Originally Posted by Chozart
I disagree with that point. Part of performance is to have RAM that will allow yuo to push that FSB higher when you have that low multi CPU. A divider will work, but will decrease performance! And wasn't the initial thought to increase performance?
Also, in your own post, you first state that you don't buy RAM to be able to increase your FSB and then state that you need high-speed RAM so you can increase your FSB. This is contradicting. In my opinion, it does NOT matter what the multi of the CPU is. In either case, you're better off buying faster RAM so you can push that FSB higher with RAM on a favorable divider and get the best performance. Obviously, at some point price comes into play....
I totally respect Chozart's opinion as a seasoned OCer but, unless you're running 1/1 (DDR 533) this arguement doesn't really hold much water. (Beyond the fact that he pretty much says that he agrees with me in principle.) It feels more along the lines of clinging to an outdated dogma of times prior to greater than 1:1 dividers. His answer suggests that a lower multiplier chip IS PREFERRABLE over a higher multi simply because it holds ram speeds at higher Mhz @ 1:1, which I agree, is the best performance divider Mhz to MHz. HOWEVER, dividers above it (3:2 etc) still provide more raw performance to overcome the slight efficiency loss of non-1:1 ratios.
I don't have HARD Intel based numbers to prove this.... yet. AMD results have and will agree exactly with what I say, but that will fall on deaf ears here. I will, as soon as, I have time to stop worrying about some personal matters I have pressing me at the moment and get my Conroe up and
I digress a bit...
Chozart claims that memory dividers are a big hit on performance.
If you drop from a 1:1 down to 4:5 (DDR 433) I agree, that you WILL see a performance loss from both a considerable loss of bandwidth and a slight reduction from not being 1:1. However, on dividers greater than 1:1, you'll gain more bandwidth than you lose from being non-1:1. This will all be proven in time... Given that so many of us are using E6600's it appears that DDR1000 will be best done on the 3:2 (DDR800) divider for a DDR1080 at 360FSB.
Now, were you to claim that anything greater than 1:1 provides no benefit I say you're CRAZY. Reason being? I think it's VERY safe to say that technology companies are trying to appeal to the mass market. The Average Joe. Not, Joe Overclocker. Joe Overclocker is probably outnumbered 100:1 by Average Joe and is a niche market. Would Intel and AMD invest millions and millions of dollars to leap to a memory that's capable of WAY WAY beyond stock FSB speeds (DDR 533 and DDR400) if there WERE NO benefit from doing larger than 1:1 dividers? I'm not an economist, nor a businessman, but I DON'T THINK SO. ESPECIALLY, since they are sacrificing super tight timings like 2-2-2-5-(1T) 2x1 gig sticks for insanely loose ones like 4-4-4-12-(2T) at 667Mhz that is regularly seen in current desktops.
I've gotten myself lost in a myriads of reasons why I feel I'm right and it has made this one LONG post. I feel that your responses should guide the rest of my posts and save us all time. So... rebuttle?
EDIT: I agree, pricing does have a considerable effect here. DDR800 is priced much better than DDR1000 sticks. Sadly, that's not the issue I'm talking about above.