Intel Core i7-8700K Coffee Lake Z370 and Z390

reviewer is nothing but a title. making a video reviewing doesnt necessary mean they are more knowledgeable. your concept is flawed. just stop it already, making yourself look immature.

i saw your benchmark 4.9 at 1.325v thats pretty decent. a good chance for 5 under 1.4 but thats only for CB which doesnt use avx though. on der8uer w/e his username is, they selling 5.2ghz under 1.4v +/- 30mv but cost a fortune.

 

4.9ghz right now seems to be my sweet spot. I’ve been daily driving at 4.8ghz though. 5.0ghz boots fine and completes the bench but it’s obviously wanting more. Could very well be the board and it’s less than top tier power delivery. It’s not bad just not made for extreme overclocking and benching IMO.

5.2ghz under 1.4v is a golden sample.

 

I’m using the Corsair h100i V2. Thermal grease is PK-3. Pretty standard stuff.

 

Love your rig man, really do. I was wondering why the CPU was so cool on OC. Do you plan on SLI anytime soon?

Also could you give me some feedback on how the grease was compared to eventual old ones you have used in the past? I was considering the TK-3 myself since it's rather thick and probably really good for notebooks.

 

Thanks bro @Talon Nice job!! I have what I need now.

4.7GHz and 4.8GHz all cores is bankers in P870TM1 Delidded and with Liquid metal.



4.9 and 5.0GHz all cores will be Bench clocks. Maybe even 5.1GHz with a very good chips.

 

these are benchmarked from within a 870TM1? not bad for the temp. was hoping a bit more for heavier benchmark like intel XTU memory stress for higher temps, or anything that runs at least avx workload, those run super hot.

my firefox on avx/avx2 loading 3000+ tabs with v45 is more stressed than intel xtu memory bench, it often crashes when xtu doesnt passes lol.

where you got these results?

 

I looked on @Talon's numbers. I was interested in the power concumption numbers with fixed clocks. This tell me everything what I needed to know!!

 

Glad I could help you out man! Keep in mind that was with me just going through, selecting the core speed with the Asus software and setting a voltage that I figured would work based on my testing at 4.8Ghz over the last few days. There is definitely some more wiggle room for tweaking voltage down. This chip inside of that laptop is going to be beast mode, especially once you guys get your hands on it and really start tweaking it.

 

I've used a lot of pastes and found the most of the top tier ones all perform pretty much the same within a couple degrees. MX-4 IMO was the absolute worst paste I've ever used. It produces the worst temps and seems to dry out. ShinEtsu and PK-3 have been my favorites with great temps. ICD7 is also great for laptops with poor gapping between the die and heatsink or a slightly warped heatsink. Never heard of TK-3, I will definitely check that out since thicker pastes seem to work better on laptops.

Never going SLI ever again. In the past it worked great, had 6800GTs and 8800GTs in SLI once upon a time. My last trials with SLI were 970 SLI and 980 Ti SLI, both of which were promptly removed and upgraded with 980 Ti and 1080 (then 1080 Ti) respectively. YMMV but SLI and Crossfire today just don't produce the same results they once did. Hell even AMD has retired the crossfire brand. While they still support it under a different name, I think that move alone is telling.

 

As my old P870DM1 Cpu heatsink can handle +131w bench load. I expect the improved Cpu HS in P870TM1's should at least handle +-150w load who is like 8700K@+5GHz in bench (Of course not for 24/7 ).

 

131w would that be max fan? because new heatsink has similar dissipation, however its capable of holding more heat before capacity hit its limit.

btw does prema's unlocked bios have something that auto downclock a set multiplier when cpu is under avx load?

 

Maxed fans and my U3. It’s around max what’s can be done with this heatsink. Yeah, Bios has the avx offset functionality. Never used it.

 

makes me wanna buy one and try the avx offset.
btw, 131w max fan at what speed of cpu and what temperature?

 

[email protected] and I mean it was 80-82C. About 15C ambient. You find it in the Clevo OC thread (I’m on my phone). Difficult post links

 

HW News: Microcenter Loses Its Mind, Coffee Lake Prices & Thermals

 

Seems like there are some issues with the 8700K on the clevo notebook:


Guess people will have to tinker with this again until it runs properly.

I don't know why clevo just doesn't finally get ryzen into their notebooks....

 

Why post old info? And why trust old info? And why can’t you find new info? Difficult? I don’t get it!!

 

That would be a disaster. Having a CPU that struggles with 4.1GHz would suck. But, it would also solidify my decision to abandon notebooks. Owning a belly-button laptop computer that runs more or less the same as all the other cookie cutter crap would be as boring as watching paint dry. At least what we have can be made better with tinkering and tweaking. That can be fun. Click and run crap is never any fun and that would really suck. Running benchmarks with scores that are basically the same as everyone else's, or only a small fraction better, is also no fun.

 

im sure with the new cpu heatsink similar to that of p570wm + additional fan will do the trick. 6cores at 5ghz is hard to reach without a golden sample, but 6 cores at 4.8ghz seems very possible.

 

Dude, Have you skipped my post #105 ? Why should I post proper info if you don't bother read it? JK

 

those are talon's benchmark with desktop cooling right?

 

That heat sink is nothing like the P570WM heat sink. I bet it weighs less than half what that P570WM heat sink did. And, the P570WM ran super hot when overclocked, and even with AC cooling that was barely enough to avoid thermal throttling.

Edit: I still have a spare that does not fit right that I never threw away. It weighs 1 pound 4.7 ounces, LOL. Here are some pics of it.

 

Yees, But the shown power concumption numbers, tell me that those clocks should be easy to reach.

@Mr. Fox Remember the wattage numbers in bench? This is what's importent. Can the TM1's Cpu HS remove +150w heat, we will see +5.0GHz in bench with all 6 cores.

 

lol heatsink pr0n. i don't own 870 so i donno the size of the current heatsink.
its really sad, the marketing nowadays, especially msi, love to advertise how much heatpipes they've got, but the design end up killing the hype most of the time. though i must admit MSI got good work on their GT83VR GPU heatsink, too bad the vbios/bios are trash.

its not the clock, because his temp wont reflect accurately to what we would expect in 870TM laptop correspond to the same voltage/clock.

 

Interesting to note:
Ryzen 1600 at 3.2 GhZ is only 10% slower than Intel's 8700k on 3.7 GhZ.
I wonder what would happen if 1600 is clocked at the same speed.
Oh wait... here we go:


Scores 1236... which is pretty much the same.

Also, Asus created an all AMD laptop (ROG series) and offer Ryzen 1600 and 1700 (with RX 580) as options.
My guess is that since 8700k has 95W TDP, they are having cooling issues and cannot run past 3.7 GhZ without throttling.
It is after all a 6 core chip.


But Ryzen 1700 probably can't clock at 3.7 GhZ across all 8 cores if it wants to stay in the 65W TDP - that probably leads it into 1700x territory and 95W.

I'm actually confused about Ryzen boost clocks.
What do Ryzen 1600 and 1700 boost to across all cores?
The boost clocks state 3.6 GhZ for 1600, and 3.7GhZ for 1700, but its not clear if these frequencies are for just 1 core, or all 8 cores?

 

We also have to remember than 4960X in my P570WM pulled nearly 300W in Cinebench at 4.7GHz and 8700K won't hit that many watts.


And, even it easier benchmarks (like those below) it still easily pulled 200W. Was pretty awesome, actually. I love watts. The more the better. I forgot I was pulling almost 200W each with 980M. Such great memories. That was one helluva laptop in spite of being boxy and hot. Too bad it can't run 1080 SLI.

 

If the old DM1 Cpu heatsink can handle 130w load in bench from a smaller 4 core die, I expect the upgraded TM1 cpu heatsink will do a better job on the bigger 6 core die. Watt = Heat.

Where have yoo been?

 

Around. I'm just pulling conclusions based on available data.
Do you have any other data to contribute (besides sarcasm?)?

 

Haven't the P870TM1 My conclusions is based on what I have seen. Read the thread. As well the i7-8xxx Coffey Lake rewievs.

Edit. For Ryzen boost clocks Click on each model.

 

i speak overclock language, i dont speak stock language, not here.

jokes aside though, for efficiency, ryzen currently hit maybe 3.7-3.8ghz on all cores, after that we are looking at throwing it's power efficiency advantage out the window.

those cpu wattage talk isn't same as TDP though. 7700k rated at 91w thats at it's specified clock, non-avx workload. once we overclock on any cpu, ryzen/intel TDP value goes out the window. @Papusan manage to pull out 131w on 4 cores for just a CB test, i have no doubt if its P95 avx it'll go 180w easily, which is also my major concern. 5.1 at 1.35v a bit high for my preference, i'd go say 1.3v max or lower, which is why i tried to find his 5ghz and 4.9ghz with less voltage/temp screenshot but can't find it now. share your wisdom once again papusan!!

 

Here - Here

 

Here is 128W and 1.340V (load voltage) at 5.2GHz.

 

dang nice result. what ram are you using, 64gb at 3000mhz which brand. also that ring bus clock, sexy, does prema bios include all those to overclock the uncore/cache etc?

wth i just re-read your post and you said its 870TM with 8700k benchmark? how did @Talon grab hold of a sample prior to its release!! or was that his desktop benches and you're just using his voltage/clock comparison in a laptop?

 

delete double post..

 

I asked @Talon if he could run benches with 46-47-48-49 and 50x. Then I find the Power concumption numbers the cpu pulled in Cinebench. I needed to know those numbers. Then I could find out what could be possible in the new Clevo P870TM1. Without numbers, You only look in the blind, bruh And I'm quite sure the TM1 Cpu HS can handle +150w.

 

yea for your preference it'll prob work, i was looking for heavier benches. i'd say he has a decent chip, we dont have enough samples to really say for sure but i'd say its above average. toms hardware is 49x at 1.23v as I asked paul reviewer, which is better than talon's chip.

my workload would be something like 60-70% fan speed with avx workload. that would mean normally 1.3v will now need 1.34v or more at the same clock speed, imho CB workload is really light, compare to other ones out there, even many games are heavier than CB15 bench.

a realistic one incase of heavy use would be something just a bit heavier than intel XTU memory bench, really dependent on individual use case.

for example in @Mr. Fox hwinfo screenshot cpu used 1.34v for 5.25ghz on a CB run, but looking at his maximum 1.4v+ i donno if thats from his trial run at 5.3 or 5.4, or just a heavier bench at 5.2 needed more voltage. maybe mr fox can share some light.

 

Remember @Talon's chips ain't delidded and no LM. The lower temp the bigger chance for lower voltage. At least will temp go down with Delidde and LM.

 

Gamers Nexus reported a -20C temp delta when they delidded with LM. I have a tube of unused LM sitting in the drawer next to me, I am extremely tempted. If this damn CPU wasn't in such short supply right now I would probably go ahead and do it with my desktop since it's obviously a stationary device. With the power draw numbers I'm seeing, and the results you and @Mr. Fox reporting there is absolutely no concern that beast Clevo won't be able to cool it properly even when overclocked.

@ole!!! Honestly bud I was just taking a break from a late night of gaming, hit up Reddit and saw the in stock thread that was like 10 minutes old, went to Newegg and immediately checked out with overnight shipping and rush. It had shipped like 6 hours later and got to my door step the next day. I grabbed the motherboard early in the AM on the 4th from ShopBLT, also overnighted. I had my parts by 930am on the 5th. Pure luck, and well impatience got the better of my wallet.

 

Rule of thumb when ordering new hardware in short supply, always buy 2.

 

They would only let me buy 2. Honestly I figured it would sell out, just not this bad. If I had a crystal ball I would have grabbed the wife's wallet and purchased a second and shipped to her parents house near by.

 

Thanks, bro. Yes @Prema BIOS has settings for cache and uncore, System Agent and IA domains, and VDDQ/VCC IO. I am using G.SKILL 3200 16GB*4 RAM. Depending on what I am doing, 3000 is sometimes better than 3200. It is also mostly stable at 3333. But, going past 3000 has very little to no effect on benchmark scores other than RAM benchmarks. Cinebench and Futuremark results are no different between 3000, 3200 and 3333. Most of the time I leave it on 3000. Micro$loth Outlook and running YouTube videos are fine at 3000, but they sometimes get kind of sketchy with 3200 and 3333.

 

Ok I have finally figured out why my CPU is boosting itself to 4.7Ghz on all 6 cores "Stock" with optimized bios defaults. It also explains why to me when I first booted up at BIOS defaults why the voltage is already at what I considered an high overclock setting. In addition now we have an explanation why there is such a large variance in the performance different reviewers are seeing. It seems Intel has given motherboard manufacturers the ability to either default 'Multi-core Enhancement' to On or Off as "default".



Therefore my "stock" settings are going to be rerun with this option disabled to get a true performance, temp and power consumption number for you guys.

 

Actual stock score.. I should caught this earlier but better late than never. On flip side stock voltage now much lower, and therefore power consumption very reasonable.

https://imgur.com/a/9dIdL

 

Exactly where the scores should be. And Load voltage in Bench and what memory speed?
Could you undervolt down to lowest stable level for stock clocks? Then run Cinbench R15 and wPRIME v1.5 (Extract and run as adm. Select 12 threads in advanced settings. Run 1024M Stress test)

Do the same for 4.7+4.8GHz for all cores as well. Thanks, bruh

Especially the Wprime 1024M Stress test is of valuable info.

 

Awesome, glad you found it

That's why I post this info:
http://forum.notebookreview.com/thr...-coffee-lake-z370.809268/page-6#post-10612102

There's almost always some tidbit of info in every article and video, even when there isn't it's just one more step on the way to understanding and mastering the subject

 

I put it here bruh @Talon

 

Delidding a $1000 CPU - Worth the RISK??

 

From Gigabyte:
http://forum.gigabyte.us/thread/2303/intel-turbo-policy-faq

"With recent discussion about Intel Turbo on the new Coffee Lake CPU's we decided to make a quick FAQ to explain what Turbo is, what it does, and our stance on how to utilize it on our boards.

Before we get into the FAQ portion I wanted to make sure everyone is on the same page with the terminology being used. The goal here is not a deep dive into specifics, but rather a baseline for discussion.

• Default: Default is the 'out of the box' state. Another way to put it is the state without user intervention.

• Auto BIOS setting: AUTO is not a static state. It can vary depending on the inputs. A good example is AUTO Vcore. At default the AUTO Vcore will be lower than lets say manually setting your multiplier to 48x. Lets say at default your CPU Vcore is measured at 1.3V. By raising the multiplier the measured Vcore may raise to 1.4V. In both instances the setting within the BIOS remains "AUTO" however the actual output varies.

• Spec: Spec is determined by the manufacturer of the product. In this case the spec for the CPU comes from Intel. Similarly the spec for RAM comes from the RAM manufacturer.

• What is overclocking?: Overclocking by definition is running your system out of spec to gain performance.

• Q: What is Intel Turbo Boost?

Intel Turbo Boost is a feature that allows specific cores to clock higher depending on the load. The less cores that are active the higher the CPU will clock the active (load bearing) cores.

• Q: Why is Turbo necessary? Why not just run all the cores at X frequency?

The answer to this is two-fold. First running a CPU at full throttle when not being utilized created unnecessary power consumption and heat. Second not all cores are created equally. The "silicon lottery" we often refer to goes beyond the CPU as a whole. You may have core A that can hit X frequency without issue, where core B cannot. Turbo determines your strongest cores and assigns the higher frequency to that core.

• Q: What is "Enhanced Multicore Performance"?

Enhanced Multicore Performance (EMP) is a feature in the bios that pins all cores at the max turbo frequency under any load. IE whether an application uses one core, two cores, or all cores, they all run at max speed.

• Q: What is the default state for EMP?

By default EMP is set to AUTO, which defaults to OFF. We believe that defaults should reflect the specifications of the manufacturer. We offer tools to tweak and adjust these specs, however by default they should run as designed.

• Q: When is EMP enabled if I leave the setting on AUTO? What is the reasoning behind this decision?

EMP is enabled (IE "AUTO" now outputs enabled) automatically when a user sets the multiplier manually. Setting a multiplier for a CPU indicates a user wants to overclock the system and set the CPU to run at a specific frequency. Thus we force all cores to run at the user defined speed.

• Q: What if I want to keep the Turbo policy, but alter the speeds?

If your intention is to change the Turbo table, but keep the dynamic adjustments to speed, simply go into the BIOS and change the per core policy under Advanced frequency settings -> Advanced CPU Core Settings

• Q: Why not enabled EMP when I enable XMP? Isn't XMP overclocking?

There was much discussion on how and when to enable EMP. We feel that EMP falls into the category of overclocking, and thus requires some level of user input to enable. One can debate whether or not XMP constitutes overclocking... We feel that a user enabling XMP does not necessarily constitute a desire to overclock. Rather the user is trying to run their memory at the rated speed by the RAM manufacturer.

If you have any suggestions on additional FAQ items, have any questions, or want input into how we shape our BIOS policies let us know!"

 

8600K vs 7700K - Better CPU for Adobe Content Creation?

 

This was a pretty good online example (especially near the end of the video) of a 'workload/workflow' as I define them.

With a total workflow of ~4HRs a much lower end new/current platform delivers ~8% time savings (~17 minutes). In an 8 hour workday that is over half an hour saved (equivalent of effectively increasing my hourly rate by the same (~8%) time savings in $/HR.

In a 16 hour shift (not uncommon for me...) I would have saved around an hour per day - each and every time I used this system (note: stock settings only) - easily worth buying this platform at almost no matter what (total) cost.

While the overclocked results are tempting to chase (effectively double the time savings; ~16%) - when (not if) the system crashes, glitches and/or flat out dies in the O/C'd setup any productivity savings are quickly lost just trying to get the system back up to usable...

To do an extreme example:
Assuming a 5 year lifecycle @ 24/7/365 operation, there would be a savings of 3,102.5 HRs over the lifetime of this platform vs. the base i7-7700K we're assuming we currently have/use.

While just over 129 days seems like a good cushion to have for repairing an O/C'd setup over the course of 5 years; it doesn't take into consideration the time/$$$ needed to do those repairs, the loss of customer good will (if a job is late/delayed or has penalties attached) nor does it account for the fact that it may flat-out die too (as easily within the first week as it can in the next few years of it's lifecycle.

With the logic presented above - it is easy to see why the i5 platform above would be immediately skipped for the 8th gen i7 platform instead (assuming that testing actually showed gains over the i5, of course) - at almost any cost.

Five seconds, 1 minute and a handful of minutes in short/quick bm 'scores' may seem insignificant by themselves. But when applied to a real-world workflow/workload; those seconds add up and become significant to an individual or a company pretty fast.

Even for a non-paid workflow, wasting time by saving a few pennies a day (again; assuming your budget allows you to consider a higher-end platform...) over the course of ownership is not a wise course in the long run. As the video suggests; those cumulative seconds saved can easily be better used for other activities by all.

hmscott; I don't have time to see all the videos you post (not enough time it two lifetimes! ) - but I'm glad I made the time to see this one. Thanks.