• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Noob memory questions

C

commie64

n00b
Joined
Aug 12, 2009
Messages
2
Hi Folks,

I've only recently delved into the build-you-own arena and am hoping to learn a lot from all the great information here. I hope you won't mind if I ask some embarrassingly dumb questions, though...

So, I've inherited a discarded Asus P5B Deluxe with an Intel E6300 Core 2 Duo from a friend who no longer wants it. I need to purchase some RAM for it, though. Now I may overclock sometime in the future just to mess around and get more intimate with computer internals, but for now I'd just like it to run stock speeds.

Anyway my here are my questions --
  • Does the CPU FSB limit what RAM I should buy? Since the E6300 is a locked 7x multiplier with a FSB of 266 MHz does that mean if I get memory any faster than DDR2-533 it'll be a waste?
  • From my meager understanding it's possible to run the RAM at a higher rate than the FSB by modifying the FSB : DRAM Ratio, but people seem to think that 1:1 is optimal. If that's the case why is there all this dramatically faster DDR2 RAM out there?

Thanks! Looking forward to learning a ton from you guys, this stuff is really interesting.
 
If you don't plan to overclock, then there's not much use going with RAM that's faster than your FSB. However, that board and that processor tend to be great overclockers (400FSB should be no problem). So if you want to do that, you'll have to get RAM that can handle DDR2-800 speeds.


As for why there's faster RAM out there, There are a lot of Intel processors that run at 333, or even a couple that run at 400. Then AMD (and Intel's new) processors don't have a FSB as such, and can definitely benefit from running the RAM at higher speeds
 
18 is # 1 said:
Your Intel processor will give better performance with faster RAM. 1:1 is not optimal at all with Intel mobos. These are a great value right now and will run 1150+ 5-5-5-18: http://www.amazon.com/OCZ-PC2-8500-Platinum-Voltage-OCZ2P1066LV4GK/dp/B002845LDC/ref=pd_rhf_p_t_1
Click to expand...

There is a very small (if any) gain from running RAM asynchronous, 1:1 ratio is normal.

To the OP, yes, unless you plan on OCing, you don't need some super fast RAM, however, right now most of the time you can buy 800 speed RAM for cheaper than 533 or 667 etc speed RAM, so there is no reason to not at least get 800, that way you also have head room if you OC.
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
At the risk of putting words into someone else mouth I will give it a shot.

I am with Ryan, Wut ??? Intel boards where designed by Intel for the ram to run at 2x the FSB buss speed of the CPU. This 2x the clock is the speed the CPU cache runs at and so a 1:1 ratio (really ram 2x FSB ) is ideal from an engineering point of view. The memory controller does not have to do anything special with traffic control of data moving from CPU to memory and back to CPU. Everyone in the boat is rowing at the same speed. So a systems design engineer, especially an Intel one would say a 1:1 memory ratio (2x wpeed) is optimal.

However manufacturers and now Intel allow memory multipliers to be used. Basically the CPU talks to the NB (memory controller) at some speed you set and know as the FSB. But the NB/memory controller can talk to the memory faster by use of a higher multiplier/ratio. The Northbridge/MCH truely is a "bridge" between the cpu and the memory, A bridge that also has a traffic cop on the middle of it to make sure traffic flows smoothly with no crashes.

example.
lets take 5/4 ratio for example . You set the FSB at 266 and your CPU runs at CPU multiplier of 7 x 266 FSB = 1.8GHz and there is your stock CPU speed. But the memory will run at 5/4 x 266 x 2 = 667MHz . This is faster than the plain stock default 1:1 which would have given you just 533MHz.

So the memory can run faster. There are some memory operations (mainly memory to memory transfers that do not involve the CPU) that this speed will be to your advantage. Sometimes it will not help at all because if the memory contents are going to the CPU it still has to slow down to the CPU buss speed once the data is on the CPU side of the Northbridge. Also the Northbridge has to buffer and adjust the "flow" as data is coming in on one side faster than it can go out on the other. People in the boat are rowing at different speeds and if they don't watch out they will get in each others way. This is the traffic cop on the bridge he keeps things in order. Very cleverly but you never get the total added performance from the increased memory speed. But you get some, and sometimes a lot and sometimes not much at all.

So by cranking up the memory multiplier you can run faster memory and gain some performance sometimes. Benchmarks show high bandwidth etc. in the real world it helps - how much depends on the applications. Games not so much. Big memory intensive calculations - some to a lot. /shrug. One downside of the "not everyone rowing at the same speed" is sometimes it causes stability issues OCing. Sometimes it does not. /another shrug

To your original question. yes and no. And it would take several pages to go into it all. But you are thinking correctly. But I had a 6300 and that sucker was awesome, I had it up to 475+ FSB in the board I am still using in my sig. and ran it at 3.3GHz for almost 2 years, its now in my bot box running at 3GHz and still happy as a clam. With a 1:1 ratio that put the memory running at over 900MHz and I was limited because my 800MHz ram would not OC that high. I had to buy the ram I have now to see where the upper limit of my board/6300 was. Long story short, ddr2 is cheap as dirt right now and if you are thinking of OCing get at least PC6400 800mhz ram. That will allow you to OC to 2.8GHz (400 x 7) with a 1:1 ratio (everyone rowing at the same speed) and your memory running right on spec. And likely you could get another 25MHz of FSB before the memory complained about being OCed . That is the least you should do if OCing is in the future and it should be. If your case has decent fans/airflow 2.8GHz on the stock cooler is easy and is a moral imperative. You will think you got a new machine compared to stock. Totally different experience.

What I recommend due to the very small difference in price is the 1066 speed if you can swing it. The cost difference is not much. It will allow an FSB of over 500MHz (few boards will do that) and at 500 x 7 you are hitting 3.5GHz for the CPU (highly unlikely you will ever get to that) but the point is your memory will not hold you back for any OCing conceivable with your CPU and board and if you upgrade the CPU later you will still have good usable ram. And for now with the cpu stock you can try the higher ratios and see for yourself how increasing the memory sub-system speed improves performance. Typically I notice the machine is a bit "snappier" but no big deal, but everyone has different results.

So if you can swing it a nice pair of 2x2GB 5 5 5 15 or 5 5 5 18 ram 1066 speed would be perfect. If you are hurting for cash drop down to 800MHz speed but if you intend to do any OCing at all I would not go lower than 800MHz and you will find the savings is not much anyway.

So everyone is correct and if you intend to OC get the fastest memory you can afford (within reason) , its never a bad move at any memory ratio.
 
Even if it is a small gain, running asynch will in fact give a benefit over a synch aka 1:1 setting. If the system is stable and will run the memory faster than the CPU, there is certainly no reason not to do it. The benefit can range from 2%-5% depending on the applicaton or program.
 
Wow, thank you everyone for all the great input. I have a feeling I'm going to like it here ;) I will definitely be going with faster RAM now.

One followup question -- besides just looking at the parameters displayed by CPU-Z, is there a generally accepted way to measure performance gains from memory speed tweaking? I think I'm asking for benchmark application recommendations, although I know benchmarking can be tricky from some of the articles I've read. Does anything exist that can categorize the performance for difference "usage profiles" (e.g. basic office work, memory-to-memory xfers like BillParish mentioned, etc?)
 
Memtest gives a rough bandwidth score. SiSoft SANDRA and EVEREST are good tools also.
 
You must log in or register to reply here.
Back
Top