*Official* NBR Desktop Overclocker's Lounge [laptop owners welcome, too]

Ah, yes, it's a different OCCT version. Although, I can't find the version released back in early 2017. I'd be looking for version 4.4.3., and I found a version of this on a software download site, but it wouldn't run, just hang at donation screen - dodgy download or older versions don't work with current Windows 10. Prime95 is still failing though at 4.7Ghz, which it didn't before - so something has changed, not just the OCCT version.

Right, so that's very similar to what I'm seeing, after both of us updated to the latest Spectre microcode BIOS's. It'll be interesting to hear if going back to your old BIOS does indeed bring back stability. I might do some further testing to increase voltage at 4.6Ghz and check my RAM stability now that you've mentioned it. Both CPU & RAM was solid before though, BIOS is looking like the culprit given what you've said.

 

@Mr. Fox
I installed my new (1401) bios last night and reverted back last night. Everything that worked before no longer worked and I started getting extra codes on the digital screen. Windows didn't run right and it's still not running totally right either. Luckily I only ran windows 7

 

MSI just released a Bios update for my board saying "update for new intel processors", 8086K I'm guessing?

I'm holding off on this one for a while. I'm running stable 5.0 ghz @ 1.4v 24/7 with zero crashes never getting over 60c, so I'm happy! Although I havent ran any benchmarks lately to rip my OS to shreds.

@Robbo99999, might be able to try a clean install just to see if that helps your instability. Doubtful that it will but you never know. I wouldn't be at all surprised if Windoze X was the cause of instability.

 

I did a clean install just a couple of weeks ago, so I don't think it needs another clean install. Yes, I'd stay away from that new BIOS for now until more people report their findings. I may try going back to the old BIOS to see if that brings back CPU stability. I just did a round of HCI Memtest stability checking, it passed that, so it's not RAM instability, it's CPU stability then. I'm not completely sold on it being due to the new BIOS, and instead it could be motherboard VRM flakiness or CPU degredation, but Mr. Fox's recent similar experiences to mine on his latest BIOS version does seem to correlate with what I'm seeing. For the meantime I'm dialing back to 4.6Ghz until I've done further testing with the old BIOS (if I do it). But I'll report back with any findings if I do any. Interested to hear about similar issues people are having with BIOS related to the very latest Spectre protection, and interested to see if you get your stability back Mr Fox on your old BIOS.

 

I'm not happy to hear you are having issues, but glad to know it's not something goofed up specific to me. I've had errors about the ME at POST, a halt message about BOOTMGR being corrupt (not a BSOD but a total halt) and numerous BSOD about required system files being missing or corrupt and 0xC0000221 IMAGE_CHECKSUM_MISMATCH. Almost all of the BSOD reference NTOSKRNL.exe on W7 and W10.

I went all the way back to BIOS v0505 release 10/11/2017 for now. Wondering if I can downgrade to ME 11.7.X.X without bricking the board if that will help. I don't care about Spectre and Meltdown protection.

ASUS BIOS recovery is also not working now. I put a 4GB USB stick formatted as FAT in the BIOS flashing port and hold down the flashback button and it reads from the stick and never flashes. ASUS support gave me the file name to use. The light just turns blue and never changes.

I was running Legacy mode with my drives as MBR. I just restored W7 and W10 images with GPT and UEFI mode, so we will see if something in the code is messing with things worse with MBR/Legacy than GPT/UEFI.

I also noticed that the BIOS was trying to boot from my 4TB HDD formatted as GPT (due to size) and unplugging that drive--which was not even a bootable drive--stopped some of the BSOD issues, but not all of them.

Last one before I gave up trying to fix it last night was 0x0000001e KMODE_EXCEPTION_NOT_HANDLED ntoskrnl.exe 7/17/2018 10:09:35 PM.

Clean install will not help. Already confirmed that. In fact, I was having issues getting Windows Setup to even run without a BSOD with the new BIOS. The OS repair feature of Windows Setup was also throwing a BSOD. It's a firmware issue, which I why I pinged @Prema. No verdict yet if going back to UEFI/GPT disks is a fix, but I can at least boot with all 4 RAM sticks at XMP 4000 again. If the new CPU micro-code no longer supports Legacy BIOS and MBR drives, I am not going to be a happy camper about that. I do not like GPT/UEFI at all and only use GPT for disks that are too larger to work for MBR.

I agree. I violated my own rule of "if it ain't broke, don't fix it" LOL. I should have known better, with everyone seeming to be drinking the nasty firmware cancer Kool-Aid now. But, I was thinking maybe an 8-core upgrade, if it turns out to be Z370 compatible as the rumor mill suggests.

 

CPU performance is close to the same as before, but memory read and copy speeds and latency appear to have degraded.

 

So everything has been reverted to your previous except that the Intel ME won't downgrade back to the old version (you're stuck with the version associated with that latest new BIOS you tried)? And you think that might be why you have worse RAM performance? How much RAM performance percentage wise have you lost on say latency & read speeds?

I agree that not having your PC running right is a pain in the ass, I'm not happy about having to back down to 4.6Ghz, but I have to suck it up or fix it, I'm too OCD on these things!

 

Here is what the memory used to run like. It's not off a whole lot, but it is off a little bit.


Lowered RAM timings tighter back down to my standard of CL16-18-18-38 and it's still off just a tad. Maybe even within a normal margin of error. But, at least it is booting and running all 4 sticks for now. Hope it stays that way.

 

You have too remember also that when you run the bleeding edge of the CPU there is inevitably electronic degradation of the circuits. This is why you should find the limit of the overclock and drop a step or two. Even at that there may be degradation still occurring but at a much reduced rate.

 

Yes, that is true. But, I think what you might have missed is there is a clear line of demarcation that directly corresponds with a firmware "update" that caused a sudden and immediate change in system behavior, and not a good one. For three people. But, for now at least the machine is able to POST and boot Windows with the original release BIOS. Last night it could not even do that, LOL. In this case it appears the minor residual performance degradation is artificially induced by firmware cancer.

Got firmware cancer?


Edit: I am not running on the bleeding edge. I'm only using 5.2GHz @ 1.375V for the daily driver clocks. The edge of stability for this CPU is 5.5GHz @ ~1.520V. I only do that with chilled water cooling. Without chilled water, it's 5.4GHz with about 1.475V.

I'm thinking that is for the new unnamed "i9" socket 1151 octa-core CPU that is going to be officially announced in about 2 weeks.

 

Yeah, it could be, I haven't ruled that out yet...I won't know until I revert to the previous BIOS version & retest - if it's unstable then I can probably conclude that there has either been degradation of my CPU or of the motherboard VRM's (as they have shown to be thermally flaky for me in the past). I'd be surprised if it's CPU degradation though as it was 'only' running 1.4V at generally temperatures below 60 degC, occasionally in the low 60's.

 

In my case I am now questioning whether the OSes were actually harmed and if the problems I see are actually something with the BIOS corruption not being able to process things correctly any more. Windows 10 just went berserk on me after running perfectly. After powering off the system and waiting about ten seconds it booted Windows 10 fine like nothing ever happened. But, I try to reboot to load Windows 7 and it won't POST and halts with this...


I just got an RMA from ASUS for the mobo. Now the BIOS UI has lost all of the colored background and is only black without all of the pretty red stuff. I flashed back the new BIOS and it seemed OK with UEFI and GPT at first. After two or three reboots it will no longer POST and all of the problems I had noted in the prior posts have returned. As Dr. McCoy used to say, " It's dead Jim!"

Got to love that firmware cancer everyone is working so on diligently, LOL. I bought a brand new ASROCK Taichi Z370 mobo to hold me over until ASUS figures out what to do in terms of repair or replacement of the Maximus X Hero. We will see if @D2 Ultima has been recommending a good product or not. (Seems few serious overclockers at HWBOT are using ASROCK mobos, so we shall see.) When I no longer need it, somebody will get a great deal on a barely used Taichi mobo.

Here is what I mean about the black background in the BIOS.

 

I'm waiting for you to somehow decide you like the ASRock better and we laugh about it in 2.76 years when intel is on 10th gen and still 14nm++++++++ v2.5 rev. 71

Sent from my OnePlus 1 using a coconut

 

Obviously, I have an open mind to it or I would have chosen something cheaper for a temporary mobo. Whether I like it better will depend on mostly one thing, and you already know what that is. If the CPU and memory overclock better, I will like it better. If they do not, I will think it is a piece of trash, LOL. As I have mentioned before, I don't care about brand loyalty. Results is the only thing that matters in business or pleasure. If I don't like the results, then I'm done with the product or person and scratch them off of my list of the nice things that I like. They can get back on my favored list again if they try hard enough and don't screw up again. (AMD is still trying. I always forgive, but I seldom forget.)

 

Same as I feel. I may have something interesting to tell you in a week or too as well!

Sent from my OnePlus 1 using a coconut

 

ASUS will also have to handle my RMA in a timely manner, with resolution in a manner that I approve of or I will become a hater of ASUS and share the basis for that contempt often (as I do with my contempt for Alienware's broken garbage). So far, they are handling it very nicely. I am impressed thus far, but the fat lady has yet to sing.

Looking forward to finding out if you will have something interesting to share.

 

They like people who have those top tier boards. The lower end I'm not sure about xD

Hopefully it all goes well for you!

Sent from my OnePlus 1 using a coconut

 

Still having random trouble booting W7 and W10 and the (A7) ME FW - Downgrade - Request MeSpiLock Failed error during post, but BIOS v1101 (which has always been the best) seems to have performance back on track again... when it actually boots Windows successfully.

 

4000MHz 16-18-18-38

Sign me up

 

So you now overtook my Skylake score:

http://hwbot.org/submission/3418250_ajc9988_aida64___memory_read_ddr4_sdram_55851_mbs/
Still got you on latency though.

Edit: And, yeah, I would sometimes get a 1-3 training boot loop or the 55 or I forgot what the other code was. If that machine wasn't being used for another purpose, I'd pull it out and try to overtake your score again (learned a bit more with Ryzen being ram sensitive, but TR was good training).

 

Looking forward to hearing how the overclock/performance is like with your new temporary mobo vs your old one - might be interesting to see how much difference a mobo can make (or not as the case may be!).

 

Those who bin processors here home use the same MB. I look forward to some results

 

I agree. The Maximus X Hero (WI-FI AC) is very expensive for a non-HEDT platform. It will also be interesting to see if my results with it are a reflection on the fact that a comparatively small percentage of enthusiasts on HWBOT use ASROCK motherboards.

 

I'm expecting RAM overclocking/performance to be the biggest difference, as apparently the 'lesser' RAM timings when set to Auto are tuned differently from manufacturer to manufacturer - I think only the main timings are stipulated on the RAM sticks & then it's down to the mobo to decide on the rest. That's one reason why mobo manufacturers often release BIOS updates stating "RAM compatibility" for instance. I'd imagine CPU overclock to be not that different, assuming the power delivery system is good, and they're all quality boards that are being compared rather than low end budget boards.

 

Correct there, and Asus is awful at over tightening some timings, but does offer multiple modes with extremely tight, then tight. I had to use the tight (mode 2) on my Maximus VIII Extreme to get 4000 set, and it took a specific BIOS revision, and the moon and stars aligning at the solstice, and blah blah blah.

Those timings though are not just the subtimings we can play with to a degree, and I def miss some of the ram timings available on the Asus board now that I'm using Asrock, but you are correct that those can make a difference.

 

Bottom line is, I could have happily lived without this experience and had no desire to understand how another motherboard fares in comparison. But, when someone tosses a lemon in our path we make lemonade, right? I wish the new firmware was not screwed up--an unfortunate issue that plagues the industry as a whole--and wish I had no reason to waste any of my money on a temporary substitute motherboard. As they say, " make a wish in one hand, poop in the other, and see which hand gets filled up first."

I don't think I could have asked for better results than what I was getting with the Maximus X Hero, as they are phenomenal. It will be nice if the Taichi can match the CPU and RAM overclocking results, and it will be good for others to know (at my expense, LOL) since it is less expensive. As long as I get back a fully functional Maximus X Hero in a reasonable time frame I will be satisfied. That (ASUS warranty service) will be as interesting to me as whether or not the Taichi can match its performance. In a perfect world I would have no reason to find out either thing. As I was exploring options, there were not many good ones. End user reviews/results on EVGA Classified K Z370 (an option I considered) and Taichi Z370 mobos were mixed and hard to make any intelligence of. Impossible to know if that relates more to product quality control or user error and inexperience.

 

When I went from Z170 to X399, I tried figuring which boards to get. Asus has AMAZING hardware. They are screwing up on firmware. Have been for awhile and at OCN, the owners of the Zenith Extreme describe their board as perpetual beta due to the firmware. Giga has literally fried 2000 series and 1000 series Ryzens and is giga, amiright? MSI, I'm leaving it there. So, I went with Asrock as they had a decent hardware PCB and did well with the Z170 OC Formula, so I figured they were back. Except for a couple BIOS/UEFI updates, generally I've enjoyed the board, even if the BIOS is BASIC AF! That part annoys the crap out of me and they left the advanced CBS settings in hex rather than normal inputs. INFURIATING. But, aside from those things, it has performed solid for me. Would I buy Asrock again? With the info at release for X399, no doubt. But, if there was, say, an EVGA Dark X399, no contest, I'd be with the EVGA (if like the x299 Dark).

But, with what is coming, I will say I want to see the VRM setups on future TR2 boards because they are about to get beefy as hell. Also wanting to see some monoblocks shaped like a hammer to cool these chips and the upcoming 3746 or whatever pin count from Intel (although those boards and chips will be so expensive, I wouldn't recommend them to anyone but enthusiast OCers, not even for pro workstations).

Also, I don't know if Asrock did any better, but they did recently update their Z370 firmware, so if having a problem with Asus, you could either turn into the skid to see if Asrock did any better, or avoid it like the plague! Dealer's choice.

 

Been researching a little about CPU degradation due to my slight fears I might be seeing it in mine. Post #4 in the following thread is a really good account & extrapolation of some guys actual degradation tests of 3 different 7700K CPUs:
https://www.overclock.net/forum/5-intel-cpus/1673657-cpu-overvolt-death-degradation-stories.html
The crux of that post is that you will definitely lose a really tight overclock at 1.45V after close to a year (10,000hrs) because the actual extrapolation (graph) is showing a 5mv degredation at 1.45V after only 1000hrs and he was using a x10 uncertainty factor to allow himself to use the word 'definitely' - so likely to lose 5mv of stability after only 1000 hours of use. His same extrapolation graph shows 5mv of degradation after 10,000 hrs when at 1.4V. I run my 6700K at 1.4V and for just over a year, and I've seen instability recently - this makes me think I should go into BIOS and see if I've only lost in the region of 5mv of stability, ie does adding 5mv regain stability. If I test that and 5mv does regain stability then this would match his extrapolation - thereby increasing the likelihood that my recent CPU instability has been caused by degradation & less by the new BIOS - plus it's an easier test for me to perform rather than going back to the old BIOS. I'll test it now to see how much voltage I need to gain stability - well maybe not right now as it's getting late, maybe tomorrow.

The other crux of his post was that you really shouldn't run your CPU at 1.5V and certainly not at 1.55V under any circumstances.

EDIT: bloody hell, what do you know, I went ahead & did that testing, and adding 10mv in the BIOS which equates to 8mv actual increase has given me back 20min OCCT stability, which is the ultimate harshest test I've been using to dial in an overclock. Here's the screenshot:

This equates quite accurately with that guys CPU degradation test extrapolations. On the basis of this I'm long term dialing down my overclock from 4.7Ghz to 4.6Ghz with voltage decreasing from 1.4V to 1.35V during that move. Also, based on that guys extrapolation graph, by decreasing 50mv from 1.4V to 1.35V I'm reducing any further degradation rates by a factor of 10, which should make further CPU degradation a thing of the past essentially. I've done this because I want my CPU to survive another 3+ yrs without having to dial down the overclock again as the years go by. I have a feeling that staying at 4.7Ghz and 1.4V would have meant increasing voltage more & more year on year, and 1.4V was my previous max voltage limit I allowed myself so it would have been a no win situation with ever decreasing losses. I'd urge people who are using 1.4V for their CPU (Skylake onwards) to dial down to a max of 1.35V if you want to keep your CPU for 3+ years.

 

Well, maybe. I think it depends on perspective. The nice thing about running a K-series gamer-boy CPU instead of a HEDT processor is they are comparatively inexpensive, good samples overclock nicely, and they are almost always readily available at a local retail store like Fry's, Micro-Center or even Best Buy. If you're not in a hurry, vendors such B&H Photo, NewEgg and eBay (or forums like here and OC.net marketplace) can be even cheaper.

I consider the K series CPUs to be somewhat of a consumable product. If you are replacing a last generation CPU (like 6700K and 7700K - and soon 8700K) with another one just like it they are obsolete, but plentiful and relatively inexpensive. If overclocking is your thing and you really enjoy that, it amounts to about $1 per day over the course of a year if you wear it out in only 12 months. If you don't wear it out that fast even less. So, I don't worry about it a whole lot. If I were running an expensive TR or Intel HEDT processor, I would have to think harder about CPU degradation due to the financial implications that go along with it. For the same reason I am a lot more careful with my 1080 Ti than my CPU. Thankfully, now than the miner fad is fading, availability is not as problematic and costs are normalizing. But, even at normal retail prices 1080 Ti is very expensive and warrants some caution unless you have more money than you know what to do with.

 

Yes, I get that, I understand where you're coming from.


I've just now edited my previous post with that overclock testing I was talking about - to see if my potential degree of CPU degradation married with the extrapolations of that guy that did the CPU degradation testing I was talking about in my previous post - I found out it does.

 

People forget that they can sell the K-series chips they are not happy with and buy a new one (same model or upgrade if it's possible). I sold my +3 years old 6700K right before my summer vacation and got half of the price it costed new.

 

There are multiple factors that feed into transistor degredation, with voltage and heat being the primary ones. I ran my chip at 1.56 for 5.1GHz, but short runs and well cooled. Meanwhile, 4.8 going strong at 1.43 with the ring at 4.5 I believe it was. But, it is practically under no stress ever now, which I need the high frequency on single core for its primary use case. But, since never hitting full load and not really stressing even the single core, not nearly as concerned with these findings, to be honest. Might be if still using it for my workstation, though.

Edit: Nearly forgot to mention the problems of voltage spikes on Intel's side and the hard spike and dip when coming under load or finishing and the odd random spike on Kaby processors which caused Intel to say not to OC. Gotta be aware of those and the voltage changes related to LLC. It is VERY important in the degradation by voltage if trying to do too high a voltage for a daily driver.

All the more reason to buy a binned chip. Got one from SL, it takes 1.35V to do 4.1GHz on all cores. I usually run 4.05GHz on 1.2875V. Truly a beast. And the voltage is damn near linear with 1.1825V for 3.95, 1.2275 or so for 4GHz. It is keeping the heat down under major loads that is the issue if you get a binned chip, rather than the voltage and degradation from voltage. Hell, I can even keep the wattage at those voltages mostly under 250W, plus with my VRM on water, they stay around 40C on load, so I'm getting nice, smooth power delivery to boot. So, if you go for HEDT, get either the top bin or second to top and you won't be disappointed.

My next project when I clean my loop is to remove all my rads, create plastic brackets to screw them together (local library has a 3D printer at a low cost), then to attach them to a box fan and use the 5000 BTU window unit to blow directly through them with a 20" fan in pull. I'll also be reseating the CPU block, making sure the jet plate is right, etc., because my temps seem higher than others at the same frequency with them using more volts (but, my scores are equivalent with someone running 50-150MHz higher than me on all cores, so, I'm still doing something right, including the above average ram speed (mostly lucky IMC, I'm sure)). Before we got the window unit, I had to go to 3.95GHz for my daily driver to control the summer heat in this room.

 

Looking forward to seeing photos of that.

Have seen these? Might work nice with a box fan. Or, even two box fans in push/pull.
Phobya 1080: http://www.performance-pcs.com/phobya-xtreme-nova-1080-9-x-120mm-radiator.html

Phobya 1260: https://www.aquatuning.us/water-coo...ya-xtreme-supernova-1260-radiator-full-copper

 

Hadn't seen that, but I already have 3x480s (GTX Black Ice Nemesis on all three of the 480s). If I had something like that frame for 480s, I'd grab it in a heartbeat as a nice external frame. Later on I plan on getting a number of quick releases for the loop and rads so that I can just take apart any segment of the loop whenever I see fit. I think I need six of them for my loop: 2 so that I have one on each side of the rads, 2 for the graphics card, then two on the CPU block. My VRM are separate blocks in between the CPU and GPU currently, so the one on the CPU and GPU would allow me to pull that section on its own. I have 2 MCP50X pumps in the loop, so no worries on them restricting flow.

 

Hopefully the product adoption numbers are totally meaningless in terms of product performance. ASUS motherboard adoption by enthusiasts absolutely dwarfs ASROCK. http://hwbot.org/hardware/motherboards#key=z370. There are 360% more submissions with the Maximus X Hero (WI-FI AC) than the Taichi. But, on a positive note, the Taichi has more submissions that any other ASROCK Z370 motherboard. It will be interesting to find out if there are 360% more sheeple or 360% more astute overclocking enthusiasts choosing the Maximus X Hero (WI-FI AC). If you include the Maximus X Hero (not WI-FI AC version) there are 782% more submissions than the Taichi... 4075 versus 521. Nevertheless, popularity isn't a reliable measurement of product quality all of the time. Especially not in this silly Facebook generation we live in. That pathetic mentality also explains the bewildering acceptability of castrated BGA turdbooks to the retarded masses. It will be here tomorrow, so we will find out soon enough.

 

https://www.anandtech.com/show/12706/asrock-z370-taichi-motherboard-review (July 20, 2018 1:15 PM EST)

 

It has lots of pros and cons. I will make a list for @D2 Ultima when I am done testing. Not bad for the price and a truly decent value considering the low price tag, but I will be eager to get my old mobo back as long as its problems are fixed.

Had this been my first Z370 mobo and I had never seen the difference I would be satisfied with it. I can definitely see why it is less expensive. Having access to both, the Maximus X is clearly the superior product.

Their software is a real joke, but I would never buy a mobo based on that anyway. Possibly related to that in some way, the BIOS layout really sucks. Very clunky feeling and maybe done that way to create the impression of being more comprehensive than it actually is. BIOS also has a few missing features, and a really nice feature that that is pretty sweet, but it was poorly executed in terms of aesthetics.

I can see why avid overclocking enthusiasts are not all over this one. It is not really "geared" for that niche in spite of the tacky gear theme.

Memory write and copy speeds are way slower than the Maximus X for some strange reason, even with the same settings. Trial and error tinkering (not much fun) may correct that. Latency is a little better, but not enough to really matter. Memory read speed is the same. Will not boot four sticks using the DDR4-4000 XMP profile unless I give it a LOT more voltage. Tweaking it to 16-18-18-38 at 4000 requires even more memory voltage than the XMP profile with four sticks installed.

 

I think some as eg here home bin processors on this Motherboard due the low price for it (This means probably sold many of this MB due the price). And a lot people is more geared against cpu overclocking vs. overclocking of ram. It's what it is. If you can get a decent cpu overclock with it, I think many are more than satisfied.

I see this MB cost cheapest around $305.62 U.S Dollar here home. All is about the prices. In U.S you get parts like f.eks MB a lot cheaper. A computer full of the parts (finish build - notebooks) is normally ok priced here home. But the damn spare parts prices is the ****y. But the 5 years free warranty here home on such parts has it's cost.

 

How do the voltages related to the DMI, SA, and VCCIO look? If you can, or care to, if this is on your open air bench, could you grab a multimeter and compare the voltages to what is read in HWinfo64, and the same with the Asus when you get it back. The reason is that the software readings are not always correct, as seen here:


Now, reading from the back of the CPU is the most accurate, rather than any other probe points possible, which gave different analyses between Buildzoid and Chew (older OCer that does a lot with AMD CPUs) on the Taichi x370 board analyses. So, wanted to give that info as well.

There is no reason to think there isn't variance on Intel's side as well, and should be looked at as well as where to set the LLC with that information. Just some things to think about as controls. I know I sometimes gave the ram more voltage when I wasn't feeding enough to the SA or VCCIO and I had to balance where I wanted to run those for daily driving with the ram voltage. Just a thought.

As I mentioned, the Asrock BIOS is basic AF, but the BIOS often works. Asus, the component quality is excellent, and I love the options in the BIOS, the only issue is whether it works or not as of lot, which pissed me off enough to look for a competitor. If being honest, had they not bundled the useless 10Gbps NIC with the Zenith Extreme, I likely would have paid the extra $100 over my Asrock board. But $200 premium and $100 of that being a card I cannot use because my home network is all based on 1Gbps, hell no!

I also agree that the layout is awful for getting to some of the settings you want, and the overclock menu has some of what you need, just way too simple, is missing other things on that screen hidden in the normal menus, but not always where you would think, etc. That is why you should check out Buildzoid's rant about Asrock BIOSes from a couple years back.

 

I will check out those videos. I'm sure I will agree with him. When I first got into the BIOS, I'm like... WWWWUUUUUHHHT?

I cannot find an actual LLC adjustment other than the option of choosing a preset level that works the opposite of ASUS (1 being most aggressive at eliminating vdroop versus level 7 in ASUS). That's one of the features that is missing.

I saved everything to a profile that was working great when I went to bed at 5AM. Came back at noon after getting a little sleep and it took 3 tries to get the board to POST using that profile (4000 16-18-18-38 @ 1.450V).

It also keep resetting my cache ratio for some stupid reason. I have it set to 47 and it keeps flipping itself to 43 and 44. XTU does not read settings correct when launched, but applies settings correctly. But, XTU does not actually change the BIOS settings. It only seems to work temporarily, until reboot and reapplied within Windows, similar to ThrottleStop.

PCH on this board runs about 20°C hotter (65-70°C versus 40-50°C on the Maximus X Hero). I am using the same voltage (which may not be accurate as you mentioned). But, it could be the chintzy PCH heat sink they have on it.

Here is the nice latency, proper read and absurdly poor write and copy speed I mentioned.

 

Yeah, those two (IMC and core) were the only two that we got on the X399. I use Level 2 on core and level 3 on SoC (IMC and graphics, except no graphics is there for my chip).

Cold boot errors in training can happen a bit, but from the sound, it is doing it worse than Asus by a good amount.

I'm betting part are the different subtimings. But, the SA and VCCIO are the two voltages controlling the IMC on Intel chips, so playing with those could help stabilize some mem overclocks (what voltages are needed vary). DMI can have a lesser effect for obvious reasons, but some reported needing to boost that voltage slightly to stabilize (even though that wasn't my experience, still something to consider).

For resetting cache frequency, that is unfortunate and inexcusable (along with the BIOS trying to revert to the last safe settings, which then means you have to re-enter all this ****, which was part of buildzoid's rant and one of my peves at this board.

I hate XTU on principle, but definitely hear ya as the overclock software by Asus in Windows is crap. But, with AMDs side of the fence, all ryzen are done like that, not setting in the BIOS, except for maybe Asus's implementation (but RyzenMaster is just like that for software rather than in BIOS, unless having to reboot to apply certain things).

The PCH is hotter, which is a shame, but not to the point of worrying of destroying it. But, as you mentioned, the heatsink used.

For memory, that can be a number of things, including the timings you have needing changed for the subtimings they use. What is your tFAW look like on it? I know we didn't get access to the tREFI or tREFIx9 on the Asrock X399 board, but don't know on the Intel side. I've also seen positive changes playing with the termination block and CAD_BUS block, but needed guidance to mess with those (which I saw a couple mentions of Stilt's settings for those on the platform at OCN, but a mem guru came up with a good timing calculator for Ryzen that lets you punch in the type of memory being used (like B-die, or Hynix AFR, Micron A-die, etc.), first or second gen Ryzen (IMC changes), two profile versions for each frequency selected, plus a custom profile punching in the exact values from you XMP (or import from Thaiphoon Burner), and calculates a safe, fast, and extreme timings list. It doesn't always work, but gets the ryzen timings close, and then you play with it to find stable. Using it has helped give a better feel for the relationship between timings, but still there is a lot more to learn. I don't think this utility would work for Intel IMCs, but could give a bit of insight into the relationship between timings to help. A couple timing spreadsheet calculators exist out there for Intel, but no program like this that I have seen. Also, Gamers Nexus has started work on a memory series of videos to explain how ram work and how the timings interact with that function, which will eventually lead to an in depth discussion on overclocking ram. They only have one video out so far, which I don't think will be much help ATM, but something to keep your eye on.

Here is the Ryzen DRAM calc link if you just want to look at it to see the relations between subtimings (although this converts it from Intel optimized values to Ryzen optimized values). Still should be worth a look. I've found that working on getting the tFAW down can help performance a lot on ram (saw some on Intel as well), but the four activation window is a window, so you are looking for the optimal fit. If too tight or too loose, you can lose performance. Obviously you know about looking for the RFC holes.
https://www.overclock.net/forum/13-...lator-1-1-0-beta-2-overclocking-dram-am4.html


Here are the ram guys posts on CLDO_VDDP voltage on MEMCLK "holes:
https://www.overclock.net/forum/26457559-post292.html
Effect of temp and voltage on mem stability
https://www.overclock.net/forum/26472770-post445.html
And effect of CPU freq. on stability of ram:
https://www.overclock.net/forum/26902089-post1334.html

I did briefly look up mem calcs for intel, but found more excel calculators. But, something I didn't give nearly as much time trying to find because I didn't OC the ram for the skylake build (3200MHz now and trying to leave that way to avoid bitflips, etc. for its use).

 

Thanks for the technical info. I will dig deeper as time permits. Unless something goes wrong with the TMA I probably won't burn many calories on tweaking the Taichi.

I think part of the issue with this is there is no option to disable ring down-clock. I also cannot find an option to disable AVX offset, and that sucks. Not having those options is a problem if you like overclocked benching.

As Bill showed in the video, VTT voltage is screwed up on Intel side same as AMD. It will not boot with 4 sticks at 4000 using Auto for VTT and changing DRAM voltage does not change VTT. I have to set VTT manually or it will only boot to safe defaults.

My impression is this is a good budget-friendly board for gamer-boys that do not overclock, only on tinker with modest overclocking or simply run with factory overclock profiles.

 

For X399, the Asrock board automatically takes VTT off of auto and does one set of channels (a&b) to exactly half of the dram voltage, while setting the other set of channels (c&d) to 20mV higher. Might be just what they did for the quad channel boards, but going through, since you don't see auto anymore, it can draw your attention to the fact some voltage is set you didn't set yourself.

I think one OCer back when I was using the 4133 kits with the skylake mentioned going slightly higher than half (closer to 0.8) on his VTT to help with stability, but I cannot remember which OCer it was (it was a well regarded one) or the other conditions (like if it was going for high volts instead, so the half wouldn't apply, with CL12 timings), so, that might be something to try to find who did it. That was for B-die samsungs specifically, though, and I still have not looked up the general ram dies used on the kit that you have to dig deeper.

 

Yes, I go one step above half the RAM voltage. If that does not boot or has errors, I go up one more step, so that is consistent with what you are saying.

 

Disappointing RAM performance then, disappointing resetting of cache clock, and disappointing permanent AVX offset - not a good advertisement, ha! Surely there must be an option re disabling the AVX offset though - not being able to disable that could hamper your max overclock you can use on AVX I would have thought, because it would mean you'd have to run your CPU at boot with 200Mhz of extra frequency to counteract the AVX offset (which would mean it might not boot to start with)?

 

I read the Z370 review that Papusan posted; they turned off AVX offset and ran OCCT, so it must be possible. I don't know about the dropped ring clock, though.

But I don't have one so I couldn't tell you where those things are. Just that it does exist.

 

The only way I can set the cache ratio is in Windows using XTU or the A-Tuning tool. In the BIOS it shows the target settings, but they are not applied as shown.

I think the sloppy BIOS UI has wrong descriptions and I believe setting the AVX offset to "Auto" disables it. It should say "Disabled" instead of "Auto" LOL.

And, yeah... I have tweaked the ever-loving snot out of the RAM timings and the write and copy speeds absolutely suck. The read speeds are excellent.

All of this suggests ASROCK had idiots that do not know what they are doing for BIOS developers.

It's not a horrible product. This is a really nice gamer-boy motherboard, but it's really not worth a damn if you are serious about overclocking. So, while it might be great for the kiddos that think running their 8700K at 5.0GHz using the easy OC preset in the BIOS is something really special it's probably totally awesome. But, for me it sucks. I understand why hardly anyone at HWBOT runs ASROCK boards now. The Maximus X Hero is expensive in comparison, but worth every penny. Assuming I get my ASUS board back in perfect working order or a replacement, this one is going on eBay.

The only features it has that I really like is three M.2 slots and tons of USB 2.0 and 3.0 headers. It has two more SATA ports than my ASUS board, but that's irrelevant because the third M.2 slot disables two of the 8 SATA ports and there would not be enough SATA ports if all three M.2 slots are filled.

Brother @D2 Ultima probably isn't going to want to hear this, but it is what it is. It's not even close to competing on the same level as the Maximus X Hero.

 

You must know some other D2 Ultima, I like getting things actually confirmed. I understand where you're coming from and I always knew that the ASUS had many more options in their BIOSes.

But then when are you getting that Maximus Apex?

 

Oh I know what I wanted to ask too - is CPU OC Frequency & Voltage stability the same as your Asus board? (As in do you need less/more/same voltage for any given CPU OC frequency in comparison to your Asus board?) Oh yeah, and is the VRM cooling ok on that board?

 

According to the review Papusan posted, 5.1GHz+ was fine and VRM cooling was sufficient, but it needed more voltage by a little bit than the Apex (which I understand, the Apex's counterpart is the OC Formula, not this, and the Apex isn't even in the class of the Hero either).