9900K/9700K Overclock performance review

[Guest]

I haven't seen any tests with non ES version of the i9 9900K?  Have retail versions hit the market yet?  I pre-ordered the 9900K about 5 seconds after they were available for Pre-Order and haven't received shipping notice with a promised ETA of today (10/19/2018).  But I suppose within the 5 seconds it took to complete my PreOrder there could have been 100,000 entries before me … worse than ComiConn 🙂 

The overclocking tests I've see of the 9900K ES variants has been with HT enabled and on 370 boards.

I do know the 9900K will generate a lot of heat hence why I got a chiller for it, the 9700K results on ES are around 5.5 Ghz basic cooling, so I still think/expect 5.6-5.8 Ghz on the 9900K with HT off good RAM and my chiller setup will be achievable.  I don't know how low I'll be able to drop the chiller before condensation forms (around critical components) … humidity is low in my part of the world so I'm hoping around 10C (50F) but I honestly don't know the limit.

Anyway, I'll reserve judgement until I actually receive my 9900K and start the process of OC along with Asus traditional rapid release of "performance improvements and stability" EFI/BIOS updates.

Cheers, Rob.

 

 

[Bob Scott 7,719]

6 minutes ago, martin-w said:

Rather ironic that the community complained about TIM, claimed Intel were ripping them off with TIM, demanded solder.... and now they get solder and its no better than the TIM. Perhaps Intel knew what they were doing all along and as I've  argued in the past, TIM wasn't just about cost cutting. Its hard to solder modern chips without cracking the die and other issues.

As de8auer pointed out, the die is now much thicker, hence the higher temp. Perhaps a thicker die is a requirement if you solder to prevent die damage. 

Like they say... be careful what you wish for. 🙂

Well, the 9900K has 33% more cores, 4.8MiB more on-die cache, and (according to the opening volley of reviews) still overclocks at par with the 6-core 8700K.  Who's to say that the STIM isn't doing a much better job than the paste would be doing in the same scenario?  The chip is kicking out a lot more heat...no surprise thermal output is a challenge, regardless of the thermal interface in use.

The real question for use with P3D is whether these new chips can be depended on to hit that 5 GHz mark...right now there's just a few datapoints that suggest they can get there or close, but that's potentially skewed if Intel cherry-picked the samples they sent to the guys submitting the first reports we're reading now.  I think a 5 GHz 8-core 9900K would be a real beast in a sim rig...but if a significant percentage of these processors can't get to 5 GHz on all 8 cores without doing a Chernobyl on us, then we're back to the choice between something like the 8086K, where virtually all of those chips can hit 5 GHz on all 6 cores, or running the sim on a potentially slower 8-core 9th-gen processor, knowing what we do about dependence on single-core speed on the main P3D thread.

Right now I'm torn.  Newegg appears to have taken my preorder without having stock to back it up, so no shipping notification has appeared yet, and Microcenter up in Denver has some 8086K CPUs in stock.  I must admit, I am really wavering at present...

Regards

 

[martin-w 3,235]

2 minutes ago, Rob Ainscough said:

I haven't seen any tests with non ES version of the i9 9900K?  Have retail versions hit the market yet?  I pre-ordered the 9900K about 5 seconds after they were available for Pre-Order and haven't received shipping notice with a promised ETA of today (10/19/2018).  But I suppose within the 5 seconds it took to complete my PreOrder there could have been 100,000 entries before me … worse than ComiConn 🙂 

The overclocking tests I've see of the 9900K ES variants has been with HT enabled and on 370 boards.

I do know the 9900K will generate a lot of heat hence why I got a chiller for it, the 9700K results on ES are around 5.5 Ghz basic cooling, so I still think/expect 5.6-5.8 Ghz on the 9900K with HT off good RAM and my chiller setup will be achievable.  I don't know how low I'll be able to drop the chiller before condensation forms … humidity is low in my part of the world so I'm hoping around 10C (50F) but I honestly don't know the limit.

Anyway, I'll reserve judgement until I actually receive my 9900K and start the process of OC along with Asus traditional rapid release of "performance improvements and stability" EFI/BIOS updates.

Cheers, Rob.

 

 

 

 

Tom Logan at OC3D didn't seem to be too worried about temp.  He ran all 8 cores at 5 GHz and with a 240 rad was seeing quite reasonable temps. 

 

 

 

[martin-w 3,235]

25 minutes ago, w6kd said:

Who's to say that the STIM isn't doing a much better job than the paste would be doing in the same scenario?  

 

 

 

No, de8auer de-lidded it and replaced the solder with liquid metal TIM. The temp dropped by ten degrees. He then lapped the die to reduce the thickness and lowered the temp even further. Pretty much the same temp drop we have seen when delidding Intel chips with paste between die and IHS. Given that solder has better thermal conductivity than liquid metal TIM, that shouldn't have transpired. Roman hypothesises that it's the thickness of the die and the amount of indium solder required that reduces thermal conductivity. 

 

Quote

The real question for use with P3D is whether these new chips can be depended on to hit that 5 GHz mark...right now there's just a few datapoints that suggest they can get there or close, but that's potentially skewed if Intel cherry-picked the samples they sent to the guys submitting the first reports we're reading now.

 

Tom Logan from OC3D doesn't seem to be having the same relatively high temps as everyone else. 9:20 in the video i posted he's saying a 280 AIO or top end air cooler like the NH-D15 is sufficient for 5 GHz on all cores. 

Edited October 19, 2018 by martin-w

[Bob Scott 7,719]

22 minutes ago, martin-w said:

No, de8auer de-lidded it and replaced the solder with liquid metal TIM. The temp dropped by ten degrees. He then lapped the die to reduce the thickness and lowered the temp even further. Pretty much the same temp drop we have seen when delidding Intel chips with paste between die and IHS. Given that solder has better thermal conductivity than liquid metal TIM, that shouldn't have transpired. Roman hypothesises that it's the thickness of the die and the amount of indium solder required that reduces thermal conductivity. 

Yep, I saw that, but remember that we have a sample size of one.  Was this an early engineering sample, or a production chip?  I've seen all sorts of issues over the years with bad solder joints in electronics production...so was the process (temperature, pressure, positioning etc) well-enough established and controlled to properly bond the case and the die?  Does Intel expect the solder joints to case/die to cure after some use due to the low reflow temps of indium-based solder?  Lots of questions remain...I don't think we know enough just yet to declare either victory or defeat here.

Interestingly, Tom Logan's video above seems at odds with most of the other reviews I've read today, which mostly suggest this CPU is a real heat monster that requires a substantially above-average cooling solution (and no, those conclusions weren't coming from somebody selling water cooling gear).

Regards

 

[brucewtb 249]

3 hours ago, w6kd said:

Right now I'm torn.  Newegg appears to have taken my preorder without having stock to back it up, so no shipping notification has appeared yet, and Microcenter up in Denver has some 8086K CPUs in stock.  I must admit, I am really wavering at present.

Having RMA’d an 8086K cpu a couple of months ago I would have egg on my face if I went back and reordered it.  It seems my retailer has sold the one I sent back but are expecting more stock soon.  

I will postpone wrist slashing until I see real world non ES testing of the 9900K/9700K preferably with P3D but not looking good.

Bruceb

 

 

[Guest]

Five years ago I was able to run my i7 4770k at 5ghz with HT off and although it never quite managed to complete all the 24/7 stability requirements I never got BSOD's in FS. I built that system while I was still using FSX and jumped to P3D V2 a few months later. I used nature to cool my system. I would consider a new build at this point. This is the longest I have ever waited but unless I see benchmarks in the 5.4+ range I won't bother!

[phiriz 0]

Hi guys.

I have not done any flight since earlier in the year. My old PC had i7-4790K overclocked with a 1080Ti. Looking at getting a new PC that I can use for that and some other work as well. I was considering getting one of the new intel CPU (either 9700K or 9900K) but those prices are dear - may still take the plunge.

I have read a few threads where people are claiming to be running Ryzen CPUs with no issues at fairly high settings (apparently P3D is utilizing more cores which I did not think it did) and started to consider perhaps going with the 2700X, 1080Ti (20 series doesn't seem justified at this point) and perhaps get a 4K G-Sync monitor.

Anyone have any experience with the Ryzen 2700X please let me know your experience. If anyone else has some advise for what would be a good setup I'd like to know.

 

@Rob Ainscough, I'd be very interested in knowing how your i9-9900K testing goes especially with reviewers not seeming to get much more than 5.1GHz on all cores.

[martin-w 3,235]

11 hours ago, w6kd said:

Yep, I saw that, but remember that we have a sample size of one. 

 

 

 

Oh yes, absolutely. As I said, Tom Logan didn't seem to be having any issues. Roman does comment that many of the reviewers are seeing high temps, so a small sample size yes, but not one. My speculation re temps is based purely on the information we have now. Having said that, it would seem to be a reasonable hypothesis that the thick layer of indium solder  Roman encountered, compared to previous generation soldered CPU's, not to mention the much thicker die, could result in higher temps.

 

I

Quote

've seen all sorts of issues over the years with bad solder joints in electronics production...so was the process (temperature, pressure, positioning etc) well-enough established and controlled to properly bond the case and the die? 

 

True, and soldering an IHS to a die is a complex process, fraught with difficulty. But as I say, it appears to be more than one reviewer that is seeing this and Roman's hypothesis does seem to make sense.

 

Quote

Does Intel expect the solder joints to case/die to cure after some use due to the low reflow temps of indium-based solder?  Lots of questions remain...I don't think we know enough just yet to declare either victory or defeat here.

 

I've not heard of cure time being required for indium based solder. Don't think it was the case regarding previous generation soldered chips. Not that I've come across anyway. I think that would have occurred to an expert like Roman instantly if it were the case. 

I'm not declaring victory or defeat, just speculating in regard to Roman's findings and the high temps he and other reviewers are seeing. So in regard to your comment... "Who's to say that the STIM isn't doing a much better job than the paste would be doing in the same scenario?" That question is present tense, so I answered it in the present tense, so I would say no it isn't, given the information we have now from Roman and other reviews. But of course, all might change in time when there are hundreds of chips in the wild. And of course chips always improve in quality with time as the manufacturing process is fine tuned. 

Edited October 20, 2018 by martin-w

[joemiller 269]

Well, so far so poor. Not encouraging at all. Primarily we need the processor for flight simming, where it relies mostly on single core performance. And, perhaps it's not so much of Intel's fault (though I hate the milking trend), but our tough luck that our simulation is unable to effectively use all cores, equally. 

I wonder what AMD is brewing... or maybe both Intel and AMD will finally benefit us in another 2 generations. That's roughly 3 years from now. Or, maybe the coming cpus do the work we desperately need. Who knows.The problem I see is changing CPUs everytime a new one comes just to find out that the benefit is 1 or 2 fps.

[martin-w 3,235]

20 minutes ago, joemiller said:

Well, so far so poor. Not encouraging at all. Primarily we need the processor for flight simming, where it relies mostly on single core performance. And, perhaps it's not so much of Intel's fault (though I hate the milking trend), but our tough luck that our simulation is unable to effectively use all cores, equally. 

I wonder what AMD is brewing... or maybe both Intel and AMD will finally benefit us in another 2 generations. That's roughly 3 years from now. Or, maybe the coming cpus do the work we desperately need. Who knows.The problem I see is changing CPUs everytime a new one comes just to find out that the benefit is 1 or 2 fps.

 

 

I'm not simming these days. But what is required is a new sim, and not just Lockheed Martin improving the old one. A new sim from the bottom up that makes use of the latest coding techniques and the latest hardware. That is the requirement that flight sim fans have always desired and required. 

[Guest]

28 minutes ago, martin-w said:

But what is required is a new sim, and not just Lockheed Martin improving the old one.

Not this again ... if only I could show you a video ...

28 minutes ago, martin-w said:

A new sim from the bottom up that makes use of the latest coding techniques and the latest hardware

Specifically what "techniques"?  P3D makes use of multiple GPUs with NVLink/SLI support and VR and is 64bit based and multiple CPUs ... I think you are just having a hard time accepting the limits that "distance" imposes on flight simulators vs. the very short distances and limited environments found in 3D shooter and the like.

Maybe you should start a poll here and elsewhere to see if they are willing to toss aside ALL of their existing add-ons (the PMDGs, FSLabs, Aerosoft, Orbx, etc. etc.) and wait 2-5 years or more for the same level of quality of add-ons to be produced for a new ground up flight simulator that they "require".

I just finished the new Tomb Raider and Call of Duty ... very pretty, very limited environment, scripted, almost zero replay value ... they are sitting on my uninstall queue as I resume the more sensible world of flight simulation 🙂

Cheers, Rob.

 

[fghdgdfdfgfgf 111]

...meanwhile, back at the ranch:  Just captured a few minutes ago; sfo to lax ....flight was smooth as a baboon's behind.

 

 

 

 

 

above is for reference material

room temp 75 f  ...yeah, those are the noctcha fans coupled to the h100 v2

Edited October 20, 2018 by FunknNasty

[joemiller 269]

13 minutes ago, Rob Ainscough said:

Not this again ... if only I could show you a video ...

Specifically what "techniques"?  P3D makes use of multiple GPUs with NVLink/SLI support and VR and is 64bit based and multiple CPUs ... I think you are just having a hard time accepting the limits that "distance" imposes on flight simulators vs. the very short distances and limited environments found in 3D shooter and the like.

Maybe you should start a poll here and elsewhere to see if they are willing to toss aside ALL of their existing add-ons (the PMDGs, FSLabs, Aerosoft, Orbx, etc. etc.) and wait 2-5 years or more for the same level of quality of add-ons to be produced for a new ground up flight simulator that they "require".

🙂

Cheers, Rob.

 

Well, I am not even 1/4 in knowledge as you, Rob, or several folks here. Therefore, I'm going to wait on your findings before I make decision. I'm currently on a 7700k and I feel it's time for me to upgrade this CPU. I also keep running my 1070ti card maxing at 100% of the 8GB. I really want to upgrade the card to an 11Gb, but gosh they are expensive. These are about 16 months old.

CPU is @4.90Ghz and 3800mhz RAM.  Not bad at all. But, with better add-ons comes more system demand, and I love those settings pretty high, as I'm big on eye-candy. Of course I adjust as needed. 

So, in about 3 weeks, I will be upgrading CPU, motherboard, and graphics card. 

[Guest]

32 minutes ago, joemiller said:

I'm currently on a 7700k and I feel it's time for me to upgrade this CPU.

My beta testing PCs are 7700K and 8700K with 1080 FE and a 1080Ti all stock clock, no OC.  I try to keep it somewhat more realistic hardware combination that many have.

For my current main FS PC 7900X @ 5Ghz with 2X 2080Ti, I have my baseline data and recorded flights saved off to my NAS and Google Drive.  When I get this 9900K CPU built (hopefully the CPU will ship this week, fingers crossed but still no ETA from AMAZON or NewEgg) I'll be able to do an apples to apples compare 9900K vs. 7900X.

If the 9900K doesn't perform better, then I'll go back to my 7900X and put it on the chiller to squeeze out 5.2Ghz HT off.  My 7900X has been extremely reliable at 5Ghz so the 9900K has it's work cut out if it want's to earn a place in my SimPit 🙂  ... but keep in mind the 9900K is 1/2 the cost of a 7900X an the 9700K even less.

Cheers, Rob.