Ram Higher Frequency or Tighter Timings

[Mykvega]

Can you guys give a recommendation? Which is better between the two?

DDR2-800 running at

1. 733 MHz 4-4-4-12 CR1

or

2. 820 MHz 6-6-6-18 CR2

I have an Asus P5N-MX board and Intel DC E2200

Will higher frequencies also make me able to run the cpu at higher fsb?

Thanks in advance!

 

[outlw6669]

Running your DDR2 800 at 733Mhz, 4-4-4-12 t1, will give you better performance than running it at 820Mhz, 6-6-6-18 t2.
To run your CPU at a higher FSB, you will have to change in the BIOS your FSB and FSB:RAM divider.
DDR2 800 can run without overclocking to 400Mhz FSB or potentially 4.0Ghz with your E2200.

If you are intrested in learning the relationship between frequency and timings take a look at This and This link.

 

[Mykvega]

Thanks for your reply. I've been reading it and still trying to absorb what's on it.

While I do that, I have another question.

I seem to hit an fsb wall at cpu fsb 312.5 Mhz (1250) at 10x multiplier giving me 3.125Ghz on E2200. The computer crashes when I run benchmarks.

If I drop my ram lets say to 667 you think it would stabilize or would crash much faster?

My board doesn't have vcore options for changing.

 

[outlw6669]

Hard to say if drooping the RAM frequency alone would stabilise your system.
I would recommend testing for stability in stages.

Try dropping your CPU multiplier to 6x, raising your FSB to 333 Mhz and putting the RAM at a 1:1 ratio to the FSB (lowest possible setting).
Also, check and make sure your timings and voltage for the RAM is set to the manufacturers specifications.
Test in windows with dual instances of prime95 using small FFT's.
If everything is stable, you have just shown your motherboard to be stable at that FSB.

Next, raise your CPU multiplier to 9x and retest.
3.0Ghz is an excellent overclock for your CPU and, with out any voltage adjustments, is about as high as you can take it.
If it tests stable again, try for the 10x multi for a full 3.33Ghz.

After you have the Mb and CPU stable, you can start tweaking the RAM.
Start at 800Mhz with recommended timings and voltage.
Test with memtest86+ for stability.
When it is stable, you can start tweaking the voltage and timing settings to get every last ounce of performance from your setup.

 

[hairycat101]

Why ask that question... I'll bet you won't notice much of any difference at all. Bet you further that you won't see much difference even in synthetics. If you are really interested in it, try both and bench them both. I'll put a quarter on the 820.

 

[Mykvega]

Hehe you lost a quarter hairycat. Here are results I did after benchmarking

PCMark Vantage

820 MHz 5-5-5-15 CR2 Score: 3948 PCMarks
733 MHz 4-4-4-12 CR1 Score: 4149 PCMarks

3DMark Vantage

820 Score: P800
733 Score: P859

(..I know, scores are pathetic) Just to show the difference.

outlw I have yet to do what you advice and see if there are any improvements. currently I have just realized dropping the ram can let my fsb run stable. im now at 324Mhz at 9x. CPU frequency has dropped but lets see if the 200Mhz+ matter. I used to run my CPU at this speed before now at http://valid.x86-secret.com/show_oc.php?id=407836. I'll still try to reach 1333 fsb by dropping more ram then compare. And too bad my board doesn't have voltage adjustments Asus P5N-MX. By the way, my ram can support 667. Dropping it way below that is worth it?

 

[iluvgillgill]

select 1:1 ram multiplier and try to minimise latency. thats what you should always do to get the best performaceout of the setup.

 

[V3NOM]

^unless you have a board like mine that does not have a multiplier. -_-

 

[iluvgillgill]

^sigh.............live with it v3nom!lol

 

[sportsfanboy]

Hey all... In my testing (wish I took screen shots) ram frequency and latency times scaled linearly. Timings (not talking about command rate) had an impact as well, but not as much as I would have thought. I remember sitting at 64ns in everest latency test with 1000mhz ram and 5-5-5-15 timings. After changing to 4-4-4-12 and also lowering and tweaking some subtimings, I was able to further shorten the response time to around 61ns or so.

With the same setup, I dropped my multiplier and jacked up my fsb in order to test the 1:1 ratio times. The other setup was using a 5:8 divider just so you know. On the 1:1 set up, my ram was in sync at 734, 367FSB(my b3 q6600 gets hot in the summer). I ran the latency test again and on the 1:1 setup the latency was 78ns, quite a bit slower.

Conclusion... I believe frequency is more important to overall ram latency times, on the current c2d/c2q platform. That might change when Nehalem comes around, who knows. For now, I will be sticking with high speed, and slightly higher timings, to achieve the lowest ram latency times.

 

[iluvgillgill]

yeah the timing isnt that much different from one another at speed above 800mhz. i like to keep it loosen up a little bit and run less voltage since the gain i get will be very little and it will stress the ram faster.

 

[iluvgillgill]

as i have said many times before higher ram frequency at the same FSB will only gain your faster latency but the actual read and write speed will almost stay the same. and can you actually notice the difference in games?internet browsing?or even heavy video encoding that maybe run for 10 seconds longer?

 

[sportsfanboy]

as i have said many times before higher ram frequency at the same FSB will only gain your faster latency but the actual read and write speed will almost stay the same. and can you actually notice the difference in games?internet browsing?or even heavy video encoding that maybe run for 10 seconds longer?

Have you tested this? I could have sworn that read/write also increased/decreased linearly with frequency, I might be wrong.

 

[iluvgillgill]

yeah i have tested it.you will only see max 100mb/s increase in benchmark when in real life TRUST ME you wont notice a single increase! i said when FSB stay the same so the bandwidth is at the same level. i think people said it many times its like a waterpipe doesnt matter how wide the 2 end is.

 

[sportsfanboy]

I'm not sure the water pipe analogy is 100% accurate. I personally have noticed an increase in encoding times(cinema craft encoder), when running higher frequency settings, over the lower frequency, at the same fsb setting. Ram has it's own latency and will have an impact on system performance, even at a fixed fsb value.

I also read a while ago when the core2 333fsb came out, that their is only an average of .05 percent difference in system performance, going from a 266 fsb to a 333fsb. I wish I bookmarked it, but I'm telling you the truth. Because of that article, I started doing my own testing (mainly because of all the 1:1 fanboyism over anything under the sun), and quickly realized, 1:1 is not always the fastest configuration.

 

[iluvgillgill]

when encode because lots of memory is being read and write so latency is important thats why i said over higher frequency will gain you lower latency hence higher copy speed.

people mention 1:1 is because they judge base on the read and write speed which didnt really increase due to the higher frequency. and because the the voltage and timing they need to increase to keep the system stable doesnt worth what they will get.

 

[sportsfanboy]

^got ya

 

[iluvgillgill]

well just my view anyway. thats why DDR3 still used on 333/400FSB setup and dont get more then 10GB/s but latency is really low. there you go anyway.