OC Limits: Processor? Or Motherboard?

[BonzaiDuck]

I'm testing a dated but high-end processor with an ASUS budget motherboard: Z77-A.

The BIOS lacks the additional "fine tuning" feature of "Extra Voltage for Turbo." So manual voltage adjustments are limited to vOffset in "offset mode." There is no way to decrease PLL Voltage: it's either going to be 1.8V with "Auto" or 1.8 + 10%.

The board has 4+1 phase-power-design. Which -- by some comparisons -- is puny.

If I find that the processor won't easily go above a speed that should be easy for it -- if I can only get 4.7Ghz stable by increasing LLC to "Ultra High" or 75% -- and at that point VCORE equals VID and VCORE for the 4.6 setting under high load was 1.33 while it shows 1.40V for 4.7, then I have a question.

Is the limitation due to the budget phase-power design of the board? Or could it be the processor?

 

[BonzaiDuck]

I'm testing a dated but high-end processor with an ASUS budget motherboard: Z77-A.

The BIOS lacks the additional "fine tuning" feature of "Extra Voltage for Turbo." So manual voltage adjustments are limited to vOffset in "offset mode." There is no way to decrease PLL Voltage: it's either going to be 1.8V with "Auto" or 1.8 + 10%.

The board has 4+1 phase-power-design. Which -- by some comparisons -- is puny.

If I find that the processor won't easily go above a speed that should be easy for it -- if I can only get 4.7Ghz stable by increasing LLC to "Ultra High" or 75% -- and at that point VCORE equals VID and VCORE for the 4.6 setting under high load was 1.33 while it shows 1.40V for 4.7, then I have a question.

Is the limitation due to the budget phase-power design of the board? Or could it be the processor?

Wey-all, 'Lo there, Bonzai!! Been a my excellent compatriot's age!! How the hell are ya?!

Ah think it's the phase-power design Bonzi!! Not enough chokes for the VRM. Limpo motherboard, Duck-buddy!!

 

[aigomorla]

lmao....

talking to yourself?

It can be the board, however the board end will only merit about 30% better overclock at BEST. This is mostly due to a better regulation of power to the said sockets, and how well the bios is optimized to handle operations at such speed.

It's also the quality of the chip, how it leaks voltages, how the IMC inside the chip is handling operations.

Lastly, it also could be ram.

 

[videogames101]

what is this thread haha

could be either unfortunately! what kind of instability do you see when you push the chip further?

 

[BonzaiDuck]

lmao....

talking to yourself?

It can be the board, however the board end will only merit about 30% better overclock at BEST. This is mostly due to a better regulation of power to the said sockets, and how well the bios is optimized to handle operations at such speed.

It's also the quality of the chip, how it leaks voltages, how the IMC inside the chip is handling operations.

Lastly, it also could be ram.

Need to get out my calculator . . . misplaced the slide rule somewhere in the bathroom . . . Hmm . . .

30% of what? The base clock? 30% of the difference between turbo clock and highest stable clock? Whatever it is, I think the chip is probably in the ball-park. It may not be the best of the lottery -- can't say.

Not likely to be the RAM -- tested and perfect. Running at spec. But so far interesting. I'd done 3.5 hour tests with OCCT -- some with Linpack, some with OCCT: CPU. I tried 4-thread LinX large problem size and 90% RAM -- 20 passes. Then I'd do Prime95 Large FFT and one thread would fail.

That may be because of insufficient voltage, but the proof of the pudding is in my description of the highest clock setting. "Ultra High" LLC and the actual voltages is a bit much. Not even consistent with the lower clocks on this chip, or others I'd used. Seems I tried to increase VCSSA (or VCCIO) to bolster the IMC, but it didn't help before I raised the LLC just to "see."

They don't make these $130 motherboards for OC'ing. I don't know if the "motherboard conclusion" is a logical converse of this, but I'd looked at some comment on this board or other 4+1 boards, and the wisdom said "don't bother trying to clock it beyond the point where you need to adjust the voltage." Another review noted that the board got to 4.7 with an i5-3570K, but I don't know how that bears on this.

What I DID discover and should have known for years is this. There are resellers I'd not seen mentioned in the forums much who buy corporate "IT" assets as surplus to evaluate, sort, and either resell or recycle. A budget-minded electronics veteran I know vouched for one I'd discovered through EBAY while window-shopping. I think I just found the 8+4 phase-power board for which I'd paid $220 when it was "retail-fresh" for $85. It looks like they don't sell defective parts if they can help it.

I'm going to find out fairly soon about this for sure. And cut myself loose from socket 1155 except for replacement-free maintenance.

And the water project is still on beginning in 2015 . . .

BACK AGAIN: Yeah, I looked at a review of the board. IB i5 @ 4.7? Sure! With a voltage setting that makes VCORE 1.4 to 1.45V in turbo or load. Always read the small print. Luckily, I didn't buy this board for OC'ing anything, and I just happened to have it handy. We'll see how all this changes with the "corporate surplus" board . . . .

 

[videogames101]

Large FFT failures? I would be trying that PLL overvoltage setting, as well as the periph voltages you already tinkered with. Although be careful with PLL settings, they can die quick. Maybe try loosening memory timings a tad.

Also, aren't you hitting temp limits with IB at 1.4V????

 

[BonzaiDuck]

Large FFT failures? I would be trying that PLL overvoltage setting, as well as the periph voltages you already tinkered with. Although be careful with PLL settings, they can die quick. Maybe try loosening memory timings a tad.

Also, aren't you hitting temp limits with IB at 1.4V????

That's just it!! You get two flavors of PLL Voltage with this board: "Auto" and "+10%."

Enough forum exchanges, reviews, OC guides and other information seems to have established a strong possibility that stability can be found by lowering PLL Voltage. Some had actually lowered it to 1.55V, but this could seem impossible if there is a general standard of PLL voltage across a range of electronics components -- maybe 1.65. But the sweet spot for it with boards and relevant processors seems to be around 1.68V.

timings shouldn't need loosening; the RAM is running at spec settings. Consider that I've had much better experience with a board that has 8+4 phase power design. The only time I needed to twiddle with timings and IMC or VCCIO voltage involved either RAM overclocking or command-rate = 1.

As for the IB-K processors, I'd think that 1.44V would smoke those suckers. First -- the TIM in the fabrication process. Second, the lithography, die-shrink and corresponding voltage range. If you ask me, 1.44V is a bad idea for Sandy Bridge, but People had done it. Sandy is 32nm; Ivy is 22nm; and that would make it even worse.

But like I said, I only bought the board to fill two possibilities: server replacement, and fam-damn-ily desktop replacement. So neither a loss nor confirmation about this chip. I'll know more soon -- or so I hope.

It's just a headache dealing with white-box windows licenses knitted to hardware when you're mixing and matching components to "come up with something."

 

[PhIlLy ChEeSe]

I'm pretty sure 1.7 PLL is the sweet spot, How are you cooling the CPU? What CPU are you using? around 1.4v should get you to 4.7, may need a tad more on the v core. Should have a good fan running air across the VRM'S, I think you can tweak the PLL in windows with the Asus software. I know many refuse to use it, I never had any issue with it except it not working on certain installs. 1.45v should be around your max, of course if your using water that all goes up.
As you said it could just be the board, google will yield if that's true.

 

[videogames101]

The RAM itself is probably find at that spec, but loosening timings could help with stability on the IMC, I mean it doesn't cost much to try.

 

[BonzaiDuck]

I'm pretty sure 1.7 PLL is the sweet spot, How are you cooling the CPU? What CPU are you using? around 1.4v should get you to 4.7, may need a tad more on the v core. Should have a good fan running air across the VRM'S, I think you can tweak the PLL in windows with the Asus software. I know many refuse to use it, I never had any issue with it except it not working on certain installs. 1.45v should be around your max, of course if your using water that all goes up.
As you said it could just be the board, google will yield if that's true.

It's the 2700K I'd avoided buying just because it didn't make a lot of sense, then I lost my senses and succumbed to the temptation. Just a minute and let me look at that system's "Turbo EVO" to see if I can find the feature you mentioned: . . . .

Ha. It's like I said. The slider for PLL Voltage begins with 1.8V where it's currently set and can only go up and not down.

I know there are bad straws from the chip lottery, but I also know that SB-K doesn't need above 1.4 to get 4.7. Oh -- gee -- I suppose that's what I'm trying to find out here, but the temperatures are almost insane for an air cooler: Average of Maximums after an hour of Linpack are 5C lower than my other system at the same speed and voltage. Figure about 68C at 4.6.

The heatsinks on the VRMs of these boards are flimsy little aluminum chunks. They can say that the Z77-A board is overclockable, but I don't think it was made for overclocking. Anyway, the board will have a "use" in the scheme of things, but making the 2700K do somersaults isn't one of them.

Testing with a low-end board -- bad idea. Chunking a pin on a Sabertooth not only a bad idea, but a crying shame. I'm just fiddling with the Z77-A until I make up for the Sabertooth loss. Sooner or later, this will all prove out and I'll pull the curtain on this . . . project . . .

 

[BonzaiDuck]

Footnote.

I observe a few things.

First, you can get some good clocks with a 4+1 phase-power budget board.

Second, while people talk of a "chip lottery" which I don't deny, these old SB-K processors seem darn consistent.

Third, even the motherboards (at least from ASUS) seem consistent between upper-mid-range and budget: I get to 4.7Ghz with the same voltage equivalent settings on a P8Z68-V-Pro (2600K) and a Z77-A (2700K). But the actual reported VCORE on the 2700K seems a tad lower.

Fourth -- this dovetails with my observations about a certain EVGA ACX cooler posted in a thread touting the 212-EVO as awesome -- which it is, in the proper context. Without water-cooling and at 77F room-ambient, have we ever recorded here in the forums a case where 135W of thermal power -- 1.37V drooped load and 4.7 Ghz on a Sandy -- ever shown 69C average-of-maximums temperature running Intel BurnTest on the "Maximum" stress setting?

That used to be the sort of limp temperature I noted getting with Prime95.

Fifth, I suspect I should just satisfy myself what develops with this without Hyperthreading. Old s***!! I know!! I'm jus' still havin' fun!! That's all I can say!