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

I installed my 2nd EK PE 240mm Radiator. The drop seems to be about 10C. So prime 95 is running around 81C vs 91C and my gaming sessions dropped from mid 60's to mid 50's with overclock of 5.0 on the 9700K. I am going to add a 2080 GPU into the loop in the next few weeks. Been fun putting it all together. I think once i clean it in 6 months i will try some hardline tubing.

 

SiliconLottery isn't including delid with each binned CPU any longer, you have to specifically ask for the additional delid service, and as part of that service they give you an additional "delid binning" number - as if they were going to put it up for that speed for sale - what was your number?

If you bought a 5.0ghz bin, and delidded it, I would assume you got back a 5.1ghz or 5.2ghz speed bin result after delidding?

INTEL CORE I7 9700K @ 5.0GHZ BOXED PROCESSOR
$479.99
https://siliconlottery.com/collections/coffeelake-r/products/9700k50g

Optional Delidding:
Peak core temperatures under a heavy overclocked load typically decrease anywhere from 4°C to 8°C for Intel 9th generation CPUs. Delidding may take up to one business day to complete, depending on current pre-delidded inventory available.

Silicon Lottery Delidding and Binning
$39.99 Delid and Binning => 9700k $59.99 delid only, $79.99 delid + binning
https://siliconlottery.com/collections/sl/products/delid

Delidding with Conductonaut (73 W/m·K):

• The integrated heatspreader (IHS) is carefully removed from the CPU.
• Components under the IHS are coated with liquid electric tape, if applicable.
• Stock thermal interface material (TIM) is replaced with Thermal Grizzly Conductonaut liquid metal.
• IHS is sealed back into place, so the CPU can be treated just as if it were stock.

Binning:

• We put the CPU through our standard frequency tests to determine their bin, just like the processors we sell on our site.
• The maximum stable frequency we recommend for the processor when used with components on our QVL will be determined.
• Binning results will be included on your packing slip.
• For reference to historical bins, see here.

 

Doesn’t mean you have to leave it stock

 

Sorry, must have been something about the way i stated that comment. I did pay extra for the additional delidding. I do not see a sepearte delid binning number on the paper though. The paper reads:

9700K @ 5.0 GHz
S/N: M8LU546504335
Batch: L836E288
SKU: 9700K50G135VS

CPU Multiplier: 50
BCLK 100.0
CPU Vcore: 1.35
AVX Offset: 2

So hopefully they completed the delidding.

 

Emailed, and they sent me a new document with same information except just says 9700K @ 5.0GHz Delidded at the top but still the same multiplier, BCLK, CPU Vcore and AVX information.

 

Delidding doesn’t automatically mean you will get higher bin. The delidding of 9th gen soldered chips won’t get a huge temp difference. With older chips like the previous gen chips you could get 20C from delidding.

 

Ya i had a 6700K i delidded myself and saw around the same drops in temp. I did not want to delid this one myself as i would probably have damaged it. I still have that rockit delid tool. Tool made it so easy.

 

That's odd, you'd think Silicon Lottery would have at least a 100mhz bump up in performance from stock to delidded...also kind of surprised they don't give a package temp at those settings, for both binned before and after delidding.

Thank you for the details and update!

 

That's wild, Silicon Lottery on their Delidding page price is $59.99 for delid no-binning and $79.99 for delid with binning...the level above still says "From" $39.99 for "delid and binning", but the link for the Delid takes you directly to the delid page with the higher prices for delid with/without binning....
https://siliconlottery.com/collections/all/products/delid

 

Ya when i purchased I thought I saw some different prices than what I paid as well. May have just been a delid and they did not do additional testing. Just delid and sent it out.

 

I hope that pans out for ya!
Non Silcon Lottery off the shelf pick a box and open.
https://valid.x86.fr/daajf1

 

Very nice have not tried the whole validation just been working on a stable overclock. How about some prime 95 and some temps with that overclock?

 

This chip takes all of 5 minutes to stablize at 5.0 ghz.
Basically boot, set multi lower voltage and go.
But I will give you this hint. Do not expect to use the same voltages you do for prime95 or aida64 stability as you do for everything else.

 

My chip is not stable even in games at 1.305 Vcore at 5.0. For my chip around 1.33 gaming stability. Your right it does not take long unless you are trying to find the exact stable Vcore setting. I really don’t have much interest in getting a high validated score but I’m sure yours is on top as I have seen you get some pretty high compared to others on the board.

 

I get it. It's like I have Zero interest in stability at over clocking when I do not over clock to play games or do work. So basically a waste of time for me. I will however spend more time on memory over clocking as i'm trying to get that closer to 3800Mhz in a laptop. And that causes way more BSOD's than cpu over clocking.

 

That is a pretty nice validation though. Would be amazing to have a gaming rig running at 5384.

 

Thanks, but as you said...That is pretty useless, but it is fun to see.

You can, but it would cost a bit more in money & electricity to accomplish it.

The problem is you do not know your true top ceiling. And the only way to find out is to have your chip run at 50C or lower under load. maybe hang the rads next to an open window on a chilly night. Then you will see what the true voltage is. The hotter the chip the higher the voltage.

After it's all said and done, make your mods accordingly....

 

Yep, that's essentially what me and Falkentyne have been saying re VID and VCore. It's the VCore value that you need to keep an eye on - that's the voltage being fed to the chip. VID is what the motherboard uses (in some circumstances) to help determine what VCore to actually give to the CPU. If you set a manual static voltage then VID would have no influence. I believe if you were to choose an Offset Voltage, then the motherboard will apply your offset voltage to the VID and then load line calibration algorithm will then further modify to create the final VCore that is sent to the CPU - this is the mode I run on my PC. VID is therefore a 'guideline' created by Intel for that particular CPU you have, different 9900K's can have different VID values based on 'binning' that Intel have done for that 'manufacturing batch' of CPUs - they are aware that not all of their CPUs require the same voltage to be stable at stock operating conditions. (Someone correct me if you know I'm wrong in some of these details, because this is what I remember from reading around the subject, not looked it up now to verify).

 

@Papusan are you here?
Are my scores good?

 

Nice! Come on Falkentyne, tell us about your final build then! (You were debating on which cooler, etc.) (Ha, and what's your Vcore!?)

 

spot on from what ive seen at 5.0 ghz for a 9900k

 

Have it at 1.270v now, LLC Turbo. Got a BSOD during cinebench at 1.25v and LLC high, and WHEA correctable errors in Throttlestop TSbench at 1.25v LLC Turbo.

And Noctua NH-D15. Switching from the stock dual fans to the 3000 CFM fans dropped temps a few C. Still have no programs installed. Have to revalidate my steam folder on my platter drives.

 

That's cool, very interesting that you have the NH-D15 air cooler on it. Are you delidded with liquid metal? What are the extended long time running temperatures (say 30 mins) on something semi-sensible like an old version of Prime95 (choose v26.6 that doesn't use high power AVX instructions)? What VCore you at when the previous? What VCore you at when running AVX (e.g. latest version of Prime95)? What CPU watts are you at on old Prime95 v26.6 and latest version of Prime95? (I wouldn't recommend running the later versions of Prime95 for a long time, but v26.6 is ok because of no AVX). That's a lot of questions, but it definitely covers off my interests when it comes to your system so far!

 

I refuse to use AVX stress tests. I may try realbench later. Not going to degrade my chips anymore. Degraded two Sandy Bridge 2600K's doing AVX at higher volts. Never again. I'll let you know if it passes realbench. Prime95 small FFT with AVX disabled passes (prime95 29.8 with CPUSupportsAVX=0 and CPUSupportsFMA3=0). Throttlestop TSbench passes. It's a retail chip. I don't have the skills to delid it. Removing the IHS is easy. Removing the solder and polishing is not. Remember i have a metal bar in my spine. What's easy for most of you people here is very painful or even crippling for me, sometimes making me unable to move for days. I injured myself two weeks ago just repasting my r9 290X and cleaning the heatsink. Not asking for your sympathy but you need to realize that some people are doing VERY poorly out there.

Maybe if that Clevo Z390 comes out and I buy a second chip for it, I may experiment with a delid and cleaning of the solder. But if it took a healthy Der8auer a week to do it, it would take me much longer.

*Edit* realbench passed, no WHEA errors.

 

Yes, I don't really recommend running later Prime95 with AVX for long periods of time. What temps, VCore, and CPU watt consumption you at for non AVX loads in Prime95 after what length of time of running? In terms of AVX, I believe it is quite important to be AVX stable, a lot of programs/games use AVX. My CPU kicks up into AVX increased voltage when playing BF1 (some other games too), and also while virus scanning with Malwarebytes, so AVX is quite common nowadays. In terms of a non-aggressive AVX test have a look at the x264 program they've created here ( https://www.overclock.net/forum/5-intel-cpus/1570313-skylake-overclocking-guide-statistics.html), that's what I've used to validate my latest overclock, it's the same CPU watt producing load as non AVX Prime95. Click on the 'Stress Testing' spoiler tab in the first post, it's kinda hard to see.

 

BF1 doesn't run with reduced clocks when I set an AVX offset, so I'm pretty certain it doesn't run AVX code. Actually I don't know of any games that do besides AC: Oddyssey.

 

Hmm, interesting, maybe there's a threshold for the amount of AVX code the CPU is experiencing before it kicks it down to your AVX offset, BF1 won't be using AVX instructions for all of it's computation. My CPU kicks up to AVX voltage (1.368V) though when playing BF1, even though the load is 'only' 70W, but running Prime 95 v26.6 (non AVX) it will stay at the lower 1.344V even though it's running at a higher 100W. Running Prime95 later versions (incl AVX) or OCCT (also AVX), will kick it up to the same 1.368V but this time at loads of 130W. So, to me it definitely seems that BF1 is using some AVX - actually the average voltage during a gaming session sits inbetween 1.368V and 1.344V, but closer to 1.368V, so it's not permanently at 1.368V while playing, so I don't think it uses AVX all of the time, most of the time I surmise there is some AVX though.

EDIT: here's one guy with CPU kicking down due to AVX offset in BF1:
https://www.reddit.com/r/overclocking/comments/8dy3sf/avx_offset_being_applied_in_games/

EDIT #2: Maybe if your fps is 'only' around 60fps then the CPU won't be sufficiently loaded to bother downclocking with your AVX offset (although you're probably at 120fps considering your screen & specs). I'm running at 144fps pretty constant, so my CPU is highly loaded - the combination of relatively high load and presence of a portion of AVX might be enough to kick up the voltage to AVX levels, in my case. Perhaps not enough load on your CPU for it to bother with downclocking or perhaps just not greater enough proportion of AVX instructions to trigger a downclock (theories!).

 

This isn't really important.
Just because something is using AVX doesn't mean that the system is or is not stable.
No game or application in existence will draw what Prime 95 small FFT with AVX will draw. It's a complete utter power virus that has no basis on reality. If you can pass prime95 (CPUSupportsAVX=0) small FFT's and 1344K's without WHEA errors and realbench, then you should be ok to go. Even AVX with fixed size 1344K's is more than what a system should reasonably ever draw and thats the most I think anyone should even bother.

Don't do FMA3 small FFT's unless you enjoy degrading your CPU's. Or unless you need an extra house heater for winter time.

 

I think you quoted the wrong post of mine, that was a post in answer to yrekabakery's post. The post I quoted of yours is a bit further up, I was asking you various questions re VCore, temperatures, I won't write it out again, I'm not trying to get you to run AVX Prime, in fact I didn't ask you to run that in my last post & instead advised that I wouldn't run AVX Prime95 for a long time. Did you not want to answer my questions? (They're in my previous post to you). My questions are just a way of me seeing how your NH-D15 is handling the 9900K, and also to ascertain what kind of VCore you're needing for whatever clocks you're running - I'm just curious.

 

Sigh.
5 ghz, 1.270v, Cache=43, LLC=Turbo

Realbench I don't remember. It was easily below 80C I think
Throttlestop 8.70 TSbench was 81/80/85/81/85/82/79/80
Prime95 small FFT with AVX disabled: about the same as TSBench.
Prime95 AVX enabled (FMA3 disabled) 1344K in place FFT's: low 70's.
Cinebench (2 runs), Score 2216: 75/75/79/76/80/78/74/74

No WHEA Errors.

 

Well I won't press you anymore, because you obviously don't want to divulge info for whatever reason, what with your "Sigh" comment above combined with the overall 'strangeness/incoherence' of your previous post. But I will just point out that you haven't really answered my questions:

• length of stress run?
• CPU watts consumed during that run (can see HWInfo)?
• temperatures associated with the run?
• VCore?

I suppose you gave me temps for your TS Bench which I assume was the 1024 test, which is about only 80 seconds or so on a 9900K I would have thought. Cinebench you posted is even shorter. You didn't give me CPU Watts. And I don't think you gave me VCore, the 1.270V you say 'feels' like it's gonna be the value you asked for in your motherboard BIOS rather than the actual VCore, but I could be wrong. You're making it very complicated (for whatever reason), so I won't press you further. Pity, because I'm a fan of air cooling and you cooler is very similar to mine, I wanted to get a feeling for how it would handle the 9900K.

 

I answered everything in my last post. I'm a gamer, not an encyclopaedia.
The 1.270v is the REAL voltage. Gigabyte engineer MatthewH already explained this on the overclock.net forums. Sensor #2 is only about 5mv off from direct digital multimeter reading, with LLC set to extreme, true vcore is just at tiny bit higher than bios set vcore. I am using Turbo LLC, not extreme. Sensor 1 (mainboard) shows more vrise (1.284v to 1.296v). Sensor 2 shows 1.265v flat.

Prime95 AVX 1344K fixed FFT's was 135W. TSbench 8.70 1024M was 178W.

I honestly don't know why you are being so rude. I don't want to put you on my ignore list. You're literally acting VERY entitled instead of just saying "Thank you".

Also CPU Default VID is pre-programmed up to 5 ghz. My CPU default VID at full load with IA AC DC loadline set to "1" is 1.2695v. That means 1.27v is the default VID for my chip at 5 ghz. That's why 1.27v is stable (useful trick to know). This only works for chips that have not suffered degradation. Default VID stops scaling past the highest turbo multiplier.

 

What is this Time to move on.

Real world usage...
Blender v2.79 [Blender is a free 3D animation suite... (Good for testing—[Benchmarking], modeling, rigging, animation, simulation, rendering, compositing and motion tracking, even video editing, and game creation].

Demo file... BMW27_2.blend.zip

 

I'm sorry, but I see you as being the one to start off with the rudeness, what with ignoring questions and instead responding to a post of mine that wasn't sent to you and providing a response that was out of context/strange, and then thirdly your "Sigh" comment. I'm just curious, and you just seem....rude. I was rude in my last post because I had had enough of your 'behaviour'. I started this off a few pages back with a set of very simple & short questions, which you have managed to turn into a complicated mess with very little information in return, it really didn't have to be this complicated.

Thanks for providing the last set of information in the post I quote. NH-D15 seems to be coping quite well with the 9900K, although it's only at 178W there for you, so I'm not surprised given my NH-D14 can easily cope with 140W at lowest fan RPM from my small surface area 4 core chip. I wonder how NH-D15 would cope with a 220W load though, which is a widely quoted figure in reviews for stress testing on 9900K.

 

When you are asking someone to do something for you, you DEAL with their "behavior" otherwise you won't get anything because YOU are the one being needy.. That's how the world works.
Anything else is being entitled and selfish and elitist.
On to my permanent mute list you do.

 

Well I do see your point, but the recipient & responder has a certain responsibility too, it's not all one sided.

 

220W would be for a below average 9900K under load for anything other than kill your chip AVX P95.

I think hours of P95, Realbench Stress Test, and blender saw my chip hit around 170-200w max. I’m also at a super low voltage because I got a better than average chip.

 

My chp's default VID for 5 ghz is 1.270v, which is why i need that set in bios for stability.
just like in my testing on MSI throttlebooks, the default VID for each mhz step scales all the way from 800 mhz to the maximum turbo multiplier (in this case, 5 ghz 1 core, but doesn't matter how many cores are enabled). the load default VID is what is important.

quick and dirty default VID test with IA AC DC Loadline set to auto:
5 ghz load VID=1.271v.
4.9 ghz: idle vid 1.254v, load = 1.232v
4.8 ghz: idle vid: 1.224v load=1.173v
4.7 ghz: idle vid: 1.1945v load= 1.1637v
4.6 ghz: idle vid: 1.1694v, load=1.1320v

 

Yeah, so NH-D14/D15 should be able to cope with 9900K no problem, given that Falkentyne was in the 80 degC's at 180W on an non-delidded chip, and given that you say 220W is on the high side of normal. Would probably be ideal to delid the 9900K with liquid metal for NH-D14/D15 I would say. Not bad all in all, maybe I'll reuse my NH-D14 on the next CPU I buy - I hope so!

 

Just curious, does your voltage kick up in other games besides BF1? Because my VID is higher in lower CPU power loads like games, and lower in higher CPU power loads like non-AVX stress tests or video rendering. When running AVX code it does kick up though. It seems at least for me that in SSE/non-AVX workloads, my VID is inversely proportional to the amount of current being drawn. So it's hard for me to say that a higher voltage necessarily indicates a game utilizing AVX instructions. This could all mean nothing though because my laptop doesn't have an actual Vcore sensor plus other factors like different BIOS/LLC settings and VRM designs.

 

Vid or voltage, @yrekabakery ?
Remember VID and voltage are not the same thing unless you're using a MSI throttlebook.

VID is calibrated in 100 mhz steps from 800 mhz to 5 ghz. The IA AC setting is the operating voltage (boosted at high current) in mOhms, and the IA DC setting is the reported power consumption / VID setting to the MSR. If you are using manual CPU vcore override voltage, the VID shown will be overridden and ignored and only the CPU VCORE will be used instead.

My vcore is static at 1.270v, reported by the more accurate 2nd sensor at 1.265v. AVX or no AVX has no effect on cpu vcore.

If IA AC DC Loadline is set to 1, the CPU VID is reported much lower at idle than at load with the load VID corresponding to the true default vID. the gap between idle and load is much lower at lower mhz steps.

Example: 5 ghz: IA AC DC Loadline=1, CPU IDLE VID=1.208v. CPU LOAD VID=1.269v.

that tells me that my CPU's default VID for 5 ghz is 1.270v. Thus that's my stable manual override voltage.

 

VID, not Vcore. As far as I can tell, my laptop doesn't have a Vcore sensor. When using override mode, HWMonitor shows a value for IA and ring voltage that is always +30mV above the override voltage (so 1.205V when setting 1.175V), but it never changes under idle or load so I doubt it's accurate. From what I've gathered through testing, my VID at the highest stock ratio (46x) is equivalent to using an override of 1.215V. I'm currently using -40mV, which corresponds to an override of 1.175V, at 47x.

 

...or unless u actually have AVX use cases during day to day operations that actually need this kind of stress testing. believe it or not, YOUR computer usage does not equal everyone else's and in my case my cpu overclock was NOT stable and NOT ready for my everyday usage when testing with p95 v26.6. only p95 with avx gave me the stability i needed for all my usage scenarios.

admittedly though, i always do 1344k testing in p95, since that is much more relevant for testing stable cpu vcore than small FFTs

btw, how come ure only getting 135W on p95 avx? is ur cpu downclocking that severely? seems quite low for such a load at 5ghz

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

 

I've done some quick testing after seeing your post. It's sleep time soon, so I might update this post with more complete info tomorrow, but here's what I have so far:

• Prime95 (v26.6, non AVX): VID 1.189V, VCore 1.344V, 100W
• x264 (AVX load): VID 1.209V, VCore 1.368V, 99W
• Malwarebytes Virus Scan: VID 1.209V (some variance), VCore 1.368V (some decrease at low load)
• Dirt Rally: VID 1.209V (max), VID 1.091 (avg), VCore & CPU watts not recorded 'short run/bad data'.
• BF1: VID 1.209V (max), VID 1.204V (avg), VCore 1.352V (avg), 62W (avg)

In my case VID is not inversely proportional to the amount of current being drawn, and in the case of non AVX Prime vs x264(AVX) you can see VID gets kicked up a notch for the AVX loads even at the same CPU package watt consumption. Likewise Malwarebytes virus scanning makes significant use of AVX as seen by spending long periods of time at the high AVX style VID voltage (and likewise the high VCore too). Dirt Rally uses a very small amount of AVX by the looks of it, way less than BF1. BF1 uses a lot of AVX by the looks of it from my results above, as pretty much always at AVX voltages on VID & VCore. So yes, different games seem to use different amounts of AVX, if we use VID and VCore as a guideline from my results. And an everyday desktop program looks like it makes heavy use of AVX, namely Malwarebytes.

So this is why I think some form of sensible AVX stability testing when it comes to overclocking is sensible - I think it's used way more than a lot of people think.

EDIT: @yrekabakery , I updated my post with BF1 testing today, definitely uses a lot of AVX as VID and VCore mostly kicking up to AVX voltage levels.

 

Wasn't downclocking at all. I'll check again later. Might be some other protection setting I didn't disable. Right now I disabled hyperthreading (just a few minutes ago) to run CB so I can't check right now. Could also be some setting in the bios even though I have ICCMAX set to 255 amps. Let you know in a bit.

Ok my wattage memory was wrong. Sorry.
Prime95 1344K AVX is reporting 155W.

 

So this is funny.
Prime95 1344K fixed FFT with AVX disabled is running hotter and using more power (158W) than 1344K Fixed FFT with AVX enabled.
Also Throttlestop TSbench uses more power than any of those, only passed by small FFT (AVX disabled).

I give up. There's no throttling either.

 

VID, yes, vcore, no.
My VID at 5 ghz with default IA AC DC loadline is 1.39v at full load (!) with it about 1.32v at idle.
vcore is steady at 1.265v (bios is set to 1.270v, LLC=turbo).

With IA AC DC loadline set to 1 (0.01 mOhms), then the VID is 1.21 at idle (droop for some reason) and 1.269v (very close to my actual vcore setting) at full load.

That's how I know the default pre-programmed VID for my CPU is about 1.270v at 5 ghz, so I set my vcore there and it's stable. But remember what I told you about MSI trashbooks? You can only find the true default VID by setting IA AC DC Loadline to "1". If it's at auto the VID will skyrocket.

 

Why are we even mentioning VID? Who cares, VID is nothing more than what the CPU is asking the motherboard for at a given clockspeed. It means nothing in reality, might be useful in defining chip quality but even then the proof in the pudding and you need to test it real world. Screen shots of Vcore as measured by Hwinfo64 with CPU-Z open as well seem to take care of business.

 

VID is important.
You can use the cpu's default VID to show the voltage it needs at a turbo boost setting, up to 5 ghz. Saves a lot of time!
My CPU needs 1.27v to be stable at 5 ghz with HT on, and 1.25v to be stable at 5 ghz with HT Off. Got those numbers by setting Core IA AC DC Loadline to 1, then doing stress tests, which matched up nicely with the VID shown at full load. (this won't work if IA AC DC loadline is set to auto; the VID will be up to 100mv higher than shown, depending on current and base clocks; the higher the current, the higher the VID boost. Just like on a MSI throttlebook).

My results:
HT on: VID=1.2695v maximum. So I need 1.27v bios voltage for stability.
HT off: VID: 1.2540v. My 1.25v bios voltage setting is stable.

At each mhz below 5 ghz, in 100 mhz steps, the VID will drop substantially. So you can get a nice ballpark what you need without wasting tons of time.

It's a very fast way to see what starting voltage you will need for your chip (up to 5 ghz or whatever the maximum turbo boost is. It stops scaling past maximum turbo boost).

 

But once you do that, stop reporting things in terms of VID and use Vcore. VID is fine if your laptop doesn't have a vcore sensor and is using VID instead.