A7n266-C memory

Question... I'm gonna be buying the Asus A7N266-C to replace this POS ECS K7S5A, and I have a question about the memory. Now, to get the best performance, or to enable dual channel, will I need to run with 2 DIMMs, cuz all I got now is 1 256 MB DIMM. Also, if I should get another DIMM to enable the dual-channel memory, will I need to get another 256 MB, or could I go with just another 128 MB? Or, was having 2 DIMMs on the nForce only for the integrated graphics? Thanks for any info.

"Trying is the first step towards failure."
21 answers Last reply
More about a7n266 memory
  1. To the best of my knowledge (and I believe I also read it on Tom's site) you need to use two exact sized DIMMS on the board to take advantage of the advanced memory architecture.

    I am thinking of getting this board also. Where have you seen it? And does the board you're looking at have the onboard audio?

    Brian

    What did we do before we had computers?
  2. If you look on pricewatch.com you'll see some for $110 without audio, and some for $120 with audio. I wanna get the one with audio..

    "Trying is the first step towards failure."
  3. you can use either both the channels or a single channel. using a DIMM in both channels will get you better performance due to the Twinbank memory architecture, which is especially useful with integrated display (the GeFore2MX)the nVidia chipset 420D has.

    but your board A7N266-C has a 415D chipset that does not have a integrated display, so you would be fine with a single DIMM. its recommended that you DIMMs be of the same size, preferably same parameters.

    girish

    <font color=red>Nothing is fool-proof. Fools are Ingenious!</font color=red>
  4. I have just build a system around the ASUS A7N266E this board has both the 5.1 Dolby sound and the GF2mx video. According to the memory configuation table in the manual you can have differernt size DIMMS installed. I currently am running only 1 256MB Dimm and can see no real problems with the performance.

    My only complaint is that WinAmp 7.1 does not work well at all for MP3s or CDs. When I installed a 2nd stick of 256MB the problem DID NOT go away. Real Player works fine so I am not sure what is wrong with WinAmp (other then M$ of course).

    Sure wish I would have bought a couple of extra sticks of the Crucial memory last Oct when it was $28 for 256MB!
  5. I do not agree with what you say here. I think it is very usefull to use two dimms instead of one, even if you are not using the integrated mx2. I am sure I read a lot of reviews about the nForce chipset that were quite clear about this. And most of them didn't use the integrated video in the IGP but used a GF3-500 instead. If I 've got the time, I'll post some links to show you this.

    BTW, a friend of mine has also build a nForce-based system. Now he is running with only one DIMM, since the second high-quality Apacer memory module was broken. He will be able to see the difference between the two configurations when he gets his second module. BTW, he does not really use the integrated video, the only thing he is interested in is MPEG-4 encoding. So that virtually eliminates the influence of the IGP to the configuration. Maybe I'll post some benchmark results, too.

    One question, though: I don't know who posted this, but the person in this thread who had an nForce-sytem running with the mx2 in use, did you happen to run 3dmark 2001? My friend only managed to get something close to 2000 out of it. What is your score? He's using a XP 1900+.

    Greetz,
    bikeman

    EDIT: Hmm... I just read at AnandTech, where they were talking about the nForce 620/615, were they said performance differences between the 420 and 220 (single channel version, that is) is minimal. But the link they provided <A HREF="http://www.anandtech.com/chipsets/showdoc.html?i=1535&p=6" target="_new">this one</A>, actually), and the pages that follows, prove that there IS a difference. But the others test give no better performance when comparing the 220 to the 420. Hmmm ... And I really was convinced of the fact that the 420 outperformed the 220 ... I'll go and look some further ...

    <i>Then again, that's just my opinion</i><P ID="edit"><FONT SIZE=-1><EM>Edited by bikeman on 03/14/02 09:24 PM.</EM></FONT></P>
  6. Ok, look at <A HREF="http://www.amdworld.co.uk/nfmemory.htm" target="_new">this</A>. It's quite clear, no? I guess both AnandTech and Tomshardware made wrong memory configurations when testing the platforms.

    I rest my case.

    Bikeman

    <i>Then again, that's just my opinion</i>
  7. well, i guess we were talking about different aspects of the memory architecture.

    i said if the integrated display was present, the bandwidth would be shared and that would decrease the performance. with TWO DIMMs in two channels this performance decrease is not as much since the system can always read/write to/from the otehr channel and contention occurs only if the system neds to access the same channel where the display keeps its memory. this, is because the dual-channel memory controller can operate independently with each channel.

    now when you put in two DIMMS with a 415D chipset which does not have integrated display, with 128 bit mode you simply get a double memory bandwidth, equivalent of DDR533! and that improves the performance a lot.

    I agree using a double DIMM would benefit on a non-IGP board, but a single one at least wont hurt performance that did with the IGP.

    girish

    <font color=red>Nothing is fool-proof. Fools are Ingenious!</font color=red>
  8. nVidia just announced an update to the nForce chipset, the 620-D and the 615-D. Both of which will support PC2700 memory.

    However, according to AcesHardware, "the CPU is still bottlenecked by its 2.1 GB/s FSB".

    Would this mean the PC2700 memory is run asynchronously? What good is that other than speeding up horrendously slow onboard video?

    See AcesHardware article <A HREF="http://www.aceshardware.com/#55000466" target="_new">here</A>.

    <b>We are all beta testers!</b>
  9. I guess the other thread will answer your questions. I would like to say something, but I'll first check the other thread.

    Hehe ... yet another useless post to pump up my post-count :wink:

    Greetz,
    Bikeman

    <i>Then again, that's just my opinion</i>
  10. Oops, I missed your mention of the new nForce chipsets but at least I contributed a new link and some thoughts for the discussion.

    I also wish to add that I thought there was/is very little benefit with the nVidia memory architecture when onboard video is not used. The reason being, that even if the full 4.1 GB/sec memory bandwidth is available (and I'm not sure it is) when onboard video isn't used, system performance is bound by the fact that FSB is restricted 2.1 GB/sec. At 133 mhz and being DDR that's the limit of the bus regardless how fast the memory subsystem. If the bus is limited to 2.1 GB/sec then the memory can't be accessed any faster.
    <b>We are all beta testers!</b><P ID="edit"><FONT SIZE=-1><EM>Edited by phsstpok on 03/15/02 05:07 PM.</EM></FONT></P>
  11. yes, the processor is still limited to 2.1 GB/sec bandwidth.., unless you overclock it!

    thats the idea! i have said in a number of my posts about the FSB and core speed relation in overclocking processors section, you will have to overclock the FSB to get better memory performance. since the memory runs asynchronously (effectively *decoupled* from the CPU by the chipset) all you need to match is the processor bus bandwidth and the system memory bandwidth. so if you are using a 4.2 GB/sec dual channel DDR266, or 2.7 GB/sec DDR333, all you have to do is to increase the processor FSB so that it can actually make use of that available bandwidth. if integrated display is present, you may keep it to about 25% of the memory.

    and lastly, you also need to ramp up the core speed so that the processor is fast enough to use that heavy inflow of bytes into it! there needs to be a balance, everywhere in life, so it does between the FSB and core speed!

    girish

    <font color=red>Nothing is fool-proof. Fools are Ingenious!</font color=red>
  12. Quote:
    the processor is still limited to 2.1 GB/sec bandwidth.., unless you overclock it!

    Yes, I understand this. Until, AMD comes out with a 300mhz processor or a 333mhz processor we will have to overclock in order to utilize PC2400 and PC2700 memory. However I think there is a difference between when a motherboard officially supports the higher FSB speeds and when it does not. Knowing that Athlons already do quite well with high FSB speeds if I were going to run one at 166mhz I would prefer it if the motherboard officially supported this speed. I mean you may run some KT266A boards at these speeds but stability is far from guaranteed. So a chipset that officially support this speed combined with memory (PC2700) which also supports this speed should (and I say "should") improve one's chances for high speed <b>WITH</b> stability.

    <b>I have so many cookies I now have a FAT problem!</b>
  13. FSB overclocked systems arent that unstable, in fact a succesful overclock means a stable systems, and there are a lot of successfully overclocked systems!

    one doesnt have to wait for AMD releasing official 300 or 333 MHz processors, if they work well they can be used so. I call it level 1 of overclocking. all things in place, at stock settings, if we exploit just the safety margin the manufacturer keeps while binning the processor, its free. higher levels of overclock need tweaking in core voltage or better cooling gear etc.

    as a matter of fact, a lot of boards like the Iwill XP333 and A7V266E and KR7 are guaranteed for DDR333 operation, and as ong as the processor supports it, i dont see any reason why we should keep ourselves away from it! I would prefer the ALi board, A7A266E is good, but XP333 is better, which is most compatile and stable chipsets, just lags behind a bit in performance.

    girish

    <font color=red>Nothing is fool-proof. Fools are Ingenious!</font color=red>
  14. I guess you are right. If one is only overclocking FSB (and not the processor) then a system should be pretty stable with ordinary cooling.

    However, I wonder how many 133mhz boards can truly run stable at 166mhz FSB with just basic cooling (even if the CPU is not overclocked). Maybe I'm wrong but I just don't think enough of the boards with 133mhz (266 DDR) chipset can do this, not with basic cooling. However, if a chipset is designed for 166mhz FSB I really think it is a reasonable expectation that any customer should be able to run at the rated speed without the need for advanced cooling.

    Another expectation I have is that if a motherboard has overclocking options then that board should actually be able to overclock. There should be plenty of margin designed into board. I mean what good is an overclocking board that can only overclock from 133 mhz to 136 mhz (for example). If the design can't handle overclocking then don't bother putting in the options in the first place. Of course, this is just my personal opinion.

    <b>I have so many cookies I now have a FAT problem!</b>
  15. well, the safety margins, as high as 20~30% are present in almost all products, and they can be pushed further by voltage tweaks and improving the cooling!

    Some boards arent truely overclocking boards, but they just provide some options, like FSB increments in steps of 3~8 MHz etc. They sometimes dont even provide voltage settings!
    But there are a good number of true overclocking boards that have all the tools you need for a good overclock. Considering almost all the hardware can run as much as 25% faster at stock settings (which may be limited by a single crap part like cheap memory or a poor cooler) its no surprise almost all of these boards do well at 166 MHz. I mean they are stable, performance is not the issue here.

    In fact, some boards like the Iwill XP333/R and A7V266E have been run at 200 MHz or higher FSB (DDR400) more than once, I found one on <A HREF="http://www.lostcircuits.com" target="_new">http://www.lostcircuits.com</A> and a couple of others places I dont remember.

    To sum up, there is a distinct difference between what is a good overclocking board and what is not, namely:
    1. adjustable core voltage
    2. stepless frequency selection (preferably in 1 MHz steps)
    3. preferably jumper settings for multiplier and FSB
    4. higher PCI and AGP dividers
    5. extra FAN headers
    6. preferably CPU thermal protection circuit

    any board lacking even one out of first four features is not a overclocking board, although one could give it to a board that doesnt suppor higher PCI dividers, which is important only after certain value of FSB.

    girish

    <font color=red>Nothing is fool-proof. Fools are Ingenious!</font color=red>
  16. It looks like we are just going to argue the point back and forth so maybe we should get back to the main thing, does Ksoth need 2 DIMMS for his A7N266-C?

    I'm still under the impression that no extra bandwidth is available when onboard video is not used. I mean I thought I read that lightspeed architecture, itself, is not available when an AGP video card is used (and I'm far from certain of this). Even if I'm wrong about that I still don't see how having a single DIMM, which would only have 2.1 GB/sec of bandwidth, would be any hinderance because with FSB at 133 mhz then the CPU is limited by an equal 2.1 GB/sec bandwidth. Increasing FSB doesn't help, relatively speaking, because the bandwidth of one DIMM will always match the bandwidth of the bus.

    Unless I'm mis-understanding something, 1 DIMM is only a problem when onboard video is used. This is because both the Video and the CPU share that 2.1 GB/sec bandwidth in that case. When 2 DIMMs are installed then the onboard video utilizes memory on DIMM1 and the CPU is free to access part of the bandwidth of that DIMM and all of the bandwidth of the other DIMM. The problem is total CPU bandwidth is still limited by FSB, still 2.1 GB/sec. I just don't see how anything is gained by the memory architecure.

    Performance can only be lost as described above. It can't be gained (relative to more conventional DDR architecutures).

    <b>I have so many cookies I now have a FAT problem!</b><P ID="edit"><FONT SIZE=-1><EM>Edited by phsstpok on 03/20/02 03:37 AM.</EM></FONT></P>
  17. yes, back to the point!

    For the A7N266-C, with the 415D chipset, there is simply nothing to share the memory bandwidth. Its a well known fact that the onboard display eats into as much as 20~25% of the available memory bandwidth, which is now available to the processor. The point is, using an integrated display should gain much by using two DIMMs rather than a single DIMM, if its not used, one DIMM or two DIMMs it wont make much of a difference.

    Now this is just a off-hand speculation. Now, with 2 DIMMs, its still the same 2.1 GB/sec bandwidth on the processor bus. BUT, if 128 bit mode is enabled, (the nVidia IGP/SPP docs clearly mention the two operating modes) where the memory bus effectively becomes 128 bits wide, or twice the FSB! A sort of double channel double pumped P4 like bus! No wonder it provides the same 4.2 GB/sec bandwidth, which will obviously increase by increasing the FSB!!

    But the question is, will the CPU really make use of this immense bandwidth available to it? The answer is simple, NO. For one, its FSB is limited to 2.1 GB/sec and it cannot exceed it. Second, the core is running at a speed that 4.2 GB/sec is simply too overwhelming.

    In this case, if the board permits, one could leave the memory running at 128 bit mode, 133 MHz (or even lower with CPU:mem ratio of 133:100, since the memory bandwidth will reduce but still be 3.2 GB/sec which more than 2.1 GB/sec!) and increase the processor FSB so that its bandwidth could be increased closer to what memory is offering! It wont equal it though since the memory bandwidth will increase with increasein FSB, but by 3/4 times.

    This will protect your memory, no risk of running it too high, and whatever overclocking is done, its done to the processor. Even the PCI and AGP clocks are locked on A7N266 boards. I guess that now the bandwidth is available, one could go for higher core speeds of the processor, acheiving it by running it at higher FSB! Perhaps, the quest for better memory is over, who needs a DDR333 or DDR400 memory?

    Will check the A7N266 manuals though, for what settings are available.

    girish

    <font color=red>Nothing is fool-proof. Fools are Ingenious!</font color=red>
  18. I am planning to do a test with 2x128 MB PC2100 DIMMs and 1x256 MB DIMM on the A7N266 board with and without onboard display, to study the impact of dual channels and that of 128 bit access mode, just waiting for some funds. If somebody has already done it, please point me to it.

    girish

    <font color=red>Nothing is fool-proof. Fools are Ingenious!</font color=red>
  19. Oops, I forgot the A7N266-C is a 415D mobo.

    CPU at 133mhz and memory at 100mhz? Interesting idea! I wonder if if anyone has tried that? If possible it certainly would demonstrate the efficacy of the Lightspeed architecture. Sure would open more possibilities for overclocking. Imagine 200 mhz FSB without needing for anything more than PC2400 memory. That would sure give the nForce chipset the edge. If Lightspeed truly doubles DDR bandwidth then 200mhz CPU could even run with just PC1600 memory. Memory bandwidth would exactly match Bus bandwidth.

    <b>I have so many cookies I now have a FAT problem!</b>
  20. This is pretty much what I've decided. I don't need the extra memory for the extra bandwith, as the Athlon is maxed out at 2.1 GB/sec, which 1 DIMM would suffice. However, I may just upgrade to 384 or 512 just to have the extra memory, which should always help, especially if I get Win XP sometime in the near future. Now, I have a question. Are the 415-D boards Win 98 compatible? On one companies website for the MSI 415-D board, it said it is only compatible with Win XP. Is that true, and if so, is it only true for the MSI board or all 415-D boards? Thanks.

    "Trying is the first step towards failure."
  21. "Are the 415-D boards Win 98 compatible?"

    I have no idea but I can't imagine a manufacturer eliminating a whole segment of customers by having drivers only for XP.

    Asus has drivers for all OS's for the A7N266/A7N266-E but I can't find any information for the A7N266-C.

    You might want to download some of the mobo manuals or check the nVidia site for more information.

    <b>I have so many cookies I now have a FAT problem!</b>
Ask a new question

Read More

Motherboards Dual Channel Memory