• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

DC Question

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Power Unknown

Registered
Joined
Jun 23, 2003
i was thinking maybe we are going about the amd DC thing all the wrong way. we all have seen the tests with DC inabled and it gives only a 2-5% increase in performance. and since the athlon can't take advantage of the doubling of bandwith why not just buy ddr2100 ( or even 1600 if your only going to run at 400FSB) and just buy bigger sticks.

cuz at http://www.pricewatch.com/ a 1GB of ddr 2100 (1 dimm) from micron is $169 and if i get 2 of them it would cost $338 for 2gb of ram.

this was just a thought late at night after i turned my pc off. so i figured i would remeber it for the morning . i am not a DC basher but i am a AMD fan for life and this is mostlikely one more reason to be at camp amd.

amd always the price/performance ratio king
 
if i'm not mistake DC = Dual Channel as in the new Nforce 2 mobos, where if you use 2 sticks of memory in certain spots it make's most of your benches like 2 - 5% faster....
i believe it changes the memory from 64 bit to 128 bit?
am i correct?
 
There is no way you can run 400FSB with 2100 ram. Even 2700 would be a case of shear luck. I don't even consider the "DC" issue because 2-5% is not visible except in benchmarks and minimal even there. I believe certain innovations are simply to convince us to upgrade.
 
hold on lets see if this will help. you have timing setups in your bios for your ram. my ram is at 7/10 because it's 333mhz not 400. i would run it at 1/2 because that is what it would be set up to run as and i figured since is was DC it would run as if it were twice the speed
266mhz ram x 2 = 533mhz.

the mobo i wanted to get is capible of getting up to 250mhz or 500fsb.

i hope this helps at what i was trying to ask.
 
I think it has been reported by c627627 that (based on these analysis) PC2100 153 MHz in dual channel with FSB at 204 MHz (ratio 3:4), is better than running memory and fsb in 1:1 (SYNC) at lower FSB dictated by memory (around 153 MHz).

Link:
Dual Channel Memory (1 vs 2 sticks, sync vs async)

hitechjb1 said:
Dual Channel Memory (1 vs 2 sticks, sync vs async)

Many would say sync timing gives better performance in most MB, and dual channel is preferred for nforce2/P4 motherboard (MB). I think it's a bit more complicated, and depends on the memory speed (RAMDDR in MHz), compared to the FSB speed (MHz).

Assuming you found the top speed of the FSB, memory and processor of your system. AGP/PCI speed are assumed locked.

1. Memory speed is equal or higher than FSB. The max BW is limited by the FSB.

E.g. the MB can run at 200 MHz FSB, RAM is PC3200 (DDR 400). So you will get a max memory bandwidth (BW) of 3200 Mb/s. Actual BW ~ 3050 Mb/s

1a) SYNC with 1 stick: Can go a few MHz higher and tighter timing, may get 2-3 % better BW.

2a) SYNC with 2 sticks: May get 2-3% better BW (less overhead). Difference of 1a) or 2a) are insignificant.

For current AMD MB, there is little advantage to do dual channel since the FSB with SYNC memory can only take the bandwidth of 1 memory stick, unless you are using the integrated video. But a very different story when running in ASYNC mode with slower memory.

2. Memory speed is slower than FSB (e.g. running P4 with dual speed FSB, or reuse older/cheaper memory), so FSB can absorb most of the dual channel memory BW.

E.g. FSB 200 MHz, ASYNC 84% dual channel using PC2700 (DDR333), it will give a boost in memory bandwidth of 10% compared to running SYNC at 166 MHz. It is like running the RAM at 166 * 1.1 = 182 MHz. This is 92% maxBW w/ slower memory !!!

E.g. FSB 200 MHz, ASYNC 75% dual channel using PC2400 (DDR300), it will give a boost in memory bandwidth of 18% compared to running SYNC at 150 MHz. It is like running the RAM at 150 * 1.18 = 177 MHz. This is 87% maxBW. Best price performance tradeoff !!!

E.g. FSB 200 MHz, ASYNC 66% dual channel using PC2100 (DDR266), it will give a boost in memory bandwidth of 28% compared to running SYNC at 133 MHz. It's like running the RAM at 133 * 1.28 = 170 MHz. This is 83% maxBW. Very economical !!!

E.g. FSB 200 MHz, ASYNC 50% dual channel using PC1600 (DDR200), it will give a boost in memory bandwidth of 55% !!! compared to running SYNC at 100 MHz. It's like running the RAM at 100 * 1.55 = 155 MHz. This is 78% maxBW with slowest memory.

Dual channel can increase memory bandwidth up to 50% when the FSB data rate is 50-100% faster than the memory data rate.

Dual channel or single channel mode in nforce2 mb is not that crucial for overall performance when fsb and memory at running a 1:1 speed (or data rate). The difference is few % (for the same FSB, say 2-3%, maybe 5% for new NB stepping ?). Also single channel may let FSB to go a few MHz higher due to a smaller chance of potential dual dimm mismatch and memory controller stress at high FSB, I think. On the other hand, dual channel memory controller provides some performance advantage due to its intrinsic speculative caching capability. At this point, the little higher FSB from single channel offset the performance advantage of dual channel, and the two is within few % (depends on overclocked FSB speed, motherboard, memory modules, NB stepping, applications), I think, for AMD mb. For some nforce2 mb that have integrated video which can benefit from twice the nforce2 memory bandwidth, since the bus between the video and the memory controller has 2x64 bit bus.

In summary, for AMD MB, if the RAM and FSB are comparable in data rate, running them in SYNC mode will deliver highest bandwidth determined by the FSB (8 * 2 * FSB Mb/s).

For slower memory (reuse older memory) running AMD MB, or Intel P4 whose FSB is twice as fast, the FSB can absorb the dual channel memory bandwidth. Running dual channel would deliver much higher BW.
 
Last edited:
Please note that on a challenge from Mustanley, additional tests were conducted on page 2 of the "FSB don't mean s**t" thread:
http://forum.oc-forums.com/vb/showthread.php?s=&threadid=215500&perpage=30&pagenumber=2

where the following comments were seriously taken into account:
Gautam said:
The conclusions that I draw here are:

204 async will do better in certain focused benchmarks, but
153 in-sync will do better in real-world performance.
+
james.miller said:
Why stress the northbridge when you dont have to?

...causing me to drop the 204 FSB @ 75% memory frequency to 152 FSB @ 100%:

2100+ AIUHB 0248
256MB Crucial PC2100 + 2x 256MB Kingston (Hynix chip) PC2100 @ 6 3 3 2
[152] FSB x 15 = 2280 MHz @ 1.8 Vcore with memory frequency at 100% [152] @ 2.77 Vdimm
37(min) C to 46(max) C Winter
42(min) C to 50(max) C Summer
Epox 8RDA+
Thermalright SK-7 with variable speed 80x25mm YS-Tech FD1281259B-2F
BFG GeForce4 Ti4200 8X 128MB; Antec SX-835II case ; 350W Antec SmartPower SL350
 
better yet i should have asked will single channel ddr 400 perform faster than dual channel ddr 200? so lets say

ddr400 with 512mb in SC vs ddr200 with 256mb x 2 in DC mode.

therodically they are both at 3.2 GBits BW.
 
Back