9900K/9700K Overclock performance review

[martin-w 3,238]

Seems like the consensus so far is that 9900K was pushed to the limit to compete with AMD. Thicker die due to danger of cracks forming, softer and thicker layer of solder to avoid the solder fracturing. Result... higher temp. 

[Ray Proudfoot 8,973]

The consensus here seems to recommend HT should be disabled. Given the main difference between the 9900 and 9700 is the latter does not have HT might that be the better choice?

[joemiller 269]

18 hours ago, w6kd said:

The 8086K is a pre-binned 8700K.  No difference in the circuitry...these chips are cherry-picked off the production line, and overclock to the chip's expected limits (5.0-5.3 GHz) very reliably, and at lower-than-average voltages.  There's an outfit out there (Silicon Lottery) that tests and sells binned chips, and they reported that 100% of the 8086K chips they've processed overclocked to at least 5.0 GHz on all cores.  Many of the 8700K chips will, too, although I have also read a few articles that suggest since the 8086K hit the scene the percentage of 8700K chips that will overclock that high has dropped, which makes sense given that they're grabbing the better silicon and rebranding it.

 

I am not going to lie, as I haven't learn how to, yet (lol). But I am itching all over to go out and get the 8700K and a strong motherboard- TODAY! 

[fghdgdfdfgfgf 111]

Don't want to turn this into a overclocking thread but I do want to keep it real.

Last nite I posted a hwinfo image that had errors in it ...yeah, I tried to shoe-horn 17-17-17-38 memory timings into the 5.2 setup, and has always been the case with those timings it bite me in the rear (been OCD with those timings ever since I got the kit).

Anyway, new error free image with relaxed timings (17-18-18-38). The below image caught this moriining LAX to SFO ....over continuous cloud cover ....less cpu and more gpu on this flight.

 

https://i.imgur.com/IAx3tZX.png

[331BK 96]

i was also waiting for the 9900k but for the current prices here in Austria i can get a TR 2950X and new board.  

I know i have only 4,4 Ghz but 16 Times and when LM hopefully improves Multi Core Handling then i have more then enough power. 

But still i cannot completely ignore the 9900k.  I know Intel is the way to go but double cores for "same" price?

Edited October 22, 2018 by 331BK

[martin-w 3,238]

3 hours ago, Ray Proudfoot said:

The consensus here seems to recommend HT should be disabled. Given the main difference between the 9900 and 9700 is the latter does not have HT might that be the better choice?

 

HT off with chips that have that feature, or same architecture but no HT usually results in 10 degrees lower CPU temp. But wouldn't be much difference in single core performance compared with the 8086K or 8700K.

So then we have cost... £393 for 8086K. Whats the 9700K going to be? £499? Just over 100 quid spent for nothing! That's the Scan prioce for 9700K. No idea if they will be cheaper else where.

That's my opinion, but then if someone has no budgetary constraints and simply wants the latest, fair enough.

 

9700K has less L3 cache of course. 🙂

Edited October 22, 2018 by martin-w

[martin-w 3,238]

2 hours ago, joemiller said:

I am not going to lie, as I haven't learn how to, yet (lol). But I am itching all over to go out and get the 8700K and a strong motherboard- TODAY! 

 

You naughty little boy Joseph! In my experience an itch is usually scratched. 🙂 

[lownslo 592]

21 hours ago, w6kd said:

There is enough uncertainty as to whether the average i9-9900K will reliably reach that 5.0 mark when running all cores without overheating, so I punted on waiting months to find out when I found an 8086K on sale and on the shelf at an area retailer.  A bird in the hand...

I have three sets of 9900K/XI Hero on order... two will be built on spec for clients, the third mine.  But a few minutes ago a vendor called and offered me an 8086K for $375US.  What concerns me most about the new 9000 series is that soldered IHS... Martin may have been joking about the new solution being a step back, but there may be much truth to his perspective.  Time will tell.  And I'm ready to build a new system for myself now!  Given the IPC of the 8700K/8086K is within 1/2% of the 9900K, and given that the 9900K will probably top out at 5.0-5.1 all cores, and given that I can get a binned 8700K (the 8086K) for almost 30% less than the 9900K, that 8086K is looking very tempting... 

Greg

[Ray Proudfoot 8,973]

1 hour ago, martin-w said:

 

HT off with chips that have that feature, or same architecture but no HT usually results in 10 degrees lower CPU temp. But wouldn't be much difference in single core performance compared with the 8086K or 8700K.

So then we have cost... £393 for 8086K. Whats the 9700K going to be? £499? Just over 100 quid spent for nothing! That's the Scan prioce for 9700K. No idea if they will be cheaper else where.

That's my opinion, but then if someone has no budgetary constraints and simply wants the latest, fair enough.

9700K has less L3 cache of course. 🙂

There are no budgetary constraints within reason if the increased performance justifies it. There’s around £200 price difference between Scan and Chillblast for roughly the same kit with Scan cheaper.

After more  checking I’m not convinced the i7-9700K is any faster for P3D than the i7-8700K.

Edited October 22, 2018 by Ray Proudfoot

[martin-w 3,238]

55 minutes ago, lownslo said:

 Martin may have been joking about the new solution being a step back, but there may be much truth to his perspective. 

Greg

 

de8auer's hypothesis makes sense to me. Thicker die (for whatever reason - perhaps to counter the tendency for a thinner die to crack when soldered) thicker, softer layer of solder, to combat the tendency for a thin layer of solder to crack due to temp differential between the silicon die and copper IHS. Seems reasonable to me that a die that's twice as thick as the 8700K die and gobs of solder would result in higher temps. The heat has to pass all the way up from the bottom of the die through the die itself, through the thick layer of solder to the IHS. The thinner those layers are the better.

I suspect daring enthusiasts will be doing exactly what Roman did and delidding them, grinding down the die to the depth of the 8700K die, and applying good old liquid metal. It will be the new thing for scary people to do! They will be then boasting about their low temps and awesome overclock's. It may be that all the solder has done is make it harder to delid.  😁

I wonder how this guy would perform with direct die contact?

Edited October 22, 2018 by martin-w

[lownslo 592]

55 minutes ago, martin-w said:

de8auer's hypothesis makes sense to me. Thicker die (for whatever reason - perhaps to counter the tendency for a thinner die to crack when soldered) thicker, softer layer of solder, to combat the tendency for a thin layer of solder to crack due to temp differential between the silicon die and copper IHS. Seems reasonable to me that a die that's twice as thick as the 8700K die and gobs of solder would result in higher temps. The heat has to pass all the way up from the bottom of the die through the die itself, through the thick layer of solder to the IHS. The thinner those layers are the better.😁

 

I too agree.  Steve Walton over at TechSpot just posted this article: https://www.techspot.com/article/1732-solder-vs-paste-core-i9/

Greg

[martin-w 3,238]

48 minutes ago, lownslo said:

I too agree.  Steve Walton over at TechSpot just posted this article: https://www.techspot.com/article/1732-solder-vs-paste-core-i9/

Greg

 

Interesting article. The bottom line is that for clocks around 5 GHz... soldering offers no benefit. 

[lownslo 592]

Aaaannnd... an 8700K/8086K after delidding may have an advantage overclocking above 5 GHz (like... 5.2-5.5 with good cooling).

One of the things that concerns me about the new STIM solution is that Intel can (and possibly... OK, I'm just gonna say it, probably will) change how they solder their current IHS/CPU combos.  And doing that may change how easy/difficult delidding those critters will be.  One can only wonder how many consumers will end up with broken dies then.  Additionally, there will be "the new thing scary people do" as you noted... "sanding" the die top to be thinner.  UGH!! 

Like most folks I like the idea of a soldered IHS but given that this first round hasn't been received with any great enthusiasm, I will probably build a system based on the 8086K offered to me.  At the very least I can wait a month (my return period for the 8086K), learn more about the 9000 series by following the forums and building those two spec 9000K systems, and then make a decision.

Greg

[Bob Scott 7,721]

On 10/21/2018 at 12:40 PM, shivers9 said:

Bob I am ordering the 8086 today. Just makes good sense for P3D. What Mother board and Memory did you decide on?

Z390 Asus Maximus XI Hero for the motherboard, GSkill 3600 MHz CAS 15 (F4-3600C15D-16GTZ) for RAM.

Cheers

[martin-w 3,238]

12 minutes ago, w6kd said:

Z390 Asus Maximus XI Hero for the motherboard, GSkill 3600 MHz CAS 15 (F4-3600C15D-16GTZ)

Cheers

 

Good choices.

 

I know we enthusiasts like to overclock manually, but interestingly Asus have significantly updated the Five Way Optimisation, auto overclock utility for Z390. Still based on a stress tests, so system specific, rather than a generic fixed overclock, but based on a new algorithm, and they have added a variable frequency and voltage feature, so it monitors constantly and compensates for ambient temp, dust build up on coolers etc. Asus say it will overclock a CPU exactly the same as an experienced overclocker would, same voltage max frequency etc.

 

Edited October 22, 2018 by martin-w