haswell vs coffee lake voltages

C

chrcoluk

[H]ard|Gawd
Joined
Jul 7, 2008
Messages
1,260
Why is coffee lake voltages so high?

My haswell stock voltages were just 1.05v
I can run 4.3ghz at 1.175v

Stock coffee lake seems over 1.25v ? which seems massive.
People having to run over 1.3v for moderate overclocks.

Am I missing something, no reviews have batted an eyelid at these high voltages.

Does this mean coffee lake is natively a hot and power hungry chip with less overclocking headroom?
 
chrcoluk said:
Why is coffee lake voltages so high?

My haswell stock voltages were just 1.05v
I can run 4.3ghz at 1.175v

Stock coffee lake seems over 1.25v ? which seems massive.
People having to run over 1.3v for moderate overclocks.

Am I missing something, no reviews have batted an eyelid at these high voltages.

Does this mean coffee lake is natively a hot and power hungry chip with less overclocking headroom?
Click to expand...

So run a CFL with 2 cores disabled and clocked at stock speeds if you want it to use the same power and run at the same voltages as haswell.

1.25 isn't stock, that will get you 4.8 GHz all core on most chips just fine.
 
I think I know whats going on regarding the high stock voltages, but I am speculating, as not much info I could find.

I think because haswell had a internal voltage regulator, each chip was programmed with a stock voltage based on its binning, hence if you had a good chip you had a low stock voltage. (different people reported different stock voltages)

Now the motherboard on coffee lake has the voltage regulator, the motherboard vendor has no idea of the binning of the chip so has to play safe, hence the very high stock voltage. (everyone seems to have same stock voltage which is usually much higher than needed)

Makes sense? or am I thinking of rubbish :)

It seems I think I have a reasonable chip.

I have it at 5ghz right now and the offset at 10 (but I not even tried lower yet).

So under load with LLC max it hits 1.28v at 5ghz. idles at 0.6v
At 3.6ghz non turbo speeds the vcore is at 1.168v, thats still higher than my haswell, but not as scary as I thought it would be. Although I know I have a really good haswell chip, my voltages are lower than what most people reported on haswell.

Of course yeah the extra 2 cores is obviously a factor. :)

My concern is not so much the power usage but rather not degrading the chip from excessive voltages, if the vcore limit is higher than haswell, to match the higher vcore usage so I have the same headroom, then its all fine and dandy.
 
Voltages are all relative, you can't just naively compare architectures with different core counts separated by years of development. It's apples to oranges.
 
Well yes maybe.

My friend's amd 9590 chip which has a stock vcore of over 1.4v now has to be underclocked to prevent freezing, its less than 2 years old and only had moderate use.

I asked this question to try and find the reason why and if the operational limit is higher, sadly noone seems to know the exact reason either way, extra cores is logical, tech differences possibly, but if the voltage limit is not higher than haswell then it seems logical to conclude the newer chips are running closer to their operational limit and as such would have a shorter lifespan unless intel added some kind of endurance enhancement to them.

Just saying "they run higher, it is how it is" wasnt really an answer :)

Some people have always ran chips in the 1.3+ vcore range so probably dont care one bit, but to me this is new terriritory as I have always kept my vcore way below that. :)
 
Well, a couple things which may satisfy you then.

In general, higher clocking will require higher voltages. The transition time of logic states depends upon the voltage applied, and higher will force the transitions faster. This means the line will more likely be where it is supposed to be at the next clock edge, so you are less prone to problems.

Second, as Sancus said the effects of a vcore on one processor will not have the same effect in another. In particular, the newer process nodes are usually much more effective in combating leakage, so higher voltages have less impact than they did in the past. Less leakage, less power, less heat, and less "wear".

It is not a foregone conclusion that these newer chips with a higher voltage have less operational life than previous generations with a lower voltage.
 
Consider that since Skylake, the IVR was removed. So none of that Haswell experience we have carries over to CFL lmao.
Only Pepperidge Farm remembers VRIN scaling :p
 
Using offset I have settled on this scenario.

idle 1.184 (or 0.6v if allow cpu to idle at 800mhz instead of 3.6ghz).
load non AVX 1.29-1.31
load AVX 1.34-1.39 occasionally spiking to 1.41

I wish there was a lighter AVX load tester that emulates AVX games like GTA5.

Do these above voltages show concern for 24/7 use?

c3/c6 states actually are making no difference at idle like they do on haswell, kinda odd, but I will be disabling both states anyway.
 
chrcoluk said:
Using offset I have settled on this scenario.

idle 1.184 (or 0.6v if allow cpu to idle at 800mhz instead of 3.6ghz).
load non AVX 1.29-1.31
load AVX 1.34-1.39 occasionally spiking to 1.41

I wish there was a lighter AVX load tester that emulates AVX games like GTA5.

Do these above voltages show concern for 24/7 use?

c3/c6 states actually are making no difference at idle like they do on haswell, kinda odd, but I will be disabling both states anyway.
Click to expand...

Voltage by itself is not something you should worry about. The concern from increased voltage is the effect on temperature. If you have that under control in a steady-state, you're gold.

And yeah, there's a weird middle ground of things using AVX but in a mixed execution case which doesn't destroy your CPU like a math-only benchmark. On my system, I leave AVX in the "if fully pushed, could hit scary temps and throttle" setting, because outside something like Prime95, nothing i do actually works that way.
 
the temps are fine, even if I let it have full power so no power throttle its nowhere near the temp limits, although in open air not in case yet.
 
You must log in or register to reply here.
Back
Top