X79 and/or rampage IV OC'ing thread for those of us still left....

[Zyll Goliat]

I have read pretty much elsewhere that 1.65V is the max safe voltage. However, you should keep the gap VCCSA and DRAM voltage no larger than 0.6. So I have to set VCCSA to at least 1.05V.

No that's not actually true there is plenty of DDR3 memory modules that their default Voltage is 1,65V instead of usual 1,5V.........also it is nothing unusual on X79 IF you using ALL 8 ram slots to rise ram Voltages a bit and some people claim when OC their ram that even rising Voltages up to the 2,0V it's perfectly safe.....personally I didn't have the need for that high Voltages but my memory working 100% stable on 1,68V already few years now.......

 

[Duc996]

Hello everybody, I got a strange problem ... I only have 16gb from 32gb left?! I tried every single Memory Stick in Slot A1 ... all are working correct.
- Bios tells 16gb but looking under Dimm/Post all 4 channels are in use.
- CPU-z --> 32gb Quad Channel
- Task Manger --> 16gb
- Aida 64 --> 16 gb but shows all 4 x 8gb Sticks??
- Running a memory benchmark --> Quad Channel but with less performance

I just removed my Graphics Card a few days ago (just for cleaning and Paste and Pads renewing) and refitted it today. I didn´t touch anything else ...
Maybe someone has an idea ...

Best regards
Thomas

 

[lexluthermiester]

I have read pretty much elsewhere that 1.65V is the max safe voltage.

That is a common opinion and there is a bit of merit to it, but I'm a very cautious tech geek. If this is going to be your daily driver, it's best to yield on the safe side.

However, you should keep the gap VCCSA and DRAM voltage no larger than 0.6. So I have to set VCCSA to at least 1.05V.

I'll go along with that being a safe way to do it.

Hello everybody, I got a strange problem ... I only have 16gb from 32gb left?! I tried every single Memory Stick in Slot A1 ... all are working correct.
- Bios tells 16gb but looking under Dimm/Post all 4 channels are in use.
- CPU-z --> 32gb Quad Channel
- Task Manger --> 16gb
- Aida 64 --> 16 gb but shows all 4 x 8gb Sticks??
- Running a memory benchmark --> Quad Channel but with less performance

I just removed my Graphics Card a few days ago (just for cleaning and Paste and Pads renewing) and refitted it today. I didn´t touch anything else ...
Maybe someone has an idea ...

Best regards
Thomas

You could be having a motherboard problem. Maybe a BIOS setting issue? You might even need to update your BIOS.

 

[Zyll Goliat]

Hello everybody, I got a strange problem ... I only have 16gb from 32gb left?! I tried every single Memory Stick in Slot A1 ... all are working correct.
- Bios tells 16gb but looking under Dimm/Post all 4 channels are in use.
- CPU-z --> 32gb Quad Channel
- Task Manger --> 16gb
- Aida 64 --> 16 gb but shows all 4 x 8gb Sticks??
- Running a memory benchmark --> Quad Channel but with less performance

I just removed my Graphics Card a few days ago (just for cleaning and Paste and Pads renewing) and refitted it today. I didn´t touch anything else ...
Maybe someone has an idea ...

Best regards
Thomas


View attachment 239156

Hey...Is your memory overclocked maybe or working on default speed and timings?It is absolutely necessary that your BIOS see ALL the ram first properly so that could work as it should in your system....I know from my experience that IF I tried to overclock my ram to much or if I mess with timings that certain sticks could not work properly but also I could see them in CPU-Z but not in bios.....My advice is check YOUR BIOS SETTINGS see If you have latest BIOS or even try to RESET your bios on default........GL

 

[Duc996]

Thanks for your answers ...
Ram is running it´s XMP Profile --> 2133 @ 1.65v with related timings
I created 3 different OC Profiles over the last months which were all running fine. Fail safe settings won´t work either.
I tried a BIOS update without success. I guess it´s a hardware damage. Seems like two memory channels are no longer working. I only wonder why the system is working correct though!?
When I have more time I´ll switch CPU´s. I have a 3820 and a 3960 laying around ... just to confirm that the CPU´s memory controller is not broken.

Damn ... Rampage 4 Gene prices are way too high. I don´t wanna spend ~ 200 € for a 9 year old MB!

 

[wanna_buy]

Thanks for your answers ...
Ram is running it´s XMP Profile --> 2133 @ 1.65v with related timings
I created 3 different OC Profiles over the last months which were all running fine. Fail safe settings won´t work either.
I tried a BIOS update without success. I guess it´s a hardware damage. Seems like two memory channels are no longer working. I only wonder why the system is working correct though!?
When I have more time I´ll switch CPU´s. I have a 3820 and a 3960 laying around ... just to confirm that the CPU´s memory controller is not broken.

Damn ... Rampage 4 Gene prices are way too high. I don´t wanna spend ~ 200 € for a 9 year old MB!

Try cleaning the CPU contact pads and the DIMM contacts with >99% isopropyl alcohol. Also try cleaning the motherboard's RAM slots with a hard plastic toothbrush dipped in isopropyl alcohol. If this does not help, then you have either a bad CPU or motherboard.

 

[dalekdukesboy]

I did an extensive testing using Prime95 v26.6 which is the last non-AVX version. I decided to test DDR3 2400 stability with VTT manually set to 1.10V and VCCSA to 1.05V. The CPU was at stock, but with the XMP enabled it was running at 3.9 GHz Turbo. Prime95 Blend test ran without any errors for more than 2 hours.

Previously I have used Cinebench R15 for quick stability testing. While the RAM was running at stock 1333 speed, my 1680V2 could pass Cinebench R15 at 4.5 GHz at 1.23V and at 4.6GHz at 1.27V. I now reverted the RAM speed to default 1333 MHz. However, Prime95 Blend failed even at 4.5GHz at 1.27V. I upped Vcore to 1.29V and Prime95 Blend ran for 23 hours without any error. Now I am going to test both CPU at 4.5GHz and RAM at 2400.


I have read pretty much elsewhere that 1.65V is the max safe voltage. However, you should keep the gap VCCSA and DRAM voltage no larger than 0.6. So I have to set VCCSA to at least 1.05V.

Sounds like you’re doing what I would expect and do myself…I literally haven’t used my pc in weeks been at fiancés house 95% of the time lol, but anyway I can tell you I’ve used 1.75 for my micron’s and gone even higher briefly to test so caution is fine but 1.65 is hardly the most you can do safely for years worth of use. I’ve only had one cpu die and it was my first 1680v2 and it was heavily used before I got it and I got a few years out of it and it was always quirky so it may have been degrading before I even put it in my moBo.

Also for most part long as you don’t push the FSB too high your vccsa and vvt voltage can be left at auto and usually are fine/stable. I have manually set it up to 1.2 and you can go bit higher safely but least you need to be stable is best obviously.

Good convo going lot of things there for me to check out and catch up on in the thread! Basically if you’re not in the “red zone” of voltages which is higher than most people realize I advise just doing the most voltage you are comfortable with that is within reason and not so conservative it simply isn’t enough to keep your OC stable.

I own two i7-3930K processors. 1st can do 4.5GHz at 1.26V, but 4.7 GHZ at 1.35V. 2nd one can do 4.5GHz at 1.28V, but 4.7GHz at 1.32V. Which one can be considered golden chip? Which one should I keep?

Honestly I’d get a 1680v2 and sell both of those or keep one as a backup perhaps…unless the prices have skyrocketed on ivy bridge E that is your best cpu for overall performance AND good IMC to make use of the fast memory you have. Just a suggestion especially if you’re keeping this rig as your main rig for quite some time.

 

[wanna_buy]

Honestly I’d get a 1680v2 and sell both of those or keep one as a backup perhaps…unless the prices have skyrocketed on ivy bridge E that is your best cpu for overall performance AND good IMC to make use of the fast memory you have. Just a suggestion especially if you’re keeping this rig as your main rig for quite some time.

I gave one i7-3930K to my cousin and kept another as a spare just in case. I already own E5-1680V2 that can do Cinebench stable 4.6 GHz at 1.27V. But 4.5 GHz Prime95 Blend stability requires 1.29 Vcore. E5-1680V2 at 4.6 GHZ achieves the lowest RAM latency in AIDA64 I have ever seen. AIDA64's own list lists AMD Athlon X2 6400+ as having 55.2ns latency.


My main PC has E5-2690V2 and 64GB DDR3 1866 G.Skill RAM. The default speed is 1600 MHz and the XMP profile also raises VCCSA to 1.2V.

E5 1680V2 VTT runs at 1.05V and VCCSA runs at 0.9V at stock settings and DDR3 1333 speed. I did an extensive stability testing of my CAS9 Dominator 4x4GB DDR3 2400 kit using Prime95 and the CPU was running at stock Turbo frequency of 3.9 GHz. XMP profile only sets the frequency and timings, but Rampage IV Black automatically pushes VTT to more than 1.25V and VCCSA to 1.15V when running DDR3 2400 speed. I have already posted that Windows Memory Diagnostic finds no errors. Prime95 custom 448-4096 FFT run with 14GB RAM allocated failed within 2 hours. I raised VCCSA to 1.2V and left VTT on Auto, but the RAM was still not stable in Prime95. I also tested 2133 MHz frequency with CAS9 latency and VCCSA/VTT manually set to 1.10V. Prime95 custom test ran for hours without any error. I thought at this point that my CPU's memory controller perhaps could not handle 2400 MHz frequency.

I also have CAS11 Vengeance 2x8GB DDR3 2400 kit. If 2400 RAM speed was unstable even on this kit, then IMC would definitely be the culprit. VTT also defaulted to more than 1.25V and VCCSA to 1.15V by Rampage IV Black. Prime95 custom run did not show any errors in 3 hours. I decided to lower both VTT/VCCSA to 1.05V. Prime95 custom run was stable for 8 hours and 30 minutes, so I concluded that the IMC was definitely up to task.


I again reinstalled CAS9 Dominator 4x4GB kit. I loosened the primary timings to 11-13-13-31 and set VCCSA/VTT to 1.2 manually. Prime95 custom run failed after 2 hours.


My conclusion is that my Dominator CAS9 kit is faulty/degraded. What do you think?

 

[dalekdukesboy]

I gave one i7-3930K to my cousin and kept another as a spare just in case. I already own E5-1680V2 that can do Cinebench stable 4.6 GHz at 1.27V. But 4.5 GHz Prime95 Blend stability requires 1.29 Vcore. E5-1680V2 at 4.6 GHZ achieves the lowest RAM latency in AIDA64 I have ever seen. AIDA64's own list lists AMD Athlon X2 6400+ as having 55.2ns latency.

View attachment 239449View attachment 239450
My main PC has E5-2690V2 and 64GB DDR3 1866 G.Skill RAM. The default speed is 1600 MHz and the XMP profile also raises VCCSA to 1.2V.

E5 1680V2 VTT runs at 1.05V and VCCSA runs at 0.9V at stock settings and DDR3 1333 speed. I did an extensive stability testing of my CAS9 Dominator 4x4GB DDR3 2400 kit using Prime95 and the CPU was running at stock Turbo frequency of 3.9 GHz. XMP profile only sets the frequency and timings, but Rampage IV Black automatically pushes VTT to more than 1.25V and VCCSA to 1.15V when running DDR3 2400 speed. I have already posted that Windows Memory Diagnostic finds no errors. Prime95 custom 448-4096 FFT run with 14GB RAM allocated failed within 2 hours. I raised VCCSA to 1.2V and left VTT on Auto, but the RAM was still not stable in Prime95. I also tested 2133 MHz frequency with CAS9 latency and VCCSA/VTT manually set to 1.10V. Prime95 custom test ran for hours without any error. I thought at this point that my CPU's memory controller perhaps could not handle 2400 MHz frequency.

I also have CAS11 Vengeance 2x8GB DDR3 2400 kit. If 2400 RAM speed was unstable even on this kit, then IMC would definitely be the culprit. VTT also defaulted to more than 1.25V and VCCSA to 1.15V by Rampage IV Black. Prime95 custom run did not show any errors in 3 hours. I decided to lower both VTT/VCCSA to 1.05V. Prime95 custom run was stable for 8 hours and 30 minutes, so I concluded that the IMC was definitely up to task.
View attachment 239451

I again reinstalled CAS9 Dominator 4x4GB kit. I loosened the primary timings to 11-13-13-31 and set VCCSA/VTT to 1.2 manually. Prime95 custom run failed after 2 hours.
View attachment 239452

My conclusion is that my Dominator CAS9 kit is faulty/degraded. What do you think?

My first thought is…how many CPU’s do you have and how many are running at one time? But on to your post with the 1680v2, I have used regular rampage and now a rampage black and with the good memory I have the latency is crazy low and the bandwidth numbers are great as well! I would assume the dominator’s have an issue but could just be a weird incompatibility between the memory and CPU. I think however it’s just a wonky pair of memory and the IMC on your 1680 is good, never seen one that can’t run 2400 and most go happily to 2600+ with the proper memory kit paired with it. Also, most of the auto settings for voltages are good with the rampage black but the cpu VCORE and couple others like VTT etc are higher than needed and you just have to readjust as needed to not overjuice and overheat your cpu for no good reason.

 

[drizzler]

I gave one i7-3930K to my cousin and kept another as a spare just in case. I already own E5-1680V2 that can do Cinebench stable 4.6 GHz at 1.27V. But 4.5 GHz Prime95 Blend stability requires 1.29 Vcore. E5-1680V2 at 4.6 GHZ achieves the lowest RAM latency in AIDA64 I have ever seen. AIDA64's own list lists AMD Athlon X2 6400+ as having 55.2ns latency.

View attachment 239449View attachment 239450
My main PC has E5-2690V2 and 64GB DDR3 1866 G.Skill RAM. The default speed is 1600 MHz and the XMP profile also raises VCCSA to 1.2V.

E5 1680V2 VTT runs at 1.05V and VCCSA runs at 0.9V at stock settings and DDR3 1333 speed. I did an extensive stability testing of my CAS9 Dominator 4x4GB DDR3 2400 kit using Prime95 and the CPU was running at stock Turbo frequency of 3.9 GHz. XMP profile only sets the frequency and timings, but Rampage IV Black automatically pushes VTT to more than 1.25V and VCCSA to 1.15V when running DDR3 2400 speed. I have already posted that Windows Memory Diagnostic finds no errors. Prime95 custom 448-4096 FFT run with 14GB RAM allocated failed within 2 hours. I raised VCCSA to 1.2V and left VTT on Auto, but the RAM was still not stable in Prime95. I also tested 2133 MHz frequency with CAS9 latency and VCCSA/VTT manually set to 1.10V. Prime95 custom test ran for hours without any error. I thought at this point that my CPU's memory controller perhaps could not handle 2400 MHz frequency.

I also have CAS11 Vengeance 2x8GB DDR3 2400 kit. If 2400 RAM speed was unstable even on this kit, then IMC would definitely be the culprit. VTT also defaulted to more than 1.25V and VCCSA to 1.15V by Rampage IV Black. Prime95 custom run did not show any errors in 3 hours. I decided to lower both VTT/VCCSA to 1.05V. Prime95 custom run was stable for 8 hours and 30 minutes, so I concluded that the IMC was definitely up to task.
View attachment 239451

I again reinstalled CAS9 Dominator 4x4GB kit. I loosened the primary timings to 11-13-13-31 and set VCCSA/VTT to 1.2 manually. Prime95 custom run failed after 2 hours.
View attachment 239452

My conclusion is that my Dominator CAS9 kit is faulty/degraded. What do you think?

Before you test further you first should get rid of the write bug (first screen) caused by 100 bclk with the chosen ram divider. Best method from my findings is to use 125 strap and lower the bclk to round about 122. You will have to test which blck is working best for your system and is giving best ram performance.

I did some posts about this earlier in this thread.
Your write speed should be round about 65+.

 

[wanna_buy]

My E5-1680V2 just passed 60 hour Prime95 custom 8-4096 test with 14GB RAM allocated. CPU frequency 4.5GHz at 1.29 Vcore with High LLC, 1.10 VTT and 1.05 VCSSA with VCCSA 120% Current capability, Corsair Vengeance DDR3 2400 16GB with 11-13-13-31 timings.

I noticed in the results that some threads were running different FFTs. The same FFTs would repeat after 18 hours.

Corsair Vengeance DDR3 2400 kit has even lower read/write speeds.

 

[lexluthermiester]

My E5-1680V2 just passed 60 hour Prime95 custom 8-4096 test with 14GB RAM allocated.

WOW! 60 hours? Why? Stability needs only 30 or 60 minutes..

 

[Zyll Goliat]

WOW! 60 hours? Why? Stability needs only 30 or 60 minutes..

Maybe is cold where he lives so he need a heater .........

 

[drizzler]

Corsair Vengeance DDR3 2400 kit has even lower read/write speeds.

Dude, see my posting before. Do not use 100 bclk with 2400 divider, it is bugged and you will not get any realiable results.

 

[dalekdukesboy]

Dude, see my posting before. Do not use 100 bclk with 2400 divider, it is bugged and you will not get any realiable results.

THIS!


I haven’t used my rig in a while now but I found a thread that showed how to get the maximum results from your memory with this platform and I believe I actually underclocked my FSB to 98.2 or something and that fixed the memory bug and saves you the bother of having to go over 100 with the X79 platform where oc’ing that metric really puts pressure on the system and causes errors and/or a lot more voltage to get it stable.

 

[drizzler]

THIS!


I haven’t used my rig in a while now but I found a thread that showed how to get the maximum results from your memory with this platform and I believe I actually underclocked my FSB to 98.2 or something and that fixed the memory bug and saves you the bother of having to go over 100 with the X79 platform where oc’ing that metric really puts pressure on the system and causes errors and/or a lot more voltage to get it stable.

We also talked about this here, see https://www.techpowerup.com/forums/...or-those-of-us-still-left.225209/post-4554290 ;D

 

[wanna_buy]

WOW! 60 hours? Why? Stability needs only 30 or 60 minutes..

You can't ensure 100% stability with only 60 minute test. My Dominator DDR3 2400 failed Prime95 stability test after 2 hours. My CPU overclocked to 4.5 GHz failed Prime95 Blend at 1.27 Vcore after 4 hours.
It is pretty much widespread notion that you need at least 24 hour test to be sure of 100% stable overclock. Entire 8-4096 FFT range needs about 18 hours for complete run. Unstable overclocked CPU causes data corruption and Windows begins to lag/crash and the game save files refuse to load.

Maybe is cold where he lives so he need a heater .........

I have 280mm AIO so the temps under load were around 66°C, the max recorded 72°C.

 

[lexluthermiester]

You can't ensure 100% stability with only 60 minute test.

Nonsense. Regardless of cooling option, heat saturation will take place within the first 10 to 15 minutes. So if it remains stable for another 15 minutes to an hour, the OC is stable. Testing beyond that point is a waste of time and power. This is 30years of overclocking experience talking.

 

[KptRex]

So a while ago I decided to go above 1.65v that my ram was rated for and was always using. After a lot of testing managed to get it running at cl8-11-11-11 2130mhz 1.735v. It was bigger performance improvement than expected, and my 2697 seems to like it too as I'v been runing it for few weeks without problems now. Gonna start looking for second kit of those HyperX sticks, as I want to get RTX 3080 at some point after this whole shortage thing is over

So if it remains stable for another 15 minutes to an hour, the OC is stable

I second that, on all CPUs I overclocked if it was sable after 30 minutes of OCCT or Aida64 it would never cause any problems during real world use cases, even after multi hour rendering. I'v seen some people getting erros on those stress-tests after 20 hours or so but honestly this seems more like program bug at this point, and if you really need something running all the time then you gonna want to take some mhz off your max OC anyways

 

[wanna_buy]

Nonsense. Regardless of cooling option, heat saturation will take place within the first 10 to 15 minutes. So if it remains stable for another 15 minutes to an hour, the OC is stable. Testing beyond that point is a waste of time and power. This is 30years of overclocking experience talking.

You have a different definition of stability. The stable CPU is a calculator who always outputs reliable and predictable results. Unstable overclocked CPU is a random number generator. 60 minute stability test may be enough if you use PC only for gaming. But if you are rendering or doing scientific calculation for days, then 100% is absolute must.

Below are some overclock-related errors I have gotten after 60 minutes in Prime95 Blend/custom mode:

FATAL ERROR: Resulting sum was 11311727082264.69, expected: 11363793230922.76
Hardware failure detected, consult stress.txt file.

FATAL ERROR: Rounding was 0.49609375, expected less than 0.4
Hardware failure detected, consult stress.txt file.

FATAL ERROR: Final result was 38837ABD, expected: 1A1DA557.
Hardware failure detected, consult stress.txt file.

FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.

 

[lexluthermiester]

You have a different definition of stability. The stable CPU is a calculator who always outputs reliable and predictable results. Unstable overclocked CPU is a random number generator. 60 minute stability test may be enough if you use PC only for gaming. But if you are rendering or doing scientific calculation for days, then 100% is absolute must.

Below are some overclock-related errors I have gotten after 60 minutes in Prime95 Blend/custom mode:

FATAL ERROR: Resulting sum was 11311727082264.69, expected: 11363793230922.76
Hardware failure detected, consult stress.txt file.

FATAL ERROR: Rounding was 0.49609375, expected less than 0.4
Hardware failure detected, consult stress.txt file.

FATAL ERROR: Final result was 38837ABD, expected: 1A1DA557.
Hardware failure detected, consult stress.txt file.

FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.

That information means nothing out of context. And if those happened after 60 minutes of testing, your OC failed. Change your setting, try again. 60 hours of stress testing is a waste. 30minutes to an hour is all that is required. 2hours at the most if it's a fiddly OC. If you experience minor problem after that, minor tweaking of the OC may be required. The Sun will freeze over before I will ever advise someone to spend 60hours stress testing an OC. It's time & resources better spent doing other things..

 

[wanna_buy]

That information means nothing out of context. And if those happened after 60 minutes of testing, your OC failed. Change your setting, try again. 60 hours of stress testing is a waste. 30minutes to an hour is all that is required. 2hours at the most if it's a fiddly OC. If you experience minor problem after that, minor tweaking of the OC may be required. The Sun will freeze over before I will ever advise someone to spend 60hours stress testing an OC. It's time & resources better spent doing other things..

So you basically agree with me that it can take more than 60 minutes of stress-testing for an unstable overclock to fail. Minimum of 24 hours of Prime95 is the platinum standard of stability testing. Google yourself and read various forums.

 

[lexluthermiester]

So you basically agree with me that it can take more than 60 minutes of stress-testing for an unstable overclock to fail.

Context is important. It depends on the OC. My point was that 60hours is NOT.

Minimum of 24 hours of Prime95 is the platinum standard of stability testing.

Nor is 24.

Google yourself and read various forums.

Thanks for the tip. Pass. Not only have I been-there-done-that, but decades worth of hands on experience is all that I need to confidently state that such notions are utter nonsense.

 

[johnspack]

Well, wonder how many x79 mobos there are still out there. Mine just died and I'm kinda f'd. I also have an 1680 v2 I could part with. Typing this on my only computer, a mobile core 2 pentium dual core. Holy crap it's slow! Anyone have any ideas?

 

[wanna_buy]

Well, wonder how many x79 mobos there are still out there. Mine just died and I'm kinda f'd. I also have an 1680 v2 I could part with. Typing this on my only computer, a mobile core 2 pentium dual core. Holy crap it's slow! Anyone have any ideas?

X79 motherboards are rare, but no so rare that it would be impossible to buy one like LGA 2011-1 (Socket R2) motherboards for Xeon E7 CPUs. There's always eBay. If you are in Canada, try winning this auction. https://www.ebay.com/itm/284696585154