MadShrimps has article on mem selection for Conroe.

[MplsBob]

Don't waste your money by fighting the basics.

The final page of this article wraps it all up.

 

[dakotagts]

So its saying that on a STOCK rig, you want to use PC4200 for the best timing?

Great!!! That saved me some money, now I may be able to step upto a better gpu or cpu X1900XT or 6600 conroe!

and it looks like the winner is PQI with the 3-3-3-12 timing

 

[MplsBob]

Yes, that's the idea.

AnandTech posted an excellent article on lower priced memory that did the job and the article vanished within a day. I don't know what happened to it. Did the memory manufacturers get upset or what? All I know is that it disappeared. Sadly, I had only given it a quick once over. When I went back the next day to give it a serious read, it was gone.

 

[Pabster]

Originally posted by: MplsBob

Yes, that's the idea.

AnandTech posted an excellent article on lower priced memory that did the job and the article vanished within a day. I don't know what happened to it. Did the memory manufacturers get upset or what? All I know is that it disappeared. Sadly, I had only given it a quick once over. When I went back the next day to give it a serious read, it was gone.

No conspiracy theories, please

Although, if you were selling $450 RAM, you certainly wouldn't want word that $150 RAM was all you needed getting spread around!

 

[dakotagts]

Now to figure out the motherboard war. and you are not talking about this one are you on the value memory page 17

http://www.anandtech.com/mb/showdoc.aspx?i=2797&p=17

Infact look at this price on the PQI Turbo

http://www.newegg.com/Product/Product.asp?Item=N82E16820141199

 

[MplsBob]

Please keep in mind that the picture presented in the MadShrimps article is not a fixed truth carved on the side of a stone mountain.

If you overclock your processor sufficently, the DDR2 533 (PC2 4200) memory will no longer be in sync. The equations in the 3rd paragraph from the end of the article need to be performed to see what your new memory requirements would be. They are quite simple and should not pose any real difficulty.

I sort of wonder whether the overclockability of the DDR2 533 memory wouldn't be sufficient to carry one through without having to buy all new memory. I would defer on this to people with more overclocking experience than I have, since I have none at all.

 

[dakotagts]

However, I am keeping a stock system. I WILL NOT OC my system, because I have too much to loose. If I fry a CPU or Ram I will not have the money to just replace it, and if I did I would have to wait for christmas to use money for it...

 

[buck]

Originally posted by: MplsBob

Yes, that's the idea.

AnandTech posted an excellent article on lower priced memory that did the job and the article vanished within a day. I don't know what happened to it. Did the memory manufacturers get upset or what? All I know is that it disappeared. Sadly, I had only given it a quick once over. When I went back the next day to give it a serious read, it was gone.

I was wondering that same thing.... WTF?

 

[Trey22]

Interesting read.

Makes sense though... glad I bought some DDR2 533 to start out with.

 

[dakotagts]

Hmm, at the same price however on the OCZ 6400 at Zipzoomfly.com on stock would that result in a true 35% increase in performance in real world apps?

If you read some of the benchmarks at stock timing it looks like that is what I am seeing over PC3200, and only a 13% increase over PC4200.

So to make the memory more usable in the future do I get higher OCZ memory at around the same price?

 

[Lord Evermore]

Originally posted by: dakotagts
However, I am keeping a stock system. I WILL NOT OC my system, because I have too much to loose. If I fry a CPU or Ram I will not have the money to just replace it, and if I did I would have to wait for christmas to use money for it...

Overclocking in and of itself can't fry your system. If you overclocked to the point that the system was overheating, due to insufficient cooling to deal with the level of overclock, it could cause damage, but this is exactly the same issue as it would be if you were simply to buy a higher clocked CPU and memory (although with an overclocked system there is more heat output from the chipset as well, which would not be the case with a higher CPU and memory stock speed).

Overclocking while you monitor the temperatures is perfectly safe, just as safe as if you were building a high speed system at stock speeds and making sure your case was adequately cooled. If you get instability, the hardware just can't handle the speed, and you cut it back. Higher overclocks than true stability can also affect data integrity.

Adjusting memory and CPU voltage in order to reach even further overclocks increases the heat output even more quickly than just the overclock, and might even result in damage just by itself, if a particular part can't handle the voltage.

 

[dakotagts]

So I am not following you completely. Does it damage things or not. In one paragraph you say OC'ing cannot fry the system, but in another you say it might result in damage.

If I only OC the processor to a certian level, will that meet the 1:1 with the memory PC6400? at which point the memory and cpu are in the same timing again.

 

[Lord Evermore]

If you increase voltages, yes, you could potentially damage the system. Overclocking does not necessarily require voltage changes, it only makes it possible to get a greater overclock in some case. Without voltage changes, most systems can overclock at least a little bit.

If you're that concerned of course, don't do it, be happy that you've got a faster system than could have been bought for any amount of money 2 years ago (consumer level anyway).

Most people aren't concerned with getting 1:1 ratios with the CPU and FSB anymore. With the Athlon64 there isn't even a FSB to match the memory to.

The Conroe 1066MHz bus is quad-pumped, so it's a 266MHz physical clock. DDR2 1066 uses a 533MHz double-pumped bus. So you've actually have to overclock PC2-6400 to get a bus clock matching transfer per transfer with the Conroe FSB. At that speed a single memory channel ought to be enough to completely satisfy the CPU bus.

I'm not sure exactly why, but dual-channel PC2-4200 isn't apparently enough to satisfy the Conroe. It should be pretty much exactly the same bandwidth as the FSB, though only one transfer occurs for every 2 transfers of the FSB. This is one of the things that all the lovely buffering and caching and stuff done on the Conroe hides, since it means latency. PC2-8500 still gets a performance gain in dual-channel with the Conroe, even though it's got twice the bandwidth needed.

 

[jmke]

increasing the speed of a GPU/CPU by itself does not damage the device, however the effect of this increase CAN cause damage, mainly HEAT, if temperatures are not monitored and the device overheats with the overheating protection kicking in too late you'll get a broken PC.

Luckily for us, with the latest hardware the overheating protection work surprisingly well, and the risk of damaging hardware has been severly reduced. just keep in mind that overclocking causes increased heat output and you'll have to invest in better CPU/GPU/Case cooling in order to prevent overheating.


About the Core 2 and memory timings, this is the answer we were looking for when we started working on the article as stated by dakotagts "So its saying that on a STOCK rig, you want to use PC4200 for the best timing? "

PC4200 for the best performance/price balance; PC6400 will give you a (very) slightly performance edge but price goes up more than performance does => worse balance.

The Anandtech article has not disappeared it's here: http://www.anandtech.com/memory/showdoc.aspx?i=2800

if you plan to overclock your Core 2 system then the old rule still applies "get higher rated memory to give you headroom for running 1:1" but that's another article

 

[Hidden Hippo]

I read somewhere else that for stock on an E6600 you should use PC5400 for best performance, as opposed to PC4200. My main question is would I see much of a performance difference between this RAM and this RAM? The first is PC6400 with 5-5-5-12 timings and the second is PC5400 with 4-4-4-12 timings.

 

[Lord Evermore]

Originally posted by: Hidden Hippo
I read somewhere else that for stock on an E6600 you should use PC5400 for best performance, as opposed to PC4200. My main question is would I see much of a performance difference between this RAM and this RAM? The first is PC6400 with 5-5-5-12 timings and the second is PC5400 with 4-4-4-12 timings.

Take a look at the article linked in the original post (skip to the last page for summary). PC4200 according to them is better than PC-5400 (or 5300 depending on brand marketing). This seems to be due to the asynchronous cycles of the memory transfers and FSB. However PC-6400's bandwidth increases are enough to make up for the difference, even though it's also asynchronous.

With those two modules, they might be essentially the same thing. You could possibly overclock the 5400 module and get it up to the 6400 speed, but have to increase the timings to 5-5-5-12. Anytime you see things that are faster, but with worse timings, the companies are essentially factory-overclocking, but of course you're at least guaranteed to reach those speeds, and the slower chips could be the ones that couldn't make the higher speeds even with increased latency.

MadShrimp's 3DMark graph shows that the DDR667 5-5-5-15 module actually got a slightly higher score than the 533 4-4-4-12 module, but slightly lower than the 533-3-3-3-8 (of course the 667 4-4-4-12 also got a lower score than the 667 5-5-5-15).

 

[Lord Evermore]

Incidentally, horrible horrible graphs and captions on them. Negative percentages of increase? It technically makes sense, but it's a very obtuse graph. Poor writing to swap back and forth from graphs that are read in negatives and positives. If you must show percentages, show "1.52% faster" and a positive graph, not "negative 1.52% speed increase". Negative graphs are usually used to show a surprising negative where the other lines are all positive.

% increase in negative numbers is like talking about slowing down with "percentage faster" in negative numbers.

 

[dakotagts]

So I am going to buy a 6400 most likely, If I OC the speed, does that mean that I am raising the 266Mhz bus level or only the Clock speed and the bus stays the same. The thing I am trying to figure out is with OC'ing for a higher clock speed will raise temps and can deal with that and not mess with the ram voltage, But will that OC be enough to really be worth it in real life applications?

...