Overclock.net › Forums › Intel › Intel Motherboards › 790i Ultra NVIDIA chipset - Still rockin' in 2015 / 2016 / 2017.. will it ever die?
New Posts  All Forums:Forum Nav:

790i Ultra NVIDIA chipset - Still rockin' in 2015 / 2016 / 2017.. will it ever die? - Page 20

post #191 of 595
Okay, that's weird, since I set it manually to 1066MHz 9-9-9-24-33-1T and checked it it with Memtest before I started the OC.
Guess I will follow your advice now and let Memtest with at least T8 run each time I change the FSB, although it's time demanding. rolleyes.gif

Thanks, will report back later and see what Memtest says.

Oh, and the 1.35V NB were fine with the C2D up to at least 410MHz FSB. I will see.

Edit: Memtest went trough without errors, this time I tried it with a higher VTT. Once I booted to Windows, it froze with a "Page Fault in nonpaged area" BSOD.

Maybe I should let it run for several hours in a loop tonight. I am just not sure if it's just a temporary "settings unstable" thing or if the last 500MHz FSB tests on the C2D previously caused something bad to the windows installation


Edit:
Did a 10 hour Memtest run last night at the previous settings and it passed without any errors. So the RAM itself isn't the problem. I'll have to check if it isn't just another form of an FSB hole, with the C2D everything below 400MHz didn't work aswell.
Or it is something hiding deeper inside.
Edited by SantasDeer - 10/11/16 at 2:19am
post #192 of 595
Hi Laotian,

It is such good to see people still around with C2Q+790I sli overclocking experience. I have tried to give my Q9550 a crack to get up to FSB1600, and after 4 hrs prime 95 test runs ok without any errors, however, next morning my system corrupted with bootmgr missing. During the course, the temperature only get up to 51 when prime is running.Could you please help me to identify what I did wrong.

My PC setup is:
CPU:Q9550 2.83The
Ram:OCZ 1333 platinum 2G*4
Cooler: cooler master v8
Motherboard: Asus striker II extreme bios 1402


AI overclocker: manual
FSB -Memory clock mode: linked
FSB-Memory Ratio: Sync mode
FSB(qdr):1600
CPU Voltage:1.27
Rest voltage (pll,vvc,dram,an,nb):auto
CPU thermal control:auto
Rest CPU configuration:disable

Never get BSOD issues, but had time that couldn't boot into bios, normally can get into the bios to reset it after few attempts, and system will run flawlessly.

Do you think I should leave fsb-memery unlinked to start with, as my ram only 1333 mhz, does 1600 fsb put stress on it when the dram voltage only set at 1.65
Many thanks
Edited by Cyanold - 10/10/16 at 10:15pm
post #193 of 595
Another small question, is it normal that you can hear your PC work?
Maybe I am just a little bit too paranoid right now, but every time Memtest wrote something to the RAM or during each test of Prime Small, I can hear my PC work.

Like you can hear the bits being written. biggrin.gif
I can't recall that it's been as "loud" with the C2D, at least I never noticed, although the PC is sitting on my desk the entire time.

Edit: Nvm I guess, could be coming from the PSU aswell.
Edited by SantasDeer - 10/11/16 at 8:52am
post #194 of 595
Quote:
Originally Posted by SantasDeer View Post

Another small question, is it normal that you can hear your PC work?
Maybe I am just a little bit too paranoid right now, but every time Memtest wrote something to the RAM or during each test of Prime Small, I can hear my PC work.

Like you can hear the bits being written. biggrin.gif
I can't recall that it's been as "loud" with the C2D, at least I never noticed, although the PC is sitting on my desk the entire time.

Edit: Nvm I guess, could be coming from the PSU aswell.

It's called coil whine. It could be mobo, PSU, or gpu.
post #195 of 595
Thread Starter 
Quote:
Originally Posted by SantasDeer View Post


Edit:
Did a 10 hour Memtest run last night at the previous settings and it passed without any errors. So the RAM itself isn't the problem. I'll have to check if it isn't just another form of an FSB hole, with the C2D everything below 400MHz didn't work aswell.
Or it is something hiding deeper inside.


Excellent, so you've got that peace of mind now that the memory itself can work with the board. What you are left to fiddle with are timings and voltage for the memory as you have proven it can be stable by itself (keep your memory at the same test speed for now, never change your memory speed once you establish stability with memory until you dial in your CPU overclock, as much as you may be tempted). You cannot forget about memory stability 100% though because as you increase the bus speed the NB is going to work harder where additional voltage tweaks could be needed for stability because the memory controller is still part of the NB. This is where we can be thrown off if not watching for it (thinking the overclock to the CPU wasn't stable after the last change when in fact it could just be the NB needing more voltage). I remember having to really kick up the voltages on the chipsets... I will always run the lowest voltage that is required so the high voltages you see in my screenshots are actually the lowest that I could use. All of the voltages used except the CPU and MEMORY were directly related to getting the board stable at a high bus speed with 4 DIMMs.

Of course overclocking 775's aren't super complex (like this ASUS RIVBE is... holy crap I'm lost in this BIOS!) but it still can be very challenging and will take a lot of patience. I am sure there are many others that know a lot more than I do about overclocking this platform but I can tell you that if you don't having an understanding of this "memory situation" it can be frustrating. Once I figured out the memory and looked at it totally differently than I had before, my overclocking success was immediately improved.

A pen and a pad of paper to record your steps and make notes is critical... try this voltage and/or combination of voltages, what were the results, etc

Overall, I do recall having some difficulty raising the bus speed.... I think this has a lot to do with a couple reasons (1) Running a QUAD core and (2) Using all 4 memory banks. If my memory serves me correctly I was unable to get a super high bus (~500Mhz bus) but I was able to get ~460. I may have been able to squeeze 475Mhz for benchies or something but 450Mhz is where things start to become difficult. The FSB hole is definitely real so I would shoot for 450Mhz-460Mhz bus as the maximum bus speed during initial overclocking attempts because they are most likely "obtainable" without too much effort and keeps you away from the FSB hole. You may need to drop your multi a notch or so for now.

An oddity that I noticed sometimes also was when constantly making BIOS changes is that there is a BIOS bug I struggled with...This issue occurred on my XFX 790i board most often but since it's a reference design it may happen to others. I really could never figure out exactly how to reproduce (I think it was related to loading a saved BIOS many times) but it's happened to me more than once... The symptoms are when trying to overclock the CPU higher and no mater what you do it isn't stable when you start backing things down to PREVIOUSLY KNOWN STABLE points and it STILL isn't stable... that's when I realized the hard way (this was noticed by accident because I had pulled out the CMOS battery during troubleshooting) that re-entering all of the same identical BIOS settings again suddenly worked again. So in other words I didn't load a saved BIOS profile I manually went through all settings of the BIOS and set them 100% identically one at a time to what they were just set to (my previously known stable point) and it started working again. This is a rare occurrence but if you start pulling your hair out and things all of a sudden don't make sense, try this tongue.gif Disabling spread spectrums and things like that help stability.

Here are pic links to the BIOS screenshots (for hi-res) biggrin.gif

Gallery View -> http://www.overclock.net/t/1547699/lightbox/post/25058806/id/2752650
Click arrow right for additional BIOS screenshots


Quote:
Originally Posted by Cyanold View Post

Hi Laotian,

It is such good to see people still around with C2Q+790I sli overclocking experience. I have tried to give my Q9550 a crack to get up to FSB1600, and after 4 hrs prime 95 test runs ok without any errors, however, next morning my system corrupted with bootmgr missing. During the course, the temperature only get up to 51 when prime is running.Could you please help me to identify what I did wrong.

My PC setup is:
CPU:Q9550 2.83The
Ram:OCZ 1333 platinum 2G*4
Cooler: cooler master v8
Motherboard: Asus striker II extreme bios 1402


AI overclocker: manual
FSB -Memory clock mode: linked
FSB-Memory Ratio: Sync mode
FSB(qdr):1600
CPU Voltage:1.27
Rest voltage (pll,vvc,dram,an,nb):auto
CPU thermal control:auto
Rest CPU configuration:disable

Never get BSOD issues, but had time that couldn't boot into bios, normally can get into the bios to reset it after few attempts, and system will run flawlessly.

Do you think I should leave fsb-memery unlinked to start with, as my ram only 1333 mhz, does 1600 fsb put stress on it when the dram voltage only set at 1.65
Many thanks

I've been called worse lachen.gif NP JK tongue.gif

The fact that you were able to run stable for so long is the part that doesn't make a lot of sense.

In your case specifically after looking at your voltages I wouldn't be surprised if increasing voltages did not help the chipset a bit.

Did it happen just once or only when the PC is cold? Sometimes all you need to do is hit reset. Perhaps if you have an ability to add a delay to your boot process it could help.. I'm not sure if you had to rebuild your O/S or if it was just a one-time fluke. This is one of the reasons I never liked using onboard SATA because O/S corruption is a real risk when overclocking bus speeds and pushing the chipsets. Running Prime95 for so long is also beating the crap out of your CPU.. I wouldn't recommend it at all... I think it's extreme and way overkill to prove stability. For the same reasons I would not recommend Kombustor or FurMark on a GPU... no matter what the argument on this one, nobody can dispute that when running Prime95 you are not stressing all components of the system (primarily PSU and GPU). Prime95 can overheat things..and it's not very entertaining.. I prefer the "play a game @ 4K max settings" test myself biggrin.gif JK you need more than that but I really don't ever use Prime95. I like AIDA64 system stability test as a quick "does it crash in 30 seconds", completing a full run of Cinebench R15 CPU test is a nice quick and easy test to give you a basic idea as well as RealBench. There's a lot of ways to stress your system not only the apps/games that are used but whether or not these test stress multiple components in your system simultaneously or not. This is where you'll see PSU issues that would be masked in many other tests.

Since memory issues are usually the source of problems I would definitely suggest that you run in unlinked mode to start with. If you look into the "debate" between unlinked and linked/synched it really comes down to this:

(1) Yes linked and synched is faster but it's not much faster.. You would only see the difference in a memory intensive benchmark.
(2) Anytime that you overclock your FSB in linked/synched mode it will ALSO be overclocking your memory.
(3) The most important reason, is that most people will NOT be able to run memory at the highest bus speed that they can run the CPU at in linked/synched mode in 95%+ of the cases. This is because you will likely exceed the memory speed capabilities (anything over 400Mhz bus is going to push memory too much). There are some divider options that allow some different memory speed options but they usually won't match up exactly to your highest stable FSB. The reward for being able to use a much higher FSB and CPU overclock easily diminishes the loss of the slight memory performance from not using link/synched mode.


Note that you can get some added stability by viewing the given "selected" voltages by the AUTO setting and manually defining those same values. This way your MB doesn't "AUTO" decide to change voltages at a bad time. wink.gif
post #196 of 595
Hi Laotian,

It is such good to see people still around with C2Q+790I sli overclocking experience. I have tried to give my Q9550 a crack to get up to FSB1600, and after 4 hrs prime 95 test runs ok without any errors, however, next morning my system corrupted with bootmgr missing. During the course, the temperature only get up to 51 when prime is running.Could you please help me to identify what I did wrong.

My PC setup is:
CPU:Q9550 2.83The
Ram:OCZ 1333 platinum 2G*4
Cooler: cooler master v8
Motherboard: Asus striker II extreme bios 1402

Quote:
Originally Posted by Laithan View Post


I've been called worse lachen.gif NP JK tongue.gif

The fact that you were able to run stable for so long is the part that doesn't make a lot of sense.

In your case specifically after looking at your voltages I wouldn't be surprised if increasing voltages did not help the chipset a bit.

Did it happen just once or only when the PC is cold? Sometimes all you need to do is hit reset. Perhaps if you have an ability to add a delay to your boot process it could help.. I'm not sure if you had to rebuild your O/S or if it was just a one-time fluke. This is one of the reasons I never liked using onboard SATA because O/S corruption is a real risk when overclocking bus speeds and pushing the chipsets. Running Prime95 for so long is also beating the crap out of your CPU.. I wouldn't recommend it at all... I think it's extreme and way overkill to prove stability. For the same reasons I would not recommend Kombustor or FurMark on a GPU... no matter what the argument on this one, nobody can dispute that when running Prime95 you are not stressing all components of the system (primarily PSU and GPU). Prime95 can overheat things..and it's not very entertaining.. I prefer the "play a game @ 4K max settings" test myself biggrin.gif JK you need more than that but I really don't ever use Prime95. I like AIDA64 system stability test as a quick "does it crash in 30 seconds", completing a full run of Cinebench R15 CPU test is a nice quick and easy test to give you a basic idea as well as RealBench. There's a lot of ways to stress your system not only the apps/games that are used but whether or not these test stress multiple components in your system simultaneously or not. This is where you'll see PSU issues that would be masked in many other tests.

Since memory issues are usually the source of problems I would definitely suggest that you run in unlinked mode to start with. If you look into the "debate" between unlinked and linked/synched it really comes down to this:

(1) Yes linked and synched is faster but it's not much faster.. You would only see the difference in a memory intensive benchmark.
(2) Anytime that you overclock your FSB in linked/synched mode it will ALSO be overclocking your memory.
(3) The most important reason, is that most people will NOT be able to run memory at the highest bus speed that they can run the CPU at in linked/synched mode in 95%+ of the cases. This is because you will likely exceed the memory speed capabilities (anything over 400Mhz bus is going to push memory too much). There are some divider options that allow some different memory speed options but they usually won't match up exactly to your highest stable FSB. The reward for being able to use a much higher FSB and CPU overclock easily diminishes the loss of the slight memory performance from not using link/synched mode.


Note that you can get some added stability by viewing the given "selected" voltages by the AUTO setting and manually defining those same values. This way your MB doesn't "AUTO" decide to change voltages at a bad time. wink.gif

Thanks for your reply, and its really helpful. I ended up install OS but wasn't feeling too bad due to the fact that i haven't reinstall like 5 years with stable win7, now i just had chance to upgrade to WIN10. I run couple trials again for these couple day after I got win10 and have some interesting findings and more questions.

First time reset bios, use Auto clock to overclock by 10%, past the bios but failed go into the desktop. System actually went into self repair, lucky didnt end up with reinstall

Second time reset bios, put overclock to "manual" and "UNLINK" the memory, set FSB at 1600, leave memory as it is. Boost into the system, can feel lots more responsive and run CPUZ test for 1 hr and play dota2 for 2 hrs no issues at all, and I have deliberately restart the computer few times just to confirm it doesn't have issues with the setting.

Now it looks like it comes to the conclusion, setting FSB-Memery auto or link has corrupted my system and made it unable to boot. My question now is :
Is my OCZ DDR3-1333 capable with FSB 1600 when linked ? If not what minimum speed would require in stock for ram that can handle fsb1600.

Best regards

Many thanks
post #197 of 595
Probably not...

I've never had luck OCing memory. You should probably look into picking up some 1600-2000mhz memory.

In simple terms, you need the original higher voltage ddr3 1.5v-2.0v. Even then, it may have compatibility issues.
post #198 of 595
Thread Starter 
Yes and to add, running memory at 1600Mhz is considered "high end" for this chipset. MOST people were running 1066Mhz/1333Mhz. 1600Mhz is really fast and the 2000Mhz DIMMs are super special, extremely rare AND only run at 2000Mhz with 2 DIMMs installed. When running 4 DIMMs, 1600Mhz would be considered very high and not easy to get stable. You're better off keeping the ram UNLINKED/1066Mhz for now and focus on getting your CPU overclock stable first @ 1600Mhz BUS and a couple months later mess with the memory once you're done with the CPU, especially if the ram isn't using the "high voltage" RAM the chipset likes.

I highly recommend GHOST or ACRONIS to make yourself some disk images so that O/S restoration is possible... (or use an add-in PCI-e controller to avoid O/S corruption.

Be sure to COMPLETE a MEMTESTx86 before booting into windows! This can save your O/S because if the memory isn't stable you risk O/S corruption.

Good luck!
post #199 of 595
That sound about right. I cannot see myself would invest more money into this platform as I have maxed out the memory with 4*2 GByte. I probably will just leave the memory at stock and unlink to avoid the fuss of data corruption as overclocking memory is way too complicate than cpu and also it is dangerous. I have been running 3.4 Ghz for 3 days now, no issues at all, somehow I set the voltage at 1.27 v but in CPUz it shows 1.31 v. Not sure it's something wrong with the software.

Do you think i will get benefits from pushing the CPU clock higher or at what stage is too much for the platform regarding the memory ddr3 1333 i have.

Cheers
Edited by Cyanold - 10/13/16 at 6:39am
post #200 of 595
Thread Starter 
Quote:
Originally Posted by Cyanold View Post

That sound about right. I cannot see myself would invest more money into this platform as I have maxed out the memory with 4*2 GByte. I probably will just leave the memory at stock and unlink to avoid the fuss of data corruption as overclocking memory is way too complicate than cpu and also it is dangerous. I have been running 3.4 Ghz for 3 days now, no issues at all, somehow I set the voltage at 1.27 v but in CPUz it shows 1.31 v. Not sure it's something wrong with the software.

Do you think i will get benefits from pushing the CPU clock higher or at what stage is too much for the platform regarding the memory ddr3 1333 i have.

Cheers

Well done! thumb.gif

That's a nice voltage for that overclock also. I'm pretty sure you could get higher with more voltage only if you have sufficient cooling. I wouldn't go past 1.4v on AIR.

These boards are known to have vdroop and not sure if your BIOS has a control for that or not (I know the XFX boards do not). The usual behavior would be to see a higher voltage when NOT under load and then when placed under load the voltage will drop a bit. If you have vdroop control you can prevent this but if not you need to "over compensate" for the vdroop and raise the voltage in the BIOS.

Always measure the voltage when under load. wink.gif

Memory mostly isn't a bottleneck. Faster memory is faster of course but the slowest part in anyone's system is the HD (even if SSD in RAID 0). The CPU/GPU always has a much greater chance to become bottle-necked before memory would ever. I would have no problems at all staying with 1333Mhz because I won't notice a difference in games only synthetic benchmarks. The GPU does most of the heavy lifting these days and the CPU needs to keep up with the draw calls.

When I had my Q9500 I think around 3.4Ghz is near what I was able to get but you are using less voltage. Mine wasn't a great overclocker so I would expect a little more from yours. My old Core 2 Extreme edition got me around 4.2Ghz but by then I was on water. My X5470 was by far the best @ 4.5Ghz
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel Motherboards
Overclock.net › Forums › Intel › Intel Motherboards › 790i Ultra NVIDIA chipset - Still rockin' in 2015 / 2016 / 2017.. will it ever die?