Intel Core i9-9900k 8c/16t, i7-9700K 8c/8t, i7-9600k 6c/6t 2nd Gen Coffee Lake CPU's + Z390

I think the investigative journalism by Hardware Unboxed and others have been exceptional. What I don’t get is the significant disparity between reviewers using the same motherboard. Good example is OC3D. How much of the problem is due to variations in binning of the engineering samples sent to reviewers.

 

Hardware Unboxed and GN examine the hardware before reviewing it, so they know the strong and weak points - you'll notice they point this out in their reviews - and that's how they don't get caught up into the "Black Box Effect" that other reviewers, even the older more experienced ones do.

OC3D is great, but he has often been caught by similar unknown problems while reviewing, and when consulting with more technical people he gets clued in to the specific weakness behind the oddities he finds when reviewing.

That's why it's good to wait and let a wide range of experienced people get their hands on the product to see things from the wider perspective.

Even better is the trend where the youtube reviewers consult each other during the NDA silent period while putting together their reviews; where they can only talk to NDA reviewers and product companies, and with enough testing time get issues resolved before their reviews go public.

Otherwise we get this situation where some early reviews of the same product look "good", and other reviewers catch the problems while testing and come out with much different results.

Again, that's why it's a good reason to wait well past the initial release day reviews, as the product companies have figured this out - so they shorten the time in hand for new product reviews at initial release.

They don't want to give enough time for the reviewer collaboration to work out all the problems early before release - which can reduce sales - so buyers will go out half-informed and purchase with false security based on rushed reviews - the Asus Maximum Hero XI is the recent example.

The Asus Maximum Hero XI was one of the review kit pack-in boards, but boards and CPU's were delayed and the resulting reviews were shallow / rushed / light coverage, but still gave their exuberant approval even though they hadn't had enough time to really dig in for a deep review.

Note how long Hardware Unboxed motherboard in depth VRM reviews were delayed. The most important sub-system in a high performance CPU motherboards are the VRM design / implementation.

That delay from release day until today is a good measure of the minimum time to wait next time before purchasing.

From the release of the 9900k to now was about 2 weeks, so next time wait 30 days from release before purchasing to make sure you get a good view of all the products released so as to get a combined consensus rating on each product.

Who knows, maybe 30 days out from the 9900k release - and the next CPU / motherboard release from Intel / AMD - the CPU in stock levels and the price will be normalized to MSRP, as well as having in depth motherboard and CPU reviews you can depend on for purchasing decisions.

 

What performance deficit does the board exhibit? Please show me a single benchmark that shows the board is throttling or otherwise not giving full performance that most users would expect?

-- Instead you can get a great VRM with gigatrash BIOS.

 

It's maybe beyond your understanding to notice so far and right now, so keep reviewing the videos and articles discussing the power limits of the Asus Maximus Hero XI over other motherboards that deliver full power and performance.

I have mentioned it to you before, 195w isn't full power draw. Look at other reviews where the power draw is much higher, and their resulting temperatures are much higher.

The 5.0ghz, 5.1ghz, and 5.2ghz examples in various reviews I have already posted and directly pointed out via text and images - if you were paying attention to those posts you would have known this a while back.

It looks like Hardware Unboxed will have more in depth coverage and testing results, that should help.

You'll figure it out eventually. Or maybe not, it depends on if you take off your rose colored glasses. I'll leave that up to you for now, no sense trying to show something to someone that doesn't want to see it. It's already clear from what has been posted here in this thread.

The Asus forums have discussions on this subject, here's one thread and quote:

Can Someone Clear Up Misinformation on the Z390 XI Hero's Power Phasing?
https://rog.asus.com/forum/showthre...n-the-Z390-XI-Hero-s-power-phasing#post740087

"My Hero XI is sitting here as well unopened. The large temp differences observed between some of these review sites is alarming. From what I am hearing is that unless you allow the mobo to provide more than 95W the chip will throttle and affect performance. Reviewers that saw better temps used boards that had the 95W limit kept in place and resulted in much cooler 9900k chips....and lesser performance. 7% lower in cinebench for example. That in itself is not a big concern as you can adjust that wattage cap. However, the REAL concern was that you have some folks saying that removing the 95w cap was going to overwork certain boards with "lessor" power delivery set ups. The Hero XI was singled out as an example. I believe the Hero XI is a 4 x 2 +2 set up using doublers but the photo above does seem to indicate a different layout?? "

Edit - Steve from HU is sticking to his claim even 19 hours after his initial comments about the Hero XI and getting backlash:

He points to a Reddit comment that says:

"Sorry, he's bang on correct about it. Proof of phase count. Controller is an ASP1400CTB (rebranded IR35203 or 35201, an 8+0 phase controller) , there is only one voltage controller on the board, and there are "10" total phases, there are also no doublers on the board, which makes it a fake 8 phase, aka a 4 phase. mosfets are SIC639's (50 amp phases). ( https://www.reddit.com/r/intel/comme...mance/e88fj4t/)

Pretty much what Jackflynn is showing above with the pics.

Asus needs to step in here and add some clarity. I would find it hard to believe Asus would put out a board without ample power delivery in the ROG line so I suspect this is no concern. But the hubbub on the net about the Hero XI (and it's siblings) is somewhat alarming. The 9900k is a power hog for sure and I want something bullet proof."

There are more posts, and more threads all over the internet, get to reading.

 

https://www.google.com/amp/s/wccfte...g-strix-z390-e-gaming-motherboard-review/amp/

 

This is why I don't like trying to discuss things with you. I go through the trouble of explaining things, providing quotes and images charts and details, along with the source links and what to look for specifically.

Then in answer you post a link with no title and with no details on what you want me to gather from the link - what's your point? If you won't make the effort at detailed communication, why should I?

Lets just stop here, and I'll let you independently figure it out from my posts and others.

 

Edit:

Agreed. Let’s end it.

 

Well, here is 170W with just a 6 Core chip.

 

Do you have the 9900k installed in a desktop motherboard? The P870TM1 was showing a maximum of about 170w as well with the 9900k, or was it higher?

I'm sure you can match this guys efforts showing 293w power draw with one of your full power desktop motherboards:

9900K OC 5.2 With AVX2 load for 24hr on a Z390 Aorus extreme
http://forum.notebookreview.com/thr...e-lake-cpus-z390.811225/page-85#post-10813565

There are other points in the video with a bit clearer focus, that's what I grabbed from early in the video. Check out the video for more info.

195w in comparison from the Asus Maximus Hero XI vs 293w from other motherboards.

I'm sure some tuning can be done to balance power draw vs performance for the higher power delivery motherboards as well, but the Asus Maxims Hero XI doesn't seem to be able to play in that range - maybe it can be tuned to allow full power - maybe not?

Maybe that 4 phase design worked fine for the 8700k, but the 9900k is pushing past that? Maybe it's a firmware limitation / bug?

 

@hmscott

This question is all over the map. The following video provides a list of some of the recommended motherboards. But, the Asus XI hero is not one of them. The reddit articles appear divided. Is the XI Hero good for Overclocking the 9900K to 5ghz+? GN’s Overclocking live stream seemed to suggest that with proper cooking, the XI would be fine.

 

The XI is fine. Listing a random YouTube video showing 293w shows is nothing. It shows us you’re doing extreme AVX loads, higher voltage and binned chip and using your with custom water cooling**

GN, Tech Yes City, OC3D, Guru3d have all stayed it’s a great board lol regardless of using a “twin” 4 phase or 4 phase with doubles to make a virtual 8 phase.

I’ll take my 250w+ of capability and a few extra degrees on a VRM any day over those savings and a **** BIOS.

 

No such thing as a bug, but this is unknown territory in a laptop.

 

Maybe you can put the minds to rest in this thread, bro. They're freaking out about internet rumors, and you're proof that they need to chill.

Can someone clear up misinformation on the Z390 XI Hero's power phasing?

---

der8auer



ASUS MAXIMUS X HERO (Z370) Running 9900K @ 5.3GHz and DDR4-4200 should not be an issue. 8700K pulled 210W without missing a beat.

Benchmarks 9900k and Maximus X Hero Z370 https://www.reddit.com/r/intel/comme...s_x_hero_z370/

 

Ah, no I meant the Asus Maximus Hero XI lower power usage - power limits - as reported by many suggest it's either a configuration setting, BIOS limitation, or just a design limit compared to the power draw by other motherboards. The highest I've seen is 195w package power draw.

Even your P850TM1 9900k example draws 239.5w package power on a laptop motherboard.

I'm sure the Asus Maximus Hero XI runs fine, even if it potentially has the lowest phase count / power limit of the z390 motherboards.

As @Talon points out there are other reasons to get a particular motherboard besides power delivery, it's only that the 9900k can be such a monster power draw when OC'd.

If you are into stock operation or moderate OC, then I'm sure it's fine.

Looking forward to the incoming motherboard reviews rating the z390 motherboards to find the highest power delivery, and see how that OC's the same 9900k.

It might be that highest power delivery won't be the only determining factor for which board OC's the best.

So have you given up on the 9900k in the P870TM1? Or waiting for BIOS tuning? Or just checking out the rest of the new CPU's?

 

Ah, I see now. My mistake. I see where you were talking about the Hero now.

I haven't given up on testing my 9900K, I just have quite a few things going on at the exact same time. It's already a fact it runs in a P870TM1, but what needs to be adjust is how much the cpu "should" draw in terms of power. And whether or not that power is wasted watts or what.
The TM1 is already drawing 240W.

 

Right, that's what I meant about tuning for performance - even the high power draws we see with ThreadRipper and Skylake-X can be moderated depending on actual performance increases.

Same for the RX / Vega GPU's. AMD could have done a better job of scaling the power usage vs performance a bit better at first. Now people undervolt and detune the power and get even better performance for both AMD and Nvidia GPU's.

It's all about the tuning.

But, you've got to have the raw materials there to work with - limiting power limits your freedom to tune - even if you need find out on your own that the higher power doesn't really bring more performance, or that the added performance isn't worth the trade off.

Kinda like the decision between choosing an 8700k or 9900k or AMD Ryzen / ThreadRipper - tuning the solution for your budget first, then adjusting your desires for power accordingly.

 

In other words, the extra 80 to 100 watts is pretty much considered wasted watts is what i'm leaning to right now.

 

Yeah, but wouldn't that count out the power hungry 9900k in general?

I think we should wait for comparison tests to show tuning and power usage benefits before guessing away potential performance.

I've always looked for the highest power delivery when deciding on a motherboard, depending on the application, not just for performance but for headroom / efficiency reasons.

I think Hardware Unboxed mentioned efficiency ratings based on % power used vs peak power, which would be less for a lower maximum power delivery headroom design.

 

Lol. Seems I was wrong thinking it was merely 4 phase doubled... now I am schooled that the P870 vcore vrm is a "twin 8 phase" ;p

 

i9 9900K LOTTERY (100+ people lining up!!!!)
In Full Effect
Published on Nov 3, 2018
The 9900K was officially released on the 20th October here in Japan but there has been ZERO stock so far! Just 3 days ago big PC chain here, TSUKUMO, tweeted out they would get a small delivery but they would need to do a lottery since they only got a few units (~14!!). Check how many people lined up!!!!!


It was fun to see the part of the video with the line outside the store. Start @ ~01:50 - 05:00, the rest is babbling on about how happy he is.

He never mentioned the price... ... maybe in his next video, the build video?

The "Jewel of the Nile" money shot at 04:56

So, 9900k's are a paper launch... Everywhere.

 

Not sure what your talking about.

No need to wait, I have shown two prior examples. 1 at 185W with a better score than one at 239W. That is pretty much self explanatory i'm guessing. I'm all about efficiency over bragging about how many watts i can pull from a wall. or how high my multiplier is, but yet people with multipliers 2 to 4 bins lower score better numbers. That my friend is proof that more does not always = better.

And since it is already just about on par with @Talons desktop version. The TM1 is right where it needs to be.

PS: Will not be running prime95 24 hours

 

That's no where near a large enough sample pool of test comparisons to say such a thing is resolved.

What do you attribute the 55w increase over 185w power draw to then?, poor tuning?, poor utilization?, inadequate cooling?

What were the settings for each extreme for comparison?

Do you really think you've optimized performance by limiting power draw to 185w with the 9900k...?

That really doesn't sound right to me. I'll wait for some more reviews using full power desktop motherboards with a wider range of tests than just 1 comparison on a laptop motherboard.

As I recall @Mr. Fox spent weeks tuning with Prema on adding support for higher power LGA CPU's in previous releases. Has that much effort been put forth with the 9900k?

 

I am thinking people need to stop watching YouTube videos and stop listening to people that don't own one endlessly debating the unknown in forums and, if they want a 9900K, just grab one and start tweaking and tuning them like you and Brother @Talon are doing instead of freaking out and stressing over things so much. All the hypothetical stuff and the ongoing drama is getting to be a bit much. This always happens to varying degrees whenever something new is released, or even planned for release, (and it applies to hardware, software/OS, and even malware,) but the real world results never actually surface until that starts happening. Watching YouTube videos can be useful and entertaining, but it's a lot like the news. (Even the word news probably needs to be in quotation marks since we can't take most of it seriously any more.) Every channel has a different spin on things in one way or another based on one person's experience and opinions.

 

So we should ignore your videos, reviews and tuning postings because it's a waste of time too?

Results from various sources are exactly what is needed to cut through the complacency of some and inexperience of others.

The release is new with discoveries and tuning being worked through, which will take time to resolve into a consensus.

As I mentioned in my last post you spent weeks tuning previous CPU releases on the Clevo LGA firmware, and now in 1 week all's well with tuning the 9900k?

Doesn't sound right to me that you both are ready to "give up" so quickly hardly pushing performance with the 9900k.

Let's watch and see how things fall out over the next month or two.

 

I didn't say ignore them. Nobody should take everything they read or hear as being gospel or absolute truth in today's world, including things that I post, because it's almost always based on limited information, a special interest, or slanted in one direction or another.

So, reviewer A gets a dud sample of product X, so does reviewer B, but reviewers C and D fare better in the silicon lottery and everyone accepts what their favorite person says on face value. Just like the news, LOL.

I'm not sure what you mean by giving up. I don't own a 9900K and probably never will. Doesn't seem like @Talon or @Johnksss are giving up. They're just barely getting warmed up from how it looks from where I am sitting right now. I'm sure they're just scratching the surface. If not for their examples, (at least in this forum,) we would still be limited to accepting what the talking heads have to say.

 

I never said it was resolved, but making a mountain out of a mole hill is counter productive.

I know what i'm doing so I know how to manipulate the system to pretty much read what ever.

I'm not testing for optimization i'm testing for temps.

You are more than welcome to youtube your self to death on information, I'm not mad at you. I'll stick with hands on every time.

All the effort they put into the LGA, you still had something to say negative.
Sorry, but I wont be burning up my system running Prime95 for 3 trillion hours to prove it's stable.

 

Getting modded firmware fully functional and ready for public use it totally different than retail products being sold to consumers. When you buy a retail product, you still need to do tuning and tweaking to get what you want from it, but all of that back-end stuff that takes such a long time is already done. Or, at least from the perspective of the OEM/ODM it's done, otherwise it wouldn't be available for purchase.

A good example is the RTX cards that are dying. That happened in the hands of consumers, not YouTube reviewers. Same thing needs to happen with 9900K for us to find out the real scoop on it.

 

This. The only way I was able to generate over 250w of CPU package draw on my 9900K @5Ghz with Maximus XI was by running Prime95 29.4 small FFTs.

 

Did you post screen shots with power draw and temps with hwinfo / stress tool screens like @Johnksss ? Sorry if I missed it, link? Key bios settings would be nice too, in case you later tune it differently please post those too and their effect in comparison.

 

tl;dw: BIOS update(s) for EVGA z390 FTW fixes auto-voltage Vcore too low.

EVGA response: Auto-voltage for Vcore was too low for some CPU's, 2 BIOS revisions later Joker's EVGA z390 FTW is stable.

It Finally Works!! - EVGA Z390 FTW UPDATE
Joker Productions
Published on Nov 4, 2018
Update on my EVGA Z390 FTW motherboard with the i9 9900K I was seeing issues with on Frame Raider 2.0. Its finally working now thanks to a BIOS update.

 

I think extreme power draw is kind of neat because it carries a sense of awesomeness in a world overrun with pathetic low TDP garbage, but it doesn't scale much with performance past a certain point. It is also dependent on workload, as you mentioned with Prime95 testing.

I think way too many calories are being burned on that. It is similar to clock speeds. While "clock is king" still holds true to a very large degree, there is a point of diminishing return unless you start to tweak other things to coincide with the increased clock speed.

For example, my 7960X benchmark scores at 5.0-5.1GHz are the same, sometimes slightly better, than 5.2GHz. However, when I bump up the memory clock and tighten up the timings, add more memory voltage, etc., only then does 5.2GHz actually produce significantly better benchmark results.

 

I think you are saying what I was trying to convey earlier, I think trimming off the "excess power" and poo-pooing the power limit headroom issue is "giving up too soon".

Once you get the power draw, additional tuning is needed to get the full performance. But, if you didn't have the power delivery headroom to push things to the limit, you wouldn't get the opportunity to do that tuning - and leave performance on the table.

I get it that the laptop environment even in the awesome P870TM1 is going to be limited due to cooling, so you do need to leave power headroom and therefore leave some performance behind, but for a desktop there is no reason to accept the lower power delivery - find the highest stable VRM power delivery design and implementation and go from there.

Otherwise you are wasting that $600 CPU's full potential.

 

This has always been true to some extent. I still stand by my belief that doing more with more trumps doing more with less. Doing more with less is a cop-out and lame excuse that is overused by dufus corporate executives that want to work the guts out of employees without having to pay for it, and that flawed philosophy gets applied to technology as well.

I think where we have the disconnected is the assumption that 9900K needs to draw more power running things like 3DMark and Cinebench, and because it doesn't there is something wrong. I do not agree with that. The results will stand on their own and the power draw associated with those tasks is nothing but data. If the results are what you want and expect, then the data can mostly be ignored by the end user. The people that need to look at the data are those building the platform to support it so nothing burns out or catches on fire. Some people burn too many calories analyzing numbers instead of celebrating results. Anyone that works in a huge corporation knows exactly what I am talking about. Pencil necks kills the rest of us slowly, with millions of paper cuts, LOL.

 

XTU will generate serious watts as well. Try messing with the CPU 1 aux voltage if your board has that. Along with cpu input voltage. Along with the other 4 or 5 settings in the digi menu in the bios.

 

Nah, you're going off the path into the weeds now, no need to denigrate imaginary scenerios, just stick with the tune for maximum power and tune for maximum performance, doing more with more, that's the core thought I was trying to put across.

How is it the 2 guys that should be pushing for maximum power delivery for tuning maximum performance gains are backing off on this approach for the 9900k?

That's why I said it seems odd you two are giving up so quickly on getting maximum performance from the 9900k vs every other time spending weeks tuning for highest power - removing the gimp limits imposed by the BIOS - and tweaking infinitesimal performance out of the firmware.

I'm sure it's only a momentary setback given the cooling limitations of the P870TM1 with the 9900k... don't give up, there will be a new design forthcoming, maybe not right away but eventually, it's not like Intel is able to give up on the 14nm process any time soon, and so the next couple of CPU releases are going to continue to crank up the heat.

Clevo's gotta build a new "cooker" to contain and cool the power.

 

Okay, you're cracking me up. I'm already at 5.0 to 5.1 ghz in a laptop on a 9900K, exactly where are you at in that? I do not see a board, chip or anything resembling you doing anything in the way of finding out performance your self? I further find it funny because i can cancel out quite a bit of heat, but that's not the point right now. The point is letting people know where it stands, then we get to where it can go....

 

Well, "leaning" isn't "committed", so you had never really "given up"?

Glad to hear you are back to turning watts into higher benchmark numbers.

 

Well, @Johnksss and @Talon are showing us that the headroom is there and gets used when the workload demands it. And, like always, headroom can only be used when the right settings are applied and limits are raised or removed. There is a lot of pressure from the tree-huggers to care about power consumption, so thank God the settings can be tweaked to make that metric an irrelevant one. I look at power draw when (a) my results are less than expected; (b) something goes wrong; or, (c) I am curious. I find extreme power draw to be entertaining as well as insightful. But, when I am getting the results I am looking for then I don't check because I don't care how high or how low it is. It is a byproduct of workload and the most important thing is having the ability to use as much as you need and removing the limits so they do not interfere with progress and so you don't have to think about them.

A simple example of when I had to care about power consumption was my recent upgrade from a 1kW PSU to a 1.6kW PSU. It was not for fun and it sucked having to spend money on that, but I had to because the 1kW PSU was being exhausted and holding me back. That's a good example of when I pay attention to power draw, and making more headroom so I could push harder was the solution rather than wasting time trying to curtail or mitigate the power consumption.

Yeah, it will be interesting to see what kind of new thermal management techniques are going to be used to deal with this in a laptop. It's going to take a lot more effort and thoughtfulness than what has been demonstrated to us so far in that area. This has needed to be an area of serious effort for a while now for all of the manufacturers. None of them offer excellent thermal management solutions. If they leave the platform as is, or don't put a lot of effort into enhancements, it's going to be pretty messy.

 

Buildzoid checks in with ASUS Maximus XI Hero VRM info...

tl;dw "it's a 4-Phase VRM", "Not Impressed", "Asus considers this a $300 mid-range board for running the 9700k." " It's easier to learn the Gigabyte BIOS than it is to replace the Asus VRM's".

Spoiler: Video Time Index

VRM Explained on ASUS Maximus XI Hero | Technical Deep-Dive
Gamers Nexus
Published on Nov 4, 2018
Buildzoid analyzes the ASUS Maximus XI Hero VRM to talk about the "twin-8 phase" design and what it really means. Buildzoid explains the VRM design on the ASUS Z390 Maximus XI Hero to talk about whether it's actually bad, why ASUS went this route, and what the alternatives are. This ASUS Maximus XI Hero motherboard plays some tricks with its VRM layout and is advertised as "twin-8" phase, but the reality is that it's something a bit different.

 

I will never ever really give up! Unless of course I no longer own the hardware to further push the boundaries of that particular system.

Also to note. As everyone already knows, you included. I buy my laptops to bench and help unlock their real potential. Period. And when I want stability, I just set it back to stock and have a nice day. I do not in any way shape or form need to over clock anything to play games or do work. And no one in the real world cares about how many ghz you are running in a working environment. Or how optimal you are running in a game. Especially if your KD is .28 or 6000 Kills and 45000 deaths. So over clock numbers only matter to overclockers which is far different that stability overclockers or gaming overclockers trying to save a buck by cheating and over clocking to the next company's level part.

PS LOL @hmscott and that Rep

 

Prime 95 v26.6 is quite a nice one to run, because it's the last version before they started including AVX, so it's a bit easier on your system, maybe more realistic. I used that to tune my CPU overclock, as well as OCCT (this is a bit more severe though!).

Yep, I agree & see what you mean, and you helping to unlock laptops is great for the community too. I'm falling into that last bracket, I care about overclocking to save a buck and get as many fps & game setting quality as I can - I overclock all my stuff from Day 1, I see it as a waste to not do so, missing out on free performance, long live the overclock!

 

they should have just continued the p570wm line to house HEDT cpu, with exception of 18 cores and 20 inch laptop . now they'll realize they gotta still redesign a whole bunch of thing just to get 8c to work, or they'll simply give us the bs excuse "9900k is still 95w TDP listed by intel so we'll stick to same heatsink".

from the looks of 7700k to 8700k, they changed it a bit so i guess the engineers do have some decency to give us some improvement, i can only hope it applies to this time too.

 

That's why I do not knock this class. Hell, I was right there doing the exact same thing back in the day. Till those people I was trying to catch said....What, you think you are the only one who gets "free performance" and pretty much showed me I was a peasant trying to soar with the Eagles. So that was the end of that.

@ole!!! I will tell you this much, you can not just run some of the settings you would normally run on a 8700K/8086K if you attempt to do so, you will brick your system. And that my friend is a fact!

 

Another test, but under windows 7 and max fans.

 

That's good, that's pretty much desktop 9900K performance: https://www.gamersnexus.net/guides/3385-cheating-or-optimized-z390-motherboard-bclk-comparison. 2025 score vs approx 2040 as seen in that gamers nexus link. Although, that Gamers Nexus link was stock performance, and you're at 4.8Ghz (+100Mhz). I suppose the better RAM in the desktop systems is making up for the 10oMhz difference in CPU clock?

 

Yes, that's funny. Me testing 4.8 Ghz for throttling vs GN at 4.7ghz and cache at 4.3 while mine is 3.6 (700 Mhz faster than mine) Along with my ram running at 2400Mhz and his 3200 Mhz.I can see how he might score higher.

 

Exactly. I’m really confused why some people are asking the CPU to draw more and more power. It’s not like there is some magic dial on the front of our computer that cranks the power up. If we set the bios to use essentially unlimited power it will, but only if the application needs it. Cinebench at a given clockspeed is going to draw a certain amount of power every time you run it. What we’re looking for is the end performance result. If something is off then dig deeper and look for thermal or power throttling.

Although I’m not home right now, the only performance “bench” I could get to draw ludicrous power was that AVX loaded prime95 version. It’s unrealistic of any workload I’ve ever seen or would ever need. It’s hammering AVX again in an unrealistic manner and I haven’t seen a single reviewer use it for any reviews. Probably because it’s not relative to 99% of users. Realbench stress test is more appropriate as it uses AVX but loads and unloads.

Unless you’re going LN2 the 9900K thermal limitation is going to be the crux of any overcloxk fat before a power or VRM limitation will ever kick in.

 

And yet here is a Asus Maximus Hero Z390 running 6.8ghz with LN2.

 

Someone needs to make a video like this one, but for VRM's, PC's and budget shopping over reseller prices.

 

In that video he said he was only showing LN2 @ 6.0ghz because he didn't have a XOC Voltage BIOS for the Asus board, so the Asus Hero z390 wasn't running at 6.8ghz.

The motherboard he showed in that video @ 02:10 was the Asrock Taichi, I don't see verification that 6.8ghz happened at the Intel show, he mentions at ~01:16 "...the other day I had it about 6.8...":


He wasn't using the Asrock Taichi during GN's video coverage due to problems with condensation, and he switched to an Asus Hero z390 so GN could shoot the video.

Switching to his channel:

I think this is the first time he hit 6.8ghz on the 9900k @ 08:45 (see Chat), and he mentions the motherboard is the "Apex" - Asus Maximus Apex @ ~16:50 - it's a z370 as the z390 isn't available yet @ 25:30...

Bearded Hardware LN2 Live Stream - 6.8+GHz Cinebench R15 on Intel 9900k
Bearded Hardware
Streamed live on Oct 27, 2018
Bearded Hardware LN2 Live Stream - 6.8+GHz Cinebench R15 on Intel 9900k


Here's a brief breakdown of the Asus Maximus Apex z370 VRM design
https://www.tweaktown.com/reviews/8...pex-intel-z370-motherboard-review/index3.html

Spoiler

"The VRM on the Maximus X Apex is in an 8+2 phase mode. It also uses ASUS's 10K capacitors and microfine alloy chokes.


The VRM on this motherboard uses ASUS's ASP1405I, which is probably like an IR35201 fully digital PWM controller that offers up to 8 phases, but here it's outputting 4+2 phases. Each PWM phase (for the VCore) from the digital PWM is doubled by an IR3599, and we find four of these doublers on the topside of the PCB. Each doubler then outputs to an IR3535 driver, and we find ten of them on the backside (8 for CPU, 2 for iGPU). The MOSFETs are Infineon OptiMOS BSG0812ND, which should be rated for 50A. The VRM here is very similar to the Maximus IX Apex, but it uses different power blocks.


The memory VRM and CPU secondary rail VRs (VCCSA and VCCIO) are exactly identical to that used on the Maximus IX Apex. The memory VRM uses a rebranded two-phase controller for the memory PWM (seems like it has integrated drivers), with two N-channel MOSFETs. The VCCSA and VCCIO get power from a Texas Instruments TPS51362 10A converter and a single phase VRM controlled by an Anpec PWM and PowerPAK MOSFETs."

As described it's a beefier implementation of the previous 4+2 phase design described by buildzoid