Core i7 Overclocking by Batch #

I am not stable at the same settings with everything mounted in the closed case. Not even with all fans on max, but I'm in the process of upgrading my fan collection. :devil:

So far I've had to back off to 21x180 to get stable with 3 x 2gb @ 3.780ghz. I did however lower the DRAM voltage and tighten the timings to 7-8-7-20-1T because the RAM is slowed down by 80mhz. I also raised vCore from 1.184v to 1.200v to help with stability in the case.

The system passed LinX and 11 hours of Prime blend @ 1.184v load @ 19x200 when the board was laying on the table with 20C ambient. I checked my TRUE-120 mount and it seems good. My core temps are pretty even and I get no reduction in temps from pressing on it during load or idle. The temp inside my closed case is 30C at best with the window open (much warmer outside now) and more like 35C with the fans running slow and the window shut for hours.

I'm not trying to get help, just providing information. My chip looks very good because I gave it every advantage in testing on air. If you are subjecting your board and CPU to much higher temps than I was you are going to hit the wall much sooner and you will require a bit more voltage due to the temps also.

Note: The ambient/case temp readings are from the analog sensors that come with the mCubed T-Balancer BigNG fan controller. The mCubed digital sensors read about 6c lower than the analog sensors.

The setting of BCLK is an implementation detail and therefore is of no importance, all that matters is CPU clock.
I think that anyone who can do 24/7 without crashing, lockups, etc. at under 1.35 Vcore and at least 4 Ghz can feel good about their chip, even if it can't pass
prime or Intel Burn Test. My 24/7 number is 4011 Mhz, which I've passed both prime and IBT at a mere 1.184Vcore using my Tuniq Tower 120 w/hand made hold down. Because I generate so little heat, I run my CPU fan at only 930 RPM so it is virtually inaudible, even during games. That's the *practical* advantage of an exceptional chip... it buys you a performance-quietness budget that you can spend in any way you choose, I'm spending mine mostly on quietness.

Quote:

---

Originally Posted by roller11

The setting of BCLK is an implementation detail and therefore is of no importance, all that matters is CPU clock.

---

Except that now we have an IMC and unfortunately quality differences there (and physical module configuration differences) at high memory/interface speeds are going to have wildly varying results as to what is stable at any setting and how much vCore and VTT required to be stable at specific settings.

Also there are performance gains to be had by increasing the Uncore clock to keep up with the core clock speed as you raise it. My point is we need to keep track of both Core and IMC quality across batches.

Also, you reminded me of another good point. It may be interesting to see if some chips can drop more voltage than others when disabling hyperthreading. I can drop much more voltage by disabling HT at high clock speeds than slow clock speeds and that's probably the norm but I'm not sure. You can significantly increase the max stable OC you can get by disabling HT and see no performance hit whatsoever for the software you are running if it doesn't use more than 4 threads.

Quote:

---

Originally Posted by JWMc

Except that now we have an IMC and unfortunately quality differences there (and physical module configuration differences) at high memory/interface speeds are going to have wildly varying results as to what is stable at any setting and how much vCore and VTT required to be stable at specific settings.

Also there are performance gains to be had by increasing the Uncore clock to keep up with the core clock speed as you raise it. My point is we need to keep track of both Core and IMC quality across batches.

Also, you reminded me of another good point. It may be interesting to see if some chips can drop more voltage than others when disabling hyperthreading. I can drop much more voltage by disabling HT at high clock speeds than slow clock speeds and that's probably the norm but I'm not sure. You can significantly increase the max stable OC you can get by disabling HT and see no performance hit whatsoever for the software you are running if it doesn't use more than 4 threads.

---

I agree +1 :)

my chip must have a weak IMC then. My ram should do ddr3-2000 but i cant even get it past ddr3-1800 no matter how much qpi voltage or dram voltage.

Quote:

---

Originally Posted by liftedcj7on44s

my chip must have a weak IMC then. My ram should do ddr3-2000 but i cant even get it past ddr3-1800 no matter how much qpi voltage or dram voltage.

---

Read this:

http://www.xtremesystems.org/forums/...&postcount=877

Read it again:

http://www.xtremesystems.org/forums/...&postcount=877

Print it out:

http://www.xtremesystems.org/forums/...&postcount=877

........and read it when you go to bed. When you wake up, read it again, go to work, and come home and read again. Always read it before you retire to bed, and quote it line for line whenever asked about it :)

Do this for as long as it takes to understand, then read it again for good measure. :up:

so basically i need a new motherboard(i cant select different QPI link speeds) and i need to run ram at a lower ratio?
This is all pretty confusing and is about to force me to get rid of my i7 setup.
ok since my motherboard offers no help then maybe you guys can
QPI speed in cpu-z is qpi bus speed(no brainer)
so if i was running 4ghz
211x19
this gives me a QPI link speed of 3806 and a uncore speed of 3380.
Is this withing the range its supposed to be??

ive read it 4 times now and it still makes no sense honestly.

my 3836A690 is stable at 3.6GHz with 1.21vCore with HT/Turbo ON, so its 3.6 or 3.78 depending on how you look at it

prime temps are 60-65C

In about 2 weeks you should see my results from batch# 3839A457.... fingers crossed to it being a good batch!

Ok got mine from theegg and it is a 3839A628. Does anyone have this batch? what do you think?
I'm starting my build now :) but still waiting on my new TRUE. for now will use the stock cooler.
thanks

My 3841A i7-920 is running 3.6Ghz stock cooling and stock voltage with mid-40's idle and mid 60's load, HT on. Lots of headroom to play with it seems. Only thing I set other then bclock is is lowered the memory multi to keep ram below stock 1600mhz and locked the vcore to 1.25 so it won't go up on it's own like it does with Auto.

Can anyone tell me which Voltage is the most important ? The value that you set it BIOS or the the value that is showing up in CPU-Z ?

I guess lot of people who have posting results are relying on CPU-Z voltage and not the one set in BIOS.

Quote:

---

Originally Posted by icecold316

Can anyone tell me which Voltage is the most important ? The value that you set it BIOS or the the value that is showing up in CPU-Z ?

I guess lot of people who have posting results are relying on CPU-Z voltage and not the one set in BIOS.

---

CPU-Z is more relible than what you set in your bios. That is the voltage after V-droop and V-drop.

What you guys think?
According to Anandtech... Batch # 3841 are good for them... so it can depend on the motherboard too?
Currently i am NOT very happy with the Gigabyte x58 boards... to much vdroop... also the PCI Express Layout make my vcard scream for more space between each other... The one located at the top is having 85C degrees under load and the bottom one only 70-75C degrees under load...

I am considering to keep my 920 and buy another mobo... i am looking at the eVGA x58... but i think to use both PCI Express in x16 you have to use the 1st and 2nd PCI Express slots... i am right?

Quote:

---

Originally Posted by kpablo

What you guys think?
According to Anandtech... Batch # 3841 are good for them... so it can depend on the motherboard too?
Currently i am NOT very happy with the Gigabyte x58 boards... to much vdroop... also the PCI Express Layout make my vcard scream for more space between each other... The one located at the top is having 85C degrees under load and the bottom one only 70-75C degrees under load...

I am considering to keep my 920 and buy another mobo... i am looking at the eVGA x58... but i think to use both PCI Express in x16 you have to use the 1st and 2nd PCI Express slots... i am right?

---

Yes, but the EVGA is out of the question cause it doesn't support sleep mode.
I saw this as a problem 2 months ago when I realized if the middle slot is used for secondary PCIe card when doing SLI, it would choke the air supply and at best, cause the fan to run full blast with the irritating noise that results. My own testing bares this out. As far as I'm concerned, the middle slot cannot be used cause of the increase in noise, so the bottom slot must be used if doing SLI. The only question is what is the ramification of using the bottom PCIe slot? Some mobos offset the bottom slot towards the bottom so that the vid card will interfere with the case (gigabyte, MSI I believe). The DFI and EVGA position the bottom slot more favorably so that the card may not hit the case. The EVGA is out cause it doesn't support sleep mode, so that is a non-starter. In benchmark tests, I found there to be virtually no difference in graphics performance using a 16x/16x config vs 16x/8x, like maybe 1.5 - 0 % depending on the benchmark. I like the giga so I modified my case by cutting out the area of the case that would cause the interference problem, so I'm now running SLI with the giga ud5 using the top and bottom slots and seeing no performance loss to speak of with the 16x8x config. The DFI x58 board has the advantage of favorable slot placement and 16x/16x config. I had this board for a few weeks, returned it cause there was static on the sound card.

Quote:

---

Originally Posted by kpablo

What you guys think?
According to Anandtech... Batch # 3841 are good for them... so it can depend on the motherboard too?
Currently i am NOT very happy with the Gigabyte x58 boards... to much vdroop... also the PCI Express Layout make my vcard scream for more space between each other... The one located at the top is having 85C degrees under load and the bottom one only 70-75C degrees under load...

I am considering to keep my 920 and buy another mobo... i am looking at the eVGA x58... but i think to use both PCI Express in x16 you have to use the 1st and 2nd PCI Express slots... i am right?

---

I am running a 3841A385 on a BloodRage @210bclk HT on (x21). Board is 3d stable and load temps are low 60's on H2O 1.34 vcore. I think this a better than average chip and I just got another 3841A456 from Tiger today.

Quote:

---

Originally Posted by tripgood

I am running a 3841A385 on a BloodRage @210bclk HT on (x21). Board is 3d stable and load temps are low 60's on H2O 1.34 vcore. I think this a better than average chip and I just got another 3841A456 from Tiger today.

---

Is it 1.34 in CPU-Z or BIOS? What is you QPI/VTT voltage ?

Quote:

---

Originally Posted by icecold316

Is it 1.34 in CPU-Z or BIOS? What is you QPI/VTT voltage ?

---

Tested with DMM on board. VTT 1.48.

Quote:

---

Originally Posted by roller11

Yes, but the EVGA is out of the question cause it doesn't support sleep mode.
I saw this as a problem 2 months ago when I realized if the middle slot is used for secondary PCIe card when doing SLI, it would choke the air supply and at best, cause the fan to run full blast with the irritating noise that results. My own testing bares this out. As far as I'm concerned, the middle slot cannot be used cause of the increase in noise, so the bottom slot must be used if doing SLI. The only question is what is the ramification of using the bottom PCIe slot? Some mobos offset the bottom slot towards the bottom so that the vid card will interfere with the case (gigabyte, MSI I believe). The DFI and EVGA position the bottom slot more favorably so that the card may not hit the case. The EVGA is out cause it doesn't support sleep mode, so that is a non-starter. In benchmark tests, I found there to be virtually no difference in graphics performance using a 16x/16x config vs 16x/8x, like maybe 1.5 - 0 % depending on the benchmark. I like the giga so I modified my case by cutting out the area of the case that would cause the interference problem, so I'm now running SLI with the giga ud5 using the top and bottom slots and seeing no performance loss to speak of with the 16x8x config. The DFI x58 board has the advantage of favorable slot placement and 16x/16x config. I had this board for a few weeks, returned it cause there was static on the sound card.

---

Suposedly eVGA's are working on fix that issue...

And could you take a shot of what you modded?

i7-940 batch 3838B125

WoW! Lukeee. What cooler are you using. Nice OC. What is ur qpi/vtt in Bios?

Quote:

---

Originally Posted by icecold316

WoW! Lukeee. What cooler are you using. Nice OC. What is ur qpi/vtt in Bios?

---

Thanks. It's watercooled. I am using an Apogee GTZ extreme with a double and triple rad.

qpi/vtt is 1.35

yeah, high temp was about 67 or so. I think I need to reapply TIM though. I don't think it is even.

My 920 couldn't hit 210 bclk unless I uses 1.75+ qpi/vtt. It was a great chip if you wanted to use max 200 bclk.

So far I am happy with this chip. I am still trying to press it and see where it will go. :)

My stable setup on Scythe Mugen 2.

http://coding-forum.com/prime-i7-3.9ghz.jpg

Your minimum is 64C???