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

they can use AC or whatever thats their choice. i mean desktop got water and chiller so what is your point you're trying to make? if laptop boards has less phase and less cooling capability, then tone down the clocks, thats all there is to it. this is still the best laptop that can house the most powerful existing consumer CPU.

also im not ignoring a possible issue, the issue will not be one dependent on your scenario and how you apply it. take johnkss for example, he benches on AC, when using it for other tasks he runs it at stock.

im not him and i do neither, i bench and run daily tasks under room temp, so my results will reflect just that. if i have more conditions that need to be met such as fan noise etc then i need to make adjustment to my oc settings even more.

 

So, with that, found the data sheet, and his analysis is roughly the same as mine.
http://www.ti.com/lit/ds/symlink/csd87350q5d.pdf
The chips used are 25A efficient, max 40A, are pushed with a single driver that doesn't double it, so the two fets working together allow for the 80A per phase. I don't know if there is current or temp balancing, but probably not. Yes, that is better than some lower end boards having 40A 4-phase systems. He is right there. But, I'm not sure that post took into account information that may have come out later, like many of the mid and high end boards having 12 or 16 phase true phase boards with current balancing and 40A fets, meaning they can drive it even harder.

Also, the data on exactly how much that chip actually pulls wasn't out yet during his post. Considering the 9900K pulls consistently in the 100s and around 200W at around 5GHz. So, power delivery on these machines means it may be able to handle it better than lower 4-phase boards, but it says I need to review the exact power delivery of the boards used in those reviews to see how much more the laptop can supply vs the desktop boards that had throttle.

Even with that, you have 4 phase 40A low quality boards, you have 4 phase 80A on the Clevo (double), you have 40A (or 45A) 12 phase with doubling (about 33% more than Clevo) and 60A 12 phase or similar with 16 phase boards. So, even 60A fets on a true 8 phase deliver more, and there is a chance some of those may get a bit hot. Sure, spreading it out over more fets can help with heat density. But that doesn't mean his analysis wouldn't change with the new information, which even he said "I'd only be concerned when overclocking an 8-core, which aren't even out yet."

@Khenglish - would you be willing to chime in here? We are discussing the heat issue of the 9900Ks from reviews and the video back a page from Hardware Unboxed discussing the reviews saying you don't need better cooling for the 9900K having throttle events, including a replication mentioning the 9900K current throttling down to 4.2GHz. This is leading to a discussion of how well the P870xM boards can potentially handle the power delivery of the 9900K in light of this information.

 

Run 5.0GHz overclock is out of questions with 8 cores 9900K but you should be able to run a nice clock speed anyway. And the 4th fan header is already soldered on the MB on the models for 8700K (intended for cooling the grills right above the power delivery). What changes Clevo will do for the Z390 models we don’t know yet.

 

When considering only power delivery directly, while the p870 has a high cpu current rating, it is a dense setup where the inductors and FETs have to dump the heat to a small board area. Is this a concern? It would be if the CPU heatsink could cool a CPU holding 200W+, but it can't. The 9900K will run over 200W until thermal throttling, at which point it will be cooling capacity limited like the 8700K and will settle at the same power draw. The 9900K during this initial burst phase is still well within the p870 limits.

Overall I see the CPU heatsink cooling capacity as the greatest concern when running a 9900K in a p870 system. There may also be a hardware set current limit on the CPU VRM that is below what a 9900K can pull that will need to be raised. I can check schematics for that.

 

My point wasn't about new models. You can redesign new models as needed. My point was about existing boards, and there is still a question of throttle events even trying to run those chips at 4.7GHz all core boost. At some point, there is enough throttle where the overclocked 8700K may overtake a throttling 9900K. That was my point. If you wind up with current throttle, it means that you either exceeded the power delivery available, or if it overheats the phases, it will clamp down on current or shutdown the laptop to protect itself. It also raises questions, as Khenglish pointed out, of warping the mainboard due to the heat, etc.

So the point was not about what the new boards designed directly for these chips will be and can do, rather it was about those looking to upgrade their CPUs inside existing machines.

Thank you for that insight. I would greatly appreciate it, as would the community I'm sure, if you could check the schematic for that.

 

Re-read my post. I mentioned both options. Both for old and coming models. And both is valid for the thread.

The 4th fan header is there already and fan will most likely come.

 

this machine would not be as good without the fan control software, really made a difference of passive, low fan profile cooling. at 20% fan speed i can barely hear the noise, while it keeps fresh air flow for moderate loads. golden!

 

i9-9900k Low Temperatures - Using the Same Motherboard and Settings as Linus
Tech YES City
Published on Oct 22, 2018
The Intel Core i9-9900K has just been released, and Hardware Unboxed made a great video explaining the differences between why tech reviewers may be getting different results, especially when it comes to temperatures, in particular a lot of people have noted how Linus Tech Tips achieved only 58 degrees celsius and got a mammoth cinebench score of over 1900 points, so today I investigate.
Having the same Z390 Maximus Hero XI and a i9-9900k on hand, I decided to test some results and try to replicate the settings stated in Linus' review in the comments section, and the results are interesting...
Linus' Video - https://www.youtube.com/watch?v=_1Cb_...
Hardware Unboxed Video - https://www.youtube.com/watch?v=NGHiR...
OC3D Article - https://www.overclock3d.net/reviews/c...


Another Z390 Maximus Hero XI build...

Z390 & 9th Gen CPUs Step by step, in depth overclocking guide for 9900K, 9700K, 9600K
Newegg Studios
Published on Oct 22, 2018
JJ from ASUS walks you through a full overclocking guide covering how to overclock the latest Intel 9th gen K series of CPU’s. If you are wondering about which options to use in the UEFI BIOS, confused about ratios, voltages and bus clocks, load line calibration & more; worry no more as all this and more is covered in depth in this full guide. Additionally, users looking for easy ways to enable their XMP memory on Z390 motherboards or looking to automatically overclock their 9th gen K series CPUs are also covered with specific information on how to accomplish both ensuring a easy, effective and stable overclock. Take your 9th Gen CPU to 5GHz+

areczek1987 5 hours ago
"That freqency drops on this light load of cinebench is a thermal or power limit? Its not looking soo good..."

NILOCOgov 4 hours ago
"Right? Second run was like 400 pts lower."

 

SL testing and found some weird results. so far sanding off .2mm reduced by 1-2C, going further to .3 or .4mm will probably double that. less than what d8 found.

also they are getting different overclocking peak results on a z370 hero board vs a z390 hero board.

so much excitement!

 

"Right? Second run was like 400 pts lower." If that's the case my 8700K which runs at a measly all core 4.3Ghz beats the 5Ghz 9900K in consecutive Cinebench runs.

 

its shocking how shallow more and more tech reviewers handle their newly acquired hardware without even trying to scratch the surface and interpret their review results... good to have people like GN and HU to cut through the crap and really show us whats up before we can do our own hands-on research

 

Many many yotubers are only people of marketing, so they aren't reliable

 

https://www.overclock.net/forum/27679368-post2489.html


Msi engineer toppc lin talks about "can 9900k really handle 5ghz easily?"

1. yes, it is easy but also depends on the silicon lottery of the cpu and what software for testing 5ghz, and what is 5ghz like all cores 5ghz or few cores at 5ghz only

2. msi version of cpuz can reflect the true cpu voltage reading, while the normal version cant

3. 1.35v 5ghz is already above the average of the silicon lottery.

4. cpu usage is an essential figure to see whether the cpu is throttling or not, 100% is a must to show the cpu is running without throttling

5. >10c difference among cores in 9900k is normal. ~20c is acceptable.
on x299 we can adjust independ voltages of each core to reduce the temp differential between cores.

6. becuase of the high temp differential among cores in 9900k, intel allows the standard of 9900k tjmax to be 115c, those z390 positioning as overclocking having 100c thermal limit are silly.

7. in godlike once users tweak the cpu ratio, all limitaion of power will be removed correspondingly.

8. if you are afraid of 115c cpu temp, and you are not luckly enough/rich enough to get the best silicon lottery's 9900k, then
dont use 9900k. the stereotype of 100c is no longer suitable for 9900k.

9. cpu cooling is very very important, as well as the cpu lottery. then you need a good mobo with good vrm.

10. intel turbo is good for general users, dont chase for all cores 5ghz for nothing and without considering your constriants/usages.

11. different prime95 version would give you different cpu temp and cpu power consumption.

12. godlike is powerful.

 

most of what he mentioned should be considered as common knowledge for overclockers. im interested in those 2 quoted. will try a msi cpuz vs traditional cpuz and then test difference with ac/dc at default value and at 1.

that 2nd point especially makes me wanna go for x599.

 

Wow, really, I didn't know that 9900K throttle temperature stipulated by Intel was as high as 115 degC! That's good news, that means that they have confidence that it can run hot & survive. Although this prompted me to Google the "9900K Intel Ark", and it's showing 100 degC as max allowable temperature: https://ark.intel.com/products/186605/Intel-Core-i9-9900K-Processor-16M-Cache-up-to-5-00-GHz-
What's all that about?

EDIT: ( https://www.tweaktown.com/reviews/8768/intel-core-i9-9900k-9th-gen-coffee-lake-review/index7.html) seems like Intel have added the ability to increase TJMax by 15 degC to 115 degC up from 100 degC. So even though the stock TJMax is 100 degC as seen in the datasheet above, this can be changed to 115 degC in motherboard BIOS - something that Intel is allowing. I'm not sure what to make of that - praps it means that 115 degC is deemed pretty safe, but might reduce chip longevity slightly although not significantly - just my musings.

I find it hard to believe that say 1.4V at 115 degC would be safe, I also find it hard to believe 1.35V at 115 degC. High voltage & high temperature = massively shorter life. I can probably imagine 115 degC at stock voltage being safe, but not when increasing voltage significantly. Thing is it's probably when you increase the voltage that you need the extra temperature allowance up to 115 degC given how difficult these chips can be too cool!

 

I'm sure the different MB engineers have talked with Intel about reliability with higher temp than Intel's stated specs ark showing 100C. Adding very high voltage will probably not be the best but ok voltage with binned chips (lower voltage) running right above 100C shouldn't scare you. Worse than average chips running +100C and high voltage is another matter. Running max overclock near max temp limits will probably make your overclock “flaky” and not so stable.

From Z390 overclocking guide - Gigabyte

 

i would take high voltage low temp the reverse. looking at LN2/HE OC for example, they go to like 1.7-2v just fine and thats because temp is under control.

also looking at ram as an example, in old days we had 1.8v OC on ram ddr2 as long as doesnt over heat. then now we have 1.2v default ddr4, even if cpu and ram are made different. so temperature is more important in this case imho.

 

Ofc he was talking about average joes, not extreme overclockers, who don't run CPUs and RAM at those extreme volts for long-term anyway, only short benches.

 

no i meant in general, high temp at over 100c is way worse than having high voltage. of course voltage would have to be under operating conditions, just like temp would need ot be under operating conditions as well.

the trend of ddr2 ram 1.8v down to ddr4 1.2v as long as operating temp is good. if ram is at 100c no doubt ram will be put to a even lower operating voltage range. temp just not good with electronics.

 

You forget that DDR4 is made on a much smaller process node than DDR2, resulting in lower operating voltage and higher efficiency.

 

From https://overclocking.guide/the-risks-of-overclocking/

The influence of high voltage and current

Todays CPUs are being manufactured in structures close to sizes of atoms. The structure of Intels Haswell CPUs is only 22 nm thick whereas an iron atom is only about 0.25 nm thick. At this level, quantum physics is very important and there are several effects causing the degradation of a microprocessor.

Electromigration is already quite well-known in the overclocking community and caused by high current densities in very small traces. Since the traces are made out of copper, an raised temperature will increase this effect even more. Eventually, the electromigration will cause damage in traces and will lead to a permanent malfunction of your processor.



However, this effect is always present and will eventually kill the CPU no matter if it was overclocked or not. It depends on the production tolerance if it appears sooner or earlier. Some CPUs will last 15 years, others only 3 years. You never know how fragile your CPU is, so just give overclocking a try and don’t worry. It could be that your CPU lasts 10 years even with overclocking. And who keeps a CPU for such a timeframe?

 

Lower voltages make higher temperatures safer (to a point). Conversely, lower temperatures make higher voltages safer (again, to a point).

Higher temperatures are not necessarily worse than high voltages, it completely depends on what temperature and what voltage you are talking about. There are more safeguards built in to temperature control than voltage control, so voltage is what will usually kill a CPU, not temperature. In addition, using the 1.7-2V pushed through CPUs on ln2 is a bad example, as most of those CPUs are killed during those sessions.

 

I don't know if there is a laptop made that can actually supply enough voltage to fry a desktop chip. Most laptops have a voltage capacity that holds back full overclocking potential, but it is a safeguard whether intended or a result of engineering shortcuts.

My last two ASUS motherboards (Maximus X Hero and Rampage VI Apex) and many other enthusiast boards have a voltage limiter that has to be turned off on purpose to push the voltage to scarier levels. If you do not disable that limiter, it will fail to boot with an overvoltage error message, and an invitation to press F1 to change BIOS settings.

 

yes, but would ddr4 work with 1.8v? i'd assume it'll work, would ddr4 work with 100c, i wouldn't dare, though it might work.

yes higher voltage will result in higher temp, then requires more voltage for cpu overclocking, its the cycle of death, for cpu overclockers.

but in reality we dont actually know how it'll affect hardware. a cpu at 1.5v running at 50c 24/7 for 5 years vs a cpu at 1.3v running at 100c 24/7 which might not run after 3 years. my experience tells me from my old sony vaio, is that cpu voltage is regulated by laptop itself, but having constant 85-90+ c killed the laptop, have about 4 years.

 

On liquid nitrogen.

 

i havent seen anyone oc hardcore on ram, or maybe i have. those 5ghz ram oc probably on LN

 

I don't really agree with that based on what I learned from reading around the subject some time ago - I think voltage is more harmful than temperatures, with extreme low temperatures with LN2 cooling being the exception. With normal water cooling & air cooling the voltage is more dangerous than the temperature in terms of degredation of the chip, but of course a high voltage & a high temperature combined is the worst case scenario, which was my initial point about the 115 degC throttle point of the 9900K. Given that Intel have allowed 115 degC for their 9900K I would imagine that is pretty safe at stock voltage, but I doubt that it is safe when you start increasing the voltage if you're at 115 degC. If I'm running 1.35V I'd want to see temperatures below 80 degC, and if I was running 1.4V I'd want to be running below 60 degC, in fact I probably wouldn't run 1.4V for a 24/7 overclock because I saw about 10mv degredation of my i7-6700K when running 1.4V for a year at temperatures pretty much always below 60 degC when gaming - I think 1.35V is a safe maximum voltage for long term, but I'd want to keep it below 80 degC for sure when doing so, if not even lower. I remember finding some testing that someone did with Skylake and 1.35V was fine for longevity, 1.4V some degredation likely over a year or so, and above 1.4V degredation happened quickly, 1.5V within hours.

EDIT: oh yeah, found the thread where the guy did his testing & extrapolations, it matches the slight degredation I saw of my CPU too:
https://www.overclock.net/forum/5-intel-cpus/1673657-cpu-overvolt-death-degradation-stories.html
And here's my initial post in these forums where I talk about some more detail re my CPU degredation and how it relates to the thread I linked above:
http://forum.notebookreview.com/thr...ers-welcome-too.810490/page-296#post-10766430

 

FWIW, max operating voltage increased from 1.52V on 8th gen to 1.72V on 9th gen, per Intel's datasheet: https://www.intel.com/content/dam/w...y-datasheet-vol-1.pdf#page=123&zoom=100,0,220

 

I wonder if that means we can assume that 9900K has +0.22V extra protection against CPU degredation in comparison to Skylake/Kaby Lake/Coffee Lake. I find that surprising though because it's still based on the same 14nm silicon node, so I would have thought that the voltage safety ranges would be the same.

 

Core i9-9900K performance, Intel cancels 10nm?? Q&A | The Full Nerd Ep. 73
PCWorld
Streamed live 17 hours ago
Join The Full Nerd gang as they talk about the latest PC hardware topics. In today's episode we're going to cover the Core i9-9900K reviews and performance metrics and the report (and rebuttal) of Intel cancelling it's 10nm process. As always we will be answering your live questions so speak up in the chat.
Start- 3:14
Core i9-9900K - 4:51 - Certificates are awarded.
10nm cancelled? - 34:55
Q&A - 50:06


9900K Hardware Unboxed Response
OC3D TV
Published on Oct 23, 2018
https://forum.overclock3d.net/showthread.php?t=90750

 

FYI
https://www.overclock.net/showthread.php?p=27682624

 

we cant relaly trust intel though thats the thing. what if they want us to kill our processor?

 

Says a lot that the Asus Maximus X Hero cannot handle the power delivery for the 9900K!


Sent from my SM-G900P using Tapatalk

 

Of course, they are a corporation that wants you to buy more. Hence saying use monster voltages, use higher temps. Only way for them to offer the higher performance.

Instead, just be reasonable on voltages, keep your temps in check for 70s to low 80s tops for daily driving, try to keep voltages to 1.4-1.45 unless your cooling can handle it, preferably less considering I've seen between 1.25V and 1.35V to hit 5GHz to 5.2GHz (seen in reviews), and call it a day.

I wouldn't recommend the 9900K for a laptop, or at all, but if you get one and put it on a desktop, then consider 280 or 360 AIO CLC or building a custom loop.

Edit: GN vid with more on delid


Sent from my SM-G900P using Tapatalk

 

grab a 9900k and turn off HT, power consumption should be similar to that of 9700k, which only uses tad bit more power than 8700k. but again i need to see the binning result first from SL.

 

I still feel if you are turning off HT, then why not get a 9700K? If planning to move the chip to a differnet machine in the future, that is fine. But the prices may come down by the point you do that. That is kind of my problem with it overall.

Also, as I mentioned in another thread, I don't think needing a beefier power delivery is a bad thing. Nothing has changed on the VRM front in a long while. Now, we have a mainstream chip needing HEDT type power delivery (even the 4 phase doubled to give 8-phases, which is fine, is having trouble with these chips), we have the incredible VRM on the MEG Creation board and the boards for Intel's 28-core monster, and we should consider at some point doing better for X299 or its successor. But, I do want to make sure consumers know the limitations of their current hardware so they don't buy a product then get less than they thought and blame Intel on bad performance when it is the board not being capable of doing its job that is the problem. Sadly enough, bringing it up HELPS Intel in PR, rather than harms them (the price of the chip did enough on that front)!

 

Well I checked the P870TM1 schematic and the current limit is set to 138 Amps for the CPU. This is 186W at 1.35V, so the TM1 should auto shut down a 9900K at stock. This is strange because the TM1 is already being sold with the 9900K, so I don't know what shenanigans were done to make it run, or if end users will just have their 9900K systems crash all the time. I would think they would have had to impose an extra power limit somewhere.

It is easy to mod the current limit to be higher. Lower the resistance of PR188 on the motherboard from the 110k default.

UPDATE:

This limit is just what the VRM reports as the current limit, and the system will be limited to this, but it will run and be stable.

 

I appreciate @bennyg 's post with part of this info, but the whole quote from Silicon Lottery answers some of the questions I am seeing posted after, so some didn't go to the site and read the whole quote, here it is:

"10-24-2018, 01:06 AM - Thread Starter

Silicon Lottery
Customer Support


More random tidbits ( all notes without delidding).

The 9900K runs very warm under stress tests, but in gaming/encoding/benchmarks temperatures are more realistic. The 9900K shines under a custom loop with a good waterblock. We'll be binning under the conditions of 240mm AIOs, but you can expect to stretch a 9900K's legs a bit further under custom water.

The 9700K doesn't really have the same thermal issues, and is just a bit warmer compared to a delidded 8700K. The advantage to the 9700K here, is it can use much higher voltages under stress tests than the 9900K.

What were considered fairly good motherboards on Z370 (Taichi, Maximus X Hero), are not making the cut. Overclocking headroom is reduced on these boards with a 9900K, with a tad of VRM throttling (especially the Taichi here.)

The Z390 boards we're going through are looking pretty promising, you can take a look at the old Coffee Lake QVL to see what Z390 boards we are attempting to validate with.

With the higher current, Vdroop can be pretty extreme so LLC settings will need to be changed on most boards to make sure the proper voltage is being held under load.
www.siliconlottery.com "
https://www.overclock.net/showthread.php?p=27682624#post27682624

So it's unclear what the "conversion" factor will be from stock 9900k stock "paste" IHS on 240mm AIO SL testing to Custom Loop + delidding.

If these guys don't do delidding service, I don't see as much value from their services, especially with the 9900k / 9700k.

I can understand why they wouldn't want to get into grinding the die - for a total temperature drop of 15c-20c - but at least fix the bad STIM situation with LM or even a non-conductive TIM under the die.

 

But will you need 1.35v for stock clocks?

 

Please check your PM

 

I made that up as an example. Just do V * I for power limit.

 

Yeah, I know sir. This was only to show a good binned cpu should be able to run on those specs.

 

Not necessarily. I saw at stock (non-undervolted), they saw between 100 to 138W pull. This suggests that as long as you can keep the chip from thermal throttle, you could potentially run that chip at stock.

But, when OCing to 5.0GHz all core, then you wind up exceeding the wattage allowed with some samples (Hardware Unboxed's sample was tested at 1.35V@5GHz, whereas Gamer's Nexus was in the low to mid 1.2 voltages doing their live overclocking session for around 5GHz-5.2GHz).

That means a good binned chip may run stock or low OC, but above that, if trying to do all core 5GHz, it will be pushing that board to the limit (I cannot remember seeing a chip run 5GHz in the reviews at less than the 190s, but could be wrong there).

Still, 4.7GHz all core with an undervolt to try to help on power specs and heat should be possible, even if hot. And the lower Watts at stock suggest it could run those chips at stock. But, I know a lot of people want to overclock their hardware here, which is why I've been pushing to think deeper on it.

 

Yeah, I know 5.0GHz is too much on older boards (probably also for the cooling).
Handbrake from Tweaktown review.

 

Looks good on a 12 phase board and getting to 5.3GHz on 1.344V (don't know his droop or what he set for LLC, but still, or if that could actually run the tests at 5.3 at that voltage with his sample, but still falls right in line with what I said on low 1.2V up to 1.35V for 5GHz to 5.2GHz, depending on sample quality).

Edit: Especially with the 172W on CPU package, which many other reviewers reported higher (that also can vary by load used to stress the chip)

 

Test done with 5.2GHz Read the text in red. With lower clocks than this the power consumption will decrease due lower needed voltage.

 

Yes, and we don't know that the voltage used was or was not 1.344V at 5.2GHz, or whether the Handbrake test really pushed it as hard as you get from Blender or other tasks. WATTAGE DRAW VARIES BY LOAD AND INSTRUCTION SET. He likely used Handbrake for AVX instruction and use, and we don't know if he applied an AVX offset for that matter. There is a lot left to know from that image that the red DOES NOT tell us. I know you are more knowledgeable and a better OCer than that to not understand what I'm saying in this instance.

Edit: We also do not know the length of time run on handbrake.

 

Alright, I checked the VRM schematic and the 138A limit is for what the VRM reports to the rest of the system. The VRM does not have a hard current limit programmed at all. I talked to Prema some and confirmed this.

 

The system doesn't shut down if you hit the 138A board limit, the CPU will just current limit throttle itself. I've tested this on my system. It's pretty easy to draw 138A in AVX workloads, even on a 6-core chip without Hyper-Threading. On the one hand, this current limit keeps the 8-core chips about as cool as the 6-core chips when used in this motherboard, on the other hand it limits their clocks more due to the 2 extra cores requiring more current. Intel's spec for the 9th gen 8-core S-Processor line is an ICCmax of 193A. An updated motherboard with a higher current rating, and improved cooling, is needed to get the most out of these 8-cores processors.

 

https://www.kitguru.net/components/...9900k-review-mighty-processor-mighty-price/6/


https://www.hardwareinside.de/intel-core-i9-9900k-coffee-lake-acht-kerner-im-test-36861/3/


With a good binned chip, ok voltage and cooling it could be a better choice than 8700K. All depends if you can cool it. But forget decent overclock. Most likely better with coming Clevo with Z390 MB.