• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

VRM Spread Spectrum - What does it do? What platforms?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

wingman99

Member
Joined
Dec 10, 2003
Earthdog said:
I am starting this thread so I can take some off topic posts out of another and get to the bottom of these questions. There are a couple of links below we found to help us answer it, but it still isn't certain.

The questions are:

What EXACTLY does VRM Spread Spectrum do? Does it change frequency or the actual voltage for additional stability and EMI reduction?
Is this a Haswell/Haswell-E feature due to the FIVR? I have only seen it on Z97 and X99 in my quick google searches...


Because of the chronology of these posts, wingman's is first... see above for 'back' information.... :)










When you 'disable' spread spectrum it only affects the BCLK (that we can see)... It does not tighten up voltage tolerances that I have seen. Its about the clockgen and signal integrity/EMI prevention.


Really, these are, to me, typical voltage fluctuations and nothing to worry about.

My sandy-bridge vcore voltage was rock solid with the GA Z68 board. With my new GA Z170 HD3 I think they use spread spectrum on VRM voltage because it varies and you can't disable it.
 
Last edited by a moderator:

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
Voltage varies from what is set to actual regardless of spread spectrum settings... not to mention, you are using software to read your voltage which is notoriously inaccurate in the first place.

Setting spread spectrum to fix voltage fluctuations seems to be akin to fixing your brakes and expecting your AC to work better.

Some links:
https://en.m.wikipedia.org/wiki/Spread_spectrum#Spread-spectrum_clock_signal_generation
Spread-spectrum clock generation (SSCG) is used in some synchronous digital systems, especially those containing microprocessors, to reduce the spectral density of the electromagnetic interference (EMI) that these systems generate. A synchronous digital system is one that is driven by a clock signal and, because of its periodic nature, has an unavoidably narrow frequency spectrum. In fact, a perfect clock signal would have all its energy concentrated at a single frequency (the desired clock frequency) and its harmonics. Practical synchronous digital systems radiate electromagnetic energy on a number of narrow bands spread on the clock frequency and its harmonics, resulting in a frequency spectrum that, at certain frequencies, can exceed the regulatory limits for electromagnetic interference (e.g. those of the FCC in the United States, JEITA in Japan and the IEC in Europe).

Spread-spectrum clocking avoids this problem by using one of the methods previously described to reduce the peak radiated energy and, therefore, its electromagnetic emissions and so comply with electromagnetic compatibility (EMC) regulations.

http://electronics.stackexchange.co...age-controlled-oscillator-improve-emi-on-your
Spread spectrum improves EMI by spreading out the peak emission. Instead of one very strong spike at a high energy level, you end up with a wider "peak". The power output is the same, at the cost of frequency accuracy.

The benefit to this becomes apparent when you have specific emissions requirements to meet. You may not be permitted to emit more than xmW at a given frequency. Without a spread spectrum clock, you might not come close to meeting this requirement, but if you spread the emission frequency out a little, the power at any specific frequency in that range might fall below the maximum allowed, and now you pass.


If you can find something that stated otherwise, I'm a ears. :)
 
OP
W

wingman99

Member
Joined
Dec 10, 2003
Voltage varies from what is set to actual regardless of spread spectrum settings... not to mention, you are using software to read your voltage which is notoriously inaccurate in the first place.

Setting spread spectrum to fix voltage fluctuations seems to be akin to fixing your brakes and expecting your AC to work better.

Some links:
https://en.m.wikipedia.org/wiki/Spread_spectrum#Spread-spectrum_clock_signal_generation


http://electronics.stackexchange.co...age-controlled-oscillator-improve-emi-on-your


If you can find something that stated otherwise, I'm a ears. :)


CPU Voltage Frequency can to be set to Manual to allow selection of a fixed operating frequency for the Extreme Engine DIGI+ III. The higher the switching frequency, the faster the transient response, which yields a more stable delivery of CPU Input Voltage. This may help to yield just a little more BCLK O.C. margin for the CPU used. The effect of high CPU Voltage Frequency may vary with respect to the CPU used. It is highly recommended to Enable VRM Spread Spectrum or Enable Active Frequency Mode when not intending to set the CPU Fixed Frequency to the highest level to allow less emission of electromagnetic interference or better power saving. https://rog.asus.com/10062013/maxim...ngs-for-overclocking-maximus-vi-motherboards/

Like I was saying my sandybridge did not have VRM Spread Spectrum and my skylake does.
 
Last edited:

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
You have a setting in your bios for vrm spectrum? I don't get what you keep on trying to say...

So that board has it, does yours?? (Just because one board has it doesn't mean all do or its 'enabled automatically' on your board - that is a ROG board versus your budget one... as well as being haswell versus your skylake). However, it is just as I said, for frequency and EMI reduction. It does not seem to have an effect on the actual voltage value as you seem to infer in post 19. Did I misread your passage or meaning?

I also wonder, since what you quoted is for Haswell, is because of the FIVR. Skylake went away from that model and are more like SB in that respect...you can see why Haswell would have it with everything packed into the die, but not Skylake.
 
Last edited:
OP
W

wingman99

Member
Joined
Dec 10, 2003
Just because there is not a setting for VRM Spread Spectrum does not mean it's not there, I can see it working with my new board the voltage goes up and down unlike the PSU voltage. It is a Intel specification fore the motherboard manufactures to use it and adhere to.

I know you like to say great motherboard have clean VRM Voltage so I know this is hard for you to digest. Where would we be if you purchase $400 Asus board and Asus says the VRM section is great however you voltage will fluctuate for stability.
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
That is true that if there isn't a setting perhaps its automatic...or perhaps its not there. At the same time, if I am thinking correctly, you also can't see it working. All I know, from what I linked and reading yours, is that it changes the FREQUENCY, not the actual voltage. For example, you can have 1.2V at 61Hz and 59Hz with it 'enabled' but without it would hold at 60Hz 1.2V with more EMI. Again, from what I have read so far, I don't see where it changes voltage, just the frequency to reduce EMI. The PSU voltages have nothing to do with it...

You would think if it was automatic on some boards, common advice all over the place on it would say to enable it. However, most of what I have read, say to leave it disabled, like CPU spread spectrum. Every time I googled I came across it on Z97 boards and X99 (both Haswell and Haswell-E chipsets). That could be a coincidence though...

I am still open for links that support what you are saying...hopefully your question at the Intel forum gets answered since you are guessing at it like me: https://communities.intel.com/thread/105982


Its only hard to digest because of the lack of supporting facts from you in this discussion. ;)


EDIT: Ok, its in Skylake too. I see an option in my board. :)

In my MVIIIE it shows 'reduces the magnitude of peak noise from the VRM. Enable to reduce peak noise, Disable this setting when overclocking'. So I tested my voltages using a multi-meter and did not find a difference in actual voltage set or fluctuations with it enabled or disabled. That said, I am not sure there is a way to find out without asking Gigabyte if your motherboard uses it without having that option in the BIOS. But so far, it appears that it does NOT touch the voltage as you state.
 
Last edited:

JrClocker

AKA: JrMiyagi
Joined
Sep 25, 2015
The best way to think about digital clocks is with the concept of a Fourier Series:

https://en.wikipedia.org/wiki/Fourier_series

Basically, any non-sinusoidal signal can be represented as a sinusoid at the base frequency of the clock, plug a sum of the harmonics of the base frequency...with each of the signals have a different magnitude and phase.

For example: A perfect square wave (which is what you want in a digital system) at 100 MHz will be composed of a sine wave at 100 MHz, 300 MHz, 500 MHz, etc.

However, it's impossible to generate a "perfect" square wave as this would require a circuit with infinite bandwidth.

The faster the rise or fall time, the more higher frequency content is required.

If you look at a frequency spectrum plot of a square wave, you will see the largest peak at the fundamental frequency, and peaks of reducing amplitude at multiples of the fundamental frequency.

A common mistake (that my engineers just made last week) is to look at a 50 MHz square wave signal on an oscilloscope with a 200 MHz bandwidth...and then spend hours troubleshooting the circuit as it's ringing. I smack them up side the head, remind them of their buddy Fourier, have them get an oscilloscope with a 1 GHz bandwidth...and solve their non-existent problem.

Spread spectrum is a technique where you vary the frequency of the source signal (square wave, sine wave, etc.) a few percent (to 10's of percents...depends on the application). When you look at this sort of signal in the frequency spectrum, you will see a peak that represents a "spread" of frequencies for the fundamental and harmonics. The amplitude of these peaks will be lower, but the energy will still be the same. (The energy is calculated by integrating the signals...or calculating the area under the curve in the frequency spectrum.)

EMI testing is all about keeping your peak energies below the required values...and less about the total spectral energy your circuit produces.

I hope that helps explain stuff...it's kind of hard to break down high level electrical engineering concepts when you may not have a lot of the necessary physics and calculus background...if I confused anybody, let me know and I will try to explain differently.

(A common mathematical technique used in electrical engineering is to do a Fourier Transform on a signal in the time domain. This will produce the frequency domain content of the time domain signal. Consequently, you can do an Inverse Fourier Transform on the frequency domain content to get the time domain content. The reason you do this is that you can multiply subsequent circuit transfer functions in the frequency domain...but would have to do a rather ugly convolution in the time domain to get the same result. An FFT is a Fast Fourier Transform (and IFFT for Inverse FFT) that uses digital sampling techniques to produce an approximate of the mathematical Fourier Transform calculation.)
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
That was a great and detailed explanation Jr, thank you. :)

From your understanding, VRM spread spectrum affects the frequency, not voltage, correct? Just as cpu spread spectrum allows the BCLK to 'float'. In 'extreme' overclocking, it has always been recommended to disable cpu spread spectrum as that 'float' can cause instability.
 

JrClocker

AKA: JrMiyagi
Joined
Sep 25, 2015
Correct...just the frequency.

However, if I was to put an oscilloscope on the signal (with enough bandwidth), I may see a lower peak voltage value as the frequency is bouncing around.

For anything extreme, I would highly recommend turning off the spread spectrum...as the constant changing in frequencies could cause chaos.

The "goodness" of digital systems is measured by creating an "eye diagram". Basically, you run an infinite persistence time-domain trace of the digital signal at a fixed trigger voltage and/or time. If the signal source is stable, you'll see a plot that looks like an open eye...with the rising and falling edges creating the picture. If the 'eye" is too closed, your high and low signal levels are not far enough apart and/or you have ringing issues because your circuits do not have enough bandwidth. If the "eye" is too narrow, then your frequency is not stable and/or you have ringing issues because your circuits do not have enough bandwidth.

If you try to make an eye plot on a spread spectrum signal...you get complete garbage as the frequencies are bouncing all over the place. The subsequent ringing at different frequencies can "add up" and cause instability...kind of like when many small waves are floating around the ocean, but can combine "just right" to make a huge Rogue Wave.

Make sense?
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
It does indeed to me. :)

The lower peak voltage value one may see, you mention "with enough bandwidth"... so that would mean our software which refreshes at what .5s? 1s? would not capture such changes, correct? You would need the order of 1GHz bandwidth/updating to see the peak voltage actually lower? How much? Are you talking .01V? .001V? How much may it change if it changes and is caught with high end tools?
 

JrClocker

AKA: JrMiyagi
Joined
Sep 25, 2015
Correct...but for different reasons.

What you are referring to is taking a very slow speed sampling. To do this, you have to run the signal through a low-pass filter. This low pass filter will eliminate a certain amount of high frequency content. For a slow sampling interval of 0.5 sec to 1 sec, you are basically looking at an RMS (or DC value) of a very fast signal.

Apples to oranges.

If I had a fast enough oscilloscope, I would see a "square wave", but it would be flickering all over the place as the frequency is constantly changing. I would have to trigger the scope at a certain voltage level, and turn on the "persistence" function of the scope. This would yield a picture that would be "square-wave-ish", but would have very fat rising and falling edges as the frequency is constantly changing. To verify that the voltage levels are the same, you would want a very thin line for the top and bottom horizontal parts of the square wave. If the top part is "fat" too, then the voltages are not the same for the different frequencies. If the edges of the top are "fat" but the center is thin...then your voltages are the same and you are seeing ringing.

Sooooo hard to explain without a picture...and I doubt I could sketch something in MS Paint to do it justice!
 

EarthDog

Gulper Nozzle Co-Owner
Joined
Dec 15, 2008
Location
Buckeyes!
Gotcha... In other words, we won't catch it without some very accurate and likely expensive equipment. ;)
 

JrClocker

AKA: JrMiyagi
Joined
Sep 25, 2015
Expensive is relative - hehe...but yeah.

As a general rule for digital signals, you want your scope to be AT LEAST 10x the frequency you are trying to measure. So, to measure a 100 MHz digital clock, you want at least a 1 GHz bandwidth scope...that will run you at least $10 K.

Then, you need scope probes that have enough bandwidth...add another $2K or so.

Then, you need to know how to properly measure the signal (the ground clips and/or wires you solder into the circuit to make measurements will cause issues if you don't connect properly - i.e. you make a low pass filter without even trying to). Making these sorts of measurements can be difficult.

And then...you can change the signal just by trying to measure it as your scope probe puts a load on the circuit.
 
OP
W

wingman99

Member
Joined
Dec 10, 2003
When you have 1.3v DC with a CPU vcore VRM how is there any frequency in that for Spread Spectrum? Unless it is a Digital DC signal that you don't need to power the processor transistors.
 

JrClocker

AKA: JrMiyagi
Joined
Sep 25, 2015
VRMs are basically DC-to-DC switching power supplies (probably just efficient buck converters...but I've never looked at a schematic).

The switching frequency is what is varied.


Edit: Yes, they are buck converters:

https://en.wikipedia.org/wiki/Buck_converter

Scroll down to "Multiphase buck".
 
Last edited:

JrClocker

AKA: JrMiyagi
Joined
Sep 25, 2015
+1 to JrClocker. Couldn't have said it better :thup:


Side-note: Are you still an active EE?

I'm a COO now. I keep my "EE self" busy doing side projects, but I haven't done any detail circuit designs in about 10 years...but I still write code (embedded and Windows) for fun!

In my COO role, I have manufacturing, quality, IT, supply chain, and engineering reporting to me. When I put on my "EE" hat now, it's more of a systems engineer role. But, the physics is still the same! :thup:
 
OP
W

wingman99

Member
Joined
Dec 10, 2003
VRMs are basically DC-to-DC switching power supplies (probably just efficient buck converters...but I've never looked at a schematic).

The switching frequency is what is varied.


Edit: Yes, they are buck converters:

https://en.wikipedia.org/wiki/Buck_converter

Scroll down to "Multiphase buck".

If the switching frequency is varied would that allow time for the voltage to raise and fall faster then the digital metal oxide semiconductor field effect transistor can control in time?
 
Last edited:

JrClocker

AKA: JrMiyagi
Joined
Sep 25, 2015
Don't think so...if the design is done correctly.

However, I would expect a transient as the frequency is shifted.
 

Alaric

New Member
Joined
Dec 4, 2011
Location
Satan's Colon, US
It was an available tweak on my AM3+ platform, too. (M5A99FX Pro) Asus' explanation was a simplified version of what's been posted, frequency modulation seemed to be the gist of it. I think the options boiled down to performance or stability on that board. The adjustment is bundled in Asus' AI Tweak in the BIOS and also in their software suite (which name escapes me at the moment). I think the BIOS adjustment actually gave frequencies in a sub menu to choose from. I won't have access to that rig until Monday, but I'll try to get some pics of the screen.