Jump to content

The truth about Tras and its dangers....


Guest shaolin95_merged

Recommended Posts

Guest shaolin95

Depending on your other settings 7 may or may not be risky...still at say 2.5+3= 5.5 + 2 = 7.5 so tras 7 is pretty much within the safety range but then again tras 8 is safer and works just as good if not better for me. The past example is for people using the very common 2.5-3-3-x timings.

Regards

Share this post


Link to post
Share on other sites

  • Replies 100
  • Created
  • Last Reply

Top Posters In This Topic

Guest shaolin95

I decided to play it safe by using 8 or 10 since the tiny difference you can get at least in my case is more than likely the common error margin of bechmarking.

Regards

Share this post


Link to post
Share on other sites

I have tested my settings with many Benchmarks (prime95, memtest, PCMark04) without any errors. I think 2.5-3-3-7 it's safe. I have luck but how long :D. I will test 2.5-3-3-8 soon. Will see what happen :)

Share this post


Link to post
Share on other sites

Same here, just love my 2.5-3-3-7 2.9v setting and is a little faster than my 2.5-3-3-6. 2.9v

Faster in 3Dmark03,05, SuperPi and SiSoft Sandra bandwith and CPU test.

 

Next step in testing my rig: 2.5-3-3-8 2.9v

 

All of this at 285 x 9 1.35v x 123% (1.6v CPUz and SG).

Share this post


Link to post
Share on other sites

Well I just got back and tried 266x9 with 5-3-3-3 again but with 11-15 instead of 9-13 for tRC and tRFC.

That worked fine, not as stable as my old setting though of 3-3-2-3 9-13.

 

It does seem that tRFC is off on my rig a tad, probably supposed to be 14 but that's not the whole prob for sure.

 

 

Anyways, my wisdom lol :P.

For tRAS I allways just go from rated to just above or matched with the rest of the main timings.

Seems like an edge thing I dn, was allways faster that way.

I allways bench it one step at a time ;), though I did'nt when I stepped down from 4-4-4-4 13-15.

 

For instance 3-2-2-2 was faster then 5-2-2-2 for me on my old corsair back in the day.

 

I dn what to use for adding it up really.

Perhaps CAS+tRCD-Write, just a guess.

 

For tRC and tRFC it seems a bit easier.

tRAS+tRCD(maxxed ver, say read is higher so use that value)+tRP

 

For me = 3+3+3 = 9 tRC

 

For tRFC:

tRC+CAS+CMD

 

For me = 9+3+2 = 14 tRFC

 

Just a guess.

I know they're suppose to go somehting like that.

Actually that looks wrong but ohwell, I know that the alpha timings and drive/slew rates can be a huge part of them sometimes.

 

All our stuff is diffrent, like nf2 and nf4 mobos, way diffrent because for instance they can do a low tRC and tRFC from the bios, while the nf2's are limited to 9 min from the bios(0 min with wpcredit).

The built in memmory controller on the cpu for the nf4 mobo, blah stuff like that.

 

I've seen pretty high normal timings and pretty low tRC and tRFC too on those nf4 mobos.

 

There's all sorts of things you can do if you try for a while to find a nitch.

 

 

Anyways the higher the normal timings the higher tRAS is gonna have to be.

At min it has to match the rest of the normal timings, that's at min, it can never go lower.

 

The higher all the normal timings are the higher tRC and tRFC is gonna have to be, sorta.

You up tRAS and you'll have to up tRC and tRFC if you got them tightened down to the max allready.

 

Like I said before though, you can probably throw all this out the window, you never know what can do what until you try.

Though the tRAS rule of min=matched should allways be in place.

Share this post


Link to post
Share on other sites

Guest shaolin95

Just remember one thing, that what seems faster in a benchmark could be also whats going to corrupt your driver later on cause you need to realize that a benchmark is normally so short in duration that there will be not much time for errors to show up thus giving a false impression of safety. Just my thought of course ...

Share this post


Link to post
Share on other sites

Guest shaolin95

I noticed that a lot in Super Pi where very thight timings will give me better 1MB runs but then the 32MB run will be bad compare to other slightly looser timings.

Share this post


Link to post
Share on other sites

I decided to re-check my rig for all of that stuff.

 

1st I up'ed tRAS to 15, tRFC to 31 and tRC to 15.

 

I then checked tRAS, 10 was the highest scoring.

 

I then checked tRC and 11 was the highest scoring.

Then checked tRFC and 14 was the best score.

 

I then check for stability, to see if it was higher.

At these high room temp(becoming summer) it was able to get to memtest #5 and fail a tiny bit, not 100's of errors or anything.

That was 275x5, 10-3-2-3 11-14.

 

Using my 3-3-2-3 9-11 it would lock on the mem count most of the time.

That's a sure sign of slight stability increase.

There was'nt barely or no diffrence in 266x9.5 however.

 

Though 266x9 was memtest stable like my normal timings.

 

So it's more stable that way...

I goto check the bandwith to directly cvompare with my original timings and man man man.

My old timings blow this away.

 

So I suppose there is vadility in people saying they up'ed there tRAS for more bandwith.

I've actually seen it.

It still does'nt beat good old tight timings though :).

Share this post


Link to post
Share on other sites

Hehe :), I'm on the nf2 chipset btw ;).

I dn need to worry about odd things just yet...

 

Though I am looking forward to it, odd things on the nf4 heh :).

 

This thread was defently an allright thread :), I did learn somehting new.

That using a high tRAS with the correct timings can increase stability and have decent speed.

 

I'm looks as if others here have re-checked some of there timings it seems too :).

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now

×
×
  • Create New...