Asrock Z97 Memory Voltage Issue...

[PowerPie5000]

Hi all,

I have a small issue regarding the actual voltages supplied to the memory with the Asrock Z97 Extreme4 board. First of all, does this board (or any other Asrock Z97 board) actually have real DRAM voltage sensors? I'm asking as the only utility that seems to display the correct DRAM voltage is Asrocks own A-Tuning utility, and even then I don't think it's a sensor reading... I think it just shows whatever voltage I select in the bios (not from a sensor).


Speccy shows the Memory Controller voltage as 1.696v which is too high for the 4790K (mine should be set at 1.500v). I thought it might have been a phantom/false reading, but HWMonitor also shows the same voltage as VIN1 and VIN11... No utility other than Asrocks own will show a voltage of 1.500V for the memory . I have a feeling this board is overvolting the memory and the Asrock A-Tune utility only shows the voltage number set in the bios and not the actual voltage reading itself. The Asrock utility also shows lower motherboard temps compared to other software.

[parsec]

Re: Asrock Z97 Memory Voltage Issue...

Interesting, thanks for posting this.

But IMO what you are seeing in Speccy that you think is the standard memory voltage is something else. Since your PC is a Haswell processor based system, voltages are different than earlier Intel systems.

Recall that Haswell processors have a CPU Input voltage rail that is not the final VCore voltage. Haswell processors have another input voltage for the memory interface that is also not the final DRAM voltage. Note that Speccy labels that voltage Memory Controller, not memory or DRAM voltage. I think you are seeing the memory interface input voltage in Speccy.

Which brings up the fact that ASRock does not provide a UEFI/BIOS option for the memory interface voltage, that is available on other boards. If you think your memory voltage is high, you should see mine on an ASRock Z87 EX 6 board.

Out of time now, more on this later...

[PowerPie5000]

Re: Asrock Z97 Memory Voltage Issue...

Interesting, thanks for posting this.

But IMO what you are seeing in Speccy that you think is the standard memory voltage is something else. Since your PC is a Haswell processor based system, voltages are different than earlier Intel systems.

Recall that Haswell processors have a CPU Input voltage rail that is not the final VCore voltage. Haswell processors have another input voltage for the memory interface that is also not the final DRAM voltage. Note that Speccy labels that voltage Memory Controller, not memory or DRAM voltage. I think you are seeing the memory interface input voltage in Speccy.

Which brings up the fact that ASRock does not provide a UEFI/BIOS option for the memory interface voltage, that is available on other boards. If you think your memory voltage is high, you should see mine on an ASRock Z87 EX 6 board.

Out of time now, more on this later...

Thanks for the explanation and what does Speccy show for your Asrock Z87 board?

So do you think 1.696v (basically 1.7v) sounds about right as the input voltage for the IMC on a 4790K? My RAM is currently set at 1.5v, but there's no real way of checking this as I don't think this motherboard has any DRAM sensors. The Asrock A-Tuning software also incorrectly reports the motherboard temp. The BIOS and other software will show it hovering around 32c when idle, but the Asrock software will show it at 24c... There's usually an 8c difference. It can reach around 55c when gaming (i'm guessing it's the PCH temp?).

Another quick one... My CPU Cache is set at 1.20v by default (auto) In the bios, but OCCT reports it reaching 1.29v under load. That's a bit of an overshoot! Is it anything to worry about in the long run?

[parsec]

Re: Asrock Z97 Memory Voltage Issue...

My Z87 EX 6 board, with a little G3258 processor, had that voltage set to 1.808V. That's with Samsung 'green" memory, which can operate (and is set at) 1.35V.

My Z97 EX 6 board's "Memory Controller" voltage is the same as yours:



I've done this already, but a test you can do is change the DRAM voltage in the UEFI/BIOS, and see if the value shown by Speccy changes. If it was the actual DRAM voltage, I would expect it to change, and mine did not.

Speccy is not able to read a DRAM voltage, the values is shows in the RAM section are the SPD data voltages programmed into the memory. The values did not change when I changed the DRAM voltage in the BIOS, to 1.40V during my test.

So the "Memory Controller" voltage shown by Speccy is the Haswell memory interface input voltage (Intel's term is VDDQ ). Just above that we find CPU CORE, which is really the CPU Input voltage, but the value shown has not been adjusted correctly. That is, some sensor data values must be adjusted before they are the correct value. Given your Speccy screen shot, I predict your CPU Input voltage is set to 1.808V.

If we generally equate and compare CPU Input voltage to the Memory Interface input voltage, looking at your values first, we can see they are not very different. A CPU Input voltage of ~1.800V is a common default/Auto value for a Haswell CPU running at stock clock speed. Our Memory Interface voltage is 1.696V, or ~1.7V, quite close to 1.8V. Given DRAM operating voltages of 1.35V - 1.5V+, those would be high to crazy high actual VCore voltages.

OTOH, given what the Intel specs for VDDQ seem to be, 1.7V or more is higher. I'm not sure what to think about that, Intel's numbers don't seem to make sense (VDDQ = 1.35 - 1.5V, back to DRAM voltage again!) I assume the higher voltage on our boards is for memory OC, but considering the Intel specs, I want a VDDQ setting. OTOH, the VDDQ rail is static, as is the DRAM voltage it creates, and is a low power interface (2.5A) so setting it internally and forgetting it may be Ok.

You might notice that my CPU Input voltage, translating from the Speccy value, is 1.584V. That is for stock clocks of an i5-4670K, manually set by me. Some people would think that is to low, but my CPU's stock VID is ~1.0V. Intel suggests CPU Input voltage be 0.4V - 0.6V higher than the maximum VID, which mine is. Intel states using higher CPU Input voltages causes more wear and reduced life span of the FIVR components. I've found I can turn it way down depending on the CPU speed, or really the VCore. The same should apply to VDDQ IMO.

Otherwise, VIN5 is CPU Digital IO voltage, VIN4 could be CPU Analog IO voltage. Your VIN6 value has me wondering about mine, not that either are wrong. What CPU power saving options do you use? Any C States?

The temperature readings in A-Tuning I don't use, I use another monitoring program. Speccy System Temperature matches the mother board temperature I see, but that is not normally the PCH/chipset temperature. Speccy may not be using the same sensor reading that A-Tuning does, so different temperature readings.

Your core idle temps are very low, is it cool where you live now, or do you have liquid cooling, or something else going on?

[PowerPie5000]

Re: Asrock Z97 Memory Voltage Issue...

Your VIN6 value has me wondering about mine, not that either are wrong. What CPU power saving options do you use? Any C States?

All C-States are enabled in the bios apart from the one that says "Package C-State Support". The VIN6 value is shown in OCCT as the CPU Cache.. I'm not sure how accurate it is and I might also add my CPU cache is supposed to be 1.20v according to the bios.

Your core idle temps are very low, is it cool where you live now, or do you have liquid cooling, or something else going on?

It's pretty cool here and my computer room is about 18c. I have the heating turned off in here most of the time as my PC warms up the room when gaming (mostly down to the GTX 780 exhaust )... It's only a small room and doesn't take much to warm it up! My chip is pretty good when it comes to temps... That's assuming the temps are accurate of course. My GTX 780 uses a reference blower cooler so I don't have hot air dumped inside the case. I also have 2x intake and 2x exhaust case fans.


So basically from what I can gather, this board lacks certain sensors/readings and bios adjustment options. It might also be a bit too aggressive with some of the voltages supplied when leaving everything to auto in the bios. I'm running stock and the bios still feels the need to add offset voltages to the System Agent and the digital/analog IO voltages.... I hope none of the default settings are going to degrade any part of my CPU! Oh and you're correct as the input voltage for my CPU is 1.8v. I'll probably leave that as it is as my CPU voltage hits 1.263v when it boosts to 4.4GHz!

[parsec]

Re: Asrock Z97 Memory Voltage Issue...

All C-States are enabled in the bios apart from the one that says "Package C-State Support". The VIN6 value is shown in OCCT as the CPU Cache.. I'm not sure how accurate it is and I might also add my CPU cache is supposed to be 1.20v according to the bios.







It's pretty cool here and my computer room is about 18c. I have the heating turned off in here most of the time as my PC warms up the room when gaming (mostly down to the GTX 780 exhaust )... It's only a small room and doesn't take much to warm it up! My chip is pretty good when it comes to temps... That's assuming the temps are accurate of course. My GTX 780 uses a reference blower cooler so I don't have hot air dumped inside the case. I also have 2x intake and 2x exhaust case fans.


So basically from what I can gather, this board lacks certain sensors/readings and bios adjustment options. It might also be a bit too aggressive with some of the voltages supplied when leaving everything to auto in the bios. I'm running stock and the bios still feels the need to add offset voltages to the System Agent and the digital/analog IO voltages.... I hope none of the default settings are going to degrade any part of my CPU! Oh and you're correct as the input voltage for my CPU is 1.8v. I'll probably leave that as it is as my CPU voltage hits 1.263v when it boosts to 4.4GHz!

I had forgotten about the default/auto offsets applied to the System Agent and the Digital and Analog IO voltages, I noticed that after using the board for a while. To change that, I just set the offset voltages for those options to the lowest value, +0.001V. Never had any problems with those settings.

Your CPU Input voltage is not to high given your maximum VID at 4.4GHz. I just enjoy knowing that at stock clocks and lower core voltage, CPU Input voltage can be set much lower, and I do that when running stock clocks.

I agree that VIN6 seems to be the CPU Cache voltage... well, VID of course.

A bit strange the OCCT displays CPU Input voltage correctly, but has CPU VCORE displayed as ~0.9V. As I said before, that value when adjusted is CPU Input voltage. But on other boards that value might be the correct VCore, or as close as can be determined. No such thing on most ASRock Haswell boards, so we get that value instead.

[PowerPie5000]

Re: Asrock Z97 Memory Voltage Issue...

A bit strange the OCCT displays CPU Input voltage correctly, but has CPU VCORE displayed as ~0.9V. As I said before, that value when adjusted is CPU Input voltage. But on other boards that value might be the correct VCore, or as close as can be determined. No such thing on most ASRock Haswell boards, so we get that value instead.

The HW monitor in the system bios shows 1.112v as my CPU Vcore (ignore the crazy fan speed and the spelling mistakes for "voltage" on the digital/analog IO ):

[parsec]

Re: Asrock Z97 Memory Voltage Issue...

The CPU "VCore" of 1.112V seems right IMO given what I've seen with my board and CPU. The BIOS reading tends to be lower than in Windows. But is that at 4.4GHz?

CPU Cache Voltage seems high, is that an Auto setting? Is that with the Cache at 4.4GHz?

I've seen the Cache Voltage on this screen a bit higher than the CPU VCore, with manually set Adaptive voltages that should have had the Cache voltage lower. Makes me wonder...

I don't have the Volatge typo in the Extreme 6 UEFI, that's because it costs more.

Or that could be a fancy way of saying voltage, like Vol la tah gshah

Yeah, what's with the 10K RPM fan? Is that a 60mm Delta fan that is also used in food processors?

[PowerPie5000]

Re: Asrock Z97 Memory Voltage Issue...

The CPU "VCore" of 1.112V seems right IMO given what I've seen with my board and CPU. The BIOS reading tends to be lower than in Windows. But is that at 4.4GHz?

CPU Cache Voltage seems high, is that an Auto setting? Is that with the Cache at 4.4GHz?

I'm not sure what speed the CPU or cache is running when i'm in the bios. The CPU Cache Voltage is set to auto, but it can reach 1.29V according to OCCT (shown in my last post). I don't know where or how Asrock come up with these auto figures... Surely they're not within Intel specs? The Cache is set to x44 in the bios, but the A-Tuning utility shows it reaching only 4000MHz! I think the bios or this board has some teething issues.

What does your CPU Cache voltage show when running stock/auto? Does OCCT show it going above what it shows in the bios (under load)?

I don't have the Volatge typo in the Extreme 6 UEFI, that's because it costs more.

Maybe the team working with the cheaper boards get paid less . The UEFI should be pretty similar with our boards. The Extreme6 has a couple of extra features, but the power setup is the same (VRMs, regulators, chokes etc...).

Yeah, what's with the 10K RPM fan? Is that a 60mm Delta fan that is also used in food processors?

The board just doesn't seem to like my Raijintek case fans when it comes to the readings (they appear to work fine though), and there is actually a fan connected to case fan header number 3 but the sensor reading disappears now and then (shows as N/A in the bios). I've had this issue with different fans on other boards too. My older PC uses an Arctic Freezer 7 Pro Rev 2 and it sometimes reports the fan as something like 9475385625rpm (lol) and the same thing happens with the Gigabyte R9 280X card when the fans go below 25%.

[parsec]

Re: Asrock Z97 Memory Voltage Issue...

I'm not sure what speed the CPU or cache is running when i'm in the bios. The CPU Cache Voltage is set to auto, but it can reach 1.29V according to OCCT (shown in my last post). I don't know where or how Asrock come up with these auto figures... Surely they're not within Intel specs? The Cache is set to x44 in the bios, but the A-Tuning utility shows it reaching only 4000MHz! I think the bios or this board has some teething issues.

If you check the HWMonitor screen in the BIOS, you'll see what speed the CPU and Cache are running at.

What does your CPU Cache voltage show when running stock/auto? Does OCCT show it going above what it shows in the bios (under load)?

In the BIOS at stock speeds, the CPU and Cache voltages are 0.960V. But I'm not using Auto voltage settings. I'm using Adaptive voltage, with the main voltage for CPU and Cache set to 1.20V, but then I use negative offset values to reduce it to ~1.0V.

If you have all Auto settings for voltages, you are using Adaptive, and the CPU will use (really request via the VID) whatever it wants. That is not an ASRock setting for voltage, just the Auto behavior of a Haswell CPU. Using Auto with Adaptive is known to cause higher voltages under load on Haswell processors.

The board just doesn't seem to like my Raijintek case fans when it comes to the readings (they appear to work fine though), and there is actually a fan connected to case fan header number 3 but the sensor reading disappears now and then (shows as N/A in the bios). I've had this issue with different fans on other boards too. My older PC uses an Arctic Freezer 7 Pro Rev 2 and it sometimes reports the fan as something like 9475385625rpm (lol) and the same thing happens with the Gigabyte R9 280X card when the fans go below 25%.

That's strange but can happen.

[PowerPie5000]

Re: Asrock Z97 Memory Voltage Issue...

I think the RAM is running at maximum power/voltage no matter what is selected. I was just messing with the A-tuning utility and I set the RAM to 1.2v to see what would happen... Nothing happened! I expected the PC to crash or not startup when I rebooted it! I then increased the RAM voltage to 1.7v to see if any values change in any of the monitoring utilities and guess what... Everything stayed the same (including that 1.696v reading that Speccy shows as the memory controller and HWMonitor shows as VIN1 & VIN11).

Is it possible that Asrock have set a single voltage of 1.7v (probably supposed to be 1.65v) for the RAM and it just can't be changed? It may show the DRAM voltage in the bios or A-tuning utility, but I don't think they're actual readings from sensors, it just shows the values we have selected and doesn't actually apply it. I could be wrong of course, but I feel something is not quite right here.

[PowerPie5000]

Re: Asrock Z97 Memory Voltage Issue...

Ok, I just re-flashed the bios, cleared the CMOS and also reinstalled the A-Tuning software for good measure. I then proceeded to see what would happen if I applied the lowest DRAM voltage in the A-Tuning software (1.165v)... The PC powered off and then restarted again with default 1.5v values. Now i'm getting somewhere as the DRAM voltage was obviously applied this time .

EDIT: I just dropped the DRAM to 1.250 and then 1.200v volts again and the system carried on normally without it rebooting like it did when I selected the lowest 1.165v option. Check out the weird voltage results below, particularly most of the the minimum and current values shown in HWMonitor (it even added an extra temp at just 7C for some reason!). Speccy also shows the memory controller as 0.400v when my RAM is set to 1.250v! I can't explain this

[parsec]

Re: Asrock Z97 Memory Voltage Issue...

I can't explain those readings except for what I know about sensor readings and software that reads them:

There are no standards for sensor readings, and there are multiple sensor chips available. There are some specifically identified inputs on the chips that should be used for VCore, for example, but the other inputs are labeled as VIN1... VIN11. A mother board manufacture can connect them to whatever they want. If there were standards, why do the programs reading them use the VIN4, VIN5, etc, labels if they knew that VIN4 was say DRAM voltage?

The answer is they don't know what those readings are unless the mother board manufacture provides that information. Do we know if CPUID requested that information? I know in at least some cases, requests for that information from mobo manufactures are not answered, they won't provide it.

I've worked with the programmer of one hardware monitoring program to help them label a couple sensor readings correctly, because of the things I listed above.

All of the third party monitoring programs have at least the following variables to deal with:

Processors from AMD and Intel, that use completely different methods to store temperature and voltage data.

Within each manufacture, each generation of processors can do things differently as they evolve over time. Haswell and their internal voltage regulators is a perfect example of that.

So at any time there are multiple processor models from multiple manufactures, and each manufacture has multiple generations of processors in use in our world at the same time.

Sensor data may also be provided by the board's chipset(s). Each processor generation has multiple chipsets. Some processor generations use two chipsets, others use one.

Then we have multiple mother boards per processor and chipset generation, and the different sensor chips that each of them use. Then there is the way data is connected to the inputs on the sensor chips. Almost forgot, now some boards have two sensor chips.

Fortunately the processor manufactures supply most of the information about the basic sensor data, how it can be read and how to interpret it. But each step we take down into the mother boards and sensor chip usage, the less we know.

Interpretation of the data is another subject entirely. Do the temperature sensors on processors output data that can be displayed as is, and will be "50C" for example? No, it's much more complicated.

Which is why many of the monitoring programs just don't bother, and spit out whatever they basically can read, and label it "VIN6", for example. They also don't tell us any of the other details I listed, or warn us that the values shown is raw data, that has not been interpreted into a valid value.

So we can't look at the output of these programs and decide one thing or another is not right, or something is being done or not done.

Is it impossible that your memory cannot run at any other voltage than 1.5V? Have you ever tried that before?

[PowerPie5000]

Re: Asrock Z97 Memory Voltage Issue...

Is it impossible that your memory cannot run at any other voltage than 1.5V? Have you ever tried that before?

My RAM appeared to be running ok at 1.250V without any other adjustments (which it really shouldn't be able to do)... That's if Asrocks own A-Tune utility is to be believed. Look at the A-Tuning DRAM voltage in my last post (1.250v with DDR3 1600 RAM). I set it to 1.250v and it actually changed the values in Speccy and HWMonitor (shown in my last post). Setting the RAM to 1.7 or 1.8V doesn't change any readings at all as far as I can tell. Dropping it to 1.65V (lowest) causes the PC to instantly reboot.

This now makes me think the RAM can be changed, but there must be some kind of voltage overshoot (I set the RAM to 1.250v for example and it could actually be running at 1.450v). Could the system agent offset have anything to do with this? It defaults at 0.199v when set to auto. I really wish Asrock had implemented DRAM voltage sensors (amongst a couple of other sensors). I think i'll just leave everything set to auto and i'll see over time if my CPU degrades at all. The temps seem fine so hopefully there won't be any problems using auto settings in the long run. On auto the VID hits 1.263v at stock and the CPU cache can hit 1.29V at stock (despite being auto set to 1.20v in the bios). The CPU cache ratio is also set to x44 but it never goes over x40 for some reason.

I suppose it's almost impossible to determine the real voltages (as you said before in detail) without using a decent multimeter. I've heard that fully digital boards are better with sensor readings compared to analog or analog/digital hybrid boards (like the Asrock Z97 series), not sure how true this is though? I'll try and get a full digital board next time when Sky Lake or AMD Zen arrives :).

[parsec]

Re: Asrock Z97 Memory Voltage Issue...

I saw that Memory Controller voltage reading before, but got side tracked with my previous post/blab-fest.

So I tried lowering my DRAM voltage in A-Tuning OC Tweaker, and ran Speccy afterwards. This is what I got:



I set the DRAM to 1.200V in A-Tuning, and you can see the Speccy reading for Memory Controller did not change, still at 1.696V. Why yours changes, I don't know. Why would the Memory Controller Input voltage change with the DRAM voltage? We don't expect or see the CPU Input voltage changing with different VCore settings. They could behave differently, but mine does not change at all. Is that your BIOS being different? Speccy being buggy?

After this, I restarted into the UEFI, and set the DRAM voltage to 1.200V. I'm using the Samsung "Green, Miracle" memory that can be run at 1.35V, which I do all the time. I only have it at 1600, 9 9 9 24 1T now, but have 16GB installed, 4 x 4GB. Surprises me it runs at 1.200V, with 16GB, typing this now with that voltage, so far not any problem with the PC.

That kinda makes me wonder if that is really the true voltage. But how often have we tried lowering the DRAM voltage below its spec, I never have, we usually increase it. Any OC of the memory could fail, I've had this memory up to 2000 at 1.35V with this board and CPU.

Might try CPU-Z just for the memory tab.