The truth about Tras and its dangers....

[rizorith]

I'm really interested in any more info on this. I'm running 2-2-2-2 heh. But it's a new build so I don't know if I'm getting any corruptions or not. 10 hours in prime though, no errors.

 

I bumped my tras down to 6 and tested it and my memory performance was down about 5 percent, quite a big chunk but if 2 is causing problems I'll slow it down.

[nSaNe]

i agree... for me.. every single test ive ever ran on my bh-5 tras of 5 gives the best performance... but its no where near 6% better... maybe 2% tops. i also should add.. that it in no way shape or form has affected anything stability wise and has not corrupted any files whatsoever. ive been running 5-2-2-2 since before christmas @ ddr520+ and i have had no problems at all. i think someone @ muskin mistook a memo and wrote that article.. i would say maybe there is a specific IC that the memo refers too.. i dunno but ive heard nothing from any other ram company about this at all

[`clipse]

[nSaNe']i agree... for me.. every single test ive ever ran on my bh-5 tras of 5 gives the best performance... but its no where near 6% better... maybe 2% tops. i also should add.. that it in no way shape or form has affected anything stability wise and has not corrupted any files whatsoever. ive been running 5-2-2-2 since before christmas @ ddr520+ and i have had no problems at all. i think someone @ muskin mistook a memo and wrote that article.. i would say maybe there is a specific IC that the memo refers too.. i dunno but ive heard nothing from any other ram company about this at all

 

i agree with this. especially with the 2-2-2-5 for winbond ICs such as BH5/6.

i never found that 2-2-2-10 worked better for winbond. I found that the performace was worse in the middle ground between TRAS 5 and 10, but it peaked at both, and was slightly better at TRAS 5.

[Vista]

Now you guys have me paranoid. I notice that Mushkin link doesn't work anymore LOL

[Guest Blooz1]

i dont get how there is an arguement that tras is dangerous when no test seems to show it. if this is really a problem then how come memtest does not detect it as an error?

 

'clipse, the Memtest issue was discussed earlier on in the thread.

[Dexx]

Ive tested bandwidth in Everest using 2.5-3-3-X on my TCCD. using Tras of 6 is 1.4% faster than Tras 10 on my Ultra-D

[ViperJohn]

You cant stress ram more with regular windows usage then you do with memtest #5.

If it does not fail when it is in the RAM why would it fail when i put it to HD ?

 

Yes you can. Memtest 8 of the CD version.

 

Viper

[Angry_Games]

keep in mind the memtest program stresses a lot of ram but not that much cpu (dont get me wrong, it stresses...but nothing like it would with prime95, superpi, etc, and not a total system stress like a continuous loop of UT2004 timedemos which use LOTS of cpu/ram/vid)

[Jimbo Mahoney]

If it helps anyone, I had a real-world experience of data-loss:

 

This was on an Nforce 2 platform (Abit NF7 in fact).

 

Obviously, I was overclocking and using a low Tras. Long story short, the FAT of one of my drives became corrupted and I lost all the data on that drive (a LOT of MP3s).

 

This was a real pain, although it was a long time ago so I don't remember the details. I remember also reading an article about aggressive Tras settings here:

 

http://www.lostcircuits.com/advice/bios2/7.shtml

 

An excerpt:

 

If a precharge occurs (to wipe all the information from the wordlines for the next bank activate / row access), before the signal is strong enough to restore the original content in the memory cell, "tRAS is violated", resulting in loss or corruption of the data. Often enough, the corruption of data is not enough to crash a system immediately, however, once the data is written back to the HDD, the drive content is corrupted as well which can cause failure of the operating system or even "bad sectors" on the hard disk drive.

 

It is true and will for the most part go unnoticed. That is unless a specific portion of the drive gets a bit of corrupted data written to it, in my case, the FAT

 

I can't blame people for being skeptical because the corruption either goes unnoticed and builds up over time, or never actually does any real harm. But technically, it is happening.

[Guest scott]

Does anybody know what the Mushkin article "The Truth About Tras" said?

The link in the opening post is dead. I sure would like to learn this stuff.

[Jimbo Mahoney]

I'm intrigued too.

 

Have you checked the lostcircuits links I gave? They explain Tras and why it causes data corruption.

[uwackme]

it just explained why Tras needs to be set correctly based on the other memory settings to operate smoothly. If it is too tight (too low) the memory can end up running out of "time" on an access, and generate a retry, effectively WASTING time and slowing your OC down, rather than speedin it up. but lots of SYSTEM issues come to bare and so you need to test with real results.

 

 

On NForce2, Tras of "11" was deemed the best result based on HOW the memory controller functions. NF4 is different and has a different number....7,8,9? as the best overall result systemwide.

 

etc etc.

 

but testing will reveal the best results, use a variety of benches of mem, cpu, video, disk and see how it goes.