790 corruption issues?

[Broly]

My question is, how can I get it fixed/
I have instant product replacement warranty at my store, and I want this motherboard replaced or fixed. How can I prove to these lamer techs that are going to look at my PC (They aren't very saavy with issues and these forums) that I have this issue? Should I tell them to format and reinstall 10x and see how it BSoDs? I'm getting multiple hangs that are very annoying, and sometimes a BSOD booting into windows during the splash screen.
Running vista x64 with 1600 mhz viper and qx9650 all stock

[5010]

hmmm very interesting!
could you post some more details about this?
did you adjust the voltages in bios or with a mod?

Hey Saaya,

Nice to see someone finally taking notice!

Seems to me the issue is pretty widespread, plenty of ppl over at the EVGA forums are complaining about freezes/hangs/lock-ups during video playback and during some other tasks aswell. I have multitudes of terabytes of HD content so the issue was apparent to me almost at once and I found that just sitting idle in the CRYSIS game menu's would cause the same hang-ups(during one of these hangs my CRYSIS installation was !!C-O-R-R-U-P-T-E-D!! but the rest of my OS was left intact!) I began investigating the issue very carefully and I found through trial and error (ie: infinite looping x264 + stop watch) that I could control and duplicate the system hangs and data CORRUPTION by manually adjusting the FSB + GTLREF Voltages alone.

I did my adjustments from the BIOS since this is my personal computer and I have it installed in a case nestled into my desk etc.. I do work in a small boutique and I'm resposible for system design and overclocking(all of our rigs are OC'd) but in the same breathe I've got this 790iU for myself and so that I can get to know it very intimately for all future commercial builds we will do based on it...

Cheerz

[dinos22]

can you provide your full testing notes here please

[3oh6]

To check if your system contains corrupted data:
System file integrity scan:
- search for "cmd"
- right click-run as administrator
- type in "sfc /scannow"
after scanning it will tell you if your windows is corrupted and weather it was able to fix the corrupted data(most probably not).

so whats suppose to happen here with this check? when i do it, it demands to have the XP SP2 disk in the drive then it goes through its little progress bar and the prompt disappears. nothing ever comes up after the progress bar. i am assuming then that it didn't find any errors?

reason i ask is because i am crushing XP installs one right after the other trying to push this memory kit i am working with over 1050MHz HCI stable. but when i do the file check, it does what i outlined above. i am pretty sure the borked Windows are simply from pushing memory too far and not the corruption issues being discussed here.

[5010]

Notes are far from scientific and far from complete.

EVGA 790iU, P03 Official BIOS
Q6600 @ 3200MHz/1600FSB 8x Multi, 1.3125V BIOS/1.22V load (familiar chip from several previous builds)
2GB Patriot Viper XP3-12800 @ 1600 8-7-6-1T (Sub timings Auto) 1.625V BIOS/1.82V Everest - Well tested RAM
2x 8800GTX OC2's (626/1450/1000)
HDDs: Raptor-X(OS), Seagate 500GB 7200.11, WD 500GB Raid Edition, WD 250GB AAKS (Multiple 350GB+ Multi-file transfers done without errors between
drives)
Pioneer 212DBK SATA DVD/RW

SPP: 1.35 (Less produces various BSODS)
MCP: 1.50V
Spread spectrums and all CPU features except Virtualization: disabled

Testing done using an inifinte looping 1080p x264 in Media Player Classic & a stop watch.
FSB V GTL1 GTL2 GTL3 GTL4 Time Until Hang(h.m.s)
1.40V Auto Auto Auto Auto 00.00.53 (CRYSIS was corrupted here after about 10 seconds in the game menus)
1.35V Auto Auto Auto Auto 00.01.41
1.30V 15mv 15mv 15mv 15mv 00.00.10
1.25V Auto Auto Auto Auto 00.02.32
1.25V 05mv 05mv 05mv 05mv 00.03.46
1.25V 10mv 10mv 10mv 10mv 00.03.37
1.25V 15mv 15mv 15mv 15mv 00.07.16
1.25V 20mv 20mv 15mv 15mv 00.09.11
1.25V 20mv 20mv 20mv 20mv 00.14.38
1.25V 20mv 20mv 25mv 25mv 00.21.32
1.25V 25mv 25mv 25mv 25mv 06.11.51
1.25V 30mv 30mv 30mv 30mv 48.00.00 ( Approx) | w/o loop
1.25V 35mv 35mv 35mv 35mv Current ~ > 48.00.00 | w/o loop


It's a lengthy testing procedure for the most part but it is showing a pattern as you can see from the results I transcribed off the back of an envelope where I originally wrote them down.

Symptoms and triggers of this hang are: Sudden frozen video output accompanied by a buzz/squeel or looped audio from your speakers. The hang is most often triggered during video playback but on occasion it occurs during gaming and also in the CRYSIS game menu's which seem to place an odd stress on a system.

Other notes: I have been unable to reproduce the hang while the system is loaded(Quad Prime95) but I have been able to remove load while video playback is in progress and have the system hang immediately or within 30 seconds of stopping the Prime load. SLI was disabled during all testing but the hangs are reproduceable during gaming with SLI enabled.

Page file is turned off on my system.

[dinos22]

FSB V GTL1 GTL2 GTL3 GTL4 Time Until Hang(h.m.s)
1.40V Auto Auto Auto Auto 00.00.53 (CRYSIS was corrupted here after about 10 seconds in the game menus)
1.35V Auto Auto Auto Auto 00.01.41
1.30V 15mv 15mv 15mv 15mv 00.00.10
1.25V Auto Auto Auto Auto 00.02.32
1.25V 05mv 05mv 05mv 05mv 00.03.46
1.25V 10mv 10mv 10mv 10mv 00.03.37
1.25V 15mv 15mv 15mv 15mv 00.07.16
1.25V 20mv 20mv 15mv 15mv 00.09.11
1.25V 20mv 20mv 20mv 20mv 00.14.38
1.25V 20mv 20mv 25mv 25mv 00.21.32
1.25V 25mv 25mv 25mv 25mv 06.11.51
1.25V 30mv 30mv 30mv 30mv 48.00.00 ( Approx) | w/o loop
1.25V 35mv 35mv 35mv 35mv Current ~ > 48.00.00 | w/o loop
.

have you pushed past 35 (>> interesting part is seeing the jump to 25 >> is that replicable)
are you setting positive or negative GTL values

can you please do negative value testing as well (if possible)

btw this is really good research and you might have just saved some ppl at least from trouble of corruptions


also i am surprised your CPU needs so little voltage
can you give it a bump too

[BenchZowner]

My data corruptions happened with stable hardware settings ( it could run Orthos, and any benchmark just fine ) [ and the RAMs were tested with memtest test #5 for 50 loops before I booted to windows ].

[dinos22]

My data corruptions happened with stable hardware settings ( it could run Orthos, and any benchmark just fine ) [ and the RAMs were tested with memtest test #5 for 50 loops before I booted to windows ].

Bill

so did 5050 yet he had specific problems with the HD playback method

so clearly not completely stable

GTL Ref is an area of bios which has never been covered thoroughly by anyone and probably deserves more attention seeing the changes that 5050 is experiencing

wish i had a spare week. i'd go crazy with the numbers while listening to iTunes comedy radio stations and kill myself laughing while testing

[BenchZowner]

Will check it out soon.
I'm currently memtesting the sh*t out of it
Can't do a thing here, limited by PCI SuperPi VGA ( all the cards that I have here with me are Twintech's and are incompatible )
Will have to wait some days for my VGAs package to arrive.

Edit: @5010 I'd enable pagefile just to make sure it's not windows/apps f*ing with me, because some apps have problems when you have no pagefile

[3oh6]

wish i had a spare week. i'd go crazy with the numbers while listening to iTunes comedy radio stations and kill myself laughing while testing

i hear ya, things at work are getting slow again so i might have a week on my hands soon enough. wonder what i can do with it

BTW, found a way to get POST with 1.500v set for SPP voltage and managed validation at 1100MHz...with 2.125v vDIMM set in BIOS
it froze while i was trying to run Everest bandwidth.

validation link

[5010]

have you pushed past 35 (>> interesting part is seeing the jump to 25 >> is that replicable)
are you setting positive or negative GTL values

can you please do negative value testing as well (if possible)

btw this is really good research and you might have just saved some ppl at least from trouble of corruptions


also i am surprised your CPU needs so little voltage
can you give it a bump too

These are all positive GTL's, Did some(very little) negative GTL tests but it was before I started keeping notes and from what I remember it was all basically instant hangs. Wouldn't mind probing further around but if you extrapolate my chart to include all GTLs + all FSB in all combinations its gonna get pretty huge and this is only for 1600FSB, at best it's gonna end up patchy.

This is the second Q6600 I've had which works at these speeds and volts (0.01V difference between them actually), this one I kept since it ran a heck of alot cooler(almost 10C) and with only a maximum of 1C difference between the cores Ex: 39C-38C-38C-38C.. I could bump it sure but it's run like this in two other boards for about 6 mnths, she;s pretty well proven

[dinos22]

I could bump it sure but it's run like this in two other boards for about 6 mnths, she;s pretty well proven

just for testing purposes

later you can tweak it down all you like hehehe

[5010]

Will check it out soon.
I'm currently memtesting the sh*t out of it
Can't do a thing here, limited by PCI SuperPi VGA ( all the cards that I have here with me are Twintech's and are incompatible )
Will have to wait some days for my VGAs package to arrive.

Edit: @5010 I'd enable pagefile just to make sure it's not windows/apps f*ing with me, because some apps have problems when you have no pagefile

Im on XP and haven't used a pagefile since I ran socket A :P except for when using one or two stupid programs I've come across that required one.

--

Dinos,

How well learned are you when it comes to GTLREF Voltages? I've studied the theory alot but I've never got my hands dirty like this.

[dinos22]

i hear ya, things at work are getting slow again so i might have a week on my hands soon enough. wonder what i can do with it

BTW, found a way to get POST with 1.500v set for SPP voltage and managed validation at 1100MHz...with 2.125v vDIMM set in BIOS
it froze while i was trying to run Everest bandwidth.

validation link

http://img.photobucket.com/albums/v2...00_valid-1.png

good work man

i've benched some Crucial sticks at 2200MHz CAS8 but 32M was stopping at 4th loop from memory

but with 2.275v in bios lol

[dinos22]

Dinos,

How well learned are you when it comes to GTLREF Voltages? I've studied the theory alot but I've never got my hands dirty like this.

my friend i have very little knowledge in this area and have only really read some hints from other OCers

from what i thought i knew negative values are the ones to go for when pushing high FSB but that is such a generic statement and one i can't back as i never did any actual testing to prove either way

OCing is a lot of fun and can be extremely time consuming (even more when you roughly know what you're doing lol) as every Q&A has more Qs hehehe

i think you've clearly touched onto something here and if it is not too difficult i would like ot see you replicated the results with Auto values and then going from low values to 25 and see if stability will still go from 2-20 minutes to 6+ hours

[5010]

my friend i have very little knowledge in this area and have only really read some hints from other OCers

from what i thought i knew negative values are the ones to go for when pushing high FSB but that is such a generic statement and one i can't back as i never did any actual testing to prove either way

OCing is a lot of fun and can be extremely time consuming (even more when you roughly know what you're doing lol) as every Q&A has more Qs hehehe

i think you've clearly touched onto something here and if it is not too difficult i would like ot see you replicated the results with Auto values and then going from low values to 25 and see if stability will still go from 2-20 minutes to 6+ hours

I've read similar things from others aswell but when I read over some techical papers on Gunning Transceiver Logic I discovered that it doesn't quite work like that..

Basically from what I understand it helps govern the data between the CPU and NB, where FSB VTT = the the maximum voltage for a "1/On" and the off is 0.04V. The GTLREF voltage is like a line drawn in between these two voltages, whatever signal is above the GTLREF is a "1" and whatever is below is a "0". <- BASICALLY, but its a little hard to explain it in a paragraph.

Adjusting GTLREF is very much like tuning the injectors in a car engine.

I happen to OC for a living(Or atleast a large part of my work) been at it for a very long time. Last time I fully explored a motherboard I spent nearly 3 months finding the perfect 24/7 sweet spot on the 680i and likewise it's gonna take awhile but I intend to do the same with the 790iU.

[dinos22]

great just the right person we need in the thread (and in general) heh

keep up the good work

[RLM]

Hey Saaya,

Nice to see someone finally taking notice!

5010, i did read your thread with interest, and I know it's a bit frustrating that it seemed to go by un-noticed.

Personally I tried your values and it appears very similar variations, when that didn't work. i.e. +15mv - +30 mv, with similar FSB voltages.

The reason I did'nt acknowledge your post is that as mentioned it didn't make any difference to my stability issues.

I know from using the DFI in the past how important the GTL refs can be with Quads, and also how working values can vary even with the same hardware. However until some of the other issues are sorted bios wise, it's a bit of a fruitless exercise it seems. Especially taking into account EVGA have acknowledged a Q6600 issue.

RLM

[saaya]

Who knows, maybe that BIOS is the miracle evga users are waiting for and it deals with the corruption too.
But now that I have seen the poll results and with that many people NOT having corruptions at all, I would REALLY like to try this board on my own and test it myself.

i dont know man, according to the poll its still 50/50 for the ref boards and 60/40 for the asus boards. if half the people DO have corruption issues thats a very large number. what concerns me is that some people do not have any coruption issues whatsoever even when overclocking. so it sounds like only some boards are affected. if thats the case then it really sucks...

i wanna try 790 too to make a map of the stable and unstable fsb/mem combinations

Hey Saaya,

Nice to see someone finally taking notice!

Seems to me the issue is pretty widespread, plenty of ppl over at the EVGA forums are complaining about freezes/hangs/lock-ups during video playback and during some other tasks aswell. I have multitudes of terabytes of HD content so the issue was apparent to me almost at once and I found that just sitting idle in the CRYSIS game menu's would cause the same hang-ups(during one of these hangs my CRYSIS installation was !!C-O-R-R-U-P-T-E-D!! but the rest of my OS was left intact!) I began investigating the issue very carefully and I found through trial and error (ie: infinite looping x264 + stop watch) that I could control and duplicate the system hangs and data CORRUPTION by manually adjusting the FSB + GTLREF Voltages alone.

I did my adjustments from the BIOS since this is my personal computer and I have it installed in a case nestled into my desk etc.. I do work in a small boutique and I'm resposible for system design and overclocking(all of our rigs are OC'd) but in the same breathe I've got this 790iU for myself and so that I can get to know it very intimately for all future commercial builds we will do based on it...

Cheerz

nice! if its gtlref and vtt then this makes sense! that would explain why for some people the board works and for others it doesnt since every cpu reacts different to gtlref and vtt

so for some cpus those volts are too high, for some cpus they are too low...

i have a feeling your instabilities might have been cpu related though and not chipset/mem related like the corruptions most people here experience.
but then again, who knows, maybe the boards are different and so everybody is actually experiencing cpu instabilities cause on the 790 boards you need other gtl and vtt voltages to reach the same cpu speeds stable...

from what i thought i knew negative values are the ones to go for when pushing high FSB but that is such a generic statement and one i can't back as i never did any actual testing to prove either way

i think thats only the case on asus boards, they have relatively bad gtl implementation and have way too high tgl once you increase vtt.
gtl is supposed to be 66&#37; of vtt, but with higher vtt, its more like 80% or even more.

remember fsb swing? its a really low voltage, ive been thinking this could be the 0.4v "line" that is used to measure a logical 0. fsb swing is supposed to be 1/4 vtt. so we have vtt, gtl ref, and fsb swing, which from my understanding are used to interpret a 1 or 0.

high = vtt
ref = cpu gtl
low = fsb swing

1200mv
800mv
400mv

i dont know how exactly gtl works, but i assume a "wave" has to be between 800 and 1200mv to be detected as a logical 1, and between 800mv and 400mv to be detected as a 0. if we increase the clockspeeds the transistors will push the voltage up and down faster, meaning the "waves" will not be as high and the valleys between the waves will not be as low anymore. so then its hard to tell when its a 0 and when a 1. so in this case reducing vtt would make it easier to detect 0s and 1s again. i think thats why intel reduced vtt to 1.2v in their latest cpus from 1.4v originally.

i dont think we need fsb swing and vtt, what we really need to adjust is just gtl voltages. as long as a "wave" is above gtl, it will be a 1, and if its below it will be a 0, it doesnt matter how close it is to the vtt line and the vfsb swing line. ...i think!

but having vtt and gtl adjustable is nice, cause we have 2 ways to adjust gtl then. we can either widen the window which will increase gtl milivolt wise and push the window up, or we can adjust the ratio, increasing or decreasing gtl in rough steps in milivolt. but of course, the best would be if we could adjust gtl ref in very fine steps with dividers, since there are other voltages related to vtt, such as nb gtl

5010 from your results it looks to me like your board has too low gtl settings, its most likely below 66%, and when you increase vtt its probably even more off and higher, just like on the asus boards i meassured. are you using the asus S2E or a 790 ref board? asus right?

[lopri]

Let me post the way I would reproduce the corruption on 680i. I know that this thread is about 790i, but I suspect the root cause could be similar. If you think my example is not applicable for 790i please feel free to ignore this post.

Under XP:

- Need 2 partitions (physically different drives or same drive)
- Copy over files/folders that are 15GB+ from partition A to partition B
- immediately after start another copy procedure of 10GB+ files/folders from the other partition (partition B) to the partition A
- Now open the built-in defragment in Windows (or any other defrag utility) while the copy procedures are taking place.
- Start defraging. Once finished, immediately re-run the defrag. You can alternate partition A and partition B to your liking. Keep defragging for 10 times. (shouldn't take much time after the first defrag)

This almost always resulted in lock-ups or BSOD in my 1st AND 2nd 680i boards, after which event everything went south.

I know the above scenario was somewhat extreme, but after experiencing random corruptions I had to make sure things are stable for a long-term usage.

[eternal_fantasy]

so whats suppose to happen here with this check? when i do it, it demands to have the XP SP2 disk in the drive then it goes through its little progress bar and the prompt disappears. nothing ever comes up after the progress bar. i am assuming then that it didn't find any errors?

reason i ask is because i am crushing XP installs one right after the other trying to push this memory kit i am working with over 1050MHz HCI stable. but when i do the file check, it does what i outlined above. i am pretty sure the borked Windows are simply from pushing memory too far and not the corruption issues being discussed here.

I have only performed the check on Windows Vista, so I'm not sure what the result will be under XP. But doing a quick google search for "check corrupt system file xp" should find some ways to do the same in XP.

The result under Vista after typing sfc /scannow if there are no corruption found:


And when there is:

[slim142]

i dont know man, according to the poll its still 50/50 for the ref boards and 60/40 for the asus boards. if half the people DO have corruption issues thats a very large number. what concerns me is that some people do not have any coruption issues whatsoever even when overclocking. so it sounds like only some boards are affected. if thats the case then it really sucks...

i wanna try 790 too to make a map of the stable and unstable fsb/mem combinations

Thats what I cant understand neither. I dont understand how some people are overclocking 500Mhz+ and are not getting corruptions. They're probably messing with GTLRefs and they dont think it is actually saving them from the corruptions?

Btw, what board do you plan on getting?

[RLM]

Isn't it a case of certain hardware/brands working better in these boards than others?

For instance looking through posts it seems 8400's are doing well currently.

Or is that not the case?

RLM

[saaya]

lopri, so while it still copies you defrag the partitions you were copying to?

slim, im planning on getting the ref board, and maybe try out the asus board too, but for now ill try the ref board first. the asus one seems to be worse, so if the 790 ref is ok, then i might move on to the asus. if the ref board already gives me a hard time i know i dont wanna mess with the asus one

[3oh6]

I have only performed the check on Windows Vista, so I'm not sure what the result will be under XP. But doing a quick google search for "check corrupt system file xp" should find some ways to do the same in XP.

thanks for the screens. in XP i just get a progress bar window saying it is checking and then it disappears at the end. in the CMD prompt, it never says anything. will have to look into it i guess for XP.