View Full Version : should i go dual channel ddr support? ...
chaego
May 24th, 2005, 11:04 AM
hi im running a althon xp 2500+ on a asus a7n8x-x mother board with 1.5gig ddr pc3200 ram ...that's 3 stick of 512 ddr pc3200 ram
...unfortunately my mother board doesn't support dual channel ddr....so im thinking of upgrading to a motherboard that does but right now im tight on budget
so my question is ...is it worth it? does it make any significant difference that shows if i were to use a mother board with dual channel ddr support?
if not i rather just stick with what i have and save the cash for something else
here is my system specs
Athon Xp 2500
Asus A7N8X-X mobo
1.5gig DDR 3200 / 400mhz
geforce 6600GT 128mb
120gig hd
windows xp sp2
so what you guys think should i upgrade the mobo? ...meaning to do it just for better gaming experience
willy
May 24th, 2005, 11:18 AM
The answer is no. Keep your system the way it is now and make the big leap later.
With 1.5GB RAM already in place, it appears the CPU will be the first bottleneck. Unforunately, there is very little headroom for socket-A CPU upgrade from your 2500+ and the cost nowadays is inflated somewhat.
PennyArcade
May 24th, 2005, 11:23 AM
No.
Dual channel would not increase your performance all that much. It is better to wait until you have more money and get a mobo and cpu together.
Carnage
May 24th, 2005, 11:34 AM
Dual channel has next to no performance gains for AMDs.
The only CPUs that it's useful for are Athalon64 and P4.
And even then it's not a huge speed boost. Don't go out of your way to upgrade to dual channel, you won't notice the difference.
Amourek
May 24th, 2005, 11:40 AM
Save your cash for a real upgrade. i.e. A64 S939
chaego
May 24th, 2005, 12:41 PM
cool thanks alot guys
Cafe_333
May 25th, 2005, 11:36 AM
:eek: ?!?!?!
Didn't we have this discussion before?? Dual channel *is* the way to go. You'll get 20-25% system performance gains, period. That's huge. Why else was this technology so widely adopted, preferred, and accepted? When this technology was first introduced, it made AMD mobos SO popular and faster forcing all Intel boards to eventually adopt this technology as well. Nevermind what anyone else tells you otherwise. While you will get little to no gains in real world benchmarks, who really cares about that anyways? Most RFDers are not limited to just real world benchmarks.
And your mobo is an NF2 chipset mobo. It already does support dual channel. Just stick two similar bars of 512's in the blue slots and you're set. You cannot use all 3 dimms in conjunction with dual channel. Then stick your remaining bar into another rig or sell it. I would even say dual channel is faster than single channel with more ram - but of course it depends what you will be using it for. If you only do typical apps and surfing, then you won't see any benefits. Anything beyond this including games, you WILL see gains.
EDIT: just found out your mobo is actually not dual channel. read below posts.
Fightguard
May 25th, 2005, 11:45 AM
And your mobo is an NF2 chipset mobo. It already does support dual channel.
Asus A7N8X-X uses the nForce2 400 chipset (not Ultra 400), it's single-channel only. I'll have to side with the others here, a mobo upgrade isn't worthwhile just for the sake of dual-channel...make a bigger leap down the road.
Cafe_333
May 25th, 2005, 11:48 AM
ah... i just saw nf2 but didn't read closely into it. so it actually is a single channel mobo. My opinion was based on the presumption that he could go dual right away. Well in light of this, I do agree that making a big leap later is the better option, especially a mobo upgrade just for the sake of dual channel, is definately not worth it. I do stand by that dual channel is superior and faster than single. Definately something to consider for the future.
Carnage
May 25th, 2005, 11:56 AM
:eek: ?!?!?!
Didn't we have this discussion before?? Dual channel *is* the way to go. You'll get 20-25% system performance gains, period. That's huge.
The performance gains are not that high. Unless you're talking marketing hype.
And it's been proven before that AMD(with the exceptino of the 64) can't take advantage of the increased memory bandwidth. Who cares if the memory is faster if you can't use it.
It's like redlining your car between stoplights.l You'll get there first, but the car beside you doing the speed limit will still catch you at every light.
scottrempel
May 25th, 2005, 11:56 AM
:eek: ?!?!?!
Didn't we have this discussion before?? Dual channel *is* the way to go. You'll get 20-25% system performance gains, period. That's huge. Why else was this technology so widely adopted, preferred, and accepted? When this technology was first introduced, it made AMD mobos SO popular and faster forcing all Intel boards to eventually adopt this technology as well. Nevermind what anyone else tells you otherwise. While you will get little to no gains in real world benchmarks, who really cares about that anyways? Most RFDers are not limited to just real world benchmarks.
And your mobo is an NF2 chipset mobo. It already does support dual channel. Just stick two similar bars of 512's in the blue slots and you're set. You cannot use all 3 dimms in conjunction with dual channel. Then stick your remaining bar into another rig or sell it. I would even say dual channel is faster than single channel with more ram - but of course it depends what you will be using it for. If you only do typical apps and surfing, then you won't see any benefits. Anything beyond this including games, you WILL see gains.
LOL...I don't think you said one thing right in that entire post!
- Performance gains are very minimal. The only area DC helped on NF2 was with the IGP. NF2 just can not make use of the extra bandwidth. Intel is starved for bandwidth.
- As already mentioned, the -X left off DC support. Why? Saves money and really doesn't have a benefit. DC doesn't help the S939 platform much either, which is why S754's performed similar. DC on AMD's is more of a marketing gimick then anything.
- You are also wrong about not being able to use DC with all 3 512MB sticks. If he had a DC board, he would be running DC...it just would not be perfectly matched, but no performance would be lost.
Would you like to try again? :lol:
Cafe_333
May 25th, 2005, 12:03 PM
Like i said, the gains are minimal to none in real world benchmarks. This same point was brought up to my attention in the last discussion too. So with that respect you guys are right - this I have not disputed, so why are you all picking a fight with me on that????? but beyond this there ARE gains - wherein lies my point. No one, I repeat no one, can argue that you get absolutely no gains going dual channel. You *cannot* tell me that going dual channel does nothing. You *will* get gains any way you want to look at it - it's just a question of wether or not you can make use of the gains it's situated for. And my point about not being able to use all 3 sticks was in regards to an NF2 dc mobo - so get your facts straight. I am correct in that statement as well.
Dual channel is a proven technology. Not just marketing hype. No manufacturer would put forth a technology that didn't work. And your analogy with the cars seems contradictory to your position. Doesn't getting from point A to point B faster just glorifies my point?
scottrempel
May 25th, 2005, 12:39 PM
And my point about not being able to use all 3 sticks was in regards to an NF2 mobo - so get your facts straight. I am correct in that statement as well.
I was talking about NF2 mobos...and you are still wrong. So, get your facts straight!
I never said there were no gains. I said they were very minimal.
You said "You'll get 20-25% system performance gains, period. " Please note the "period".
Please show me where you get these amazing gains...I was not participating in this other discussion thread you speak of, afaik. The way you make it sound, there must be several areas where the DC gives 20-25%...so list some so I can go do my research.
Carnage
May 25th, 2005, 12:55 PM
I was talking about NF2 mobos...and you are still wrong. So, get your facts straight!
I never said there were no gains. I said they were very minimal.
You said "You'll get 20-25% system performance gains, period. " Please note the "period".
Please show me where you get these amazing gains...I was not participating in this other discussion thread you speak of, afaik. The way you make it sound, there must be several areas where the DC gives 20-25%...so list some so I can go do my research.
I think he was trying to say that the 20-25% performance boost is only doable in non-real life benchmarks.
Which makes the entire point meaningless seeing as any normal person wouldn't make use of said performance boost.
Cafe_333
May 25th, 2005, 01:03 PM
Then please, point me to any link of any NF2 mobo that can use *all* 3 dimm slots for dual channel. I have never heard of any nf2 dc mobo being able to do this, which is why the dimms for dual channel come in different colors on these boards. My nf2 mobo certainly can not - and the manual clearly states it as well. If you can then I stand corrected but not refuted - as not all nf2 mobos can use all 3 dimms to dual channel.
As for the 20-25% performance, I don't want to do the research for you, it was a long time ago when I did mine. What I will do however is refer to this THG article from the nf2 era:
"The benchmark results prove that, with its Athon XP 2800+ plus the nForce2 chipset, AMD can take back the performance crown (Personal Addition: this was pitted up against P4's 2.8GHz flagship of the time). Above all, the CPU can take full advantage of the higher bandwidth of the nForce2 chipset with Dual DDR333. This offers a theoretical bandwidth of 6.4 GB/s (PC3200 x 2).
Furthermore, note that the Front Side Bus clock and memory clock now communicate synchronously. Here, we refer to our lab findings regarding the new 166 MHz FSB clock: when the currently fastest VIA KT333 chipset is used (KT400 is much slower than KT333!), the higher FSB clock results in an increase of just around 5% in the total performance. But it's a different story when Nvidia's nForce 2 enters the scene: here, the performance increases by 20% to 25%.
Nevertheless, the performance of Dual-Channel DDR333 is convincing because it brings system performance up to the level of a top P4 model."
http://www.tomshardware.com/cpu/20021001/xp_2800-23.html
Cafe_333
May 25th, 2005, 01:39 PM
That goes back to the question to raid 0 or not - because real world benchmarks have shown too that you would see little or no gains. But one thing you need to keep in mind about these tests: while they show there is no real benchmarks while you are running real world apps like MS Word, there are definately gains to be said in other areas. Obviously there is no gains while you are in MS Word when it comes to reading and writing to harddrives - or memory in this case. So obviously... if you are limiting benchmarks to situations where faster speed is not designed for, then you are obviously not going to see any real gains. These tests are only used to inform users of the limitations of dual channels and what they should really be setting their expectations for. Here lies the popular 5%, or the 'hardly little to no gains' that many users talk about with dual channel. Everything else, the gains are 20-25%
Dual channel is as much of a proven technology as Raid 0 can give performance gains. In single channel mode, data is written to the first bar of ram before it gets to the second bar of ram. In dual channel mode, data is written to both rams in synchronous mode.
I think he was trying to say that the 20-25% performance boost is only doable in non-real life benchmarks.
Which makes the entire point meaningless seeing as any normal person wouldn't make use of said performance boost.
scottrempel
May 25th, 2005, 01:57 PM
The THG stuff you quoted does not say anything about NF2 DC compared to NF2 SC...but rather was comparing the NF2 to other platforms. I'm also sure a lot more was understood after such an early article was printed. If that is where you are getting 20-25% better then it really means nothing. If you have some form of example where a certain application or something performs 20-25% better then let's see it.
I'm not saying don't run DC...it's there...might as well use it. I'm just contesting this 20-25% percent stuff...there are tons of tests out there showing the typical 5% gain, but I haven't seen any showing that DC on NF2 is 25% faster then SC on NF2...that's all I want to see.
And since you still haven't shown me anything, what makes you think I am going to bother showing you that DC can function with 3 sticks? If you want see for yourself go read the many discussions over at nforcershq.com.
ShadowVlican
May 25th, 2005, 02:31 PM
That goes back to the question to raid 0 or not - because real world benchmarks have shown too that you would see little or no gains. But one thing you need to keep in mind about these tests: while they show there is no real benchmarks while you are running real world apps like MS Word, there are definately gains to be said in other areas. Obviously there is no gains while you are in MS Word when it comes to reading and writing to harddrives - or memory in this case. So obviously... if you are limiting benchmarks to situations where faster speed is not designed for, then you are obviously not going to see any real gains. These tests are only used to inform users of the limitations of dual channels and what they should really be setting their expectations for. Here lies the popular 5%, or the 'hardly little to no gains' that many users talk about with dual channel. Everything else, the gains are 20-25%
Dual channel is as much of a proven technology as Raid 0 can give performance gains. In single channel mode, data is written to the first bar of ram before it gets to the second bar of ram. In dual channel mode, data is written to both rams in synchronous mode.
great analogy to raid-0... i just don't understand why people use it when there's little benefit and double the normal risk of losing your data
back to dc-ddr... THG isn't the stuff i'd be reading to learn about anything... imo their reviews are horrible...
dual channel doesn't bring anything close to 20% performance increase, probably for memory BENCHMARKS, but is that all you're gonna run? :lol: you're right about the IGP though, helps me more than it helps normal nf2
i'm still on the AthlonXP platform, and i can say that dual channel isn't worth buying a new mobo, but if it's there, might as well use it right? :)
oh, and you can run DC with 3 sticks, but i'm sure it has to balance like this for example:
channel 1: 512MB
channel 2: 256+256MB
scottrempel
May 25th, 2005, 03:16 PM
oh, and you can run DC with 3 sticks, but i'm sure it has to balance like this for example:
channel 1: 512MB
channel 2: 256+256MB
It does not have to balance either actually...although the example you gave would be optimal. If you have 3x512MB, for example, you will still get dual channel, but 512MB of the memory in the one channel will not really be operating in DC....the rest will.
Carnage
May 25th, 2005, 03:35 PM
Dual channel is as much of a proven technology as Raid 0 can give performance gains. In single channel mode, data is written to the first bar of ram before it gets to the second bar of ram. In dual channel mode, data is written to both rams in synchronous mode.
I accept that dual channel is a proven technology, and that it does give performance boosts(though how much is debatable).
My only argument is that the performance gains are not enough to make it worth an upgrade simply to go to DC. If upgrading your system, definetly go DC there's not really much reason not to. But I wouldn't suggest someone buy more ram just to switch to DC.
Just like I wouldn't suggest someone upgrade their hard drive to SATA, just to get SATA.
ShadowVlican
May 25th, 2005, 07:11 PM
It does not have to balance either actually...although the example you gave would be optimal. If you have 3x512MB, for example, you will still get dual channel, but 512MB of the memory in the one channel will not really be operating in DC....the rest will.
ah. thanks for telling me that, i was not aware dual channel can operate with unbalanced channels :)
audit13
May 25th, 2005, 09:55 PM
I'm using 2 x 256mb and 1 x 512mb ddr ram (all pc2700) on an Abit NF7-s and the boot screen says that dual channel is enabled using all 3 slots. The 256mb sticks are installed in the slots farthest away from the cpu and the 512mb stick is installed in the slot closest to the cpu. Is the bios wrong ? Is there a program that will test for dual channel mode ?
Cafe_333
May 26th, 2005, 01:32 AM
I have to state that my argument was based on that 3x512mb would not have all 3 dimms running in DC mode, and it doesn't. But yes if the channels were balanced as ShadowVlican puts it then it would be fine.
audit13, your channels are balanced, it's working fine.
Carnage: I agree with you as I stated in an earlier post, upgrading just to go DC is not worth it, but something to consider for the future when overhauling his entire rig. My posting was only refuting your comment that my point was enitrely useless, which I don't believe it is. ;)
Cafe_333
May 26th, 2005, 01:34 AM
We've hijacked this thread completely, :D So I've decided to continue this in a new thread:
http://forums.redflagdeals.com/showthread.php?t=164899
Please feel free to comment and post your thoughts in this ongoing informative debate!!