Are memory errors concerning?

[elputohuma]

are memory gpu errors concering while overclocking? i got some memory erros on hwinfo with the memory overclocked, not much around 80-100, with the memory overclocked i gain 5fps on heaven benchmark so im not losing performance, i just want to know if its ok having some errors or should i drop the OC

 

[Peter Martin]

errors are generally bad. 5fps is nothing as far as a gain. and for what, sacrificing stability and proper performance over a long life. I have to ask you, is it worth it for 5fps?

ERRORS ARE BAD.

just take a peek around here with all the OC going on and all the massive amounts of GPU problems people CAUSE THEMSELVES

 

[Darkbreeze]

ANY memory errors, whether they are system memory (RAM) or GPU card memory (VRAM) are bad. Same applies to CPU overclocking. ANY errors indicates instability and you should either increase the voltage and re-test or decrease the overclock setting(s) and retest. If you cannot eliminate the errors, then go back to the standard configuration. You do not ever want to continue running the system, other than for testing, if there are any kind of CPU or memory errors occurring.

 

[kerberos_20]

ram is storage, if u have errors there, those data will get corrupted, so u may experience game crashing, system freezing, visual glitches, performance degradation, and overaly it can get damaged due to excessive clock/voltage

 

[elputohuma]

i heard its not concerning unless you have thousands of errors per second

 

[kerberos_20]

but u dont know if u have thousands of errors or not
gddr5 cards got ecc so u will see less errors than with gddr3 (at the cost of performance when it kicks in)

 

[elputohuma]

well i drop the clock to 2100 and still have some errors 1-2, only 2075 gave me 0 errors

 

[Darkbreeze]

i heard its not concerning unless you have thousands of errors per second

THIS, is patently false. One "recognized" error, that can be detected by a stress utility, memtest or Windows itself, likely indicates that the current level of stability is most probably introducing tens or hundreds of micro-errors into the system file structure including OS files, personal files, game files, etc.

It doesn't take major errors that cause blue screens or freezing to create havoc. A one here where a zero should be, a couple of zeros there where it should have been a one, and before you know it the whole works is corrupted.

Obviously, this is not entirely the same as it would be if it were system memory that was involved, but there are a great many games and applications that make use of the graphics GPU for other things than simply rendering game data, so it can be a bit more far reaching than simply what you see in the game or on your screen. If you do much, or any, encoding, transcoding or other GPU intensive tasks it's a really, really bad idea to allow any instability at all.

If all you do with your graphics card is play games, and you have full on backups of all your game data elsewhere and don't mind reinstalling games or the OS occasionally when unexplainable problems start showing up, then perhaps the errors aren't as critical for you. For most people, that will not be the case though.

One error is exactly the same as one thousand errors. You don't have to cut off all your fingers in order to bleed to death. Just one finger cut off might do it.

 

[elputohuma]

i overclocked all my gpus and this is the first time i get gpu memory errors thats why i ask, when a memory oc was unstable on my old gpus i got a grey screen crash with sound loop but never had errors, btw i did an RMA today, because one fan is making clicking noise between 50%-60% speed, maybe, also im a bit tired of the random screen flickering of idle with AMD, maybe the next card overclock more than this one xD

 

[Darkbreeze]

Exactly. The fact that you are getting errors might suggest that there was a problem with the card anyhow. It's never a good idea to ignore errors for any reason. Errors don't exist on correctly functioning hardware, at least not to a level that could be detectable through most testing utilities. Everything might error out eventually, but getting errors within the standards that are commonly accepted as acceptable risk or not acceptable risk, are to risky. LOL.