4x1gb vs 2x2gb

Joined
Mar 10, 2008
Messages
59
There's a great deal right now at newegg for some corsair 1gb ddr2 800 sticks.

I'm wondering how much of a problem running 4x1gb sticks compared to 2x2gb stick will be with my 6600 build. I plan on overclocking and have heard that 4 sticks puts more stress on the northbridge. Is there much truth to this? I'd like to save the money and get the 4x1gb sticks, but if it's going to hurt my performance too much or stress out my mobo I might reconsider.

:D
 
If you have a P5B deluxe like me I wouldn't go for the 2*2gig sticks....

:(

They broke my overclock.
 
How well do the patriots OC? Is it 4-4-4-12? I'd like to keep the cas at 4 so that I have the option to step it down later during the OC if needed.
 
If you have a P5B deluxe like me I wouldn't go for the 2*2gig sticks....

:(

They broke my overclock.

But so can 4 sticks of memory of any size. You see, modern memory controllers work best with only 2 sticks of memory. I briefly had a P5B series motherboard which could not run 4 sticks of memory regardless of memory size even at stock speeds (I had to significantly underclock its E6600 CPU from 2.4GHz all the way down to 1.8GHz -- and not by multiplier adjustment at all; I also had to lower the processor's FSB just so it could run memory at PC2-3200/DDR2-400 speed -- just to even run stably at all).
 
Nope, officially CAS 5, but they'll do CAS 4. If you want CAS4 rated sticks it's $100 shipped for 2x2GB of the Patriot Viper DDR2-800. Can probably do better than that; I know the DDR2-1000 CAS 5 G.Skill kits are like $100 shipped at Newegg.

4GB DDR2-800 CAS5, $67 AR kit
4GB DDR2-800 CAS4 Viper, $100 shipped kit
Viper DDR2-1066 kit is $105 AR at Newegg.
G.Skill DDR2-1000 kit is $100 shipped at Newegg. That's probably your best bet if you're concerned about OCing.
 
I've had great luck in the past with OCZ I think i'm going to go with their reaper HPC model for 109.
 
No, I've reverted back to a 2 gig kit in my sig. The P5B deluxe isn't really good with 2gig sticks. In the manual it says it doesn't run the sticks above 667 reliably.... and they are very correct. With different brands and revisions of ram I was unable to even get it stable at 800MHZ at 5-6-6-15...

I know the performance of the ram in my sig is good. Great in fact. But it's not the case with the 4 gig kit. Trust me on this it is the board. That ram is doing fairly well on a P35 board right now.

Max stable overclock with the 4 gig kit was 2.66GHz on my P5B deluxe.
 
I am running these now (4 x 1GB):

http://www.mushkin.com/doc/products/memory_detail.asp?id=350

Would I be better off running 2 x 2GB to be able to achieve 1066MHz?

Also, is there any point in getting DDR2-1000 or DDR2-1066 if there is no JEDEC spec for it? Aren't those memory timings/MHz rating just OC'd/Overvolted DDR2-800 and I am just paying more for someone else to OC it for me?
 
What CPU will you be pairing the RAM with? There's no performance benefit to running outside a 1:1 ratio on Intel platforms anyway.

To find what RAM you need for a given OC:
speed of chip at stock / ( speed of quad pumped FSB at stock / 4) = multiplier
OCed speed you want / multiplier = real FSB speed required for OC (x4 if you want to get quad pumped number, though pointless in most cases)
real FSB x 2 = RAM speed required for 1:1 ratio at OCed speed

So with a Q6600, it's rated at 2400Mhz at 1066Mhz quad pumped FSB.
1066 / 4 = 267 real FSB at stock
2400 / 267 = 9 multiplier
If you want to run 3600Mhz,
3600 / 9 = 400 real FSB required for OC
400 x 2 = 800 so DDR2-800 requred for OC at ideal 1:1 ratio

The chips used in DDR2-1000 and 1066 stick are better than those in most DDR2-800 sticks, so if you need much more than DDR2-800 for you desired OC according to the above equations you should seriously consider buying higher rated sticks.

Burner27, are you OCing at all? the chip is rated for 1066Mhz quad pumped FSB at stock, so I'm a bit confused.
 
Not OCing but I'd like the memory to run at the same FSB as my CPU. Isn't that what I should be shooting for?
 
Not OCing but I'd like the memory to run at the same FSB as my CPU. Isn't that what I should be shooting for?

No, because FSB as advertised by Intel is "fake" quad pumped. To link and sync a Q6600 at stock (1:1 ratio with RAM) all you need is (1066 / 4) * 2, or ~DDR2-533. Buying any less than DDR2-800 in today's market doesn't really make sense though, in case you end up wanting to OC in the future, cause it's rarely more than $2-5 different from the really weak stuff (DDR2-533 or DDR2-667) to DDR2-800.
 
Should I dump my current memory in favor of 2 x 2GB of PC2-6400 then? Or should I just run the ram at its advertised specs of 4-4-3-10 @ 2.0v @ 800Mhz FSB? Is 2 x 2GBs faster than 4 x 1GBs when they are run at the same timings? (ex:5-5-5-15)

Right now I am running them at SPD which sets them to 5-5-5-15 @ 1.8v @ 800MHz FSB
 
Should I dump my current memory in favor of 2 x 2GB of PC2-6400 then? Or should I just run the ram at its advertised specs of 4-4-3-10 @ 2.0v @ 800Mhz FSB? Is 2 x 2GBs faster than 4 x 1GBs when they are run at the same timings? (ex:5-5-5-15)

Right now I am running them at SPD which sets them to 5-5-5-15 @ 1.8v @ 800MHz FSB

If you're not OCing there is no reason to go to 2x2GB. The only advantage of 2x2GB over 4x1GB is less stress on the MCH, which only becomes an issue at high FSB. Also, run your RAM at spec rather than from SPD. You paid for performance you're not getting, how ever minimal the change will be.
 
What was funny about this ram was when i plugged it in and booted it up the first time, it booted at 1066 settings. What I mean is that when i set the BIOS to "Auto" for detecting type of memory I have it thought the memory was DDR-1066 but had the voltage set to 1.8v. It never made it into windows (BSODs). But when I left it at auto and upped the voltage to 2.0v (+.2v increase) it booted into windows no problem. CPUz had it at 533MHz. What do you think of that?
 
It probably fell back to failsafe when it failed to boot the first time and you raised voltage with the failsafe timings in place. Set it to 4-4-3-10 2T 2.1vdimm and make sure it's running at DDR2-800, and you're good.
 
Setting it to 4-4-3-10 @ 2.1v @ DDR2-800 yielded very little performance gain.
 
Setting it to 4-4-3-10 @ 2.1v @ DDR2-800 yielded very little performance gain.

That's why we regularly tell people not to worry about timings (so long as it's ~5-5-5-21 and below). Doesn't matter in much save for benchmarks.
 
So should I get faster ram and bump up the FSB or am I only going to see that gain in synthetic benchmarks?
 
Only in benches. Zero reason to replace your current RAM unless you intend to run phase change or some other exotic cooling allowing a Q6600 past ~3.6Ghz (400Mhz FSB). I don't want to make too many assumptions, but I'd guess that's not in the cards at present.
 
Only in benches. Zero reason to replace your current RAM unless you intend to run phase change or some other exotic cooling allowing a Q6600 past ~3.6Ghz (400Mhz FSB). I don't want to make too many assumptions, but I'd guess that's not in the cards at present.

You would be correct!!!! Thank you for all of you advice. I really do appreciate it!
 
That's why we regularly tell people not to worry about timings (so long as it's ~5-5-5-21 and below). Doesn't matter in much save for benchmarks.

Not questioning your expertise, mind you, but I did an experiment. I ran my current memory at DDR2-1066 settings with 5-5-5-15 timings at 2.0v and the memory bandwidth is much higher and the latency decreased. Would this benefit me doing video encoding?

I was using Everest Ultimate to benchmark the memory.

Thx
 
Something to consider:

A lot of "Dual Channel" memory controllers only work properly with exactly 2 sticks of RAM on the 1st and 3rd slots. Anything else will slow your memory.
 
Something to consider:

A lot of "Dual Channel" memory controllers only work properly with exactly 2 sticks of RAM on the 1st and 3rd slots. Anything else will slow your memory.


Where is this written? I would like to read that article.
 
Not questioning your expertise, mind you, but I did an experiment. I ran my current memory at DDR2-1066 settings with 5-5-5-15 timings at 2.0v and the memory bandwidth is much higher and the latency decreased. Would this benefit me doing video encoding?

Not substantially, no. Again, never said you wouldn't see a difference in benches -- it's outside of benches that it's unsubstantial. You're welcome to run an encode of the same material at both speed/timing sets, but I'll bet it's not more than a few seconds different, if that. CPU speed is far more important than RAM speed when encoding.
 
Not substantially, no. Again, never said you wouldn't see a difference in benches -- it's outside of benches that it's unsubstantial. You're welcome to run an encode of the same material at both speed/timing sets, but I'll bet it's not more than a few seconds different, if that. CPU speed is far more important than RAM speed when encoding.


Do you agree with ROCKSTA107's comment about 2 x 2GBs vs 4 x 1GBs?
 
I got mine at 962mhz@1.925v cas 5,5,5,18. 2GB*2 mushkin ddr2 800 ram. Pretty good ram for the price.
 
I have buffalo memory PC-6400 2x1gb micron D9's and it is the best memory I have ever used and it was cheap too like $50 for 2gigs. I run it at 4-4-4-12@1000mhz at default voltage 2.1v.
 
Back
Top