I9 9900k Official Reviews from Anandtech, Tomshardware. Add your own links to others !

[SPBHM]

I wonder why they are using the 9900K die for the 9600K? I guess just for salvaging those dies that didn't make the cut? and they save the good 6c dies for laptops or something? or because it's designed for the soldered heatspreader and the other is not?
or is the native 6C die simply EOL!?

I don't know, impressive CPUs, but the 8700K beats the 9700K fairly often and is looking better in other aspects...
it's impressive that they run clocks so high, but the temps and power draw are really showing it, it's a CPU running near the limit at stock settings...

 

[The Stilt]

The power figures stated in the review seem pretty funky to me, as a "package power".
As a whole system power draw it is plausible, but not as the consumption of the CPU alone.

Pov-Ray 3.7.1 b9 with "pvengine64 /benchmark" command.

The power limits, voltages and the overall behavior were set to bios defaults.

 

[Chaotic42]

Yikes, this thing is pretty disappointing. I was hoping to have an obvious upgrade from my 4790K, but I'm not so sure.

 

[majord]

I'm a bit lost with the whole TDP situation here..

seems Compubase are the only ones that looked into this and tried actually limiting to 95 like it should be? https://www.computerbase.de/2018-10/intel-core-i9-9900k-i7-9700k-cpu-test/3/ and get notably lower clocks and performance as you'd expect - not just AVX but across the board on any threaded workloads.

Why are Intel labeling these 95w CPUs? They even state a 130w Cooling solution as a minimum? So even if you argue TDP's true meaning - Thermal design Power , how can you then go ahead and specifiy Thermal solution well beyond that

 

[jpiniero]

Why are Intel labeling these 95w CPUs? They even state a 130w Cooling solution as a minimum? So even if you argue TDP's true meaning - Thermal design Power , how can you then go ahead and specifiy Thermal solution well beyond that

You could still use a 95W rated cooler I imagine and still keep it well above the base clock. But you wouldn't be able to hold the FCT obviously.

 

[Abwx]

Computerbase numbers at 95W are still not valid because the CPU still has an upper limit that is way above 95W as long as the temp doesnt reach a given level.

This way you can bench CB at 115W for the first run and then average the power it used over say 10-20 runs.

We ll know more once people get regular products instead of the cherry picked ones sent to the press..

 

[formulav8]

Why are Intel labeling these 95w CPUs?

Intel says 95w TDP because they use the better case scenario. They base their TDP on the base clock only. So when it comes to the 9900k, they only give you the numbers based on a frequency of 3.6 ghz and not on the 4.7ghz-5ghz turbo clocks.

 

[krumme]

95 because it's less than 105 and under 100. And similar to 8700k. And on paper fit for old 370 boards.

Its not about tdp or watt or freq but a decision made by sale and marketing before the cpu final base and turbo freq was decided. As stilt says the vrm is pressed to the max. Its part of the problem. As with skl x they probably did not intend to take perf this far. The 95 number just had to stick no matter what. Marketing shareholder value whatnot.

 

[Timmah!]

I think its about the VRM efficiency. It is above sweet spot and the board just must handle the load current.

This reminds me of the Prescott days. The room heater.

But absolute performance of the chip is excellent. the 7900X as 10C is tied, except pure AVX workloads.

For me from pure performance this the the real upgrade for my 6600K@4.4GHz. Higher single threaded enough to be visible, multithreaded no comment. Games the same top performance.
I would even run through this power with avx workload but I don't want a water cooler.

Yeah, 8C at 5GHz being equal to 10C at 4GHz, who would have known BTW, Skylake-X AVX performance in that 3D particle movement :-O Leaving all the other CPUs in dust. If only it was actually useful in regular real-world apps.

 

[jpiniero]

AT appears to have redone the power consumption test with a different board and it's much lower. 9700K does 124 and 9900K does 168.

 

[Arachnotronic]

I wonder why they are using the 9900K die for the 9600K? I guess just for salvaging those dies that didn't make the cut? and they save the good 6c dies for laptops or something? or because it's designed for the soldered heatspreader and the other is not?
or is the native 6C die simply EOL!?

I don't know, impressive CPUs, but the 8700K beats the 9700K fairly often and is looking better in other aspects...
it's impressive that they run clocks so high, but the temps and power draw are really showing it, it's a CPU running near the limit at stock settings...

Spectre and Meltdown.

 

[rainy]

AT appears to have redone the power consumption test with a different board and it's much lower. 9700K does 124 and 9900K does 168.

50W less as result of changing motheboard?
That's quite a lot.

 

[jpiniero]

50W less as result of changing motheboard?
That's quite a lot.

That's the quadratic effect of lower voltage.

 

[slashy16]

This is a gaming CPU and people are going crazy that it draws a lot of power running benchmarks which do not reflect the real world.
It looks like while gaming it's going to draw(while overclocked) slightly more power than an 8700k and 1/4 the amount of power my
2080ti sucks up. I have read nearly every review now and the 9900k wins every benchmark. It's the best 8 core CPU
in the world hands down. A lot of reviewers are comparing it to the 2700x, sorry but you can't make that comparison.
The 9900k is on another level of performance.

 

[Abwx]

AT appears to have redone the power consumption test with a different board and it's much lower. 9700K does 124 and 9900K does 168.

Less TDP headroom should result in lower scores, so AT benchs are at PL3 max power and power measurements are made on another MB set at PL2 max...?

FTR Golem.de stated that when set to 95W the chip will still get to 119W for the first 30s, in these conditions the CPU score 1901 pts in CB R15.

https://www.golem.de/news/core-i9-9...te-5-ghz-kerne-sind-extrem-1810-136974-5.html

 

[Head1985]

Temps and power.Default 76°C with water cooling and 145w on CB 10s load(prime 100°C).After oc to 5Ghz 180w and 95°C after only 10s load(prime BSOD)

 

[IntelUser2000]

Why are Intel labeling these 95w CPUs? They even state a 130w Cooling solution as a minimum? So even if you argue TDP's true meaning - Thermal design Power , how can you then go ahead and specifiy Thermal solution well beyond that

If you see Anand's and Computerbase's numbers, you can see Kabylake with the 4 cores is the last generation where TDP was as stated.

It seems with the 6 and 8 cores they have changed the meaning of PL1 and PL2 values to get more performance. So they increased power use to get more performance. It's shady if you ask me.

Even in laptops, cooling solutions have become better and more beefy in some cases so they try to use the same nonsense to improve performance at the "same TDP". You'll see most modern systems use the 25W cTDPup value rather than the 15W nominal one. Though of course they can't mislead as in desktops because TDP = Power use for thermally constrained environments like laptops, and you have to dissipate that somewhere.

Nothing has changed. 486 CPUs barely needed a heatsink. Alongside process and architectural improvements, CPUs/GPUs have improved performance by increasing power use by 20x. Looks like they are trying to increase it further.

 

[jpiniero]

Less TDP headroom should result in lower scores, so AT benchs are at PL3 max power and power measurements are made on another MB set at PL2 max...?

Huh? You don't get worse performance with less voltage. Unless it crashes of course.

 

[Thunder 57]

I'm glad we seem to be getting to the bottom of the power usage issue, as AT has redone their numbers and it makes more sense. It is interesting to see that HT adds about 44W. Maybe that was part of the reason for the removal of HT, and not just segmentation?

Also the conclusion for Tom's review is a bit deceiving. They keep quoting the 2700X at $378 when it is $305, and the 9600k at $263 opposed to $279. Didn't check the others. They also use those numbers in their graphs. I get that a lot of times places like to use launch prices for consistency but this is rather disingenuous at best. I've never considered them as credible as AT though I still read their reviews. But between things like this and the "just buy it" regarding the 2080 Ti, I think they have lost just about any credibility they had left.

 

[epsilon84]

This is a gaming CPU and people are going crazy that it draws a lot of power running benchmarks which do not reflect the real world.
It looks like while gaming it's going to draw(while overclocked) slightly more power than an 8700k and 1/4 the amount of power my
2080ti sucks up. I have read nearly every review now and the 9900k wins every benchmark. It's the best 8 core CPU
in the world hands down. A lot of reviewers are comparing it to the 2700x, sorry but you can't make that comparison.
The 9900k is on another level of performance.

I'd agree with this, the 2700X is a closer competitor to the 8700K / 9700K - usually comes out on top in highly threaded apps but lightly threaded apps and gaming will favour the CFL chips. They are also much closer in price.

The power draw in gaming seems reasonable considering how high its clocked and the voltages required to attain 5.0GHz.

Reviewers will compare it to the 2700X because it is an 8C/16T CPU, but in terms of 'aggregate CPU performance' its more in line with a TR 1920X in terms of MT performance and obviously ahead in ST.

TPU actually rates it as 'equal' to a 2950X across their CPU bench suite - though I think this is rather misleading because the 2950X would have a big lead in MT but an equally big deficit in ST performance:
https://www.techpowerup.com/reviews/Intel/Core_i9_9900K/19.html

 

[epsilon84]

Maybe Mark can update his OP now? It's not 221W under load... it's interesting how nobody thought the Anandtech numbers were off, but instead thought the other reviewers showing 'proper' power numbers were the 'outliers'.

Notice: When we initially posted this page, we ran numbers with an ASRock Z370 board. We have since discovered that the voltage applied by the board was super high, beyond normal expectations. We have since re-run the numbers using the MSI MPG Z390 Gaming Edge AC motherboard, which does not have this issue

 

[IEC]

I'd expect retail chips to be closer to the original numbers. 1.25V for 5GHz across 8c/16t is a little hard to believe as a typical result. Especially considering my i7-8700K requires more than 1.25V for 4.7GHz...

Probably hand-picked golden samples for reviewers, as usual.

 

[epsilon84]

I'd expect retail chips to be closer to the original numbers. 1.25V for 5GHz across 8c/16t is a little hard to believe as a typical result. Especially considering my i7-8700K requires more than 1.25V for 4.7GHz...

Probably hand-picked golden samples for reviewers, as usual.

Wouldn't you expect a 9900K to be more aggressively binned than a 8700K though? But I agree 1.25V for 5.0GHz is probably not realistic except for golden samples, I would think 1.3V+ would be the norm.

 

[Elfear]

This is a gaming CPU and people are going crazy that it draws a lot of power running benchmarks which do not reflect the real world.
It looks like while gaming it's going to draw(while overclocked) slightly more power than an 8700k and 1/4 the amount of power my
2080ti sucks up. I have read nearly every review now and the 9900k wins every benchmark. It's the best 8 core CPU
in the world hands down. A lot of reviewers are comparing it to the 2700x, sorry but you can't make that comparison.
The 9900k is on another level of performance.

Guess it depends on what resolution and settings you play at. According to TPU, at 1440p a 9900k is 4% faster than a 2700X and at 4k that dwindles to 1.6%. If you game on a higher res ultrawide (3440x1440), that puts the 9900k at ~3% faster. I would venture to bet that 1.6-4% difference is indistinguishable on average. The 9900k's lead will increase with a 2080Ti (TPU used a 1080Ti) but the difference won't be huge.

But I agree, if you play at 1080p on medium settings with a 2080Ti, the 9900k is the chip for you.

 

[epsilon84]

Guess it depends on what resolution and settings you play at. According to TPU, at 1440p a 9900k is 4% faster than a 2700X and at 4k that dwindles to 1.6%. If you game on a higher res ultrawide (3440x1440), that puts the 9900k at ~3% faster. I would venture to bet that 1.6-4% difference is indistinguishable on average. The 9900k's lead will increase with a 2080Ti (TPU used a 1080Ti) but the difference won't be huge.

But I agree, if you play at 1080p on medium settings with a 2080Ti, the 9900k is the chip for you.

Just on this point - not everyone games at 'ultra' settings. I personally don't, because the performance vs IQ trade off isn't there. I'll normally bump the IQ settings down a couple of notches to maintain 100fps+ whenever possible, and to keep min fps above 60fps. High refresh rate gamers (144Hz or above) are also likely to tweak settings to be less GPU bound and to get better framerates closer to their monitor refresh rate. This is where a faster CPU also comes into play.

By the way, why is everyone pretending the 9900K is the only competitor to the 2700X? I find that rather perculiar. There is the 9700K/8700K which by virtue of pricing alone (as well as overall performance) is a much closer match to the 2700X.