CS...

I have a nForce2 ultra chipset and 3 dimm slots populated running in dual channel mode. Verfified in CPU-Z.

WinXP does support 4GB of RAM, not 3. 4GB is overkill at this point IMHO.
http://www.microsoft.com/whdc/system/platform/server/PA...
January 31, 2006 2:29:22 PM

The 1T vs 2T with 4x512 thing:
1T is fine at 333 (i'm running currently at 2.5 3 3 1T 333 with ch 2x512 sticks and tccd 2x512 sticks), and with TCCD ram only, some people have achieved 4x512 at 400 and 2T. In my primary machine i have 2x1024 at 3 3 3 8 233fsb 1:1.

Microsoft's 'Supporting of 4gb' means that it won't crash if u have 4gb in...however the above still stands. Windows won't recognize 4gb. It recognizes up to 3gb.

Quote:
The virtual address space of processes and applications is still limited to 2 GB unless the /3GB switch is used in the Boot.ini file. When the physical RAM in the system exceeds 16 GB and the /3GB switch is used, the operating system will ignore the additional RAM until the /3GB switch is removed. This is because of the increased size of the kernel required to support more Page Table Entries. The assumption is made that the administrator would rather not lose the /3GB functionality silently and automatically; therefore, this requires the administrator to explicitly change this setting.

The /3GB switch allocates 3 GB of virtual address space to an application that uses IMAGE_FILE_LARGE_ADDRESS_AWARE in the process header. This switch allows applications to address 1 GB of additional virtual address space above 2 GB.

The virtual address space of processes and applications is still limited to 2 GB, unless the /3GB switch is used in the Boot.ini file. The following example shows how to add the /3GB parameter in the Boot.ini file to enable application memory tuning:

January 31, 2006 2:35:37 PM

That Is If Your Running The 32Bit Version Of XP, The 64-Bit Edition Will Recognise Alot More (If Im Not Misaken Up To 8GB)
January 31, 2006 2:37:54 PM

Quote:
Microsoft's 'Supporting of 4gb' means that it won't crash if u have 4gb in...however the above still stands. Windows won't recognize 4gb. It recognizes up to 3gb.


I have had 4GB in an old IBM 4500R server with XP and it recognized all 4GB. Now, as for if it uses it all that's definitely another issue. :wink:

Edit! I'm not arguing mind you, just what I've seen. :D 
January 31, 2006 2:52:07 PM

You might want to double check that. Do some searches online, you won't find anyone using 4gb, where it shows in windows using 4gb, it will show 3gb every time.

Edit: Actually I saw an instance of this recently...hold on i'll find a link.

http://bink.nu/forums/8805/ShowPost.aspx
http://forums.pcper.com/showthread.php?t=407270&highlight=4gb


At the bottom of the pcper, there is a intel document, stating again what windows xp is capable of. It says in it it's capable of 4gb of physical ram...I didn't get finished reading though, lunch is here. =)


I think i may be wrong...what i was thinking of may have been that an individual program can only address up to 3gb of information...I'd rather be wrong, and be corrected then wrong and think i'm right. =) That said, i'm not convinced from the information so far, that i'm incorrect. All i have to go on, is past experiences. I've never seen Win Xp show 4gb in windows. And i've run in to issues with this on both amd and intel systems. Gotta run, foods gett'n cold, i'll chk this in a bit.

Edit 3: Ok...i found what i was looking for.
http://download.microsoft.com/download/3/8/1/381c074b-b5d9-4054-a909-fbba470af276/Benefits_of_Windows_x64.doc
See page 3 at the bottom...under 32-bit Pain Point...

Quote:
The 32-bit versions of Windows use a flat, 32-bit virtual address space, limiting the amount of virtual memory that can be addressed directly to 4 GB (232). By default, this 4 GB is divided into two equal buckets: 2 GB that can be addressed by a process, and 2 GB that is addressable by the operating system and shared across all processes. These buckets can be reconfigured to give 3 GB to an application by using the boot time /3GB switch, thus limiting the operating system to only 1 GB. This requires that an application be compiled to take advantage of the extra space. Squeezing the operating system into only 1 GB is a significant constraint, and this will not be an improvement for all workloads or programs.


Basically restates again, that 4gb is available to the OS, but in a way that leaves no room in my mind for any doubt. Man that really grates me, i hate being wrong. =P [doubt creeps up heh...Unless it doesn't show the memory it's allocating to the OS...which i'm 99% sure it does...grr]

After a bit of reading...The /3gb switch is for individual programs, that are capable of using up to 3gb of ram. Which leaves whats left of the memory to be used by the OS (and whatever wasn't eaten up before getting to windows). As stated above, fitting the OS in to the 1GB is an issue, but there is a way around it by changing the switch to a smaller amount manually. Also, in XP 64-bit, a 32-bit operation can use up to 4gb (with again the proper switch...), if the 64-bit OS has more then 4GB of ram.

And some more reading...The reasons i have been seeing 3gb in Windows, according to one of Intel's sheets, is specifically because of PCI-Express systems, and the amount of RAM accessed by the pci-express card (256mb and 512mb pci-e video cards...taking memory away from the available/addressable amount to the OS). I assume the same would be true for AGP slots as well. Subtract 64, 128, 256, 512, etc depending on how much memory each graphics card has. So, after you take away the memory for the graphics, the memory on the page file, and memory taken by other hardware (pre-OS), i end up with around 3gb. OR, there is a hardware limitation of some sort (be that memory controller limitation or other).
January 31, 2006 3:59:27 PM

The old server I built has only 2GB now and I can't add 2 more to check because they're actually using it at the moment. Pfft! :roll: I only ever looked at the task manager performance though and since that does combine the page file and the rest of system memory my own memory could be spotty on recollecting what it reported. I do know they were only using about 1.5GB, that's why I took the other 2 out to put to better use.

Besides that mach1 wanted to know if he would get the same timings and performance and I believe we have unanimously decided that's a no. :) 
4GB in windows XP right now is total overkill anyways, regardless of what it reports. I say wait for Vista in 64 bit to make use of it.

Hope your lunch was good, I'm off to eat mine now!
January 31, 2006 4:31:01 PM

Yea 4gb is overkill. The only thing i can think of that would use 4gb off the top of my head (on a workstation anyway...), would be some* CAD based software.
February 17, 2006 8:27:13 PM

Just barging in to set some stuff straight here.

Indeed the /3gb switch means that each individual program may be allocated up to 3GB of RAM. Normally a program can grab 2 at most.
This does not influence the amount of RAM available whatsoever.

Also, WinXP uses Physical Address Extension to allow memory mapped above 4GB to show up, but due to the wonderful security updates in SP2 that doesn't work anymore.
BTW when I say mapped above 4GB, that means the last gigabyte or so of your RAM. Standard procedure is for a PC to start at the top (4GB) to allocate stuff like BIOS, PCI stuff, video card memory, etc. So they work downwards. Memory works upwards. If RAM encounters the reserved space it is either cut off from there or remapped above that limit. New boars like this ASUS remap it, but then you run into windows. Use x86-64 instead. I know I do.

Lastly, the speeds are NOT due to controller limitations. For some reason the ASUS board cannot cope with the signal. They released an update (1013) to make it default that the memory clocks down when 4 sticks are used. Like indicated above, it looks like a signal issue.
The 2T thing IS a limitation of the controller.
February 18, 2006 1:54:09 AM

Question, I will be using 4x 512 Corsair XMS 3200XL. On a EPoX 9NPA+ Mobo. Will I be able to OC to override the limitation or did I waste some money on an extra 512 of ram? If you have a different solution please let me know.
February 18, 2006 9:37:27 AM

You can OC the memory/CPU if components allow, and factually it will negate the penalty for using four sticks of RAM.
On the other hand, it is known that using 4 sticks overclocks the memory less favourable than using 2 sticks.

It is a compromise. Generally, the more memory you use, the less it will overclock: either four sticks of lower density (512MB) or 2 sticks of high density (1GB), both don't OC like a champ, in general.

On the other hand, according to some articles and benchmarks I read, the limitation of 4 sticks is only to reduce command rate from 1T to 2T, and this does not overly affect bandwith. Personally, I think the gain from more memory is worth the slight loss in performance (which might even be bridged, like I said, by a modest OC)
February 21, 2006 12:58:49 PM

KK, the 4gb thing i think we have a handle on now (in regards to xp 32-bit lol)...


As to the speeds...I don't think the limitation of 1T vs 2T is any different then it's issues with running at 400 vs 333. Unless there is a hardware (memory controller AMD) issue specifically where it won't handle 2gb @ 1T and 400(which i know there isn't...i have 2gb 2x1024 @ 500 and 1T and a machine with 4x512 @ 333 and 2T meh).

There really isn't any difference in 2x1024 and 4x512...except where the ram is located (i'm not sure how unclear i just made that =/...). Basically the difference is the connection at the motherboard which creates latency issues. The greatest amount of latency is created by the non physically complete contact between the ram and the mobo, so with 4 sticks there is an increased latency of 4 from 2. Which is why you can run at 400 and a 1T command rate with 2x1024, but you can't with 4.

Further more, some ram will run at 400 and 2T, but only tccd ram (that i've seen). So the latency can be overcome by some especially low-latency ram, that has been specifically designed to work with the controller (or just happens to work better with the AMD controller then it works with p4 ddr memory controllers). So is that an issue with the controller? Not really, it's an issue of the mobo and latency imo. Additionally, some people when using the same 4 sticks of ram (and some without), have been able to get 333 @ 1T cmd rate. So to me, that says the same thing as the 400 @ 2T...That it's not a controller issue, but a latency issue, where the physical connection just isn't good enough.

Any thoughts? Holes in my psedo-logic?
February 21, 2006 3:38:12 PM

My memory (TwinMOS stuff, I am not sure of ramchip type) runs fine at 400MHz, 2T. 4 sticks of a GB each. It ran on a DFI LanParty UT NF4 SLI-D board. The asus won't do it, though. As far as I know mine is not spanking high end RAM in terms of clockspeeds, so I think the limitations of the A64 memcontroller aren't that large. Not, at least, on E6 revision chips (like my X2 3800+)

The difference between 1T and 2T timings are minimal, even negligible. Check the following thead @ HardOCP:
http://www.hardforum.com/showthread.php?t=917047

On the other hand, the difference between DDR333 and DDR400 is quite a bit more obvious:

PC-2700: DDR-SDRAM memory module specified to run at 166 MHz using DDR-333 chips, 2.667 GByte/s bandwidth per channel
PC-3200: DDR-SDRAM memory module specified to run at 200 MHz using DDR-400 chips, 3.200 GByte/s bandwidth per channel

This means a difference of 533MB/s, being 1/6 of PC3200 total bandwidth. A performance loss will probably be noticable, especially when using dual core processors etc. Note that it says per channel. In a dual channel setup the total bandwidth will thus increase to 6.4GB/s, with the difference increasing to 1066GB/s
February 21, 2006 5:45:27 PM

Quote:
My memory (TwinMOS stuff, I am not sure of ramchip type) runs fine at 400MHz, 2T. 4 sticks of a GB each. It ran on a DFI LanParty UT NF4 SLI-D board. The asus won't do it, though. As far as I know mine is not spanking high end RAM in terms of clockspeeds, so I think the limitations of the A64 memcontroller aren't that large. Not, at least, on E6 revision chips (like my X2 3800+)


If they ran fine at 400 2T on the DFI but not on the Asus, that doesn't sound like a cpu issue at all...Sounds like a better(closer? more accurate?) implementation of the ram lines to the cpu. But who knows for sure.

To quote Wusy from the forums chatting at me in my 400 and 1T or 2T thread
Quote:
..."Apesoccer, there is probably someone somewhere who is running 400 and 1T...".
And he's right, so there's no doubt people without tccd ram are running 400 and 2T just like you. But they're exceptions to the rule rather then the rule.


That's a great thread, i've added it to my Fav articles. That said...it was run on a socket 754...Let me add a couple of links to your pile. I'll add them as i find them.

...actually how about just one:

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2069&p=8
Quote:
The Importance of Command Rate

Socket 754 Single-Channel motherboards performed best with a memory Command Rate setting of 1T in BIOS, but that generally was a stable option with only one DIMM. 2 or more DIMMs normally required a 2T Command Rate setting for most stable performance. There was a performance increase at the 1T Command Rate setting, but the real performance increase was very small.

Socket 939 Dual-Channel motherboards were found to exhibit a very wide performance difference between a Command Rate setting of 1T and a setting of 2T. The impact on memory bandwidth is dramatic between these 2 settings. In SiSoft Sandra 2004 standard buffered Memory Benchmarks, a 1T command rate showed a Sandra bandwidth of 6000 Mb/sec, while a 2T rate with the same 2 DIMMs in Dual-Channel mode was only 4800 Mb/sec. This is a huge difference in memory bandwidth and the Command Rate setting definitely impacts performance test results on Socket 939 motherboards. All AnandTech benchmarks were run at a Command Rate setting of 1T. This includes all benchmarks that were run in the CPU tests, as all benchmarks were rerun in the CPU tests as soon as we had verified the performance impact of Command Rate settings.


As to the 400 being better then 333. That kinda goes without saying... I knew that wouldn't come out right...When i was referencing 333 and comparing it to 400, i didn't mean that 333 1T was faster then 400 2T. I was just saying that 333 and 2T were both conditions created by latency issues in a flawed system. ...not that any system is perfect lol. Omg...i just reread that...sounds like i got my head stuck up my ass. lol. Man, i really need to get out more. [cry=apesoccer]Doh[/cry] heh.
February 22, 2006 12:48:00 AM

Quote:
Well not sure to what degree it would be noticable but, the dual channel only works on the two blue slots, so if u fill the wrong slots or in the case of 4x 1gb strips u fill all of them then the dual channel function will not work, so in theory it will slow you down, i have 2 1gb KVR strips on my A8N-SLI, and its all i need, theres no point getting more unless u do video editing and graphics work, in my opinion anyway.



sorry to burst your bubble, but my current config consists of four dual channle sticks, and cpuz says its dual channle, boot says dual channle too(when its in auto mode) but it doesnt matter as i overclock and run at the same speeds and faster and allways its detected as dual channle.
so i guesse i proved you wrong, then again, countless others have too
February 22, 2006 1:03:16 AM

also, the mobo i use is asus a8ne, which does infact when set to auto, sets the ram frequency to ddr 333
with all 4 sticks.
thing is, i set the fsb to 245 from 200
and set the ram timings to 3 3 3 8, and my system runs fine with a clock speed of 2.2 and actual ram speed of 200mhz, which is where it would be at stock with ddr 400 timings

so the solution is what i did basically, just overclock untill you are at the required speed anyways. works good
February 23, 2006 1:53:34 PM


That's a great thread, i've added it to my Fav articles. That said...it was run on a socket 754...Let me add a couple of links to your pile. I'll add them as i find them.

...actually how about just one:

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2069&p=8
Quote:
The Importance of Command Rate

Socket 754 Single-Channel motherboards performed best with a memory Command Rate setting of 1T in BIOS, but that generally was a stable option with only one DIMM. 2 or more DIMMs normally required a 2T Command Rate setting for most stable performance. There was a performance increase at the 1T Command Rate setting, but the real performance increase was very small.

Socket 939 Dual-Channel motherboards were found to exhibit a very wide performance difference between a Command Rate setting of 1T and a setting of 2T. The impact on memory bandwidth is dramatic between these 2 settings. In SiSoft Sandra 2004 standard buffered Memory Benchmarks, a 1T command rate showed a Sandra bandwidth of 6000 Mb/sec, while a 2T rate with the same 2 DIMMs in Dual-Channel mode was only 4800 Mb/sec. This is a huge difference in memory bandwidth and the Command Rate setting definitely impacts performance test results on Socket 939 motherboards. All AnandTech benchmarks were run at a Command Rate setting of 1T. This includes all benchmarks that were run in the CPU tests, as all benchmarks were rerun in the CPU tests as soon as we had verified the performance impact of Command Rate settings.


Yeah, but then again, I read at a few places (like that thead, methinks) that Sandra really loves 1T, and that other benches and real world situations show a much smaller difference.

And indeed, I was also thinking better lane design of CPU <=> RAM bus in the DFI. Quite sad ASUS couldn't properly layout their buses.
!