CAS Latency timing question

[WINDSHEAR]

does it help to loosen CAS (as well as other timings) when overclocking? When should you try loosening it?

 

[MadModMike]

does it help to loosen CAS (as well as other timings) when overclocking? When should you try loosening it?

It helps to loosen ALL timings when overclocking. After 10%, start loosening timings or increasing voltages.

~~Mad Mod Mike, fixin' the world 1 rig at a time

 

[WINDSHEAR]

ok, I'm using Mushkin Enhanced PC3500 DDR433,... 2-3-3 rated timings (running at 2-3-3-6). I loosened up the timings to 2-4-4-8, and ran at DDR464. I then tried 3-4-4-8. It didn't seem to make a difference. I still got errors running SuperPi (I don't bother with memtest unless it passes SuperPi, I'm lazy ) I can relax the timings all the way, and max the voltage, and the stick will still not run at DDR464 without errors halfway thru the 1M, if I'm lucky. Weirder yet, I have managed to run it at 2-2-2-6 @ DDR436, boot into Windows and start having errors in 1M, after having a BSOD the first time around, and upping the voltage to 2.8. That was a couple of weeks ago. Today, I thought I'd have fun, I put it on at those timings, at stock system speed (DDR400), BSOD the first boot, and registry error the second time, which freaked me out so I changed it back to normal in a hurry. I ended up having to reinstall some corrupted software.

1T versus 2T did not seem to make a difference. I leave it at 1T. I think I actually had a harder time passing tests in SuperPi with 2T. :?

oh yeah, and this is a P4 3.0Ghz rig with that crappy Jetway mobo :roll:

Also when I was overclocking my Corsair ValueSelect (which I actually got thru a 1M test once, at DDR464), CAS latency didn't seem to make a difference. 2.5 versus 3.

Of course, I"m not sure that any of that doesn't make a difference, it probably does, but I failed to notice much, probably do to the poor memory sticks I got. I'm looking for some RAM with BH5 or CH or UTT IC's now, but I don't even know what that means, so if you can explain the difference and what they do and how they perform, that would be greatly appreciated.

 

[fainis]

yeappp.lowering those timings will give you improved performance...
those timings are related to CAS, collumn address strobe, RAS row address strobe and relations between them...there are also terms like precharge, changing the memory page......(a memory page is a group of memory rows).....

gigabyte gives you good timings when top performance is enabled.....you can edit them manualy but most likely you will end up in instability....

the most important element that ...expecially beginners have to understand.......is that this timings are related to clock period T....

so everytime we mention CAS for example as 4...is actualy 4 multiplied by T.......

another element that also beginners have to consider is that this T can actualy be 1T or 2T......

ofcourse 1T is faster that 2T....that is the command rate....



i`ll give one example .... lets say CAS (the time between collumn select and data output I0...I7) is 4.........that means for a pc4300 compliant memory module 4/533 microseconds...this will only apply to 1T command rate...for 2T command rate this will be twice..........

of course that will not necesary mean a twice worst performance ...there are many elements in here..the other timings will came ...but the topic was about CAS.........

good luck everybody.....

 

[WINDSHEAR]

Oh, you've got the original Winbond CH-5s. (Not fun to play with )
Up voltage to highest possible vDIMM(2.9V?) then try 3-3-3-7 1T, see how that goes.

I had gotten desperate and put it up to 3.05volts (the max my mobo will allow), and timings of 2-4-4-8 1T. Then tried 3-4-4-8, same difference. Yeah, I really tried when I did it. Those CH-5's go nowhere. I've yet to try 3-5-5-9 (loosest timings my board will allow) @ 3.05volts.

and thanks fainis for your explanation of 1T (I already understood the rest). (what I didn't understand was how CAS latency affected overclocking and vice versa.)

 

[Da_Banig]

How F*** up can 3-3-3-1 2T be?

 

[chuckshissle]

Is it better to lower the frequency than loosen the timings? I have my ram's frequency lowered because the loose timings make my system somewhat unstable during heavy 3D application and I get choppy screen.

 

[WINDSHEAR]

different RAM sticks (with different ICs) perform differently. The system is slightly slower with looser timings, if you really loosen them, then I could understand having it choppy, but that sounds strange.

For you, it sounds like tighter timings are better, at least until you get better RAM.

 

[WINDSHEAR]

How F*** up can 3-3-3-1 2T be?

Impossible.

Yeap, that's how F**ed your brain could be.

:lol:

Next you're gonna overclock your P2 200Mhz to 20Ghz :roll:

ok, seriously, I think it might be simply that the BIOS options are mixed up, you are misunderstanding something, or the software is misunderstanding something, because that is an physical impossibility, and besides, motherboards don't even support that. Sorry, it's just not possible.

 

[Da_Banig]

How F*** up can 3-3-3-1 2T be?

Impossible.

Yeap, that's how F**ed your brain could be.

:lol:

Next you're gonna overclock your P2 200Mhz to 20Ghz :roll:

ok, seriously, I think it might be simply that the BIOS options are mixed up, you are misunderstanding something, or the software is misunderstanding something, because that is an physical impossibility, and besides, motherboards don't even support that. Sorry, it's just not possible.

Um... I set that in the BIOS and I also run CPU-Z and both shows 3-3-3-1 2T. I even got an image of it.... and I posted this topic before too lol

http://jlovesc4ever.tripod.com/RamTiming2.JPG
http://jlovesc4ever.tripod.com/RamTiming.JPG
(Need to copy and paste)

Two different type of RAMs, one is Kingston HyperX and another one is Corsiar XMS. Both can set 3-3-3-1 T2 and can still play hours of game without a single crash. How does that sh!t work? I don't even know lol but it does work. I tried them both on my old Asus P5AD2-E Deluxe, didn't try on my P5WD2 Premium.

 

[WINDSHEAR]

it looks like it's CL5-5-5-15. Interesting, though. Then too, that "1" is only Cycle time, not Bank Cycle time, which is what the (usually) largest number usually stands for..,.. I think.

 

[Da_Banig]

Yes I know that wusy, I was just saying its possible to have my RAM timing run at 3-3-3-1 2T. Not talking about speed nor frequency at this moment

 

[WINDSHEAR]

It looks like you ain't going anywhere with high Mhz so you might as well stick with highest possible frequency at lowest timing for best performance.
From experince that's 2.5-3-2-x with ~220Mhz as maximum.

hmm as I sit here I'm doing 218 with 2-3-3-6 timings, which has worked for me error free for a week now. Could I do 2.5-3-2-6 if I upped the CAS to 2.5? I was running it at 220 (2-3-3-6 2.7v) (this was last week I did all these experiments), and after 6 or 7 hours a program failed, which freaked me out, cuz from my experience, when there's memory problems, programs start failing, and then a BSOD ensues. I can't remember if I got a BSOD from it, but I think I might have.

 

[WINDSHEAR]

Yes I know that wusy, I was just saying its possible to have my RAM timing run at 3-3-3-1 2T. Not talking about speed nor frequency at this moment

well, that's certainly interesting, lol. but it does run slower with that kind of timing. Can you run it at 1T? Now that would boost performance.

 

[WINDSHEAR]

hmm as I sit here I'm doing 218 with 2-3-3-6 timings

Yours is a better timing. Stick with your 2-3-3-6 unless you can break 220Mhz with 2.5-3-2-6.

Alternatively try 2-3-2-6 with your current clock speed and see how it turns out.

ok, now I'm sitting here with 218 2-3-2-6 1T. Did 512k, 1M, and then 2M, with no errors. I'm pretty convinced it's stable. If it passes SuperPi, it's good.

Now, would it be better to try 2.5-3-2-6 at 220? would it be faster than 2-3-2-6 at 218?

I'll do some more testing later, I guess.

 

[Da_Banig]

Dumb question, how can we find out the x-x-x-x XT ? (The bold and italic TIMING)

 

[WINDSHEAR]

thanks for hijacking my thread :roll:

 

[ZOldDude]

....I still got errors running SuperPi (I don't bother with memtest unless it passes SuperPi, I'm lazy )....

Well if it is not passing SuperPI the problem is not the RAM...it is the CPU.

Put the ram back to the higherst mhz it would run at with stock times that SuperPi first failed....raise the CPU voltage and retest!

 

[WINDSHEAR]

no, it's a mild overclock. And, I have passed SuperPi with other, better, RAM. It's not CPU, it's RAM. SuperPi is very sensitive to memory, (it stresses memory, FSB, CPU and prob a few other things too).

The CPU passes other benchmarks just fine, it is only in memory sensitive applications that I have trouble.

anyways, thanks for your suggestions