KD7 vs. NF7 vs. NF7v2.0

Which one should I get? After reading reviews and threads of this forum, I'm leaning towards nVidia chipsets... But I really can't find out about the differences between NF7 ver. 1.x and NF7 ver. 2.0... What has changed in 2.0, what's better about this board?

I have absolutely no use for the SATA and Firewire ports, so I won't get a NF7-S. I hope there is a 2.0 revision for the board without -S

I intend to do some moderate overclocking, ie to keep a 1700+ in the 2GHz range, and I won't have high-performance RAM (can't afford it) so I guess I'll be limited to 166 FSB. My PSU is generic crap 300W, but it holds very well with my Tbred1700+ OCed to [email protected] on Abit KT7 so I guess PSU won't be a problem.

The revision 2 has the C1 northbridge, which unofficially supports 200 mhz fsb. It also supposedly fixes bios corruption issues, which hasnt really been confirmed.

There is a rev 2 for the plain NF7, and it uses the same C1 northbridge that the -s uses.

Thanks for clearing that out. And what is this corruption everybody is talking about? If it's the SATA corruption, it won't affect me since I'm getting the non-S model. Is there a corruption problem with the plain NF7 too?

Theres bios corruption problems with all nforce2 boards. With the revision 2 it has supposedly been fixed. Although it still happens to some, it is quite rare now. If youre only shooting for 2 ghz with a low FSB, I dont think youll have to worry about it. Its most common with extreme OCing.

The SATA corruption is another bit of corruption, but as you said it wont matter to you anyway.

Thanks for your reply. I don't quite understand what "bios corruption" really means, but as long as I don't lose my data and it doesn't kill me (or does it?), I suppose it's all right...

I'll have to wait a bit longer, the NF7 2.0 isn't yet in Romania, we only have NF7-S 2.0... Until then, I'm curious about some other details: Do I have to buy 2x256 DIMMs, or 1x512? Is there a true performance improvement if I use dual channel? And, is nForce2 like the VIA chipsets in regard to synchronous/asynchronous FSB:memory operation? With KT400, I understand it's better to use memory at the FSB frequency, even if memory supports higher speed; is this true with the nForce2 too?

If you don't have any use for SATA, Firewire or high FSBs, then there's not much advantage to you getting the NF7-S over the KD7. The remaining advantages would be the sound quality and an overall performance benefit (it's not quite as big as everyone makes it out to be, maybe 2 - 3%). If those aren't a big deal for you and you want to save some cash, the KD7 might be a good choice (not sure what the price difference is where you are.

You don't need to buy two memory sticks, unless you want to run dual channel. Running dual channel will not usually yield much of a performance benefit, although if you have cheap ram then it might be able to run dual channel memory with your FSB set much higher (you'll see a benefit if your FSB is clocked significantly higher than your memory bus). With 512MB or less some applications that use a lot of memory would benefit more from running in single channel though.

It is usually better to run the memory and FSB in sync, except maybe in a dual channel situation like I explained above. It would still be better to just get better memory that can handle the higher bus speed and run it sync though.

I'll get two 256MB sticks of whatever brand I afford, and try running the board in dual channel and single channel configurations (I think it's possible to put two sticks on the same channel, isn't that true?)

Another thing I don't know about nForce2 chipsets: on VIA KT400 boards, in order to keep PCI and AGP at their rated frequencies, you only have three choices of FSB (100, 133 and 166); you can go higher but there's no guarantee your HDD or videocard or something else isn't holding you back. I understand that nForce2 can keep PCI and AGP locked at their specified freq (33/66) regardless of the FSB you're running; in that case, is it best to run your FSB (and memory) at the highest frequency your memory can achieve with the most aggresive timings? For example, if you have two sticks of 256 that work perfectly at 150MHz with the fastest settings but require higher latencies for running at 166MHz, is it the best to let them run at 150MHz (and use a FSB = 150 to keep it in synch with the mem)? Thanks

With regards to your first question, yes you can run two sticks in single channel (that's what I'm doing right now).

On the second question, I believe that you're right in this thinking although with dual channel this might not be the case if you have poor memory. For example, if your memory can only get up to 133MHz, then running the FSB at 166MHz and the memory at 133MHz doesn't make a great deal of sense because the FSB will be bottlenecked by the memory (although you will probably still get some performance benefit because of the caching effects). But, with the dual channel option availabe, you can run the memory at 133MHz dual channel, which is theoretically as good as running one stick at 266MHz. Thus, your FSB won't be bandwidth limited unless it's over 266MHz.

In reality, dual channel isn't quite that effecient so you will probably still be bandwidth limited at an FSB of 266MHz. But since you won't be getting the FSB that high, it's pretty reasonable to assume that an FSB at 200MHz or so will be a pretty good match for dual channel at 133MHz.

My suggestion would be to get the best memory that you are willing to spend on, and then work from there. Keep in mind that if you run dual channel, then you're losing the advantage of having that much memory (so running two 256MB sticks at 133MHz dual channel is theoretically equivilent to running ONE 256MB stick at 266MHz, not 512MB at 266MHz). This is probably obvious but I just thought I'd point it out.

You can also try running single channel with the buses async, which probably won't help much, but experimenting is part of the fun, so you can always try.

I have no idea what kind of memory prices you're faced with, but for me the Kingston HyperX PC3000 wasn't a great deal more than most PC2700. With a bit of extra voltage, I am able to run this memory at 200MHz with pretty aggressive timings 2-2-2-6, and I'm probably able to get a bit more if I tried. But I'm happy with running the buses sync at 200MHz, so I'd recommend getting this memory if you have the opportunity and it's in your price range.

Captain Hilts said:

Keep in mind that if you run dual channel, then you're losing the advantage of having that much memory (so running two 256MB sticks at 133MHz dual channel is theoretically equivilent to running ONE 256MB stick at 266MHz, not 512MB at 266MHz). This is probably obvious but I just thought I'd point it out.

Click to expand...

Heh, it wasn't that obvious to me , thanks for enlightening me...
This obviously mean I won't run dual channel, as I strongly need 512MB of memory and cannot afford buying 1GB RAM even if I choose the lowest possible quality...

The only HyperX I can find in Romania is PC2700 CAS2, no PC3000 or 3200. They stand at $96 for a 512MB stick, and that's $10 more than a 512MB PC3200 from KingMax I was thinking of getting. For $108 I could buy 512MB of Corsair XMS2700, but that's a little more than I'm willing to spend on memory. Which one do you advise me to buy? Thanks.

Captain Hilts said:

Keep in mind that if you run dual channel, then you're losing the advantage of having that much memory (so running two 256MB sticks at 133MHz dual channel is theoretically equivilent to running ONE 256MB stick at 266MHz, not 512MB at 266MHz). This is probably obvious but I just thought I'd point it out.

Click to expand...

This is not true at all. I'm not here to dispute, so I'll put it simply. Running 2 256 mb sticks is like having 512 mbs. One dimm isnt mirroring the other dimm. They are splitting the data in half and each still using their full 256 capacity. It is basically like a striped array in a raid configuration for HDs.

2x256=512, dual channel or not.

Okay, thanks for the correction. But isn't some of the advantage of having 512MB still lost though? Why is it that there is more use of the swap file when running dual channel? I have had this happen myself when testing single vs. dual channel in my system.

As for the memory, that sounds kind of expensive for the PC2700 HyperX, the Kingmax sounds like a better deal but I don't know too much about that memory.