TPU's Core i7 Overclocking and Feedback

[t_ski]

hahaha, yeah so true. I think you are doing well. Your VCore is basically your limiter at this point for a stable OC.

And temps. I have parts on order for my EK to go 1366. As soon as I can swing it I'll get her under water, which should help with the load.

 

[Binge]

And temps. I have parts on order for my EK to go 1366. As soon as I can swing it I'll get her under water, which should help with the load.

I stand corrected lol. You're the first person to put that chip on air. It's been on water since I took it out of the box She likes it.

 

[t_ski]

Lol she likes it wet

 

[t_ski]

Currently going for 3.8GHz

"Overclocking Navigator"
Intel Speed Step: Disable
Ratio CMOS Setting: 20
CPU Frequency Setting: 190
QPI Link Speed: Full Speed
DRAM Frequency: 1540
Command Rate: 2T

"DRAM Timing Configuration"
8-8-8-23-8-5-10-5-5-68 (IIRC) - same for all three channels

"Clock Gen Configuration"
(All Default)

"All Voltage Configuration"
CPU Vcore: +0.240V = 1.392v
CPU Vtt: 1.28V = 1.264v
CPU PLL: 1.8v
DRAM Voltage: 1.660V
NB Vcc: 1.20v
SB Vcc: Auto
QPI/PCIe PLL: Auto

"CPU Configuration"
C1E Support: Disable
Hardware Prefetcher: Disable
Adjacent Cache Line Prefetch: Disable
Max CPUID Value Limit: Disable
Intel Virtualization: Disable
CPU TM Function: Disable
Execute-Disable Bit: Disable
Intel HT Technology: Enable
Active Processor Cores: ALL
A20M: Disable

OCCT is peaking around mid-70's, while CoreTemp is peaking in the low 80's Not sure which one to trust yet.

 

[Thrackan]

I've been running some tests yesterday on the lower end of the spectrum to see what is the lowest vCore my chip takes on stock.
Settings: 20x133, HT on, Turbo on
The Gigabyte UD3R defaults vCore at 1.29375 on Auto settings, which is about 1.25 (idle) in CPU-Z. I have been testing about 5-6 (CPU) vCore settings below that.

I found out 2 things:
- vDrop (BIOS volts vs. CPU-Z volts) is somewhere around 97% for all voltages I tested. So CPU-Z reported voltage on idle is 97% of the voltage set in BIOS.
- I could run 3+ hours OCCT without a single error on 1.0375 vCore (which vDrops to 1.008). Every setting below that had BSOD's or boot problems BEFORE stressing it, and every setting below that vDropped to under 1.000 volts. I guess 1 volt is the absolute minimum for my chip.

Why I wanted to test this:
- First of all, I wanted to see if there was a number I could stamp on the vDrop. I still have to test that on higher voltages, but it looks like 97% is the number I was looking for. It's just plain annoying to set a value in BIOS and getting another value in CPU-Z all the time (Or EasyTune 6 for that matter).
- Second, I plan on doing the same tests with a couple of higher speeds to see if I can find a connection or formula to calc how much vCore I need for which speeds.
- Third, it was bloody cool to see my chip hanging around 40C load for a change

 

[t_ski]

- Second, I plan on doing the same tests with a couple of higher speeds to see if I can find a connection or formula to calc how much vCore I need for which speeds.

Usually voltage is not linear (it's an upward curve). I can take stock volts up to 3.2 GHz or so, then a little more is needed to go to 3.4. However, the amount of voltage needed to be increased for each bump in speed gets bigger as the speed goes higher.

 

[Thrackan]

Usually voltage is not linear (it's an upward curve). I can take stock volts up to 3.2 GHz or so, then a little more is needed to go to 3.4. However, the amount of voltage needed to be increased for each bump in speed gets bigger as the speed goes higher.

Thing is, the chip can do stock speeds on lower volts than the stock (auto) BIOS setting of 1.29375 without failure. I'm not looking for stock volts, but the lowest volts a chip can run a stress test at, because that's the same criteria I use when OC'ing.

But whether it's linear or a curve, we'll know after testing

The plan is to clock purely with BCLK and vCore from 133 BCLK up to 166 in steps of 11 (133, 144, 155, 166) and get the absolute lowest vCore for those settings. I'll be using the same method of testing stability I did earlier to get (relatively) correct results.

Last time I went beyond 166 BCLK, more QPI volts were needed and that will influence the results. This is purely (and only) for vCore.

If the results aren't clear enough I might have to throw back the BCLK to 122 and lower, but I have a hunch my chip won't run properly at any speed when fed less than 1 volt.

 

[Tatty_Two]

Thing is, the chip can do stock speeds on lower volts than the stock (auto) BIOS setting of 1.29375 without failure. I'm not looking for stock volts, but the lowest volts a chip can run a stress test at, because that's the same criteria I use when OC'ing.

Agreed, I can get 3.6gig with HT enabled on "auto" VCore, I might be able to get more IDK as I havent tried, I have also found with my chip, for 24/7 settings (currently at 4gig on 1.25V ) that using the 19x multi rather than 20 really does allow me to lower the voltage and therefore temps. I can now OCCT on air at 4gig for 1 hour and my highest temps are 56C without HT and 63C with.

 

[Thrackan]

Agreed, I can get 3.6gig with HT enabled on "auto" VCore, I might be able to get more IDK as I havent tried, I have also found with my chip, for 24/7 settings (currently at 4gig on 1.25V ) that using the 19x multi rather than 20 really does allow me to lower the voltage and therefore temps. I can now OCCT on air at 4gig for 1 hour and my highest temps are 56C without HT and 63C with.

Do keep in mind that the 'Auto' setting gives me more volts at 3.8Ghz than it does on 2.66Ghz obviously...

I might try pumping BCLK with a 19x multiplier too in a later stage.

 

[Binge]

T-Ski have you tried to do x21 with a lower BCLK? What were the results of that approach?

 

[N4cot1c]

I got my stuff today from PPCS. Took almost a week to get my stuff and I live in the same state..... Thank God for USPS Priority Mail.

I got 3 88 CFM Yate Loons. Should I have them pushing or pulling?

 

[Binge]

Pushing. I've had the best static pressure with Yates pushing.

 

[freaksavior]

I think im going to try and get a 975... but i really dont want to pay the extreme price. maybe i'll wait for th i7 950's

 

[Binge]

A 950... you could do better than that with a 920 >.>

 

[freaksavior]

A 950... you could do better than that with a 920 >.>

why you say that? just bc qpi will max?

 

[trt740]

D0'S are flat out unreal MINES DOING 4.OGHZ ON AIR WITH ONE 120 mm FAN AT 1.33V, FIRST TRY!!!! 21X191 Results will be forth coming.

 

[Binge]

why you say that? just bc qpi will max?

Yep. QPI becomes the limiting factor in a lot of things.

 

[freaksavior]

then why the hell didn't intel realize this before they released the chips?

 

[Binge]

Why wouldn't Intel realize what they're doing? It's like a gateway chip to the extreme. Give you ONE more multi and then restrict the QPI. You could do so much more with more of both! In that instant some hesitant guy with deep pockets bought a $500 and then a $999 chip.

 

[trt740]

D0'S are flat out unreal MINES DOING 4.OGHZ ON AIR WITH ONE 120 mm FAN AT 1.33V, FIRST TRY!!!! 21X191 Results will be forth coming.

Make that 1.296v and dropping, much better than any C0 I have seen on air.

 

[Binge]

It's exciting isn't it?

 

[freaksavior]

Why wouldn't Intel realize what they're doing? It's like a gateway chip to the extreme. Give you ONE more multi and then restrict the QPI. You could do so much more with more of both! In that instant some hesitant guy with deep pockets bought a $500 and then a $999 chip.

so your say they did it on purpose? or am i missing it?

 

[trt740]

It's exciting isn't it?

now testing 1.280v and it's priming and yes it is very exciting. I would say for me to get impressed it takes a lot and I am very impressed.

I believe this voltage will go lower. This is using my Noctua U12P air cooler and one 120mm fan with the case side on.

 

[Binge]

so your say they did it on purpose? or am i missing it?

Yes I believe they're doing it on purpose. Not to overclockers but to mainstream people.

now testing 1.280v and it's priming and yes it is very exciting. I would say for me to get impressed it takes a lot and I am very impressed.


http://img.techpowerup.org/090429/do.jpg I believe this voltage will go lower. This is using my Noctua U12P air cooler and one 120mm fan with the case side on.

My 4.0 with the D0 was 1.216V I wouldn't be surprised if you could get that.

 

[Baer]

Amazing. I think I am going to sell my CO 920 at a bargan rate and get a DO. Very impressive.