9900K/9700K Overclock performance review

[shivers9 1,174]

1 hour ago, martin-w said:

so it monitors constantly and compensates for ambient temp, dust build up on coolers etc

LOL that sold me on the z390. How can you turn down Auto overclock that is kitty cat resistant. My new 8086 build is in transit as we speak. Still up in the air as to 1080TI or 2080 for GPU.

Edited October 22, 2018 by shivers9

[brucewtb 249]

Deleted - previous content posted in error.

bruceb

 

 

Edited October 22, 2018 by brucewtb
Repeat posting error

[brucewtb 249]

Interesting video from Hardware Unboxed  suggesting that the Asus Maximus XI Hero had too weak VRM for the 9900K ( and presumably the 9700K) at only 4 phase.  Very odd as this is not a cheap board.  Hardware Unboxed  also suggest that this is the reason some reviewers are reporting much lower temps and power consumption when testing with 4 phase MBs - they are throttling the 9900K so it runs at a much lower frequency.  Thoughts anyone?

bruceb

 

 

 

 

 

[shivers9 1,174]

56 minutes ago, brucewtb said:

Interesting video from Hardware Unboxed  suggesting that the Asus Maximus XI Hero had too weak VRM for the 9900K ( and presumably the 9700K) at only 4 phase.  Very odd as this is not a cheap board.  Hardware Unboxed  also suggest that this is the reason some reviewers are reporting much lower temps and power consumption when testing with 4 phase MBs - they are throttling the 9900K so it runs at a much lower frequency.  Thoughts anyone?

bruceb

 

 

 

 

 

Well my thoughts are that I am glad that I decided to stay very very far away from the 9xxx mess. If it is true that the Hero boards are the cause of the weird numbers then until proven otherwise, I have to believe that its not the fault of ASUS. I have done more research on this processor than anything I have ever bought computer wise and I have bought a bunch. No where and I mean no where did the 2nd graders who are running Intel say that to use this over heated boat anchor would require a Godlike $600.00 MB. Well that is my thoughts at least as far as I can post here.🤬

EDIT: It is like buying a Corvette and you pick it up at the dealer only to find out that the new one only have a 4 cy engine and no cooling system and oh by the way they cost 40% more than last years model.

Edited October 22, 2018 by shivers9

[lownslo 592]

1 hour ago, brucewtb said:

Interesting video from Hardware Unboxed  suggesting that the Asus Maximus XI Hero had too weak VRM for the 9900K ( and presumably the 9700K) at only 4 phase.  Very odd as this is not a cheap board.  Hardware Unboxed  also suggest that this is the reason some reviewers are reporting much lower temps and power consumption when testing with 4 phase MBs - they are throttling the 9900K so it runs at a much lower frequency.  Thoughts anyone?

More info here, Bruce: https://www.reddit.com/r/intel/comments/9qdemh/masssive_psa_if_you_want_full_stock_performance/

To cut to the chase about the Maximus XI Hero, see the post by "petascale" with this link to the Z390 power delivery tiers: https://www.overclock.net/forum/attachment.php?attachmentid=223640&d=1539304287

HTH,

Greg

[brucewtb 249]

Thanks for the links.  Hardware Unboxed seem to have opened something of a Pandora’s Box (no pun intended). Unsure what to make of it all with so many conflicting opinions.  Was about to order a Hero now not so sure.

bruceb

[penta_a 429]

On 10/21/2018 at 10:55 PM, Rob Ainscough said:

 I do most of my testing of P3D V4.x on an 8700K (no OC)

Cheers, Rob.

🤔

[shivers9 1,174]

12 hours ago, brucewtb said:

Thanks for the links.  Hardware Unboxed seem to have opened something of a Pandora’s Box (no pun intended). Unsure what to make of it all with so many conflicting opinions.  Was about to order a Hero now not so sure.

bruceb

It just seems that the info we have at this point leaves some of us in between a rock and a hard place. If you feel that you just have to have a 9xxx at this point then the safe thing is to pay the $500.00 or $600.00 US for a top of the line board. It sounds like they may force enough juice through these things to make them clock slightly faster. The trade off is that now you have a computer that doubles as a furnace to warm you home and grill your hotdogs. At that point you may be looking at $400.00 or more to cool it back down. Something just does not seem right with this picture.

[martin-w 3,235]

Quote

It sounds like they may force enough juice through these things to make them clock slightly faster.

 

Clocking "slightly" faster would make hardly any difference to frame rate. 

 

Quote

$500.00 or $600.00 US for a top of the line board.

 

Maximus X1 Formula is £407 in the UK. Maximus X1 Code £359. Don't know if it's the same, but the the code was always the same board as Formula but without the built in water block. I have the Maximus X Code... awesome board. 

 

Edited October 23, 2018 by martin-w

[fghdgdfdfgfgf 111]

And let’s not forget the i59600k vs the 8086 battle ....lol

[HighTowers 1,434]

The benchmarks I have seen with overclocking have 5ghz as a sweet spot. There is reasonable performance gain from 4.5-5ghz.  Above that point, the performance tails off quite a bit. Therefore the benefit is only for bragging rights and not worth the tradeoff of extra heat, and more expensive supporting components just so you can hit 5.1 or 5.2.   Therefore reasonably good board and respectable 240-280mm AIO should be good to keep a stable 5ghz and prepar3d or whatever sim reasonably happy.  

If we cant get some better performance after all this, then theres gotta be some extra fine print in the intel legal wording that somehow detects flight simulation and immediately scales back the performance. 

[Guest]

If I can get 5Ghz on most of my cores with a 7900X on water cooling, I'll be disappointed if I can only get 5.1 Ghz on a 9900K with a water chiller setup ... hopefully when the retail CPUs are released I can find out ... Nov 21st isn't that far away and gives me time to start the PC assembly/fab process so I can get everything in place and tested while I wait for the CPU's arrival.

If that does prove to be the case with the 9900K, I'll put my 7900X back in, remove the IHS and get EK's mount for direct contact to EK block and chill it from there ... relegate the 9900K to test PC duties IF that turns out to be the case ... but that's all part of the fun for me.

Cheers, Rob.

[fghdgdfdfgfgf 111]

36 minutes ago, HighTowers said:

The benchmarks I have seen with overclocking have 5ghz as a sweet spot. There is reasonable performance gain from 4.5-5ghz.  Above that point, the performance tails off quite a bit. Therefore the benefit is only for bragging rights and not worth the tradeoff of extra heat, and more expensive supporting components just so you can hit 5.1 or 5.2.   Therefore reasonably good board and respectable 240-280mm AIO should be good to keep a stable 5ghz and prepar3d or whatever sim reasonably happy.  

If we cant get some better performance after all this, then theres gotta be some extra fine print in the intel legal wording that somehow detects flight simulation and immediately scales back the performance. 

I agree. Orbit nor-cal and flight beam sfo, in fact all us orbix regions with all the payware airports in said regions -less lax area- then no meaningful difference between 5.2 and 5.0 ...well, that is unless you count the 12c thermal difference.

but one with 5.0 and 3200 c14 mems still has two options to get a 10% FPS boos at lax: 

a) overclock to 5.5

Or

b) install 4133 c17 or faster memory

Note: don’t have the numbers, but I did have the sense that the cockpit and outside views were a lil more perky with  5,2, but at the same time the Pmdg 747 fmc seemed a lil laggy with 5.2 <me thinks memory was dropping to secondary timings, could be fixed with tweaking, me also thinks.

 

edit: me thinks the fmc lag might also be due to voltage deficiencies 

Edited October 23, 2018 by FunknNasty

[martin-w 3,235]

33 minutes ago, Rob Ainscough said:

If I can get 5Ghz on most of my cores with a 7900X on water cooling, I'll be disappointed if I can only get 5.1 Ghz on a 9900K with a water chiller setup rom there ... relegate the 9900K to test PC duties IF that turns out to be the case ... but that's all part of the fun for me.

Cheers, Rob.

 

I predict you will get 5.2 max, at a push, just about stable. Or 5.3 with HT off. And I will expect a special prize if I'm right. 

Seems to be the same situation as with paste, there's a point of diminishing returns. You'd think a custom loop and chiller would do the trick, but the  gobs of solder and thick die are the bottleneck.

"relegate the 9900K to test PC duties as Martin's prize" 😁

Edited October 23, 2018 by martin-w

[Guest]

22 hours ago, lownslo said:

with this link to the Z390 power delivery tiers:

The Formula Z390 is 8 + 4 … I think the Hero Z390 is the same?  The Extreme Z390 is 8 + 8 (as far as I know the Extreme Z390 isn't available yet).  Unless one is running LN2, I wouldn't be concerned with power delivery.

It's hard to say why some reviews are so negative about the 9900K … on paper the 9900K looks good and the L3 cache at 16MB ("smartCache") which is larger than my 7900X 13.75MB with 10 cores … the Intel UHD Graphics 630 is a waste of space and I'm assuming the reviewers are disabling?

If I can hit 5.5Ghz with HT off and the chiller, I'll stick with the 9900K, but I'll be sure to run various frequencies against my P3D V4.x base and see how it stacks up against my 7900X @ 5Ghz.  My Thermaltake case is like a mini-fridge on wheels 35" H X 19" W X 27" D … my wife just rolls her eyes and says "men!"

5 hours ago, martin-w said:

You'd think a custom loop and chiller would do the trick, but the  gobs of solder and thick die are the bottleneck.

There is much science behind thermodynamics … with water cooling setup and IHS there first phase is conduction and the second phase is convection … there will always be conduction even without an IHS.  Conduction works best with more physical surface area for heat to transfer … if Intel had to strengthen the silicon to deal with solder IHS then that could reduce surface area and localize heat transfer resulting is the same or less rate of transfer performance.

Copper, Gold, Silver, Aluminum (1050A) have very high rates of heat transfer, BUT all require far too high a temp to turned into a liquid that could be applied to a CPU.  Solder (Tin/Lead) is NOT great for heat transfer but has a very low liquid temp.

As you can see, Silver being the best … "rate of heat transfer" (but who sells silver water blocks and with copper a close 2nd one can see reason why).  So the use of Solder (Tin/Lead) should be better than Silver paste which is around 8W/mK vs. Solder around 35-57W/mK … "liquid metal / Galinstan" is around 13W/mK … so I guess that begs the question why are these CPU's so hot internally if heat transfer is 3-7X faster?  And why the variance in reports?

I think the prudent process here would be to "wait" for the "retail" versions and more user testing, updated BIOS/EFI's released, etc. etc. … I know reviewers like to be "first" but there is obviously lots of inconsistent data across the reviewers.

Cheers, Rob.