A8n-SLI (Single channel vs Dual Channel Memory)

[Fent]

Ok, odd's are my "not on the list ram" will not work in dual channel. But I thought I would ask the dumb question others might wonder as well.

What makes the memory work in dual channel vs single channel, how can you tell one from the other.

Do you have to do somthoing other than plug 2 simms into A1 and B1 on the A8N-SLI?

What are some real world benchmarks of single channel vs dual?

What is the best way to "tell" what mode your ram is in? (softsandra?)

FYI I have Kingston HyperX KHX3200UL/512 512MB DDR400 PC3200 Ultra Low Latency

Seems to be single channel mode.

 

[Tripp17]

First of all, do you have your memory in the blue slots....lol...i know that might sound dumb, but you never know. Also, the benifits are amazing in dual channel. I get about 7,200 in Sandra for my memory. If i had single channel it be about 3,500. So it almost doubles your mem performance. I always buy dual memory kits.....mine is the OCZ El Platinum Rev.2 (TCCD) 2x512 sticks. At stock they run 2-2-2-5 at 400. Mine is at 490 1:1 and my settings are 2.5-3-3-10. I hope some of this helps you if not now but in the future.

 

[Fent]

yes, they are in the blue slots.

And they are the same type, brand speed.. but were not a "kit"

and your right your speed is showing a lot faster than what I get so how do you "fix" it?

SiSoftware Sandra

Benchmark Results
RAM Bandwidth Int Buff'd iSSE2 : 4929 MB/s
RAM Bandwidth Float Buff'd iSSE2 : 4907 MB/s

Int Buff'd iSSE2 (Integer STREAM) Results Breakdown
Assignment : 4765MB/s
Scaling : 4779MB/s
Addition : 5085MB/s
Triad : 5089MB/s
Data Item Size : 16 byte(s)
Buffering Used : Yes
Offset Displacement Used : Yes
Bandwidth Efficiency : 77% (estimated)

Float Buff'd iSSE2 (Float STREAM) Results Breakdown
Assignment : 4774MB/s
Scaling : 4743MB/s
Addition : 5058MB/s
Triad : 5053MB/s
Data Item Size : 16 byte(s)
Buffering Used : Yes
Offset Displacement Used : Yes
Bandwidth Efficiency : 76% (estimated)

Performance Test Status
Run ID : VIDSLI on Monday, January 24, 2005 at 10:16:06 AM
Memory Used by Test : 512MB
NUMA Support : No
SMP Test : No
Total Test Threads : 1
SMT Test : No
Dynamic MP/MT Load Balance : No
Processor Affinity : P0
Page Size : 4kB
Use Large Memory Pages : No

Chipset 1
Model : Advanced Micro Devices (AMD) Athlon 64 / Opteron HyperTransport Technology Configuration
Front Side Bus Speed : 2x 1005MHz (2010MHz data rate)
In/Out Width : 16-bit / 16-bit
Maximum Bus Bandwidth : 8040MB/s (estimated)

Logical/Chipset 1 Memory Banks
Bank 0 : 512MB DDR-SDRAM 2.0-2-2-2CL 2CMD
Bank 1 : 512MB DDR-SDRAM 2.0-2-2-2CL 2CMD
Channels : 1
Bank Interleave : 2-way
Speed : 2x 201MHz (402MHz data rate)
Width : 128-bit
Maximum Memory Bus Bandwidth : 6432MB/s (estimated)

Features
(W)MMX Technology : Yes
SSE Technology : Yes
SSE2 Technology : Yes
SSE3 Technology : No
EMMX - Extended MMX Technology : Yes
HTT - Hyper-Threading Technology : No

 

[imported_Xan]

I was also wondering how you tell if the RAM is running in single or dual channel mode. I have "Corsair VS1GBKIT400 1GB Kit DDR400 PC3200 CAS2.5 Value Select Memory Retail" in the blue slots.

 

[Tripp17]

the major issue i see is that my ram is at 91% efficiency and yours is 77%. That'll do it everytime. Also, remeber that i am overclocking my memory by almost 100. Normally id get about 6,000. Make sure you have the ram at 1t.

 

[mrscintilla]

Originally posted by: Xan
I was also wondering how you tell if the RAM is running in single or dual channel mode. I have "Corsair VS1GBKIT400 1GB Kit DDR400 PC3200 CAS2.5 Value Select Memory Retail" in the blue slots.

Use cpu-z, it will show if it's running dual.
For comparison, my hyperx (2-3-2-6) 1G in 2 blue slots give me about 5250 in Sandra at stock speed

 

[Fent]

Odd cpu-z says dual..

 

[resStealth]

id like to know what those timings mean. and how do you overclock the ram, change the voltage in cmos?

 

[mrscintilla]

Those timings don't mean s** other than bragging rights. Do a little digging in other forums you will find an article that shows you real data about how overclocking RAM offers little very very very little benefit. Don't get caught up by the low latency and DDR9999999 hype.

 

[Fent]

Looks like most of the sandra readings say 1 channel.

Speed comes from OC'ing.

I wonder if no ram is working in dual channel on this board?

 

[mlc]

Fent...

Just built my machine this weekend... When the system starts up.. and shows the amount of memory installed.. on that same line during the post it will state "dual channel"....

 

[sreams]

Originally posted by: Fent
Looks like most of the sandra readings say 1 channel.

Speed comes from OC'ing.

I wonder if no ram is working in dual channel on this board?

Dual-channel definitely does work. Sandra memory scores nearly double when running dual channel here.

-S

 

[meolsen]

I have the same Sandra results with 4x512 on this board, running 70% efficiency. Eager to hear what people are thinking on this. Running stock speeds at ddr 400 with Mushkin 433 memory.

 

[meolsen]

bump...?

 

[imported_2x]

I had Crucial Ballistix 2x256mb and I was getting in Sandra - RAM Bandwidth Int Buff'd iSSE2 : 5325 MB/s RAM Bandwidth Float Buff'd iSSE2 : 5315 MB/s at 83% efficiency. Wanted more ram so I got Kingston HyperX LL 2 x 512 and I am getting the same scores. Ram timings are manually set to 2-2-2-5-1T. Sandra says I am running single channel at 2T, CPU-Z says I am in Dual mode at 2-2-2-5-11-16. Motherboard says I am in dual channel. In other words, there is no reliable way to know what the heck is going on. Maybe someone does?

 

[GuitarDaddy]

Sandra was developed for Pentium, and is best suited for Pentium, it reads many things wrong on AMD chips. 1 being dual channel, I guess since no pentiums are dual channel.

Sandra shows channel =1 on my setup, but all other utilities show it a dual channel. My performance and benchmarks all show numbers consistent with dual channel. Sandra shows my memory bandwidth as 6860 buff'd, I don't think a score this high would be possible on a single channel.

 

[karmadharma]

before my system toasted cpu-z showed dual channel but sandra single and it gave me about 3700 (4x512 Kingston low latency sticks, running at 2-2-2-5). Looking at the asus QVL there doesn't seem to be any CL2 stick of memory which will work in a 4-sticks configuration in dual channel: had I figured that out earlier I probably would've bought cheaper RAM.

[edit] btw, I think I have the same sticks as the original poster (KHX3200ULK2/1G), when emailing kingston about getting them to get together with Asus to support them in dual channel with 4 sticks that's what I was told:

========
All Kingston memory is built to meet or exceed the specifications for each
motherboard. HyperX memory (which you are using) is high performance memory
that is geared toward users who overclock their systems. It is not tested
or guaranteed to work in any specific motherboard.
......
the HyperX memory is configured to a "failsafe" setting which
usually is compatible with most boards at factory default. We do not have
tested settings for any specific motherboard.
.......
You will not
achieve maximum performance from the hyperX series unless you are
over-clocking the motherboard. If you are not over-clocking, then it is
recommended that you use the Valueram series.
........
The recommended part # for your motherboard is KVR400X64C3A/1G however I
do see in our database that engineering is still researching support for
your motherboard
.........
Dual Channel is a feature of the motherboard. The only thing required of
the memory is that they be identical modules. If the memory is purchased in
kits then it is guaranteed to be identical modules. If the memory does not
come in a kit, then it cannot be guaranteed to be identical. At that point,
if the modules are identical, and the memory does not run in dual channel,
that is an issue with the motherboard and it's manufacturer.

 

[imported_2x]

Fent-

Looking at your Sandra scores jarred my memory this time. When I had my 2 HyperX set on Full auto in the bios, it ran around 4800-4900 like yours. I went to the Bios, Changed the settings to (from top to bottom) to
Auto-1T-2-2-5-2-disable and that notched me up to 5250/5250 with 83% efficiency. Tonight I am gonna try settings of DDR400-1T-2-2-10-2-disable in the dram menu because I read an article that A64 love the Tras at 10. Hopefully that will push me up to the 6000's like it should be.

 

[Thorsson]

1. Memtest86 will give a better reading on your memory bandwidth than Sandra and is also a good way to test your memory. It also saves booting into Windows each time. Use test 5 for a quick test.

2. The timings do make a difference, but not always in the way you think. Anyway get A64 tweaker - this will show you what the auto settings are. You can also tweak you memory in Windows, but will need to use the BIOS to make the changes permanent. I changed my settings to 2-7-9-2-2-5-2-2-2-2, the biggest benefit coming from lowering TRTW (or TRWT as it is in the BIOS). My bandwidth improved 11% at stock. Changing TRAS between 5 and 10/11 made little difference (5 was marginally faster). Lots of people were claiming that 2-2-2-10/11 was faster than 2-2-2-5. Just goes to show that they only way to know is to test it yourself.

3. RTFM folks. NF4/NF3 chipsets run 4 DIMMs at 333 - they don't support 400. If you force it chances are you'll crash and burn.

4. QVLs are always out of date - you just cannot go by them. They'd have you running PC2100...