Announcement

**parsec** · 05-26-2014, 09:38 PM

Re: Asrock Z87 Extreme4 - Adaptive w/ negative offset? Don't understand these settings BIOS v2.70

Originally posted by eatmyturf View Post

Hi all,

I've successfully OC'd on Override (fixed) Voltages for my CPU and Cache/Ring/Uncore.

Question: Given my attached (well, linked below) images from the BIOS, How can I switch now to Adaptive on this mobo and achieve the desired result of preventing over-voltage on turbo/full load? (Or can I?) I'm already happy with my override/fixed voltage at full load from stress testing, and now want to capitalize on power-saving features of my Haswell CPU (scroll down for hardware details).

Parsec provided interesting insights here on Extreme6 vs Extreme4 back in December 2013, so I'm not overly optimistic that Asrock has given me the options I need to meet my desired outcome, but it's already almost June.

...

Your VCore Adaptive Voltage and VCore additional offset settings are identical to those on my Z87 Extreme 6 board, and have not changed since they were introduced. Of course there is no reason to change them since they work fine. What I wrote in the thread you linked to above, still holds today. So IMO you have the options to meet your desired outcome.

Of course some explanation of how they work would help, right? I may see why you are skeptical about these settings, given the description for Vcore Adaptive Voltage, "Configure the voltage added to the CPU when the system is under heavy load". IMO that is a very convoluted way of saying what it really is, the base maximum CPU voltage. I say base because the Vcore additional offset further adjusts the maximum VCore that will be used when the VCore voltage mode is set to Adaptive.

Adaptive voltage mode allows the VCore to drop during low CPU load. How low is also tied to the CPU power saving options used, SpeedStep and C-States (from C1E to C7), and the Windows Power Plan, Processor power management setting, Minimum processor state specifically. You cannot specifically set the minimum VCore voltage, you get what Adaptive mode plus CPU power saving options plus Minimum processor state give you.

Keep in mind that our ASRock boards only provide the CPU core's VIDs, not the actual VCore. Fortunately, the actual VCore is always less than the maximum VID. I believe Gigabyte added a second monitoring chip on their Haswell boards just to allow a true VCore (if it is, I don't know either way, but seems to be right) but hardware monitoring programs must be custom matched to these boards in order to display the true VCore.

The problem with Adaptive Voltage is if left on Auto for the Vcore Adaptive Voltage and Vcore additional offset settings, the resulting VCore (VID) is much more than necessary, as you know. My goal was to achieve the lowest VCore (VID) I can for a given OC. That is CPU over clock 101 material of course, just saying this in case your goal is different. I'm not a hardcore CPU OC person, so my few settings are not a guide for high OCs.

Also understand that the way I adjust Vcore Adaptive Voltage and Vcore additional offset to result in a final maximum Adaptive VCore voltage is not the only way to accomplish this. More on this later.

The main point here is that the combination of Vcore Adaptive Voltage and Vcore additional offset will result in a final maximum Adaptive VCore.

I still use the same settings I described in the thread you linked to above. I arbitrarily chose 1.200V as the Vcore Adaptive Voltage, as I did not want to exceed that voltage, due to CPU cooler limitations (I use a Corsair H80) and the well known CPU temperature realities of Haswell processors. I started with stock multipliers, and later increased them.

I then set Vcore additional offset to -0.100V. That negative offset with the 1.200V Vcore Adaptive Voltage gave me a maximum VCore of 1.100V. I would then use the PC and run CPU stress tests. If no BSODs, etc, I would increase (increasing a negative value, if that makes sense to you) the Vcore additional offset by 0.025V to -0.125V, for example, and repeat the process of testing and increasing the Vcore additional offset until a BSOD occurs. I would then reduce the negative offset level until no BSOD. Standard OC testing.

I also had all CPU power saving options Enabled (not Auto), SpeedStep and all C-States including C7, and Minimum processor state set to 0% (overkill, seems no different than 5% - 10%.) At idle with core clocks at 800MHz (normal for Haswell processors) the VCore (VIDs) would be ~0.650V.

What I meant by other ways to do this, is for example in your case set Vcore Adaptive Voltage to 1.180V, and set Vcore additional offset to Auto (apparently equal to 0.000V), or set it to -0.010V, for a maximum VCore of 1.170V. You could also try a Vcore Adaptive Voltage value of say 1.100V, with a Vcore additional offset of +0.080V, and just check that the maximum VCore (VID) is 1.180V.

As I recall, choosing a low Vcore Adaptive Voltage of 1.000V or less with a small, positive Vcore additional offset (say +0.100V) would result in a boot failure.

Bottom line, a fixed Adaptive Voltage setting with a negative Offset Voltage works to both limit the maximum VCore and allows low VCore at idle and low load. If your board's UEFI does not behave as mine does, then it needs a fix, but I'm quite satisfied with the way it works on my Z87 Extreme 6. I've used multiple different hardware monitoring programs (HWiNFO64, AIDA64) to check the VCore (VID) and they all agree and the value is never above my Vcore Adaptive Voltage and Vcore additional offset combination.

**eatmyturf** · 05-27-2014, 12:56 AM

Re: Asrock Z87 Extreme4 - Adaptive w/ negative offset? Don't understand these settings BIOS v2.70

Originally posted by parsec View Post

Your VCore Adaptive Voltage and VCore additional offset settings are identical to those on my Z87 Extreme 6 board, and have not changed since they were introduced. Of course there is no reason to change them since they work fine. What I wrote in the thread you linked to above, still holds today. So IMO you have the options to meet your desired outcome.

...Of course some explanation of how they work would help, right?
...The main point here is that the combination of Vcore Adaptive Voltage and Vcore additional offset will result in a final maximum Adaptive VCore.

Of course! I missed grasping this. The word "added" in the help statement of Vcore Adaptive Voltage really threw me off. Also, because in the other thread I referenced there was no explicit mention of setting names and how they were applied, I just couldn't make the connection to the first setting being the base and the second being the offset.

I think I read in other places too about folks seeing only a minimum additional offset, which doesn't support negative offsets, and I was also equating that to my scenario and to the Vcore Adaptive Voltage since it only allowed 0.8V and up. Again, that's the base max voltage, so I misunderstood. :) Parsec, thanks much.

Other details:

I'll post more info about my OC in another reply to myself here, for posterity and possible help to others. However, I took a similar OC approach as you, using fixed voltages. I started with Vccin 1.9V, Vcore 1.20V, and Vcache 1.15V. This was based on others' results online. And I conservatively jumped only to a 42x multiplier for my first test. That alone seemed to run hotter than I expected (peak of 85 C, average around 70+) even with the efficient but not beastly 212 Evo cooler. Dialing down voltages further allowed me to drop temps to 80 C peak and avg of 64-65 C. Then I upped to 43x, tuned down voltages more, and I'm happy.

I need to further tune down my voltages actually, to see my first BSOD (haven't had one!). And I'm convinced I can do at least 44x, but I bought a silent case (as I see you did also) for a reason -- I don't want sounds like an airplane taking off. I already have 4 fans, which exposes more holes in my case for noise to escape (but also cools better), and I have taken off the ULNA / LNA voltage adapters on my Noctua's for stress testing. I prefer to put back the LNA (slight voltage restrictions) on those to dip max RPM's from 1100 to 950. It was a beautiful (non)sound with the ULNA's and stock voltages/multipliers, let me tell you. :D

Thanks again, for the long explanation and time you gave me. Btw, I'm up to speed on the power saving features, except for one question. My BIOS has all Cstates enabled, and C1E / Speedstep(EIST) are enabled. However, there's one setting which is disabled -- "Package C State Support", defined as "Enable CPU, PCIe Memory, Graphics CState Support for power saving." Any thoughts why? Should I just enable that as well - it sure sounds nice, except I don't understand Cstates feature impact on GPUs?

**parsec** · 05-27-2014, 12:22 PM

Re: Asrock Z87 Extreme4 - Adaptive w/ negative offset? Don't understand these settings BIOS v2.70

I completely agree about using the word "added" in the help text, it makes you think that is an offset voltage, I thought the same thing!

The terminology "base maximum voltage" is something I made up, it's not an official term AFAIK, but seems to make sense IMO.

Forgot to mention in my novella (glad it helped you), that the two Adaptive voltage settings for the Cache work the same way.

The Package C State Support option I don't completely understand, but I use it and have had no problems. If we consider the other things it covers, PCIe and Graphics, all parts of the CPU, it must allow those things to enter low power usage states. It can affect video cards too since the PCIe lanes for graphics cards are part of the CPU. Just try it and see how it affects the PC, if at all, it doesn't cause problems for me.

I decided to embrace Haswell's low power usage capabilities which may make the OC crowd cringe, but given what you need to do for high OCs on Haswell (delidding...), and given the GFLOPs I get at 4.2GHz (over 100), why not. Plus monitoring the CPU power usage difference between no power saving options enabled at idle vs with them enabled at 4.0GHz (~40W vs ~0.5W), what is the point of using 80 times the power at idle, particularly if you leave the PC running when you sleep, etc.

I've never had a BSOD using all the power saving options, that apparently happens to some users (on other boards...).

I'm trying the quiet case thing which can work fine, but I don't suggest it for heavy duty gamers. You can use fan speed control on the board or external fan controller for use with hot running video cards, but at some point they are not enough.

I would be interested in how high you can go on lower voltages, so post that please. IMO, Adaptive Voltage is not complete without an offset setting that allow negative voltages to keep it under control.

**eatmyturf** · 05-27-2014, 03:55 PM

Re: Asrock Z87 Extreme4 - Adaptive w/ negative offset? Don't understand these settings BIOS v2.70

I'll try the Package C State Support once I'm done tuning my OC voltages, then (to reduce number of variables I'm tinkering with). And I'm a fan of using Haswell's power mgmt capabilities -- after all, I paid for it. Plus, I'm not the only PC user (i.e. family), so I need to think beyond just computing performance, but noise/heat as well.

An aside on your mention of BSOD's w/ power saving options. I encountered that before OC'ing. In the end, a Windows re-install fixed it. But I checked all my Wake settings, sleep/hibernate settings, power plan settings, and on down the line, and nothing mattered -- but a simple re-install made things right, ha!

Regarding quiet cases, I can see your point. Case design (soundproofing, proper airflow, removable cages, etc) does help. But you eventually still face a "golden triangle" of sorts with performance vs heat vs noise. I like a balance. Though I'm thinking an H100i would have been fun, even at the extra $40+ cost, since it can do 3-7 C cooler than my $30 Evo 212. :)

Originally posted by parsec View Post

I would be interested in how high you can go on lower voltages, so post that please. IMO, Adaptive Voltage is not complete without an offset setting that allow negative voltages to keep it under control.

I'll try to see how low I can go. I'll post results-to-date later today or tomorrow, and I'll probably be better dialed in by the end of the week so I can switch to Adaptive w/ negative offset (to control the over voltage under load). My chip seems to enjoy sipping power, as others online suggest starting higher at a Vccin of 2.0V, Vcore 1.25V and Vcache 1.15V.

I hope it helps someone, I'll post up images of my BIOS and its settings. One thing I found hard was to find people who would clearly document not only their success, but their complete settings, AND how they applied those settings to their vendor-specific mobo. This would have helped me in my understanding of the Asrock UEFI implementation. Also would have helped me get off the ground faster, as popular Haswell OC how-to's were done against Gigabyte or Asus. But you helped me out with my (hopefully-)last issues.

**eatmyturf** · 05-29-2014, 12:14 AM

Re: Asrock Z87 Extreme4 - Adaptive w/ negative offset? Don't understand these settings BIOS v2.70

Originally posted by parsec View Post

I would be interested in how high you can go on lower voltages

I tuned as far down as I could go on voltages, staying at 43x cpu multiplier, and raising cache multiplier to 40x (the latter may not be at its ceiling).
Vin Override 1.87V
Vcore Override 1.15V
Vcache Override 1.115V

Any lower on those and I will BSOD. (that was fun, I got weird looks from family when I happily reached this point, lol). It was obviously not going to work without more voltage, but I also tried 44x at these voltages, and BSOD'd. Out of haste, I wanted to see what I could do with Adaptive before a 2-day business trip.

On adaptive, again with 43x/40x multipliers:
Vin Override 1.87V
Vcore Adaptive Vcore Adaptive 1.155V
Vcore Adaptive offset -0.070V
Vcache Override 1.115V
Vcache Adaptive offset Auto

I tried to do just -0.005 Vcache offset, albeit with -0.08V Vcore offset as well, and BSOD within 5 minutes of stress testing. I'm clearly not at my lowest stable point yet, but I like the results -- sipping a tiny, min Vcore of 0.695 at idle w/ 800Mhz, and max of 1.175 at load w/ 4.3GHz (which is 0.25V higher than when on fixed/override). This makes sense since people remark that adaptive under load causes +0.10 or +0.11V spikes, and I only offset -0.07V. Max temps are about 2 C higher (86-87 C) than on fixed voltage testing, and I'm thinking it's the Auto VCache offset. I hope to put this back negative with more testing and to lower the heat.

My next steps are to document my tests in detail, return from work and optimize my adaptive settings at 43x, and then I may try to see at what voltages I can stably do at 44x next week -- though I don't intend to run at 44x long-term, I like having some buffer room, protecting my CPU from high temps, and protecting unsaved work from possible BSOD.

**parsec** · 05-29-2014, 01:28 AM

Re: Asrock Z87 Extreme4 - Adaptive w/ negative offset? Don't understand these settings BIOS v2.70

That sounds very good! Getting a 4.3GHz OC at 1.175V Adaptive is very good IMO. While it may not be the ultimate OC for Haswell processors, IMO an OC at 4.3 at under 1.200V is good. Plus in my experience with Ivy Bridge processors, the actual VCore is less than the VID.

I found since the cache multi is less than the core multipliers (as you know for OC stability), you can go lower with the cache voltage, I have my negative offset at -0.100V - -0.125V, with a Adaptive Voltage of 1.200V.

I played with stock multipliers, and Adaptive voltage, and got the idle VCore/VID down to ~0.635V. I also had CPU Package Power at idle/800MHz, below one Watt, actually dipping to ~0.250W, which is unheard of before Haswell.

That's great, smiling at a BSOD and just thinking, "... well that's to low, take it up a notch". Great attitude, don't lose it!

In my experience with Haswell, the 4.3GHz to 4.4GHz transition is where things really begin to change with voltage and heat, mostly heat. I have never had CPU throttling kick in ever, until I ran AIDA64 with my i5-4670K at 4.5GHz. One core hit 100C using Adaptive on Auto, and I bailed out. Still even under control of the voltage, heat will be the limiting factor. You can use your CPU normally a 4.5GHz, but stress testing is where your cooling will fail.

**eatmyturf** · 05-30-2014, 03:13 AM

Re: Asrock Z87 Extreme4 - Adaptive w/ negative offset? Don't understand these settings BIOS v2.70

I'm going to begin detailing my testing and configurations, intended for all.

This will include a table of test configurations and temperature (and/or blue screen of death) results*, at least where I tested longer than a hour. Many OC'ers wouldn't even say an hour is enough to say one can trust temp results, and it's true that there's still a margin of error and potential for undiscovered BSOD. I haven't yet tested beyond 2 1/2 hours, btw. My final test this weekend or next will be somewhere between 4 and 8 hours -- just to be sure I don't surprise myself or family with a BSOD while working on unsaved data in day-to-day use -- homework, work, video editing, etc. But I don't have the time to run that long of a test for each iteration, it would have been a waste in my mind until I got close to the limits of voltages or multipliers.

This will also include some additional images with early (non-final) configuration values in my BIOS. And it will include some other comments that may help those new to overclocking, particularly on Asrock Z87 motherboards. So, here we go.

Images first, pictures are worth a thousand words. Here are the "OC Tweaker" page settings from my Asrock Z87 Extreme4 UEFI BIOS. I repeat the third page to show the difference between override (aka fixed) and adaptive modes:
Page 1: CPU and Cache (aka Ring/Uncore) multipliers
Page 2: RAM / XMP profile settings (manual values could be set inside the DRAM Configuration link)
Page 3a (override mode): fixed voltages for CPU and Cache, and also for CPU Input (vccin), plus LLC setting
Page 3b (adaptive mode): max desired voltages for CPU and Cache, and offsets which can be used to deduct voltage under load (to counter the CPU's requests to go beyond max levels).

Discussion of values seen on the above pages, are listed below. I will admit what I don't understand -- and you're free to respectfully tell me where I'm wrong or if you disagree with me. The assumption is that any recommendations I give are for entry-level OC'ers (like me), not pro's who are pushing the bleeding edges. I'm not going to provide an overclocking guide. There are a handful of very good guides already for Haswell. I will attempt to give added context.

CPU Ratio: Set to All Core, so it applies your ratio (multiplier) to all cores equally. You can do per core, if you are so inclined. And from limited reading and my own test results, the first core gets the heaviest work and heat, the second core will usually be one or two degree C behind, and the 3rd and 4th cores get less work. If you use the Asrock automatic OC settings (I advise against, it will overvolt and often give unstable results), you'll see it will set a multiplier for 3rd and 4th cores, then the 2nd core is +1, and 1st core is +2 -- i.e. 43x 42x 41x 41x.
CPU Ratio All Core: Start at 42x or 43x. Like Parsec said before, there is a noticable leap between 43x (4.3GHz) and 44x (4.4GHz) on 4670K and 4770K cpu's. I have seen this from others' results and comments online.
CPU Cache Ratio: Start at 35x. Don't use Auto, set a known baseline to control this. You can actually stay with 35x and be stable at least to 43x cpu multiplier. This is because unlike on Sandy/Ivy Bridge, where the ring bus ran at the same frequency as the CPU, on Haswell the ring bus has twice the bandwidth as before and therefore doesn't have to run as fast and is a decoupled setting from the CPU ratio. With that said, overclockers recommend to eventually tune this higher, within 300 or 500Mhz (3-5x) of the CPU frequency (ratio) to be assured that it isn't a bottleneck on CPU performance. Works for me. I ended at 40x here, with a 43x CPU multiplier.
BCLK/PCIE Frequency: Leave this on Auto so your base clock runs at 100Mhz. First, it's easier to multiply the CPU ratio by 100. Second, other frequencies may cause difficulties in finding stability with RAM and Video Card.
BLCK Ratio: Leave on Auto. Because I didn't play with these, I don't understand the difference of the Frequency and the Ratio setting. I would think the Ratio (aka Gear Ratio), which could be 1.00, 1.25, or 1.67, would imply the BCLK frequency.
Spread Spectrum: Set this to disabled. Otherwise it will enable fluctuation of BLCK frequency to adjust for electromagnetic interference by the CPU. And this will impact OC'ing.
CPU OC Fixed Mode: Set this to Auto (enabled). Disabling will disallow throttling and remove the power management options (next two items).
Intel Speedstep Technology: aka EIST. Leave this Enabled to allow the CPU to underclock itself to save power. It won't affect stress tests, which will constantly demand the full multiplier and power throughout your test, unless you're doing something really specific and know for a fact you need this disabled. Else you might forget to re-enable later.
Intel Turbo Boost Technology: Set this to Auto (enabled). Allows the turbo multipliers beyond your base. On stock, this allows my 4670K to go from 3.4GHz to 3.8GHz under load.
Filter PLL Frequency: Leave on Auto. I don't fully understand this, though it is suggested that raising this to High BLCK (1.6x / 1600) will provide better OC results. I haven't tried. I need to research more.
PCIE PLL Selection: Leave on Auto. And I don't understand this.
Long Duration Power Limit: And Long Duration Maintained, Short Duration Power Limit, Primary Plane Current Limit. Leave on Auto. Some OC'ers recommend to set high limits (i.e. 1000) for these, but in my mind, this is for extreme OC. It will reduce/eliminate the ability for the CPU to clock back down after prior load increases. I feel like you pay for this feature, so allow it to work to throttle down when voltage exceeds the demanded levels (i.e. you go to high load scenario, then to idle/low load).
Load XMP Setting: Setting to Auto will cause the JEDEC values to be used -- I recommend this for early CPU overclocking. Overclocking your RAM may/will affect your CPU overclock performance, so limit variables and just start with CPU overclocking. Regarding JEDEC, those are the timings that the memory stick manufacturer writes to a small area of ROM, which are generally safe for all users (typically 1333Mhz or 1600Mhz, 1.5V, and CL11). Additionally, Intel XMP v1.3 allows manufacturers to also write timing configs which are more aggressive and pre-tested for stability for most users. Most sticks that support XMP will have only one profile, but you see here my Kingston HyperX Beast 2400Mhz sticks have two. I can choose either 2133 or 2400Mhz with a simple selection, and I avoid manual overclock / timing adjustments and tests. After I got my lowest voltage, stable 43x overclock, I re-enabled XMP for 2400Mhz and continued with Adaptive mode tests.
... skipping other settings for RAM and FIVR (integrated voltage regulator, which you should leave on Auto)...
CPU Vcore Voltage Mode: Test first on Override (aka fixed) until you get the highest overclock / lowest stable voltage. You can jump to Adaptive first, but then there's a fluctuating variable of additional voltage that the CPU requests when you go into high load / turbo, pushing past your desired max level. I found it beneficial to get my max overclock on fixed voltages, then to take that as a baseline for Adaptive mode adjustments.
Vcore Override Voltage: Many overclockers suggest to start off at 1.25V, paired with 44x cpu multiplier. That's actually a lot of power. This page guided me in starting at 1.19V. Look on that page for the graph titled "Haswell Voltage @ Multiplier @ Watts". You can see the author considers a multiplier of 44x dictates a minimum 1.212V on Vcore. I started testing at 42x, so 1.19V (sub-1.2V) seemed safe. And it was (I ended up with 1.15V on 43x as lowest-stable Vcore).
Vcore voltage additional offset: I didn't play with this on Override, but in that scenario, leave it on Auto. For Adaptive Mode, we talked all about this in prior posts on this thread.
CPU Cache Voltage Mode: Like CPU Vcore, but for the cache/ring/uncore. Start with Override. The page I linked in "override voltage" above also talks about the FIVR and the role of the ring bus on Haswell.
CPU Cache Override Voltage: Set to 1.15V to start, but this will likely be high and get tuned down toward 1.10V in your testing (I ended up at 1.115V at 40x cache multiplier). It isn't really until you start increasing your cache multiplier above 35x that you should expect to deliver more than 1.10V.
CPU Cache voltage offset: Like I said above, for Vcore offset.
System Agent Voltage offset: Leave on Auto. It's not expected you'll need to tune this higher unless doing some OC'ing on RAM. I didn't have to with my RAM overclock to 2400Mhz. However, going beyond that, you may need to up from 1.15 to 1.2 or 1.25V, according to others. I wouldn't know.
CPU Analog IO and CPU Digital IO Voltage Offset: Leave on Auto. I don't understand these, because I didn't find it necessary to.
CPU Integrated VR settings: Same.
CPU Input Voltage Mode: Set to Fixed Mode.
CPU Input Fixed Voltage: Start with 1.900V. You could test as low as 1.850V, in .01 or .005 increments, and for very high overclocks (45x or higher), you may be stuck at 1.9V or need to go higher, like 2.0V. This is the Vccin / Vin, the voltage given to the FIVR, which then regulates/distributes voltage to the other components, like CPU voltage, cache voltage, system agent, etc.
CPU Load-Line Calibration: You can leave this on Auto (no correction), or set to Level 3 (50%). Setting a non-Auto value helps correct for Vdroop, which is the CPU's attempt to lower voltage while in high load / turbo. In overclocked scenarios, this drop in voltage is too steep and undermines (destabilizes) your system. The LLC setting will adjust the amount of Vdroop that actually occurs, which keeps you stable.

Any other settings, consult your manual and/or Google. :)

*I said I would provide a table of my test efforts and resutls to-date, however I've had too much fun (time) writing this, and it's past 1am my time, and I have early-morning work to worry about. I'll provide my results soon. So far, I've run 18 tests, anywhere from 30mins to 2 1/2 hours each.

**eatmyturf** · 05-31-2014, 12:42 AM

Re: Asrock Z87 Extreme4 - Adaptive w/ negative offset? Don't understand these settings BIOS v2.70

Wow, I encountered an instability by running 7-Zip against a large (7GB+), split compressed RAR (i.e. decompressing). I increased my Vcore by .005 to 1.160V, which made it last longer the next time, then a BSOD occurred again. I've now pushed my negative offset on the Vcore adaptive settings from -.070 to -.060V. This worked (for now). I might want to instead test with LLC level 2 (less Vdroop); I'm currently at 3.

If it's a voltage drop issue, then that might be advantageous over higher initial voltages and reduced negative offsets under load/turbo. I'd rather keep voltages low on idle and keep more aggressive negative offsets, while compensating more for Vdroop. If it's an issue with non-turbo/extreme range, meaning LLC isn't kicking in yet, then obviously I need the higher Vcore and more cautious offsets.

**parsec** · 05-31-2014, 02:28 AM

Re: Asrock Z87 Extreme4 - Adaptive w/ negative offset? Don't understand these settings BIOS v2.70

You are now encountering the slight down side to this method. You can tweak the negative offset to the point where 99% of the time it's fine, but then there are those rare situations where the CPU usage goes from almost nothing to 90%+ one one or more cores, with an OC, and if the VRMs and PSU can't keep up (they just can't, a processor is just to fast) you'll get a BSOD.

IMO, that is why the default/automatic Adaptive VIDs we see are so high, to handle these cases. The processor really needed more than the maximum Adaptive VCore you defined.

Why does a lower fixed voltage work in that same situation (assuming it did)? That would then be a voltage speed issue, either the voltage needed was not supplied fast enough when using Adaptive, or having a maximum voltage ceiling causes a slower response from the VRMs.

This is also the place where you can start trying other settings, like Additional Turbo Voltage, or increasing the current limits, which are usually fine on Auto but are increased to high values (like 1000 Amps I think...). There are others too, like the frequency the built in voltage regulator (FIVR) runs at. But be careful with those settings, they can actually induce BSODS for some reason. Try not to mix to many tweaks together in one UEFI session, since you won't know which setting was the culprit.

Or as you said the VDroop, which is actually for what... the CPU Input Voltage. What are you using for that now? That has a bit of tweaking potential, what were you using?

Have you noticed in the UEFI if you change certain things in OC Tweaker you'll get automatic changes in other settings? I can never recall exactly what they are, but one is the CPU Input Voltage gets set higher, to 1.9V IIRC. Fan speed settings can also be changed automatically, super annoying, if that happens it's not you going crazy.

Announcement

Asrock Z87 Extreme4 - Adaptive w/ negative offset? Don't understand these settings BIOS v2.70

Asrock Z87 Extreme4 - Adaptive w/ negative offset? Don't understand these settings BIOS v2.70

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment