Which is better 3 or 4 Dimms with AMD 64 939?

[Espo]

I'm building a system with AMD 64 3000+ in a 939 mobo. I already have 2 Dimms of 256M 3200. I just bought 2 of the exact same resulting in 1024M with 4 Dimms. I hear this will limit it to 333mhz. Would it be better to just use 3 & go w/o dual channel. What if I buy a different brand of 512, resulting in 256+256+512=1024M with the 3 Dimms? Won't be doing any overclocking or gaming- just a lot of video editing.

Also, with my needs, would you go with Asus A8V or AbitAV8?

 

[Peter]

Welcome!

Generally, the 200 MHz DDR specification SAYS that two-DIMM operation cannot be guaranteed (per channel, of course). Hence, four DIMMs in a two-channel system may, but just as well may not work at 200 MHz (PC3200 speed). You may have to back down to 166 MHz. The more chips are on your DIMMs, the more likely this will become.

3 DIMMs is not a good idea because you'll ruin the symmetry required for dual-channel parallel mode.

The speed loss from stepping down 33 MHz isn't that big with an A64. Give the stuff you ordered a try.

 

[DAPUNISHER]

Looks like they haven't fixed that bug with nF4 either, at least according to t-break

One of the problems we faced with the nForce3 chipset was selecting DDR400 speed from the BIOS with four memory modules- it just didn?t work. Unfortunately, that still remains an issue and we?re pretty sure that this is a chipset and not a motherboard related issue as we have also received the Gigabyte nForce4 board that exhibits the same problem. Selecting the memory at DDR333 and then overclocking it way past DDR400 speeds doesn?t seem to cause an issue.

So while the bug is there, if you are willing to tweak the bios a little you can get far more than 400DDR with 4 sticks, have a look *Text


*ChineseDemocracyGNR found this review.

 

[Peter]

The. RAM. Controller. Isn't. On. The. Chipset.

And once again, this is how PC3200 (or "DDR400" as many prefer to falsely call it) has been calculated, defined, and engineered at the JEDEC consortium. One DIMM per channel. PC2700 lets you have two, PC2100 lets you have three.

This happens to be the one cause why server/workstation systems use "registered" DIMMs. These impose a smaller capacitive load onto the main memory bus(ses), and let you have more DIMMs with less signal degradations, and consequently, higher frequencies.

Some BIOSes might have other issues like those you describe, e.g. problems during initial frequency setup - but that doesn't get to the root of it, which is:

Two DIMMs per channel at 200 MHz is not even supposed to work.
If it does, be happy. If it doesn't, you have no point to complain about.

 

[DAPUNISHER]

Good stuff Peter, now if you could just learn to be less of a caustic ass when you school us that would just be great :disgust:

 

[fbrdphreak]

Originally posted by: DAPUNISHER
Good stuff Peter, now if you could just learn to be less of a caustic ass when you school us that would just be great :disgust:

:thumbsup:

 

[Peter]

Originally posted by: DAPUNISHER
Good stuff Peter, now if you could just learn to be less of a caustic ass when you school us that would just be great :disgust:

Let's see how much steam comes out of your rear end after you've been posting the same stuff ten times a day only to have the next poster confuse it up again, or contradict what you said obviously without even considering your post.

 

[DAPUNISHER]

I simply quoted what the reviewer stated. Feel free to e-mail and school him about why the chipset can't be the issue and his conclusions are wrong Also, your first post just said may or may not work, and didn't provide the very welcome technical reason for it the 2nd did. So I linked that article because they had a workaround for the problem. Had you included that in the first reply I would have just offered it as a potential workaround for the OP and not quoted the comment about the chipset that evidently infuriated you. You seem to expect us all to be as knowledgable in this area as yourself, but the fact is we aren't. I also tend to scan and don't always read all of every post so I missed where you stated this before I guess.

Now, while I can certainly understand your fustration, there is no need to act out Why not just point out that the reviewer was mistaken in believing the chipset could be the source of the issues? Nothing is accomplished by lashing out with acerbic replies. I doubt it made you feel any better and it certainly didn't make my day to be the recipient of one. I do thank you for the excellent information :beer: and I hope you keep providing it! I often learn something from you, but like many teachers, you need to work on being more patient with the ignorant :light: