Tight timings vs higher clock speed for RAM?

[Krushull]

Just wondering if its best to have your RAM set to all the fastest settings and o/c it as much as you can or to lower all the settings and try to reach the highest clock speed?

 

Spent awhile getting my RAM to the fastest timings and pretty much all the fastest settings in the expert bios for RAM at 440mhz but wondering if its worth trying clockspeed at all?

[HokieRif]

It's best to read the overclocking guide if you haven't done that already - it'll answer most if not all of your questions...

 

http://www.dfi-street.com/forum/showthread.php?t=28049

[Krushull]

I did read most of it today actually must have not gotten to the RAM part though

[Guest Dynamic]

What are the settings? Vcore/Chipset/HTT/LDT voltages/VDIMM? You're running two sticks of 1GB ram, which will not hit higher more likely then regular 2X512 DDR400. What kind of ram is that by the way? BH-5,TCCD,TCC5, any clues?

[Redbeaver]

http://www.dfi-street.com/forum/showpost.p...2&postcount=204

[Krushull]

Sorry its 2x 512 they are TCCD sticks, im pretty impressed with how they run on tight timings and I have a feeling I can clock them fairly high just wondering if its worth the effort

 

Im allready at DDR440 with all the fastest settings any higher and I get corrupted graphics in POST though, by the looks of the results of the above post it doesnt seem worth it really but Im hoping I get can get much more than 466

 

All voltages are stock except the RAM I set to 2.8v I do have to up the vcore if I divide the RAM again and up the CPU but dont run it at that until the temp reader in BIOS is fixed

[Nogin]

Angry_Games did some testing on this too

 

Lookie Here

 

Nogs

[colelt1]

My chip can do 2.5+ but it needs 1.6v, so I was shooting for 2.4-2.5ghz on 1.45-1.55v. I cant get my memory controller to do DDR600+ (300x8), so I gave up on that last night. So now im shooting for the DDR550. My ram is rated for 550 2.5-3-3-7 and 600 2.5-4-4-8, but the guide I used said if my memory controller can run them at those settings, then I have to use the slower setting (I even tried everything on higher numbers/auto). So Now im worknig on 275x9 = 2.475ghz and will tighten up my memory also.

 

The point I am trying to make, is follow the overclockers guide, find your max cpu overclock, then your max ram speeds, the find a set of those numbers that you are comfortable with, then tighten up your ram.

 

I would rather have my ram at DDR550 at 2.5-3-3-6 than to have my computer struggling with DDR600@ 3-5-5-10 (plus I get an extra 75mhz out of my cpu).

[Krushull]

yeh im sitting at 292x9 2.6ghz at 1.6v and RAM at 1.84v after some tweaking today bit scared though seeing as I cant measure the cpu temp grr, seems stable though no crashes in BF2 which is very sensative whith the new crappy patch

 

I dont think I will bother with max RAM clock seems like theres not much of a boost over my max clock with best timings, sounds like a 2.0L Turbo car vs a V8

[Thasp]

I did some tests myself and came to these two conclusions.

 

a) In the VERY limited applications where timings/bandwidth makes a difference more than 5%(winrar and quickpar), timings mattered more.

 

B) RAM overclocking is not worth all the time you'll put into learning the settings for the perfrmance increase.

[smolt]

B) RAM overclocking is not worth all the time you'll put into learning the settings for the perfrmance increase.

now thats a good one thats why there a data base so you can use someone else hard work

[nYdGeo]

I did some tests myself and came to these two conclusions.

 

a) In the VERY limited applications where timings/bandwidth makes a difference more than 5%(winrar and quickpar), timings mattered more.

 

B) RAM overclocking is not worth all the time you'll put into learning the settings for the perfrmance increase.

 

I want very much to agree, but there are so many instances of testing done by so many sources showing that bandwidth does matter, especially with the A64 processors having the on-die memory controllers. This is said to avoid the memory bottlenecking seen with other hardware. From what I'm reading, even with socket 939/NF4 stuff, 300 x 10 is alleged to be the better performer over 200 x 15.

 

My problem is that these things are always done with benchmarks. Even when using real games and not things like 3Dmark'0x, they use benchmarks, which are usually standardized 'demos' for lack of a better term. These do work the graphics in a standardized way, but are they true representatives of real world gaming (as an example) where the machine isn't just replaying a recorded sequence, but the CPU is also performing all of the calculations required to play current games? it would seem not.

 

For this one reason I'v always liked the (now old) UT2004 Benchmark; you actually play the game in it, making it a more realistic benchmark of the machine's capabilities. In synthetic benchamrks, my old rig scored better at 250 x 10 than at the RAM-normalized 200 x 12.5, but my best overall in-game performance was at 228 x 11. How odd that a compromise should appear to provide the ultimate performance.