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

@Robbo99999 - this is probably more severe than your situation.

Running the display at factory pre-set refresh rates higher than 120Hz can result in noise and artifacts. I am able to reproduce the artifacts above 120Hz by changing overclock profiles. It is not related to GPU overclocking, but switching Afterburner profiles simulates what it does randomly with the GPU running stock. I even see this when navigating the system BIOS at refresh rates above 120Hz. It is totally random though. Might not see it for days or more than a week, then it starts happening again. Setting the refresh rate to 120Hz or switching to HDMI (60Hz) corrects it. Sometimes toggling G-Stink corrects it.

 

A person on the internet said something. It MUST be true....

Sent from my SM-G900P using Tapatalk

 

But, only the stuff on YouTube, Twitter, Reddit, Facebook and CPU/GPU Boss.com and Notebookcheck are infallible and should never be questioned.

AMD, Intel, NVIDIA, ASUS, MSI, Gigabyte, ASRock, and other internet sources run by big business are suspect.

 

Sent from my SM-G900P using Tapatalk

 

I know AMD use copper in their lid. Already posted over a year ago http://forum.notebookreview.com/thr...6-owners-lounge.797128/page-724#post-10571059 No point in being so set up (No need to get so excited/upset)

 

i want zen2 32 cores with IF/memory direct connection to the cores. or like adoretv jim stated, core die surround non core die. best if it comes with L4 cache

 

yep looks like gotta pay big bucks for that 8c16t with extra cache hoping a 5.2 sample.. and then turn off HT

 

9900Ks with HT on, using more power, will clock better than the mostly-thermally-limited 9700Ks?

Edit:

And yeah, that AdoredTV video shows what trouble Intel is in. This 8 core may well be their last "performance king" for quite some time... though if Zen2 is only a 100m2 die I wonder if it'll run cool enough to hit the high clocks to beat Intel's single core throughput. I hope they don't dust off the old Netburst era book of tricks and start paying OEMs, game devs, and benchmark makers to deliberately hobble AMD's more-cores advantage in various ways...... *cough*PT*cough*

 

@|BrotherTalon
Your imgur says CPU pegged 170watts; HM posted a uTube a few days ago, 5.2Jigahurtz run @295watts (per CoreTemp 1.12.1); in time & with superior cooling I hope yours smiles back & can pull another 125watts / interested to see your Asus board cooperate > godspeed

Spoiler: 295watts AVX

@25s @3m25s

 

Only about 900W away from catching up with the 7960X's power draw at 5.2GHz.

 

yep, thats why i said turn off HT. 9700k 51 bin is a disappointment. if 9900k can get a 52 bin sample then turning off HT will probably push it even further.

 

so lets go ahead, apply 3V to the core and clock that mofo into the terahertz range while cooling it with einstein bose condensates

Sent from my Xiaomi Mi Max 2 (Oxygen) using Tapatalk

 

Maybe just get a 9900K, give it the cooling it requires, then overclock the snot out of it and see what the benchmark scores look like. If it beats your 8700K's best, then it's an upgrade. If it does, then how it that a disappointment?

 

I had a look at your video there, but the video is grainy and a bit shaky anyway, so I couldn't see any difference in the quality of the display when you were flicking through refresh rate modes, I did see a brief white flickering just at the moment that the refresh rate was changed, but that's just lasting a few milliseconds when you click the 'apply' button. After upgrading that Firmware that Talon linked my monitor is ok this morning after being left on standby rather than the normal unplugging, but I think it takes a few days of not being unplugged from the electrical outlet for the problems to start occurring, but I'll let you know how it goes with my experiment of just leaving it on standby rather than unplugging (to see if the problem still exists or if the firmware update solved it).

@Papusan , have you seen this, you're not gonna like this, it's a 9900K non BGA throttle book for the privilege a cool 4000 Euro! (Both in Cinebench loops and also in Witcher 3 Stress Test)
https://www.notebookcheck.net/Schen...D-Clevo-P775TM1-G-Laptop-Review.350636.0.html
They need to step it up when it comes to trying to handle this 9900K!

 

i dont see a problem tbh. everyone knows that clevo stock cooling is crap and the machines only turn awesome when properly tweaked still looking forward to actual user data here

 

Yeah, but they didn't throttle like that with the 8700K & 7700K did they?

EDIT: Oh, right, yes, it did throttle even with 8700 non-K! https://www.notebookcheck.net/Schenker-XMG-Ultra-17-Clevo-P775TM1-G-Laptop-Review.288972.0.html
Ok, so these notebooks just suck at stock from the factory full stop - must be bought for tinkering/fine tuning by the end user! Still sucks though.

 

That is exactly the issue. Except that it only does it momentarily when flipping through the profiles to replicate it. When it is acting up it does it for more than a moment and keeps doing it. And, it does it in the BIOS even, so definitely not a driver issue. And, never does it if I reign in the refresh rate for 120Hz. If I set it to 144Hz is usually stops it as well, as does toggling G-Stink (usually, but not always by toggling G-Stink). The problem is primarily with the monitor set to 165Hz. And, as I mentioned. It might work for days or weeks at 165Hz until it randomly decides to act up.

Yup. For the most part. Without a @Prema BIOS and lots of TLC and massaging to coax them into being something better, notebooks are almost universally garbage now. The only thing separating Clevo from the rest of the trash is the hardware configuration and the potential for becoming something special. But, out of the box they're pretty much garbage like everything else. (Garbage with better specs, to be precise. But, at least the situation is not utterly hopeless like it is with a BGA turdbook.)

 

Ah, I understand, that white flickering is pretty much permanent once it starts acting up. That's quite different to the 'artifacts' I see on my monitor when it starts acting up. Mine becomes slightly grainy, and strangely the distant textures in games loose their clarity - almost exactly like anisotropic filtering not working. In fact I messed with anisotropic filtering settings to see if it would get rid of issue but no effect, it's the monitor instead, solved by unplugging monitor for a period of time. Perhaps this firmware update that Talon linked will fix the recurrent problem, I'll let you know how it goes. (Interestingly the problem was sometimes solved by enabling/disabling G-sync and also sometimes by changing refresh rates, so that's similar to your issue there, from that point of view.)

 

"I know some of you are wondering, and yes from initial results it looks like the 9900Ks will clock a little bit better than the 9700Ks."

straight from the horse´s mouth over at the SL forums. so im guessing 5.2+ Ghz bins are coming for the 9900K the mentioned rumors of Intel doing some extra tight binning on the 9900K chips seem to be true, which would also explain the price hikes and the low initial numbers of chips were seeing as actually being delivered. ETAs from the SL distributors keep being pushed back, so current ETA of Nov. 9th as stated on website will likely not be the actual one.

Link: https://www.overclock.net/forum/180...fee-lake-refresh-binning-20.html#post27693122

 

You do realize that adored mentioned 256MB L3, which is 4MB per core, which is double what Intel offers. Now, I'd like larger L3 victim cache over L4, but I also don't know what is put on the uncore chip yet. I now the IMC and PCIe controllers, which means they can bin IMC performance, scale number of memory channels, etc. And because it is hard to scale those things, this allows for those to be built on the 16/12nm node from TSMC or 14/12nm node from GF. As an example, the sample RTG leaked about was running 3600 ram XMP. So good things may come on supporting faster ram. And single block of IMC means only one NUMA address per socket, ending the local vs distributed questions. This just leaves the questions about latency and interconnect changes.

Although, I wouldn't mind HBM being added on die with an interposer.

Did you see the level 1 techs video on windows gimping AMD performance on the 24 and 32 core chips? Or windows running all chips at a 20% slower performance than Linux, including the 8700K?

And I don't think they will, although they already told HPE and others that of a customer mentions wanting AMD, offer big discounts on Xeon pricing. So....

But, one person I know mentioned their 7980XE was slowed around the release of Photoshop or premiere that introduced hardware acceleration for the 8700K, later shown to be snake oil by Puget, but that got people to switch from Ryzen and TR builds to mainstream Intel builds. Some games are already afoot.

The six core is the 8 core die for the 9600K or whatever. They make one die for the mainstream chips, then bin like usual. You keep top bin for top product, defects you disable HT and sell a tier lower, more or on cores but non-critical, 6-cores. That seems more like production issues.

Also, what I think was meant is Intel went from a 400-600mhz range on CPUs to 300, leaving less variance to differentiate and less performance on the table, a complaint levied by people against AMD as a bad thing.

Also, go back and read the Intel earnings call Q3 transcript I posted. Intel said directly Q4 constraints will hit more and is expected to continue next year. Straight from Intel's mouth. Or look up that TSMC will be making Intel atom, celeron, and Pentium silver SoC chips.
https://www.techpowerup.com/249039/...try-level-cpu-and-chipset-fabrication-to-tsmc
https://digitimes.com/news/a20181030PD205.html

Sent from my SM-G900P using Tapatalk

 

i dont care so much as to ram performance, because i know if the CPU core tech with capable IPC then i wont need to worry about overclocking ram/uncore, those don't benefit much and the power consumption jumps. (im more into having good ram capacity which any of WX CPU will give 128gb ram to me with ease)

however Zen arch is different and performance is affected by ram speed much more than intel's chip, with zen2 it may not be the case anymore if its a new redesign, we will just have to see.

hoping for a few thing if i were to switch to desktop:

- good ipc bump (a hard 10% increase)
- good clock bump (at least 10%, so a 4.5 minimum is necessary for 32 cores)
- IF being optimized or redesigned (128gb ram isnt really about high mhz)
- power consumption (should be good to see intel gets destroyed 7nm 32c vs 14nm 28c)
- storage performance/oc software are bonus but some things just have to be sacrificed

 

When the OC is held back by thermals, a bit better OC than the 9700k isn't going to help the 9900k overall for most buyers. Better to stick with the 8700k / 8086k right now, or go Intel HEDT / AMD Ryzen / AMD ThreadRipper if you need lots of cores.

The ThreadRipper physical layout and larger IHS area gives a much better cooling result for the same power output at max OC than the 9900k, comparing results for power usage vs CPU temperatures the ThreadRipper's are 25c-40c (60c's vs 100'c) cooler under load even with 2x-4x more cores/threads at high OC power draws.

Update:
10-30-2018, 10:37 AM - Thread Starter
Silicon Lottery
Customer Support

"Still having a hard time getting 9900Ks. There are some available in bulk I can bring in from a supplier in China, but the price is very harsh (~$700 ea).

ETAs from the distributors here keep getting pushed back, so don't put much faith in the ETAs we have on the site. www.siliconlottery.com"
https://www.overclock.net/forum/180...fee-lake-refresh-binning-19.html#post27692424

For those of you wondering where the 9900K topical posts went, they were last seen here - 6 pages back before the OT GPU video discussion began - there's some good stuff back there - check it out
http://forum.notebookreview.com/thr...e-lake-cpus-z390.811225/page-94#post-10814755

@Johnksss installed a 9900k into a P870TM1 laptop for some interesting results, that's back 11 pages, hopefully he'll come back with some tuned results showing the maximum the 9900k can do for benchmarks and daily use soon.
http://forum.notebookreview.com/thr...e-lake-cpus-z390.811225/page-88#post-10814155

 

Until we have certainty of the clock gen on the interconnect being divorced from the speed on ram, I will operate on the assumption that it is not. Hence why mem speed is mentioned. Also, why 4.5GHz on 32C? If we examine the numbers on HWBot, the average is 3921MHz on water. OC3D got 4GHz. So, 4.5 minimum is just you talking, as the 28-core Xeon would barely make 4.5 on all cores, if it does at all, without a chiller. If the IPC gap is closed due to it going up by 10-15%, then the 4 extra cores and not having the current limitations on memory to hop to other dies (as all have to hop dies to the uncore die with the IMC) should be enough for it to outperform the 28-core at a lower speed, all while potentially having the same sub-$2K price tag versus an expected $3500 28-core overclockable Xeon. It seems you are setting up hurdles on purpose so that you can then say, even if they check off the boxes on the other categories, that it just isn't good enough.

1. Now, with that said, the IPC bump is mostly confirmed at this point.
2. The clock bump, as I mentioned, may or may not it that frequency for all cores, it is a 32-core chip for heaven's sake. I feel more confident about saying that speed on 16-core. We should know more after Nov. 6. But, current rumors put the 64 core around the speed of the prior Epyc 32 core chip. Considering doubling the cores, it would suggest a potential for higher clocks on the 32-core TR replacement. But I'm not going to play into this it must meet this speed at minimum game.
3. As to IF, we don't know if it is even being used. We know they made changes to the interconnect. That might mean a tweaked IF or an IF2. It also could be switched out for a different data fabric, using Intel's AIB which was given to DARPA and proposed as the new standard for interconnect, or could be an active interposer which would drastically reduce latencies, or a passive interposer. We just do not know yet
4. And sure high mhz ram is about that. By increasing speed, you increase the bandwidth to the large pool of 128GB. Instead, since using an uncore chip, and with the rumor that the memory channels are increasing on Epyc platforms with a potential socket change (obviously with generational compatibility), and with Intel using 6-channel on Xeons and the 28-core part, I see more likely an increase in channels (which would negate speed), increasing mem support to 3200MHz DDR4 ECC, and ability to support faster modules. And since able to bin uncore chips IMCs, then just reduce connections with defects and slap them on HEDT or Mainstream, we should see better and broader memory support, as well as PCIe 4.0 support being rumored (no word if Intel's HEDT platform for X599 28-core Xeon will support that or not). Which 64 lanes of PCIe 4 would be awesome, depending on how they allow that to be subdivided and work with PCIe 3.0 cards.
5. Power consumption is pretty well guaranteed. Looking at the Adored video, it seems they took the majority of the savings in power efficiency rather than performance. Now, he mentioned a Zen X in that video, which may be the same chip optimized for performance gains rather than power efficiency, and those, if four of them stuck on a die, may have the speed you were looking for, but at the cost of sucking down power. I would bet if that is how they are doing it, or if they did two different Zen chips instead of one to go across all product lines, then those would be in mainstream chips, majority of, but not all, TR chips, and in a specific server lineup of chips where speed is prioritized over core count and power efficiency (Intel has a similar group of Xeon chips, so makes sense for AMD to diversify the offerings to address the market better).
   
6. On storage performance, you still have the ability to do a PCIe 3.0x16 card with 4xNVMe drives raided together where that would give you more performance and still have 48 extra lanes on the HEDT platform. If they do PCIe 4.0, and have it setup so that the PCIe 3.0x16, even though taking up an x16 slot, only takes 8 lanes of the 64, then you could double that, which would likely be a mirrored array or something of that nature, unless wanting to do a 5, 6, 7, 50, 60, 70 raid array or software raid array with something like unraid, etc.
   
7. OC software - yeah, still not holding out hope here (also not like XTU is much better, in case anyone thinks that is much of a program, although TS doesn't support AMD, so...).

I think that explains what my thoughts are on the upcoming product stack. Either way, we will have more info in a week. The Horizon event for AMD is on November 6th. Although they are focusing on discussing the Epyc 2 chips and servers, it should give more insight.

This goes along with the mention that supply constraints would be worse during Q4, the outsourcing of the Atom, Celeron, and Pentium Silver SoC to TSMC for supply, and focusing on high margin products like Xeon and iCore, which likely excluded mainstream mostly because even at the margins on the 8c/16t 9900K, it is less than the margin they get per chip on Xeons which cost upwards of $10K, or HEDT chips, so long as yields are stable enough on those products, and considering the HEDT 9 series are still Skylake-X chips (and potentially 14nm+ process) which could be on different lines altogether, the mainstream never seemed like something to give much love on when you have to make a choice on production. And $700 each for bulk from China, both laughable and shows people should consider a 1950X at that price.

 

the hurdles are there because I get it for what I need. also theres some misunderstanding here because i dont consider intel an viable option this time around.

1. i use a lot of consumer software which arent multi thread optimized
2. both IPC and frequency at this point are equally important especially IPC been ever closer to intel, snappy computing = good ipc and fast frequency
3. higher frequency also means better storage performance.. though very little, anything helps on 4k, not so much on sequential. sequential is a plus for sure.
4. im going on water with chiller, i dont see how it can be done at 4.5 with 32 cores
5. im betting big on 7nm being much more efficient than intel's 14nm, so a reduced temp even with 32 cores

lastly price, i always have the option to just not buy and wait for zen3 if zen2 comes short, intel doesnt have anything to offer until 2020 anyway so right now thats not of a concern. its not like intel gonna suddently cave in to pressure and offer their 28 cores chip at $2k to match AMD's chip.

 

Ok, so long as you are acknowledging Intel doesn't fit those criteria as well. But I must ask, if the software you use are not well multithreaded optimized, are you planning on using affinity or a program like process lasso to better utilize the processor or what?

On IP and frequency, it is the multiplicand of those that show overall performance. So why IPC specifically? Also, IPC varies by task, so although the ranges given are 10-15%, you may want to wait to see what additional instruction sets may be available and the specific uplifts in regards to the programs you use, something I recommend to everyone. But just curious.

Skipping comments on storage as I am sure you have some knowledge on that topic I do not, so I'm not the one to best address that point either way. But the reason I mentioned the raid setups before was the one analysis, I think by PCPer (they may have shilled for Intel, but their work on storage is rock solid, and people would be foolish to not take a look at it), showed that even with the added latency of the raid array, the array actually performed extremely well with NVMe drives. I think this was a couple years back in regards to Samsung's first or second gen NVMe drives.

On the water chiller point, then very likely you could get 4.5GHz on the 32 core chips. Even the 2990WX on regular water rigs are getting 500-600MHz less than that. So the added frequency between generations PLUS the chilled water, I feel much more willing to say you could likely have a 4.5GHz 32 core workhorse.

On 7nm being more efficient, I could agree depending on the meaning of efficient. I do think the power envelope will be lower, as they already offer compelling 32 core products within a 250W envelope. Total System Power Draw seen by OC3D was 776W when overclocked (415W stock). https://www.overclock3d.net/reviews/cpu_mainboard/amd_threadripper_2950x_and_2990wx_review/21 . Similar numbers were reported by TechSpot (Steve from Hardware Unboxed) with 756W for Total System Consumption (383W at stock). https://www.techspot.com/review/1678-amd-ryzen-threadripper-2990wx-2950x/page8.html . Considering the news they focused on power savings for the Epyc chips, and the 32-core being closest thing to a server chip, I would say it should be more efficient by a fair amount. Obviously the power to power the chiller would need calculated into the system power if going that direction, and you may wind up on past the efficiency point when OCing, but that is something to be seen later on after more information is known.

And I agree, Intel won't enter a price war with limited capacity, at least not a head on one, instead continuing as they are doing, which is reducing Xeon pricing for servers selectively.

 

heh its not so much as a matter of acknowledging or what not. i think its more about being obvious that intel doesnt have any advantages when comes to high core count chip hence i dont bother mentioning it.

even in the unlikely event intel somehow did offer a 28c chip at $2k, theres still heat problem. example 28c 4ghz intel chip be at 800w, 7nm zen2 on the other hand likely capable of doing it at 650w, given both system would have same cooling solution, I can push another 150w in frequency on AMD's system. IPC match, AMD's frequency higher, my consumer software wins, multithreaded optimized software also wins, my wallet wins but gets destroyed at the sametime.

 

@ajc9988 you mentioned about software on single thread, this area i think maybe the only one advantage intel has over AMD is TBM3.0 but tbh since i dont own any HEDT system i can't tell you how it works and if it really does work the way i wanted it to work. im also unsure if AMD's PB2 work the same way.

maybe @Mr. Fox could help me out on this part with turbo boost max 3.0, run the highest frequency on a single core for a single threaded software. like, his overclock every core to 4.7ghz, but with TB3 setting for 1 software to run on a single core to say 5.2 ghz for example. TS should be able to make it 1 core at 5.2, but really up to TB3 to see if it'll work properly. iirc the old Turbo boost 2.0 doesn't work, as soon as there are any sort of OC involved, it'll run all cores at the frequency we set, single core frequency OC will no longer work even if I set it to few 100mhz above all core OC.

i tested this tb2 with affinity + TS + XTU. never had a HEDT since broadwell so can't test this.

 

Turbo Boost 3.0 software has conflicts with ASUS software, so I don't use it. I played around with it a little without ASUS software installed and did not find any value in having it. Seems like gimmick software more than anything else to me. In fact, I really don't like any Intel software and avoid it as much as possible. XTU, Intel ProSET, Intel MyFi, Intel ME software (AMT) is generally worthless bloatware as far as I am concerned.

Edit: What exactly do you wanted tested with a single core overclock?

 

Now, according to the Adored leak, there is a new boost coming to AMD chips which may act like a TB3, or AMD's implementation thereof, for 2 cores on an Epyc chip to be WAY higher clocked, theoretically in the 4-5GHz range. The main problem with these boosts based on number of active cores is that if another process activates on another core, it exceeds the value of cores allowed at the specified frequency, so kicks it into another level on AMD. AMD has their boost setup so that 1 core runs at X speed, 2 cores run at X-1 speed, 4 cores runs at X-2 speed, 8 cores runs at X-4 speed, 16 cores at X-5, or something similar to that. This is to slowly step it down as increased core count goes up. They also monitor the voltage, power draw, and heat and have some algo to decide when to clock down.

Now, there are different types of precision boost with AMD: Precision Boost, X boost, and the new precision boost overdrive, which is the most advanced and gives the step downs mentioned above, but allows for higher frequencies than the original boost or X boost gave. Also, AMD has it where Ryzen Master already identified the best binned core out of all dies and best core and second best core in every CCX. This is great, if the scheduler honors the designations or AMD creates a remap of cores in Windows so that it uses the best core first, then on down the stack. I do not know that Windows does this, but with the information, theoretically you could work on tweaking affinities of some programs.

As we enter this new age of high core count and extreme core count chips throughout the stack, attention to the scheduler has been on many people's minds, especially as both AMD and Intel have plans to move to chiplet design architectures, where access to memory, die to die comms, and other similar concepts need to be considered when telling programs where and how to use the CPUs. Here is a recent article on the topic from anandtech I came accross. It shows why Wendell at Level 1 Techs found some boost in certain workloads by preventing core 0 from being used, but that it isn't universal for helping all programs, and that solving the scheduler issues which retard performance will not be easily solved. https://www.anandtech.com/show/1344...-scheduler-wars-with-amds-threadripper-2990wx

Also, side note for a laugh, as Wendell mentioned on his news post, Windows is adopting Google's Retpoline solution to Spectre 2 over Intel's solution because of less performance hit. He mentioned that it is pretty bad when Google, a non-chip company, can come up with a better solution than Intel which has access to the microcode black box of the CPU.

 

for example can you test this to check if it works. your 16 core all core overclock say at 4.8ghz, using the turbo boost 3.0 software to have CB15 run on the best core. then in TS or XTU or both, have everything set to 52 48 48 48 48 48... to 15th 48. then run CB15 on single threaded test, 1st run without affinity, 2nd run with affinity pinned to that best core. see if you get it working at 52x instead of 48x during the load.

 

my primary concern is that maybe those boost feature including TB3, will not work if another process kicks in (window background) etc and thats an issue, but i guess thats dependent on window itself how it works the hardware. im guessing it is too much to ask for software A to run on 1 core at 5ghz while software B run on remaining 15 cores at 4ghz at the sametime as software A. even if not at the sametime i still wish for it to recognize that window is barely doing anything so software A will still run at 5ghz while tiny background tasks are being performed.

as for knowing best core, TB3 and AMD has that so its great, concern is that if i run 5ghz on best core vs 5ghz on worst core, its the same performance they are both 5ghz, just one uses more voltage than the other. i need the best core to clock highest and work when i need it to, that kind of thing (liek the scenairo i mentioned above)

 

Wow, lots of variance due to the variety of power delivery on different motherboards at stock settings for the 9900k as by CB15 score:

I find it amusing he spends so much time to show minor BCLK variances when default / MCE power delivery capability and default voltage settings are the determining factor in the wider score differences at stock. The minor score differences due to BLK are less interesting.

More interesting would be to find the setting from stock to "fix" each motherboard so it delivers maximum power to the 9900k when OC'ing.

Maybe that will be covered in another video...otherwise I think he missed the point of contention that shows with the 9900k not able to pull enough power to deliver maximum performance, showing higher / lower temps and higher / lower performance. Of course, and reduce / tune the voltage to the minimum required for highest stable OC.
Optimized vs. Cheating: Z390 Motherboard BCLK Comparison
Gamers Nexus
Published on Oct 31, 2018
There's a fine line between optimizing a platform and pulling tricks to unfairly achieve higher scores. We discuss BCLK differences in Z390 motherboards today.

 

rather than cheating heres a better one. this is from elmor from overclock forum

 

@hmscott

Ugh really. This is what happens when you run out of ideas for a video. Squabbling over 60pt variance that can be explained by login time, programs in background, etc.

 

Exactly, and then he misses the really interesting problem to solve, the power delivery settings to get the most (useful) power out of each motherboard.

Also, far more annoying is the fact that the non-exact 100.0hz settings really mess up the mind after looking at the monitoring software for so long, and really wishing the 99.943hz setting was 100.0hz.

I mean is it really that hard to auto-tune for 100.0hz out of the box so all of the values relying on that number have even numbers??

 

Intel Core i7 9700k vs i7 8700k Test in 8 Games
Testing Games
Published on Oct 28, 2018
Intel Core i7-9700k vs i7-8700k Test in 8 Games

Spoiler: Details...

Games :
Assassin's Creed Odyssey
Forza Horizon 4 - 01:10
Shadow of the Tomb Raider - 02:21
Project Cars - 03:46
Far Cry 5 - 05:21
Assassin's Creed Origins - 06:28
Arma 3 Apex - 08:15
Grand Theft Auto V - 09:24
System:
Windows 10 Pro
Intel i7 8700k 3.7Ghz
Intel i7 9700k 3.6Ghz
Asus ROG Strix Z390-F Gaming
GeForce RTX 2080 OC 8Gb
16Gb RAM 3200Mhz

Core i7 9700k vs Ryzen 7 2700x Test in 10 Games
Testing Games
Published on Oct 30, 2018

Spoiler: Click to See Video

Intel Core i7-9700k vs AMD Ryzen 7 2700x (RTX 2080)
Games :
Project Cars
The Witcher 3 - 01:10
Shadow of the Tomb Raider - 02:26
Assassin's Creed Odyssey - 03:51
Grand Theft Auto V - 05:01
Forza Horizon 4 - 05:53
Far Cry 5 - 07:02
Fallout 4 - 08:08
Assassin's Creed Origins - 09:35
Arma 3 Apex - 11:10
System:
Windows 10 Pro
Intel i7 9700k 3.6Ghz
Asus ROG Strix Z390-F Gaming
AMD Ryzen 7 2700x 3.7Ghz
Gigabyte X470 AORUS ULTRA GAMING
GeForce RTX 2080 OC 8Gb
16Gb RAM 3200Mhz

Intel i9-9900K Explained: The Road to 5GHz
Engadget
Published on Oct 31, 2018
Intel's new i9-9900K is the companies first wide-release 5GHz chip, and its first 8-core consumer processor. It's certainly fast, but it's also a showcase for the struggles CPU manufacturers are having reaching faster speeds.

Spoiler: Click to See Video

 

This is a problem I have with GN as a whole. They mean well but get so deep into the trees that they end up missing the forest.

 

Okay, So this is a another Stock bios using nothing but the OS and max fans. -125/-125

 

is that 158W package power during CB15 with a -125mV voltage offset?

 

Yes. And that is actually quite low.

Was going to post a GTA Video, but it seems to only want to be 480P at the moment. Temps are well within reason using stock paste and max fans for cpu.

 

Last testing showed BSOD failure due to thermals @ 5.0ghz 9900k OC with CLC cooler, this testing replaces with a custom loop, testing @ ~03:56...conclusion @ ~10:25

Delidding 9900K - Is SOLDER Holding this CPU BACK!?
Tech YES City
Published on Nov 1, 2018
Today we go through a journey on before / after temps, and overclocks with delidding the i9-9900k. And this differs to the 8700k in one big way, and that is Intel is using Solder this time, so things are a little more difficult... Also shown is how to do this mod and what to use (things like coolaboratory liquid metal pro).

CPU VR Benchmark - 8700K vs 9900K - Does an upgrade make sense? [Pimax 8K][Virtual Reality]
VoodooDE VR - english version
Published on Nov 1, 2018
In this video I compare the 8700K CPU with the new 9900K CPU.
Does it make sense for gamers to upgrade? What about video rendering?
Benchmarks Start @ ~03:10, Results Review @ ~06:15

Spoiler: Click to See Video

The 9900K BUILD IS HERE - AND ITS PRETTY EXTREME
MXDOUT
Premiered 12 hours ago
The 9900K BUILD IS HERE - AND ITS PRETTY EXTREME - AND BUILD TUTORIAL

Spoiler: Click to See Video

 

The guy got a poor bin unfortunately. Seems to be a common theme among the "reviewers" especially those that didn't get it from Intel directly. Hardware Unboxed got his ES over a month before announcement. It's entirely possible their source provided ES versions that weren't final product SKU or the QS final chips. I've seen plenty of people online sub 1.3v, some around 1.2v like my chip at 5Ghz all cores. Most of these reviewers however have the ES chips and all have pretty bad temps. Intel even sent Joker Productions a second chip which he claims is a golden chip and he got 5Ghz at just 1.25v.

Also it seems this time around Intel is binning the 9900K vs the 9700K much more closely than the 8700K vs 8600K. Silicon Lottery is reporting the 9900K is overclocking better than the 9700K which is somewhat unexpected considering it should have higher thermals with HT. It seems the chips that can't make it to 9900K status with voltage and thermals are being made into 9700K SKU by disabling HT. It's possible they got an early sample that would have been a 9700K. This is the problem I have with reviewers using anything but a boxed retail sample. Go to the store, buy a chip or 2 and test it. Stop using ES chips as it's entirely possible they won't be indicative of the final product that consumers can purchase off the shelf.

 

is it possible to do a test with HT turned off to compare the temp/power with same setup?

also how come stock bios? didnt your 870tm have prema bios?

 

Seems Intel's supply issue on the 9900K is coming closer to an end quite quickly this year. Lot of users have 9900ks in hand, Amazon is shipping orders, BH is processing their orders, Microcenters have 9900ks on shelves. So unlike last year where the 8700K was much of a ghost for well over a month, it seems the 9900K is quickly going to be in the hands in consumers that want them. Hopefully shelves stay stocked longer than a few hours and prices will adjust accordingly for the holiday rush.

 

The from Intel evaluation parts are always suspect, so it's good to find reviews with retail (dodecahedren / OEM tray flat box) production parts.

It's always possible to get one just like his, the luck of the silicon lottery draw.

You got a good one, but that's also luck of the draw.

 

I deleted some posts, sometimes you just have to move on …

Charles

 

@Charles P. Jefferies

Glad it wasn't me this time.

 

Core i9-9900K, High-end Z390 VRM Temperature Test, Which Board Should You Buy?
Hardware Unboxed
Published on Nov 3, 2018

 

Amazingly my VRM temps shown by the sensor in the photo at 5Ghz are near bang on with what he measured with a FLIR camera.

https://imgur.com/a/sUGUSar

 

Check out the Hardware Unboxed review uploaded by them today that shows that the ASUS MAXIMUS Hero XI only has 4 + 2 phases not the advertised 8 + 2 phases - it's the worst VRM design of the bunch he's reviewed so far, the Asus Maximus Hero XI is the last introduced at 07:26 - but be sure to start at the beginning to get coverage of all the other MB's as they have much better VRM designs:

Core i9-9900K, High-end Z390 VRM Temperature Test, Which Board Should You Buy? <- Click on this link to go directly to the Asus Maximus Hero XI intro.
http://forum.notebookreview.com/thr...-lake-cpus-z390.811225/page-101#post-10816237

Also some Comments from the Tech Yes City's review you posted:

Pinned by Tech YES City
Tech YES City 3 weeks ago
"Guys made a huge mistake in this video, it is indeed the Hero 11 and not the 6 LOL.... I forgot my roman numerals - Cheers to OldNDirty for calling me out!"

SlavjanA 3 weeks ago
"It's a 4+2 phase doubled. Since the controller is most likely a re-branded (not an uncommon thing) IR35201 which only supports up to an 8+0 configuration, they wouldn't use more than 2 phases for the SoC, and (most obviously) you can see the doublers lol. That said, you're looking at choke temps. You need to be looking at the MOSFET temps, which HWInfo should provide as "VRM" temp."

felentus 2 weeks ago
"after looking through the z390 lineup, gigabyte seems to feature a 12 phase (doubled from a 6) using the same mosfets on a 170€ board (aorus pro elite) while this is 300€. Features (and depending on who you ask, bios) aside, I think that this is a bit dissapointing. In terms of VRM, I generally think that most of the z390 lineup (from msi, asus, asrock) to be overshadowed by the gigabyte boards this time, except for the extream high end."

Hyvaa Paskaan 1 week ago
"This is a 4 phase VRM board. Beware."

The ASUS MAXIMUS Hero XI a good example of why it's best to wait for all the products to get into good experienced reviewers hands to catch the problems before buying. Early reviews are often over exuberant from the flash of new cool products, but after that flash wears off the detailed examination often reveals buyer beware gotchas.

Intel used it as a review sample bundled for tech reviewers because of it's poor power delivery, limiting the full performance and power draw, showing better temps than a full power full performance motheboard would deliver.

All reviews with the ASUS MAXIMUS Hero XI are going to show skewed thermals due to the VRM limiting full power / performance.

I've already mentioned this in earlier posts.

http://forum.notebookreview.com/thr...e-lake-cpus-z390.811225/page-77#post-10812082

http://forum.notebookreview.com/thr...e-lake-cpus-z390.811225/page-87#post-10813758