ASRock Z270 Extreme4 Kaby Lake ATX Motherboard Review

Test Configuration, Results, And Final Analysis

We’re using our latest benchmark hardware and software suite to compare ASRock’s Z270 Extreme4 to MSI’s identically priced Z270 SLI Plus, Gigabyte’s higher market Aorus Z270X-Gaming 7, and ASRock’s previously tested Fatal1ty Z270 Gaming K6.

Loading...

Test System Configuration

CPU

Intel Core i7-7700K (Kaby Lake): 4.2-4.5 GHz, 8 MB L3 Cache, LGA 1151

Sound

Integrated HD Audio

Network

Integrated Gigabit Networking

Graphics Driver

GeForce 372.90

Synthetic Benchmarks

ASRock’s default base clock jumps around between 99.9 MHz and 100.85 MHz according to CPU-Z, and Sandra Arithmetic reflects that slight non-reference BCLK advantage. Sandra also shows slightly less memory bandwidth for the two ASRock boards, and that’s usually caused by a motherboard defaulting to slightly slower secondary and/or tertiary memory timings.

3D Games

Minor differences in synthetic benchmarks aren’t reflected in games, where ASRock’s Z270 Extreme4 produces overall results on par with the Z270 SLI Plus and Z270X-Gaming 7.

Timed Applications

One of ASRock’s competitors stumbles slightly in Handbrake and makes a slight comeback in After Effects, again leaving Z270 Extreme4 performance at par with the field.

Power, Heat And Efficiency

ASRock’s own software revealed a +0.20V offset in CPU Core Voltage, which explains the relatively poor power numbers generated by both the Z270 Extreme4 and Fatal1ty Z270 Gaming K6. MSI’s competing sample followed Intel’s voltage rules, while the Aorus Z270X-Gaming 7 used fixed voltage to generate even worse numbers in power and heat.

Poor efficiency numbers in the competing Z270X-Gaming 7 actually help ASRock’s Z270 Extreme4 look a little better. The normally-functioning Z270 SLI Plus takes the efficiency gold.

Overclocking

ASRock told us that it wasn’t able to replicate the behavior of our previously-tested Z270 Gaming K6 in its lab, but the firm still produced a string of firmware updates to improve its results. The Z270 Extreme4 showed near-perfect voltage regulation after installing the latest update, yet the CPU still crashed at CPU settings above 46 x 101 MHz. It also had the worst DRAM overclock of all four motherboards, and the only O/C it won was the one that matters least, BCLK.

A few years ago, motherboard companies would use super slow secondary and tertiary timings to win DRAM overclocking comparisons. We’ve gone back and forth on verifying the worth of those so-called wins with various high-data-rate bandwidth tests. Today’s Z270 Extreme4’s overclocked bandwidth results appear to be a combination of its lower overclocking capability and its lower default bandwidth, without any particularly bad settings that might have been used to boost its O/C capability.

Better bandwidth when using four DIMMs reflects the CPU’s preference for four memory banks, when using single-sided modules.

The Z270 Extreme4’s performance-per-dollar appears identical to that of the Z270 SLI Plus, but it will be tough for overclockers to call these equal values. Show system builders might prefer the Z270 Extreme4’s integrated RGB LEDs, where the competitor’s LEDs are white. The Z270 Extreme4 also includes an extra two-port SATA controller, but the competitor’s design doesn’t disable as many SATA ports for the sake of M.2. The Z270 Extreme4 does have a couple other advantages, such as the concealed M.2 Key-E slot for adding your own notebook-type WiFi card, but doesn’t include antenna wires for its two antenna jacks.  Thus, while the Z270 Extreme4 won’t climb up to its competitor’s higher level “Recommended” award, alternative features make it a lower-cost option for buyers who wanted, but can’t afford, its already-approved Fatal1ty sibling.

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

This thread is closed for comments
40 comments
    Your comment
  • Dark Lord of Tech
    Nice review , thanks Crash!
  • Why not use Windows 7 for a test as all benchmarks show better performance than on Windows 10 especially in synthetic gaming benches?
  • shrapnel_indie
    2164959 said:
    Why not use Windows 7 for a test as all benchmarks show better performance than on Windows 10 especially in synthetic gaming benches?



    You mean an OS, along with Win 8.x, that Microsoft wishes to be gone already despite still in the support cycle, and has less developer data-mining added-features than previous versions? To adopt Win 7 (or 8.x for that matter) would kill any DX12 testing that is important enough to some people, as well since "we know what you want better than you do" Microsoft decided DX12 for anything less than their newest OS would never happen.


    Besides, IF all tests are still on Win 10 instead, everything ran on Win 10 will have the same handicap and suffer the same way... keeping the results even, despite the lower numbers.
  • No DPC latency bench?
  • Onus
    For a mild overclocker (like myself), who may just want other "Z" features, this board may deserve a spot on the short list, although MSI's Z270 SLI Plus has much better efficiency, which will likely put it ahead.
  • Silent_Scone
    Thomas, your paragraph on this alleged DIMM offset voltage. What makes you think this is what is happening?
  • AsRock Extreme 4 series whether is x99, z97 or z270 is really awesome product.
  • 330834 said:
    2164959 said:
    Why not use Windows 7 for a test as all benchmarks show better performance than on Windows 10 especially in synthetic gaming benches?
    You mean an OS, along with Win 8.x, that Microsoft wishes to be gone already despite still in the support cycle, and has less developer data-mining added-features than previous versions? To adopt Win 7 (or 8.x for that matter) would kill any DX12 testing that is important enough to some people, as well since "we know what you want better than you do" Microsoft decided DX12 for anything less than their newest OS would never happen. Besides, IF all tests are still on Win 10 instead, everything ran on Win 10 will have the same handicap and suffer the same way... keeping the results even, despite the lower numbers.


    DX12 testing is not important at all since brings 0 performance improvements. It is another DX10 fiasco. Microsoft will have to do some serious rework with DX12 especially with MGPU support.
  • Pixdawg
    @Silent Scone--Did it occur to you that Thomas has voltage measuring tools ready to hand?
  • Crashman
    1688159 said:
    Thomas, your paragraph on this alleged DIMM offset voltage. What makes you think this is what is happening?
    My voltmeter. It shows all voltage settings consistent for most motherboards, except for DIMM voltage, which is only "as set" on boards that are NOT designed for overclocking. Very few motherboards actually DO show the offset reported within the "System Status" reading of firmware and/or monitoring software: Most companies choose to conceal it.
  • thebritscott
    "Gone are the second network controller"
    PROS
    Dual Gigabit Ethernet with teaming

    Is it me, or am I missing something?
    I didn't see a second Network Connector on the photo
  • Crashman
    125456 said:
    "Gone are the second network controller" PROS Dual Gigabit Ethernet with teaming Is it me, or am I missing something? I didn't see a second Network Connector on the photo
    You're right, I'll try to get that fixed.
  • thebritscott
    Nice to see the writer answering.
    Not sure how I get a negative vote for showing a mistake though... what kind of person would vote that down... :)

    Anyway, I enjoyed the article.

    8708 said:
    125456 said:
    "Gone are the second network controller" PROS Dual Gigabit Ethernet with teaming Is it me, or am I missing something? I didn't see a second Network Connector on the photo
    You're right, I'll try to get that fixed.
  • Karadjgne
    Extreme4 sounds a little misleading according to this article, not sure its gonna be high on my recommend list, not when put up against the similar priced offering from MSI
  • Silent_Scone
    8708 said:
    1688159 said:
    Thomas, your paragraph on this alleged DIMM offset voltage. What makes you think this is what is happening?
    My voltmeter. It shows all voltage settings consistent for most motherboards, except for DIMM voltage, which is only "as set" on boards that are NOT designed for overclocking. Very few motherboards actually DO show the offset reported within the "System Status" reading of firmware and/or monitoring software: Most companies choose to conceal it.


    Hi Thomas,
    You would need to measure the VDIMM at the socket with an oscilloscope, in order to check what the peak sag is under load (the voltage you see in monitoring depends on where the power line is tapped). A volt meter won't be able to show fast load transient changes. Thought that to be pertinent to mention as you are insinuating someone is to blame here, best to get all your facts straight first.
  • Crashman
    1688159 said:
    8708 said:
    1688159 said:
    Thomas, your paragraph on this alleged DIMM offset voltage. What makes you think this is what is happening?
    My voltmeter. It shows all voltage settings consistent for most motherboards, except for DIMM voltage, which is only "as set" on boards that are NOT designed for overclocking. Very few motherboards actually DO show the offset reported within the "System Status" reading of firmware and/or monitoring software: Most companies choose to conceal it.
    Hi Thomas, You would need to measure the VDIMM at the socket with an oscilloscope, in order to check what the peak sag is under load (the voltage you see in monitoring depends on where the power line is tapped). A volt meter won't be able to show fast load transient changes. Thought that to be pertinent to mention as you are insinuating someone is to blame here, best to get all your facts straight first.


    I'm measuring DIMM voltage, not return voltage from the DIMM. And I have evidence aplenty beginning with non-overclocking boards reading 1.20 to 1.22V default, a minority of motherboards reporting the same voltage as I'm reading, software reporting different voltage than hardware on some boards with one or the other being what's on my meter, voltage detection points on some boards reporting the same thing that I'm reading from the DIMM, and a CPU (my former Core i7-6600K) that doesn't work as well past ~1.365V DIMM...not working as well when the voltmeter says 1.37V or more. If I were just jumping into this now you might have had me second-guessing myself, but since this has been an on-again-off-again issue since the days of DDR2 I'm actually seeing things following historic trends. BTW, back in the DDR2 days, you'd find hundreds of bad Newegg reviews about various motherboards "not even working with my OCZ Gold DDR2-800". The consistent thing between all those complaints WAS the DRAM, and motherboard manufacturers responded by...bumping up the baseline DRAM voltage in the next firmware update.

    Oh, and here's the kicker: Back then MSI was the one who told me what they did (more voltage, 5mV at the time) and why (OCZ Gold not booting). Maybe boards are reporting return voltage? Maybe times have changed? My voltmeter hasn't, it's still reporting expected values on nearly everything but DRAM.
  • Silent_Scone
    8708 said:
    1688159 said:
    8708 said:
    1688159 said:
    Thomas, your paragraph on this alleged DIMM offset voltage. What makes you think this is what is happening?
    My voltmeter. It shows all voltage settings consistent for most motherboards, except for DIMM voltage, which is only "as set" on boards that are NOT designed for overclocking. Very few motherboards actually DO show the offset reported within the "System Status" reading of firmware and/or monitoring software: Most companies choose to conceal it.
    Hi Thomas, You would need to measure the VDIMM at the socket with an oscilloscope, in order to check what the peak sag is under load (the voltage you see in monitoring depends on where the power line is tapped). A volt meter won't be able to show fast load transient changes. Thought that to be pertinent to mention as you are insinuating someone is to blame here, best to get all your facts straight first.
    I'm measuring DIMM voltage, not return voltage from the DIMM. And I have evidence aplenty beginning with non-overclocking boards reading 1.20 to 1.22V default, a minority of motherboards reporting the same voltage as I'm reading, software reporting different voltage than hardware on some boards with one or the other being what's on my meter, voltage detection points on some boards reporting the same thing that I'm reading from the DIMM, and a CPU (my former Core i7-6600K) that doesn't work as well past ~1.365V DIMM...not working as well when the voltmeter says 1.37V or more. If I were just jumping into this now you might have had me second-guessing myself, but since this has been an on-again-off-again issue since the days of DDR2 I'm actually seeing things following historic trends. BTW, back in the DDR2 days, you'd find hundreds of bad Newegg reviews about various motherboards "not even working with my OCZ Gold DDR2-800". The consistent thing between all those complaints WAS the DRAM, and motherboard manufacturers responded by...bumping up the baseline DRAM voltage in the next firmware update. Oh, and here's the kicker: Back then MSI was the one who told me what they did (more voltage, 5mV at the time) and why (OCZ Gold not booting). Maybe boards are reporting return voltage? Maybe times have changed? My voltmeter hasn't, it's still reporting expected values on nearly everything but DRAM.


    What is the return voltage from the DIMM? It's not a power supply lol.

    You would need to measure at the socket (if you take into account losses along the power plane). This is about as accurate as you can get to showing how much voltage the memory is receiving. Like I've mentioned, you'd need a scope to measure this properly.

    The applied voltage needs to be adequate enough that it makes allowances for the peak negative undershoot. On top of this, the measurement points are tapped at different points of the power plane depending on which model you are using, so the comparison is not valid.

    I'm a little disappointed that you'd add a comment on blame basis for something like this at the end of a paragraph. Without the proper testing methodology, knowledge and data to show it.

    It seems rather petty for a long standing site. Perhaps we'll see some results in future.
  • Crashman
    1688159 said:
    8708 said:
    1688159 said:
    Hi Thomas, You would need to measure the VDIMM at the socket with an oscilloscope, in order to check what the peak sag is under load (the voltage you see in monitoring depends on where the power line is tapped). A volt meter won't be able to show fast load transient changes. Thought that to be pertinent to mention as you are insinuating someone is to blame here, best to get all your facts straight first.
    I'm measuring DIMM voltage, not return voltage from the DIMM. And I have evidence aplenty beginning with non-overclocking boards reading 1.20 to 1.22V default, a minority of motherboards reporting the same voltage as I'm reading, software reporting different voltage than hardware on some boards with one or the other being what's on my meter, voltage detection points on some boards reporting the same thing that I'm reading from the DIMM, and a CPU (my former Core i7-6600K) that doesn't work as well past ~1.365V DIMM...not working as well when the voltmeter says 1.37V or more. If I were just jumping into this now you might have had me second-guessing myself, but since this has been an on-again-off-again issue since the days of DDR2 I'm actually seeing things following historic trends. BTW, back in the DDR2 days, you'd find hundreds of bad Newegg reviews about various motherboards "not even working with my OCZ Gold DDR2-800". The consistent thing between all those complaints WAS the DRAM, and motherboard manufacturers responded by...bumping up the baseline DRAM voltage in the next firmware update. Oh, and here's the kicker: Back then MSI was the one who told me what they did (more voltage, 5mV at the time) and why (OCZ Gold not booting). Maybe boards are reporting return voltage? Maybe times have changed? My voltmeter hasn't, it's still reporting expected values on nearly everything but DRAM.
    What is the return voltage from the DIMM? It's not a power supply lol. You would need to measure at the socket (if you take into account losses along the power plane). This is about as accurate as you can get to showing how much voltage the memory is receiving. Like I've mentioned, you'd need a scope to measure this properly. The applied voltage needs to be adequate enough that it makes allowances for the peak negative undershoot. On top of this, the measurement points are tapped at different points of the power plane depending on which model you are using, so the comparison is not valid. I'm a little disappointed that you'd add a comment on blame basis for something like this at the end of a paragraph. Without the proper testing methodology, knowledge and data to show it. It seems rather petty for a long standing site. Perhaps we'll see some results in future.
    Er, blame basis? Historical context. The time when one could simply list the problem and everyone knew the name of the product has long since passed, the product is no longer produced, and that company is no longer in the memory business.

    As long as I'm getting the same reading from the slot as I'm getting from manufacturer-specified hard-wired voltage check points of high-end boards, either I'm doing it (slot measurement) well enough or the manufacturers themselves are doing it (voltage check point wiring) wrong.
  • Silent_Scone
    8708 said:
    1688159 said:
    8708 said:
    1688159 said:
    Hi Thomas, You would need to measure the VDIMM at the socket with an oscilloscope, in order to check what the peak sag is under load (the voltage you see in monitoring depends on where the power line is tapped). A volt meter won't be able to show fast load transient changes. Thought that to be pertinent to mention as you are insinuating someone is to blame here, best to get all your facts straight first.
    I'm measuring DIMM voltage, not return voltage from the DIMM. And I have evidence aplenty beginning with non-overclocking boards reading 1.20 to 1.22V default, a minority of motherboards reporting the same voltage as I'm reading, software reporting different voltage than hardware on some boards with one or the other being what's on my meter, voltage detection points on some boards reporting the same thing that I'm reading from the DIMM, and a CPU (my former Core i7-6600K) that doesn't work as well past ~1.365V DIMM...not working as well when the voltmeter says 1.37V or more. If I were just jumping into this now you might have had me second-guessing myself, but since this has been an on-again-off-again issue since the days of DDR2 I'm actually seeing things following historic trends. BTW, back in the DDR2 days, you'd find hundreds of bad Newegg reviews about various motherboards "not even working with my OCZ Gold DDR2-800". The consistent thing between all those complaints WAS the DRAM, and motherboard manufacturers responded by...bumping up the baseline DRAM voltage in the next firmware update. Oh, and here's the kicker: Back then MSI was the one who told me what they did (more voltage, 5mV at the time) and why (OCZ Gold not booting). Maybe boards are reporting return voltage? Maybe times have changed? My voltmeter hasn't, it's still reporting expected values on nearly everything but DRAM.
    What is the return voltage from the DIMM? It's not a power supply lol. You would need to measure at the socket (if you take into account losses along the power plane). This is about as accurate as you can get to showing how much voltage the memory is receiving. Like I've mentioned, you'd need a scope to measure this properly. The applied voltage needs to be adequate enough that it makes allowances for the peak negative undershoot. On top of this, the measurement points are tapped at different points of the power plane depending on which model you are using, so the comparison is not valid. I'm a little disappointed that you'd add a comment on blame basis for something like this at the end of a paragraph. Without the proper testing methodology, knowledge and data to show it. It seems rather petty for a long standing site. Perhaps we'll see some results in future.
    Er, blame basis? Historical context. The time when one could simply list the problem and everyone knew the name of the product has long since passed, the product is no longer produced, and that company is no longer in the memory business. As long as I'm getting the same reading from the slot as I'm getting from manufacturer-specified hard-wired voltage check points of high-end boards, either I'm doing it well enough or the manufacturers themselves are doing it (voltage check points) wrong.


    Well, it was yourself who posted regarding attempting to blame a single manufacturer.

    So now you are measuring from the socket? Sorry, did you measure from the socket or not? As a moment ago you claimed you didn't measure from the socket because you thought you'd be "measuring the return power from the DIMM". Which in itself shows a serious lack of knowledge (and at this point tact also).

    Only trying to offer you some advice here, you still need an oscilloscope to measure this correctly.
  • Crashman
    1688159 said:
    8708 said:
    1688159 said:
    8708 said:
    I'm measuring DIMM voltage, not return voltage from the DIMM. And I have evidence aplenty beginning with non-overclocking boards reading 1.20 to 1.22V default, a minority of motherboards reporting the same voltage as I'm reading, software reporting different voltage than hardware on some boards with one or the other being what's on my meter, voltage detection points on some boards reporting the same thing that I'm reading from the DIMM, and a CPU (my former Core i7-6600K) that doesn't work as well past ~1.365V DIMM...not working as well when the voltmeter says 1.37V or more. If I were just jumping into this now you might have had me second-guessing myself, but since this has been an on-again-off-again issue since the days of DDR2 I'm actually seeing things following historic trends. BTW, back in the DDR2 days, you'd find hundreds of bad Newegg reviews about various motherboards "not even working with my OCZ Gold DDR2-800". The consistent thing between all those complaints WAS the DRAM, and motherboard manufacturers responded by...bumping up the baseline DRAM voltage in the next firmware update. Oh, and here's the kicker: Back then MSI was the one who told me what they did (more voltage, 5mV at the time) and why (OCZ Gold not booting). Maybe boards are reporting return voltage? Maybe times have changed? My voltmeter hasn't, it's still reporting expected values on nearly everything but DRAM.
    What is the return voltage from the DIMM? It's not a power supply lol. You would need to measure at the socket (if you take into account losses along the power plane). This is about as accurate as you can get to showing how much voltage the memory is receiving. Like I've mentioned, you'd need a scope to measure this properly. The applied voltage needs to be adequate enough that it makes allowances for the peak negative undershoot. On top of this, the measurement points are tapped at different points of the power plane depending on which model you are using, so the comparison is not valid. I'm a little disappointed that you'd add a comment on blame basis for something like this at the end of a paragraph. Without the proper testing methodology, knowledge and data to show it. It seems rather petty for a long standing site. Perhaps we'll see some results in future.
    Er, blame basis? Historical context. The time when one could simply list the problem and everyone knew the name of the product has long since passed, the product is no longer produced, and that company is no longer in the memory business. As long as I'm getting the same reading from the slot as I'm getting from manufacturer-specified hard-wired voltage check points of high-end boards, either I'm doing it well enough or the manufacturers themselves are doing it (voltage check points) wrong.
    Well, it was yourself who posted regarding attempting to blame a single manufacturer. So now you are measuring from the socket? Sorry, did you measure from the socket or not? As a moment ago you claimed you didn't measure from the socket because you thought you'd be "measuring the return power from the DIMM". Which in itself shows a serious lack of knowledge (and at this point tact also). Only trying to offer you some advice here, you still need an oscilloscope to measure this correctly.
    Oh boy, just read what I stated OK? I'm getting the same reading from the DIMM slot as I'm getting from the voltage check points provided on high-end boards by the manufacturer, so if I'm taking the reading from an invalid point so are the board manufacturers. And the history of why companies have added voltage goes back to instability most notorious to a single product that is no longer worthy of discussion in anything more than a historical perspective. To imply that this would be an attempt to defame a brand is to miss the point.

    I don't know how you can justify labeling as DIMM voltage anything that's not reaching the DIMM slot, but I'm sure you'll find supporters.
  • Silent_Scone
    8708 said:
    1688159 said:
    8708 said:
    1688159 said:
    8708 said:
    I'm measuring DIMM voltage, not return voltage from the DIMM. And I have evidence aplenty beginning with non-overclocking boards reading 1.20 to 1.22V default, a minority of motherboards reporting the same voltage as I'm reading, software reporting different voltage than hardware on some boards with one or the other being what's on my meter, voltage detection points on some boards reporting the same thing that I'm reading from the DIMM, and a CPU (my former Core i7-6600K) that doesn't work as well past ~1.365V DIMM...not working as well when the voltmeter says 1.37V or more. If I were just jumping into this now you might have had me second-guessing myself, but since this has been an on-again-off-again issue since the days of DDR2 I'm actually seeing things following historic trends. BTW, back in the DDR2 days, you'd find hundreds of bad Newegg reviews about various motherboards "not even working with my OCZ Gold DDR2-800". The consistent thing between all those complaints WAS the DRAM, and motherboard manufacturers responded by...bumping up the baseline DRAM voltage in the next firmware update. Oh, and here's the kicker: Back then MSI was the one who told me what they did (more voltage, 5mV at the time) and why (OCZ Gold not booting). Maybe boards are reporting return voltage? Maybe times have changed? My voltmeter hasn't, it's still reporting expected values on nearly everything but DRAM.
    What is the return voltage from the DIMM? It's not a power supply lol. You would need to measure at the socket (if you take into account losses along the power plane). This is about as accurate as you can get to showing how much voltage the memory is receiving. Like I've mentioned, you'd need a scope to measure this properly. The applied voltage needs to be adequate enough that it makes allowances for the peak negative undershoot. On top of this, the measurement points are tapped at different points of the power plane depending on which model you are using, so the comparison is not valid. I'm a little disappointed that you'd add a comment on blame basis for something like this at the end of a paragraph. Without the proper testing methodology, knowledge and data to show it. It seems rather petty for a long standing site. Perhaps we'll see some results in future.
    Er, blame basis? Historical context. The time when one could simply list the problem and everyone knew the name of the product has long since passed, the product is no longer produced, and that company is no longer in the memory business. As long as I'm getting the same reading from the slot as I'm getting from manufacturer-specified hard-wired voltage check points of high-end boards, either I'm doing it well enough or the manufacturers themselves are doing it (voltage check points) wrong.
    Well, it was yourself who posted regarding attempting to blame a single manufacturer. So now you are measuring from the socket? Sorry, did you measure from the socket or not? As a moment ago you claimed you didn't measure from the socket because you thought you'd be "measuring the return power from the DIMM". Which in itself shows a serious lack of knowledge (and at this point tact also). Only trying to offer you some advice here, you still need an oscilloscope to measure this correctly.
    Oh boy, just read what I stated OK? I'm getting the same reading from the DIMM slot as I'm getting from the voltage check points provided on high-end boards by the manufacturer, so if I'm taking the reading from an invalid point so are the board manufacturers. And the history of why companies have added voltage goes back to instability most notorious to a single product that is no longer worthy of discussion in anything more than a historical perspective. To imply that this would be an attempt to defame a brand is to miss the point. I don't know how you can justify labeling as DIMM voltage anything that's not reaching the DIMM slot, but I'm sure you'll find supporters.



    Thomas,

    Apologies if I am rattling your cage here, but your reply "not measuring the return voltage from the DIMM" isn't something you can downplay at this point. Professionally, it shows you are out of your depth and not qualified to publicly call out practices or in this case over-voltage at this level. It's best to leave those things to people who understand these things better, and stick to things that you are better qualified to proclaim.

    The real time voltage and the VID in BIOS are not one and the same. This is because any power delivery circuit is susceptible to transients. As I've already mentioned a few times (which you've actively ignored), this is why an oscilloscope is needed to confirm whether the DIMM is really being subjected to overvoltage.

    You keep rolling back to someone must be doing something wrong, but in this case it is you who is doing something wrong. A multi-meter is not the right tool here. At 5-20mv levels, this will likely fall within bounds of the undershoot of two phase power delivery systems.

    If you really understood this, we wouldn't be having this back and forth. A multi-meter or any onboard voltage monitoring is not capable of measuring this, which is something you're failing to acknowledge.
  • anuproy
    Weill you provide in Bangladesh?
  • Crashman
    1688159 said:
    8708 said:
    1688159 said:
    8708 said:
    1688159 said:
    What is the return voltage from the DIMM? It's not a power supply lol. You would need to measure at the socket (if you take into account losses along the power plane). This is about as accurate as you can get to showing how much voltage the memory is receiving. Like I've mentioned, you'd need a scope to measure this properly. The applied voltage needs to be adequate enough that it makes allowances for the peak negative undershoot. On top of this, the measurement points are tapped at different points of the power plane depending on which model you are using, so the comparison is not valid. I'm a little disappointed that you'd add a comment on blame basis for something like this at the end of a paragraph. Without the proper testing methodology, knowledge and data to show it. It seems rather petty for a long standing site. Perhaps we'll see some results in future.
    Er, blame basis? Historical context. The time when one could simply list the problem and everyone knew the name of the product has long since passed, the product is no longer produced, and that company is no longer in the memory business. As long as I'm getting the same reading from the slot as I'm getting from manufacturer-specified hard-wired voltage check points of high-end boards, either I'm doing it well enough or the manufacturers themselves are doing it (voltage check points) wrong.
    Well, it was yourself who posted regarding attempting to blame a single manufacturer. So now you are measuring from the socket? Sorry, did you measure from the socket or not? As a moment ago you claimed you didn't measure from the socket because you thought you'd be "measuring the return power from the DIMM". Which in itself shows a serious lack of knowledge (and at this point tact also). Only trying to offer you some advice here, you still need an oscilloscope to measure this correctly.
    Oh boy, just read what I stated OK? I'm getting the same reading from the DIMM slot as I'm getting from the voltage check points provided on high-end boards by the manufacturer, so if I'm taking the reading from an invalid point so are the board manufacturers. And the history of why companies have added voltage goes back to instability most notorious to a single product that is no longer worthy of discussion in anything more than a historical perspective. To imply that this would be an attempt to defame a brand is to miss the point. I don't know how you can justify labeling as DIMM voltage anything that's not reaching the DIMM slot, but I'm sure you'll find supporters.
    Thomas, Apologies if I am rattling your cage here, but your reply "not measuring the return voltage from the DIMM" isn't something you can downplay at this point. Professionally, it shows you are out of your depth and not qualified to publicly call out practices or in this case over-voltage at this level. It's best to leave those things to people who understand these things better, and stick to things that you are better qualified to proclaim. The real time voltage and the VID in BIOS are not one and the same. This is because any power delivery circuit is susceptible to transients. As I've already mentioned a few times (which you've actively ignored), this is why an oscilloscope is needed to confirm whether the DIMM is really being subjected to overvoltage. You keep rolling back to someone must be doing something wrong, but in this case it is you who is doing something wrong. A multi-meter is not the right tool here. At 5-20mv levels, this will likely fall within bounds of the undershoot of two phase power delivery systems. If you really understood this, we wouldn't be having this back and forth. A multi-meter or any onboard voltage monitoring is not capable of measuring this, which is something you're failing to acknowledge.
    Ah crap, let me bring you down to my level, there's still a supply voltage that's supposed to be constant right? And what I'm saying, is that my meter measures correctly the supply voltage on regular office PC boards. And what I'm saying is, the same pins measure about 20mV higher on *most* boards that are designed to run overclocking RAM. And, what I'm saying, is that this is the supply voltage we're setting in BIOS, it's the voltage the motherboard itself is supposed to be measuring in the status report, and it's the voltage motherboard manufacturers are tapping with their voltage detection points. It's also the voltage that appears on several of the pins on the module, and, the measurement at that pin is the same as the measurement at the manufacturer's volt check point.

    Now, I'm saying that if a manufacturer says "Check the voltage here" and you say "That voltage doesn't mean anything", you're saying that the manufacturer is wrong.

    So, say I have some RAM that overclocks to DDR4-3502 at 1.375V, DDR4-3468 at 1.370V, and DDR4-3434 at 1.365V, and that this has been confirmed on several boards using a volt meter.Consistency.Even if the reading is wrong, it's wrong in exactly the same way each time.

    So I grab those three high end boards At default settings the firmware of all three boards reports that they're automatically configured to 1.35V, yet my meter shows that one board has 1.375V supply voltage, one has 1.370V, and one has 1.365V. Low and behold, one board supports stable overclocks to DDR4-3502, another to DDR4-3468, and the third to DDR4-3434. Yet when I manually set all of these to the voltage that corresponds to a 1.35V meter measurement at the DIMM, all three overclock that same memory to an identical DDR4-3400.

    You say "don't do that, you don't have an oscilloscope attached to the CPU socket". And I say "if it's DIMM voltage, it should be attached at the DIMM". When I say a voltage is High, I'm not talking about Signal High, I'm talking about the reference point to which signal high and signal low would be based.

    From the CPU side, I have a Core i7-6600K that I know will start to lose DRAM overclocking capability somewhere between 1.365V and 1.370V. So I put it on those same boards, and the one that measures 1.375V with my meter only reaches DDR4-3366. I drop the voltage manually 10mV, and the overclocking capability increases to DDR4-3434. The one that measures 1.370? I drop my CPU in there and it only reaches DDR4-3400. I manually lower the voltage by 5mV, and the overclocking capability goes up to DDR4-3434.The third board, which was previously measured as producing 1.365V at the DIMM, reaches DDR4-3434 with the lower-quality CPU, without manual recalibration. That's the same DDR4-3434 that I got from my other CPU at 1.365V. **This less-voltage-tolerant CPU appears to be a perfect secondary device for confirming DDR4 voltage too high, without even having the need for a meter.

    The fact that all three reach the same overclock when manually set to the same supply voltage should be indicator enough that all three SHOULD be set to the same supply voltage. Apples-to-apples, right? I mean, there are some boards with crappy trace routs that get signal degradation at super-high data rates, that's why we keep getting faster RAM to look for those limits.

    So tell me again, why am I not supposed to read the one voltage rail that the manufacturer tells me is VDIMM, and that goes to several pins on the DIMM itself?

    An oscilliscope would show me a wave form. It would let me see the peaks. But I'm not talking about peaks, I'm talking about reference points. Centers. Which should be damn close to averages. Which is what a volt meter shows.
  • Silent_Scone
    8708 said:
    1688159 said:
    8708 said:
    1688159 said:
    8708 said:
    1688159 said:
    What is the return voltage from the DIMM? It's not a power supply lol. You would need to measure at the socket (if you take into account losses along the power plane). This is about as accurate as you can get to showing how much voltage the memory is receiving. Like I've mentioned, you'd need a scope to measure this properly. The applied voltage needs to be adequate enough that it makes allowances for the peak negative undershoot. On top of this, the measurement points are tapped at different points of the power plane depending on which model you are using, so the comparison is not valid. I'm a little disappointed that you'd add a comment on blame basis for something like this at the end of a paragraph. Without the proper testing methodology, knowledge and data to show it. It seems rather petty for a long standing site. Perhaps we'll see some results in future.
    Er, blame basis? Historical context. The time when one could simply list the problem and everyone knew the name of the product has long since passed, the product is no longer produced, and that company is no longer in the memory business. As long as I'm getting the same reading from the slot as I'm getting from manufacturer-specified hard-wired voltage check points of high-end boards, either I'm doing it well enough or the manufacturers themselves are doing it (voltage check points) wrong.
    Well, it was yourself who posted regarding attempting to blame a single manufacturer. So now you are measuring from the socket? Sorry, did you measure from the socket or not? As a moment ago you claimed you didn't measure from the socket because you thought you'd be "measuring the return power from the DIMM". Which in itself shows a serious lack of knowledge (and at this point tact also). Only trying to offer you some advice here, you still need an oscilloscope to measure this correctly.
    Oh boy, just read what I stated OK? I'm getting the same reading from the DIMM slot as I'm getting from the voltage check points provided on high-end boards by the manufacturer, so if I'm taking the reading from an invalid point so are the board manufacturers. And the history of why companies have added voltage goes back to instability most notorious to a single product that is no longer worthy of discussion in anything more than a historical perspective. To imply that this would be an attempt to defame a brand is to miss the point. I don't know how you can justify labeling as DIMM voltage anything that's not reaching the DIMM slot, but I'm sure you'll find supporters.
    Thomas, Apologies if I am rattling your cage here, but your reply "not measuring the return voltage from the DIMM" isn't something you can downplay at this point. Professionally, it shows you are out of your depth and not qualified to publicly call out practices or in this case over-voltage at this level. It's best to leave those things to people who understand these things better, and stick to things that you are better qualified to proclaim. The real time voltage and the VID in BIOS are not one and the same. This is because any power delivery circuit is susceptible to transients. As I've already mentioned a few times (which you've actively ignored), this is why an oscilloscope is needed to confirm whether the DIMM is really being subjected to overvoltage. You keep rolling back to someone must be doing something wrong, but in this case it is you who is doing something wrong. A multi-meter is not the right tool here. At 5-20mv levels, this will likely fall within bounds of the undershoot of two phase power delivery systems. If you really understood this, we wouldn't be having this back and forth. A multi-meter or any onboard voltage monitoring is not capable of measuring this, which is something you're failing to acknowledge.
    Ah crap, let me bring you down to my level, there's still a supply voltage that's supposed to be constant right? And what I'm saying, is that my meter measures correctly the supply voltage on regular office PC boards. You say "don't do that, you don't have an oscilloscope attached to the CPU socket". And I say "if it's DIMM voltage, it should be attached at the DIMM". When I say a voltage is High, I'm not talking about Signal High, I'm talking about the reference point to which signal high and signal low would be based.



    1) Firstly, stop there. The voltage is never perfectly constant, so this is also incorrect. It is subject to deviations which your multi meter cannot read. This is why you are spiraling, and continually going down the wrong path...


    2) I've never once mentioned the CPU socket, this is you misinterpreting the use of the word. Anyone that understands about measurements would not confuse that in this instance.