Intel Core i9-9900K Re-Review: Performance unleashed, but also leashed

[Evernessince]

FX 9590 $900 MSRP
R9 Pro Duo $1500
R9 Nano $549

The R9 Pro Duo was a professional card with two GPUs. $1,500 given the form factor, performance and price. You definitely got more then 10% performance with that card. The only people who could complain about that card are those who wanted to use it for gaming, of which you would rely on crossfire support. That's goes for any dual GPU configuration.

The R9 Nano was a flagship card in an ITX footprint. It made it possible to have a powerful and small build. Performance wasn't the main concern, size and efficiency was of which it had it spades. If you were building high end mini-ITX at the time, that was the card to beat.

 

[pcnthuziast]

The 8700k is a better buy than both the 9700k AND the 9900k.

 

[RCantw3ll]

The results seem so incredibly odd to me. I know the chips have variances, but still...

My 9900k with the same Asus motherboard here running the Intel spec (cooled with H115i, not even in push pull), but with a few changes to prevent TDP throttling runs at 4.7 with an average of maybe 50 - 55c at full load.

I upped the boost on all 8 cores to run at 4.8ghz and upped the base clock to 101 (clocks to CPU to 4.85ghz under full load on all 8 cores/16 threads). Left all voltages on auto and used XMP for memory. Also set the cache frequency to 4.55ghz (basically just setting the highest base config I can and leaving it alone), which raises the temp far more than the base frequency does.

With those changes and disabling speedstep to keep my base clock speed at 3.6ghz when idle I am at 30c idle clocks and 60 - 65c on load (peaks for brief moments spike to the higher 60s). If just gaming and streaming instead of stress testing, but still putting 80+% load on the CPU I'm living in the 50s.

There are air coolers out there that can compete with my H115i so it can't be THAT difficult to cool. Especially since I am doing this in a mid tower case sitting in the corner my office under my desk.

 

[paul s2]

Thank you for putting out stock and OC numbers .
just need to keep everyone honest with third party reviews.
if I needed a upgraded cpu "current cpu 4770k" I would choose a 2700 series today.
Bringing light to this problem will let people by coolers based on expected thermal loads.

 

[Philip T]

Thank you for putting out stock and OC numbers .
just need to keep everyone honest with third party reviews.
if I needed a upgraded cpu "current cpu 4770k" I would choose a 2700 series today.
Bringing light to this problem will let people by coolers based on expected thermal loads.

These aren't stock numbers. These benchmarks are misleading. You don't lock your CPU to 95W. This is the AMD fan version of the Principled Technologies benchmarks, at least they had some explanation why the did what they did.

 

[Philip T]

I'm curious if the editorial staff stands by the 9900K rereview? Just to get it out there, I am not an Intel Fan boi. (Loved unlocking my dual core Athlon X2 to a full quad core). I'll use the chip that gives me the best performance at the cost I can afford. My problem with the article is the flawed premise that a processor can't be run past 95W, because Intel lists it as a 95W TDP processor. As it stands I question the technical credibility of Techspot. I would love to see you contact Intel and get an answer to whether their assigning a 95W TDP means that the processor is in only in spec, if you cap the processor at 95W.

This quote from a www.macrumors.com article illustrates the point.

"Lee went even further and tested six laptops equipped with an i9 from various manufacturers. Unsurprisingly, thicker laptops with better cooling did outperform thinner laptops, including the MacBook Pro. The Alienware 17 R4, Acer Helios 500, and Asus G703 (all very thick) saw higher average clock loads."

https://www.macrumors.com/2018/07/24/throttling-fix-2018-macbook-pro-improvements/

These laptops are all running in spec, trade offs were made that limit their performance. Intel intentionally made it so it can run at a wide range of clock speeds. It's the duty of the system builder to determine what is spec for the system they put together. I don't see how you can seriously say the benchmarks in the article are representative of a properly running 9900K.

The premise the article is based on is flawed, because of that it's conclusions are deeply flawed. Techspot is doing the tech community a disservice by hosting that article.

 

[ET3D]

Thanks for the interesting article. I assume that you wrote it without getting sleep first, or while not feeling well (which was the case for me for the past few days, and I certainly felt less focused). The text also felt like a ramble, being peppered with "okay so", "well" and "anyway". I feel that many paragraphs could have been made clearer by rephrasing them.

Still a worthwhile article, but I feel that it could use some editing to get it to the usual standard.

As for the test itself, I think it would be nice, going forward, to mention how long the test took alongside the results. Also, I imagine that some tests would have dropped even more if they ran for longer. Which, as you mentioned, would have been more realistic.

In general, this highlights the problem that tests of software used for creative work, where rendering can take a long time in practice, are tested with workloads which take too little time on current CPUs. This to me is a major takeaway from this, and I now suspect all work related benchmarks to be off the mark.

I'd appreciate an article on the topic, with various length tests to see at what point the test results settle to reveal the real long term difference in performance.

 

[Rtxman6]

So you did eveything you can to make the 9900k perform as bad as possible !

 

[Julio Franco]

Thanks for the interesting article. I assume that you wrote it without getting sleep first, or while not feeling well (which was the case for me for the past few days, and I certainly felt less focused). The text also felt like a ramble, being peppered with "okay so", "well" and "anyway". I feel that many paragraphs could have been made clearer by rephrasing them.

Still a worthwhile article, but I feel that it could use some editing to get it to the usual standard.

Honestly, that's on me. I'm not happy to admit because this came in at the last minute (Steve's completed work and benchmarks) on a Friday night, I skipped editing on the benchmark notes, where you observed the most grammatical inconsistencies. I've since edited the entire thing once again so it's not an embarrassing read anymore.

 

[Julio Franco]

These aren't stock numbers. These benchmarks are misleading. You don't lock your CPU to 95W. This is the AMD fan version of the Principled Technologies benchmarks, at least they had some explanation why the did what they did.

Exactly. These are not stock, we say so everywhere. This is an exercise of understanding why certain motherboards and some reviews (from outlets like Linus Tech Tips) were showing vastly different power consumption, thermal and performance figures on launch day.

A brief timeline:

#1
We review the 9900K. Fast gaming CPU, good for workloads, too, but runs very hot, and it's very expensive when compared to its direct competitors. We stand by our original review.

#2
We run a column sharing our thoughts on result discrepancies and why we might re-test the 9900K to make sense of the numbers. Do note, some motherboard makers (just a few, like Asus) are limiting the 9900K to the TDP, showing different results out of the box.

Now, since first reviewing the Core i9-9900K I’ve tested over a dozen Z390 motherboards, ranging from the cheapest $120 models right up to the much more expensive flagship versions. Out of the box all boards from MSI, Gigabyte and Asrock run without a TDP limit in place, even with the default BIOS configuration.

However this isn’t the case for the Asus boards I’ve used, the default out of the box of cleared BIOS configuration employs the 95-watt TDP limit.

#3
We publish our second look, showing what would be if the 9900K were limited to the official spec TDP. Out of the box, the results are unchanged from our original review. Furthermore we say this...

Looking ahead into future reviews, we plan to stick to show the typical out of the box experience, and if that means overclocked CPUs, well that’s what we will show. We'll also continue to publish additional insights and features like this one when need be.

---

BTW, about the Principled Technologies fiasco? We were one of the first to put that to light:
https://www.techspot.com/article/1722-misleading-core-i9-9900k-benchmarks/

 

[Brahman05]

FX 9590 $900 MSRP
R9 Pro Duo $1500
R9 Nano $549

Where you buying that 9590? And is it gold plated for that price?

 

[Brahman05]

I'm curious if the editorial staff stands by the 9900K rereview? Just to get it out there, I am not an Intel Fan boi. (Loved unlocking my dual core Athlon X2 to a full quad core). I'll use the chip that gives me the best performance at the cost I can afford. My problem with the article is the flawed premise that a processor can't be run past 95W, because Intel lists it as a 95W TDP processor. As it stands I question the technical credibility of Techspot. I would love to see you contact Intel and get an answer to whether their assigning a 95W TDP means that the processor is in only in spec, if you cap the processor at 95W.

This quote from a www.macrumors.com article illustrates the point.

"Lee went even further and tested six laptops equipped with an i9 from various manufacturers. Unsurprisingly, thicker laptops with better cooling did outperform thinner laptops, including the MacBook Pro. The Alienware 17 R4, Acer Helios 500, and Asus G703 (all very thick) saw higher average clock loads."

https://www.macrumors.com/2018/07/24/throttling-fix-2018-macbook-pro-improvements/

These laptops are all running in spec, trade offs were made that limit their performance. Intel intentionally made it so it can run at a wide range of clock speeds. It's the duty of the system builder to determine what is spec for the system they put together. I don't see how you can seriously say the benchmarks in the article are representative of a properly running 9900K.

The premise the article is based on is flawed, because of that it's conclusions are deeply flawed. Techspot is doing the tech community a disservice by hosting that article.

I stated this in the original article asking if a retest was needed. Every article ive ever read benchmarking processors includes total system draw. Most sites can even extrapolate just cpu power by removing the gpu usage or by testing a cpu intensive workload by just using the igpu, and it NEVER matches the white paper. Someone PLEASE argue this point to me. I feel surrealistic reading some of the comments like that never happens, from hardocp, toms hardware, wccf tech, extremetech, linus tech tips, hell my grandma lists it when she reviews a cpu and she dosent even know how to set the time on her vcr for christs sake. Needless to say I stand by every comment I have made in this thread and the previous. So stop whining or motherboard aib's will start setting the firmware in such a way that getting any extra performance outside of the whitepaper will be near impossible for fear of stupid backlash like this, WHICH IS THE POINT OF OVWRCLOCKING AND EVEN IF YOU DONT THANK THE STARS YOU GET THAT EXTRA BIT OVER THE DELL'S AND HP'S, who stick to the whitepaper like the word of god and pat yourself on the back for being prodigious enough to build your own computer in the first place.

 

[Brahman05]

Exactly. These are not stock, we say so everywhere. This is an exercise of understanding why certain motherboards and some reviews (from outlets like Linus Tech Tips) were showing vastly different power consumption, thermal and performance figures on launch day.

A brief timeline:

#1
We review the 9900K. Fast gaming CPU, good for workloads, too, but runs very hot, and it's very expensive when compared to its direct competitors. We stand by our original review.

#2
We run a column sharing our thoughts on result discrepancies and why we might re-test the 9900K to make sense of the numbers. Do note, some motherboard makers (just a few, like Asus) are limiting the 9900K to the TDP, showing different results out of the box.



#3
We publish our second look, showing what would be if the 9900K were limited to the official spec TDP. Out of the box, the results are unchanged from our original review. Furthermore we say this...



---

BTW, about the Principled Technologies fiasco? We were one of the first to put that to light:
https://www.techspot.com/article/1722-misleading-core-i9-9900k-benchmarks/

And this differs from every review I have ever read since my fx8120 how? I seem to be missing something here. Ok so the variation is greater with this chip than listed, but I dont seem to remember an uproar like this with the amd fx line, eventually upgraded that to an fx8570 and knew exactly what I was getting into from the start, thanks to comprehensive reviews, that never mentioned the issue you have brought to light here, which was(is, still running strong and just as hot) just as bad in my opinion.

 

[Evernessince]

TDP isn't supposed to be the limit though. Even at stock.

The variance is on the motherboard manufacturers, because performance can range quite a bit. MSI is the "worst" and eVGA being the "best" according to GN.

It has to be though when you allow newer higher power draw processors on a last gen platform. Those motherboards were never designed to accommodate the 9900K and will not run it to it's full potential. That should be made obvious but right now it isn't.

 

[kmo911]

Its not exspensive that i9-9900k lga 1151. but that other version of
Intel Core i9-7980XE 2.6-4.0GHz LGA2066 X299 (multicom.no) was a link but now its down
are wery DOH..
https://www.multicom.no/searchresult.aspx?q=kw:9700k;fd:c_1000804 LINK + water pump cooler
PRIME Z370-P II
1005174
NetOnNet
.no
https://www.multicom.no/intel-ts13x-liquid-cooling-solution/cat-p/c/p6482938 LINK
im gonna try out i7-9700k 32 gb 2400 vram rtx 2070 8gb version
benscmarking and so on with unigine free 3dmark vr marks hl 2 benchmark gta 3-5 far cry 1.5 crysis 1-3. still old games can get better fps. im at a prime z370 II motherboard not z390. only wifi and some better thing on those motherboards. hope ill get good fps on this specs. it would be best to run a amd 10-16 core but ill try out intel 1 st. ill maybe back (as)later whit som results.

 

[senketsu]

@Philip T
you have posted numerous confusing, seemingly contradicting things. You correctly post Intel's definition of TDP then somehow think Intel's requirement for a 130 W cooler means more than just that Intel wants a little buffer to cover their behinds due to hot climate, poor case ventilation, OC, etc.
TDP = 95 W
Required Cooler = 130 W
Is that not clear? To quote you "Notice it says 130W not 95W? Seems a bit like calling the kettle black?" Frankly I don't even know what you mean.

quote; "This page lists a 150W TDP for that processor. In no way are you running that processor out of spec if you run it all the way to 150W or even beyond. Your conclusion regarding work loads is completely wrong. Put adequate cooling on your CPU and it will be fine. The job of a system builder is to choose the right components. A $500 unlocked processor targeted at enthusiasts and high end users limited to 95W, indicates you either don't understand what TDP is or you are intentionally cobbling the 9900K so you can say the 2700X is as fast."
again, WTF? I addressed the first part, now for work loads, from your own link on processor specs under TDP "Thermal Design Power (TDP) represents the average power, in watts, the processor dissipates when operating at Base Frequency with all cores active under an Intel-defined, high-complexity workload. Refer to Datasheet for thermal solution requirements."
Is that not clear?

As far as the processor performance and cooler capacity go, from an Intel Whitepaper (Measuring Processor Power: TDP vs. ACP) "The thermal design power is the maximum power a processor can draw for a thermally significant period while running commercially useful software.....snip.... The processor thermal solution should be designed to accommodate thermal design power (TDP) at Tcase,max. TDP is not the maximum power of the processor....Due to normal manufacturing variations, the exact thermal characteristics of each individual processor are unique...….As such, no two parts have identical power and thermal characteristics. However the TDP specifications represent a “will not exceed” value...........Because TDP is a worst case value when running a “worst case” application, most processors, when running a more “typical” workload, will dissipate power that is less than the rated TDP value...…….It is important to note that thermal design power is the maximum thermal power the processor will dissipate, but not the same as the maximum power the processor can consume. It is possible for the processor to consume more than the TDP power for a short period of time that isn’t “thermally significant”. For example, a processor might consume slightly more power than the rated TDP value for say one microsecond…but then consume less power than the rated TDP value for a long period of time. Such operation is considered normal. Because the processor temperature does not exceed the specified limits during such a short excursion, the processor will continue to operate correctly.....it is possible to cause the processor to exceed the rated TDP value for a much longer, “thermally significant” period of time. To ensure the processor stays within the thermal specification under such conditions, Intel processors have a built-in “thermal control circuit” which reduces processor power by reducing the processor voltage and/or modulates the clock frequency."

don't know if the 'thermal control circuit' is the same in the i9-9900K because the processor wasn't out when the whitepaper was written.

 

[hahahanoobs]

Where you buying that 9590? And is it gold plated for that price?

*sigh*

 

[Philip T]

The conclusion of the article pretty much says don't buy an 9900K if you are going run it at 100% all the time, because it will scale back to pretty close to 2700X performance.

"Basically the 9900K is a really good overclocker, if you invest in proper cooling."

Considering there isn't a cooler in the box, you have to invest in something. I don't recall seeing any mention of thermal throttling kicking in from your original benchmarks. Something we do see in the MBP, a warrantied retail system.

"With motherboards technically overclocking the 9900K to the default clock multiplier table, 4.7 GHz as an all core for example, under those conditions there isn’t much left in it. For most 5GHz will be the limit, good luck keeping it cool past that without a serious amount of time, effort and risk. So realistically you’re talking about up to a 6% boost over what’s shown here when looking at the unlimited results, and we certainly found that to be true when attempting to overclock the 9900K in our day-one review."

First off technically changing the multiplier isn't overclocking at all. You have to change the clock for it to be overclocking. Next this is a processor that is designed to alter the muliplier based on a ton of complex settings, one of which is temperature. Let me say it again TDP is not a limit. If I hook up a phase change cooler and get my system running at 60x stable on real world workloads, that's within spec. My problem is you are taking synthetic benchmarks that your test system had no issues with, picking an arbitrary setting and rerunning the benchmarks. In the article you then extrapolate real world performance from your restricted synthetic benchmark. The only thing the new benchmarks say, is the performance you'll get if decide to implement your arbitrary limit. I'd like to see testing that backs up your untested performance claim. As evidenced by the comments people think these are stock numbers. All over the article, you conflate 95W TDP with the official settings. Please show me where you get that TDP means that's a maximum. Intel states it is an average W over a complex workload. For me the article is rubbish and misleading, because TDP doesn't mean what you seem to think it means. This is a processor that ships with no heatsink and the ability to configure it to a wide range of speeds.

"You could say motherboard makers are cheating but we don’t think that is it... we still feel Intel is cheating their own spec."

This doesn't help, that sounds like bias. This is reinforced when you can download an official Intel utility that lets you reconfigure processor settings on the fly. Please show me anywhere that TDP has the import you are giving it. Nobody that buys a 9900K is going to run their system locked at 95W TDP. It isn't meant to be run that way. Can you honestly say to me that you believe your 95W benchmarks are representative of the typical stable system in real world settings? Can you tell me how running benchmark with an arbitrary limit is any different from what Principled Technologies did? Are your 2700X benchmarks also run to ensure it was limited to it's TDP?

 

[Daniele 00]

Thanks to remind me that world is diveded into poor and rich people.

 

[nintendoeats]

First off technically changing the multiplier isn't overclocking at all. You have to change the clock for it to be overclocking.

This is not true. Modifying the external clock is no more or less overclocking than changing the clock factor. Both increase the number of clock cycles which occur in a given period of time; that is the definition of overclocking. Altering the FSB clock does have additional effects beyond changing the processor speed, but those things are incidental and do not have anything to do with CPU overclocking.

 

[Philip T]

@Philip T
you have posted numerous confusing, seemingly contradicting things. You correctly post Intel's definition of TDP then somehow think Intel's requirement for a 130 W cooler means more than just that Intel wants a little buffer to cover their behinds due to hot climate, poor case ventilation, OC, etc.
TDP = 95 W
Required Cooler = 130 W
Is that not clear? To quote you "Notice it says 130W not 95W? Seems a bit like calling the kettle black?" Frankly I don't even know what you mean.

quote; "This page lists a 150W TDP for that processor. In no way are you running that processor out of spec if you run it all the way to 150W or even beyond. Your conclusion regarding work loads is completely wrong. Put adequate cooling on your CPU and it will be fine. The job of a system builder is to choose the right components. A $500 unlocked processor targeted at enthusiasts and high end users limited to 95W, indicates you either don't understand what TDP is or you are intentionally cobbling the 9900K so you can say the 2700X is as fast."
again, WTF? I addressed the first part, now for work loads, from your own link on processor specs under TDP "Thermal Design Power (TDP) represents the average power, in watts, the processor dissipates when operating at Base Frequency with all cores active under an Intel-defined, high-complexity workload. Refer to Datasheet for thermal solution requirements."
Is that not clear?

As far as the processor performance and cooler capacity go, from an Intel Whitepaper (Measuring Processor Power: TDP vs. ACP) "The thermal design power is the maximum power a processor can draw for a thermally significant period while running commercially useful software.....snip.... The processor thermal solution should be designed to accommodate thermal design power (TDP) at Tcase,max. TDP is not the maximum power of the processor....Due to normal manufacturing variations, the exact thermal characteristics of each individual processor are unique...….As such, no two parts have identical power and thermal characteristics. However the TDP specifications represent a “will not exceed” value...........Because TDP is a worst case value when running a “worst case” application, most processors, when running a more “typical” workload, will dissipate power that is less than the rated TDP value...…….It is important to note that thermal design power is the maximum thermal power the processor will dissipate, but not the same as the maximum power the processor can consume. It is possible for the processor to consume more than the TDP power for a short period of time that isn’t “thermally significant”. For example, a processor might consume slightly more power than the rated TDP value for say one microsecond…but then consume less power than the rated TDP value for a long period of time. Such operation is considered normal. Because the processor temperature does not exceed the specified limits during such a short excursion, the processor will continue to operate correctly.....it is possible to cause the processor to exceed the rated TDP value for a much longer, “thermally significant” period of time. To ensure the processor stays within the thermal specification under such conditions, Intel processors have a built-in “thermal control circuit” which reduces processor power by reducing the processor voltage and/or modulates the clock frequency."

don't know if the 'thermal control circuit' is the same in the i9-9900K because the processor wasn't out when the whitepaper was written.

95W TDP and 130W cooler aren't contradictory, they aren't even the same thing one is an average load design point and one is the cooling ability of a heatsink. To get the performance Intel lists for a 95W TDP, you should have a cooler that meets the 130W spec. What people don't seem to be getting just because a processor lists a 95W TDP, doesn't mean that is the spec to force it to.

"The short version of this is that motherboard makers are currently getting blamed for running the 9900K out of spec, when in reality we strongly believe it’s Intel who’s cheating their own spec and pushing board partners to run the 9900K at the default clock multiplier table, rather than at the official power spec." 95W TDP, means on their test bed system 95W was the average safe load. Sometimes it will go higher and sometimes lower. Where this article loses it, is the assumption that the motherboard manufacturers settings are out of spec. If they aren't the whole premise of article is a load of bollocks. Now go read the many articles about the MBP laptop throttling. Do you think Apple is running the processors out of spec? I doubt it. I think they are using CPU settings that throttle the processor based on temperature readings from the CPU die. The cheating the spec conclusion would mean that all the MB manufacturers decided that they are willing to take on the liability of replacing damaged processors. Which is more likely? This articles belief that these corporations are subverting or The MBs are running in spec. The processors have a setting called configurable TDP for heavens sake. How could you possibly take advantage of that if you can't run the processor above 95W TDP and stay in spec. Might as well just lock that setting to 95W. Or maybe there are other factors that dictate the processor TDP, like providing extra cooling (Like it says under the configurable TDP instructions). My belief is that TDP cannot be considered a criteria for making claims for an enthusiast processor that doesn't even include a heatsink. Considering the MB manufacturers have a financial stake in this, I believe that the premise of this article is false and does nothing, but confuse people trying to make choices regarding what CPU to buy.

 

[MaikuTech]

Thanks to remind me that world is diveded into poor and rich people.

If a poor man gets the last laugh over a rich man all is good in the end.
You should be happy Intel buckled under and finally made things more fair in prices.

 

[ET3D]

Honestly, that's on me. I'm not happy to admit because this came in at the last minute (Steve's completed work and benchmarks) on a Friday night, I skipped editing on the benchmark notes, where you observed the most grammatical inconsistencies. I've since edited the entire thing once again so it's not an embarrassing read anymore.

The entire thing including my comment. No problem, that's fine with me.

I think it was an interesting insight into the creative process at TechSpot. Editors are always underappreciated.

 

[dj2017]

If AMD was doing this 4-5 years ago, tech press would have cut it in a million pieces.

Anyway, nice article. Unfortunately many people will buy the i9 on a store blindly with the 95-125W TDP cooler the sales person will suggest them. There are many out there who will ask for the best without knowing much about PC hardware.A cheaper cooler with this beast will make the A store look cheaper in the eyes of the consumer compared to the B store that would be correctly suggesting a much better cooling solution.

 

[i3kingwizardop]

NINE HUNDRED BUCKS on newegg. Hard pass.