Core i7 920 Progress

[turok_t]

Here is my Overclocking journey for my Core i7 920 thus far. My Vdimm is set to 1.6V and my VTT to 1.3V... Here are some results:

3.8Ghz (181x21)
1.35625V (Vcore)
1.30V (VTT)
Stable for 75 runs using LinX
Temp: 77, 75, 73, 71

4.0Ghz (191x21)
1.35625V (Vcore)
1.30V (VTT)
Stable for 75 runs using LinX
Temp: 78, 77, 73, 71

Since I reached my goal of 4.0Ghz, I began to drastically lower the Vcore.. With all settings remained the same as above, I began to use:

1.3375V Vcore 
Stable for 75 runs of LinX
Temp: 77, 75, 74, 71

1.330V Vcore
Stable for 75 runs of Linx
Temp: 74, 72, 72, 68

As you can see, the temperature began to decrease as Vcore decreased. So I continued to reduce my Vcore (and soon VTT once I find a stable Vcore) to determine my lowest limit:

At 1.275V Vcore, I get a BSOD (One that i couldnt find in the guides. Something like 0x000009C) when running LinX
At 1.28125V Vcore, LinX detects an error at 1hr 41 mins.
At 1.28125V Vcore and 1.35V VTT, I get the same BSOD as above, but much quicker.

So I guess bumping up the VTT didn't help, so I increased the VTT to 1.281750V and Im continuing to run LinX.

1. Any suggestions at this point? Im not sure what the usual Vcore would be for 4.0Ghz. Mind you, I still havent touched the CPU PLL or IOH Voltage and Im not sure what this should be set to. Bios reads 1.10v for CPU PLL.

2. Would lowering the multiplier to 19x and using 210 blck to achieve the same 4Ghz help to reduce the vcore needed?

Thanks for your help!

[Janus67]

You would probably need more vtt/qpi to put the bclk to 210 but you can try it. I ran at 4.0 stable with 1.325 , can't remember the vtt but I believe it was about the same. I could do 4.2 as well (200x21) but that took about 1.385-1.4 and about the same vtt. I never had heat problems (67-70c under load with my venomous-x, but I generally just ran 4c/4t)

[acebmxer]

0x09c is qpi/vtt most likely, but increasing vcore helped. In your case since you were lowering vcore while maintaining qpi/vtt most likely vcore. You can try increasing qpi/vtt.

And unless you like your cpu try to keep runs down to 20-25 runs max. no need for 75 runs you are just gonna kill your cpu.

[CryptiK]

75 runs with max mem, that is a bit overkill and must take you forever!

20 runs should be fine, its only one stress test, try using prime as well. As said above, 09C is usually Vtt but vcore can help, hard to say exactly what is causing it without experimentation.

[turok_t]

The reason why i ran 75 runs with max memory was based on a X58 OC guide I found on the EVGA Forum:

"Once you have your vcore in a good spot, run 50 - 75 Tests with ALL memory to be sure your Vcore and VTT voltage's are stable. You can quit here and risk having a crash someday, or run prime95 blend for at least 12 hours. "

http://www.evga.com/forums/tm.aspx?&m=81651&mpage=1

How long would you guys suggest running Prime 95 and MemTest (I remember there was a particular test that was very stressful)?

[turok_t]

So today, I started to test now with Prime 95. When I turned HT on with either 4 threads, or 8 threads, I BSOD right away. Here were some errors I got:

0x0000003B
0x000000101 (usually means not enough vcore, vtt, pll)
0x0000001E

My temperatures were also much higher as well. With HT off, my temps were usually mid to high 60's. With HT on, temps were mid to high 70's during testing.

Currently, I'm running 4.0 Ghz, with HT off, using 1.29375V for Vcore, and 1.3V for VTT, stable in LinX for 75 runs.

How much more Vcore and VTT is needed if I wanted to use HT? Does Hyperthreading really provide significant performance boost? It seems like it needs alot of voltage.

[zsamz_]

So today, I started to test now with Prime 95. When I turned HT on with either 4 threads, or 8 threads, I BSOD right away. Here were some errors I got:

0x0000003B
0x000000101 (usually means not enough vcore, vtt, pll)
0x0000001E

My temperatures were also much higher as well. With HT off, my temps were usually mid to high 60's. With HT on, temps were mid to high 70's during testing.

Currently, I'm running 4.0 Ghz, with HT off, using 1.29375V for Vcore, and 1.3V for VTT, stable in LinX for 75 runs.

How much more Vcore and VTT is needed if I wanted to use HT? Does Hyperthreading really provide significant performance boost? It seems like it needs alot of voltage.

it dont take much more around .02-.06 on v-core
dont need more vtt
but the temps get 5-20c hotter with ht
hotter the chip the more juice you need

75 runs lol

i test 20min linx 20min prime n gaming

[turok_t]

In regards to running stress tests for a long time, here is the question I posted on other boards:

"If I run these tests for such a long time, will it damage my CPU? I dont want to over stress the poor lil chip"

Here is the response I got:

"No, it is heating and cooling cycles that cause electronics damage - as long as you stick to the official limits. The hotter it gets, the more extreme the cooling cycle and therefore the lesser the life - but not by a big difference with std OC temps. So the length of time you run the test makes no difference. It's not like an engine, where it has mechanical wear over run time. "

[OC Nub]

Everyone will have different opinions on how long to test for stability and what test is the best to use.

There are fanatics that will run linx 500 times, and others that wont run it at all. As long as your PC does the things you use it for without issue, thats all that matters in the end.

There are other ways to test that are more fun than watching loops of LinX or blend. You could try SuperPI 32m, then HyperPI 32m, WPrime 1024m and set number of threads to 4 or 8 depending on weather or not you are using HT, run the different 3D benches, play the games you like to play. If it passes those and does everything else you want it to with no problems, who cares if someone else doesnt like it because you didnt use the exact method they use.

[turok_t]

Everyone will have different opinions on how long to test for stability and what test is the best to use.

There are fanatics that will run linx 500 times, and others that wont run it at all. As long as your PC does the things you use it for without issue, thats all that matters in the end.




Thanks for the reassurance, I kinda knew that real world environments matter the most, I just wanted to verify with u experts 

There are other ways to test that are more fun than watching loops of LinX or blend. You could try SuperPI 32m, then HyperPI 32m, WPrime 1024m and set number of threads to 4 or 8 depending on weather or not you are using HT, run the different 3D benches, play the games you like to play. If it passes those and does everything else you want it to with no problems, who cares if someone else doesnt like it because you didnt use the exact method they use.

Thanks for the reassurance  kinda knew that real world environments mean the most just wanted to confirm with you experts 