790 corruption issues?

[saaya]

I guess I need to check the spam folder.

i emailed you from 2 email addresses, please check

Answered in our quote section -
"We have two 74GB Raptors that are basically dead, they will power up but cannot be low-level formatted. The WD 250GB drive basically had the same problem. The two 500GB Samsungs will power up and repeat a click-clack pattern. The Samsungs have been returned for analysis as will the 74GB Raptors.

so its not just the wd drives suffering from problems, thanks, thats good to hear. well, NOT, but you get what i mean

I am getting ready to do the same for a pair of 150GB Raptors that I wrote off in December. However, those drives failed (no longer accessible) during RAID testing on the 780i board (usually I yell at myself when that happens as I have had far too many RAID 0 arrays drop a drive over the years). I did not think much of it until we started having these data corruption problems over the past six weeks.

so you had 2 raptors and two samsung drives fail? or even more?

Between Kris, Raja, and I there have probably been around 14~16 image reloads the past six weeks after overclocking. We fully except to trash the OS when exploring the boundaries of memory/fsb rates, but it might happen once or twice a month at best and is not limited to NVIDIA chipsets. However, all of these failures have been on the 780i/790i boards and we were not really pushing the systems except for two times when the drives failed or the images were corrupted.

yeah i remember raja having a good time with crysis
i recommended him to use ghost since i think he didnt use it by then cause he didnt have that many corruption issues before.
it def makes your life THAT much easier
well yes, thats the weird thing, isnt it? on intel and amd chipsets i blow a partition after 300-3000 boots at unstable speeds id say, and even then most of the time you can repair the install with the console or by replacing some files. i wonder if amd and intel and other chipset makers use some special ECC like technique to prevent corruption at unstable speeds or whether nvidia is maybe using bigger cashes/buffers which then dont get flushed if something goes wrong and bigger chunks of data are missing or corrupt.

The frustrating/perplexing problem is that the 790i testing with Kris resulted in some of the best overclocks we have ever experienced and they were 100% stable. We changed the settings to a normal overclock at 400FSB/1600 DDR3 and the images are corrupted or the drives went south. It is not repeatable at a given setting. We have seen results like ours in various forums so there is something amiss here, just trying to find it right now.

well thats the thing with nvidia chipsets... everybody who spent enough time with one or a few boards knows about this... from what i heard this inconsistency in something working once and then not working and suddenly working fine again is typical for nvidia chipsets ever since the 5 series. i really wonder what it could be, i can only imagine that some plls or some dividers lock every time the chipset is configured and initialized and for some reason it doesnt always happen the same way every time the chipset is reset.

but the bizarre thing really is that those problems happen at 1600, which is exactly where you would expect everything to work fine, as thats what you want everybody to run. if it would corrupt or give you a hard time at 1700, well, thats above the spec, too bad, we are working on it but for the moment only 1600 is stable. that would be fair enough. but the memory failing at 1600 while much higher speeds work fine is... odd...

i first thought maybe nvidia tuned the performance so at 1600 its tweaked really tight to win reviews and be fast at stock speeds, but if that would be the case with all those problems people are having id have expected them to release a relaxed stable bios by now. so it seems to be something else, or nvidia is really stubborn and doesnt want to losen timings to give up some extra performance

In all cases, we have had the memory settings set at something other than stock/default. I am still working with Derek as he has experienced several data corruption problems during SLI testing the past couple of weeks. I did not mention that until we figure out if his problems are related to ours."

well overall it seems quality control is falling apart industry wise... at least thats the impression i got over the past 5-10 years :/

which is really too bad... there are so many power users who are actually willing to BE abused as beta testers from manufacturers, and most users are more than happy to work with the manufacturers on a solution, but 95% of the feedback seems to be ignored or takes a month to reach the right people.