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

[arandomguy]

I think you are understating the issue, some consumers are going to be spending $500+ on this CPU, buying a respectable 120w air cooler and getting a big shock when some of the tasks are not getting full performance they expected.

Is this going to be the case though?

I think a lot of this is due to a combination of how Intel's CPU uarch, their behind the times power management/clocking as well as the test suite used and how reviewers present that data.

GPUs now heavily throttle down in something like Furmark to stay within power targets, but we don't say that they are inadequately specced and that gaming performance benchmarks are then misleading.

Using something like Prime95 leveraging AVX2 to that extent itself is misleading. If you are going to use that data then at least I feel you should provide a corresponding benchmark.

Tomshardware -
https://www.tomshardware.com/reviews/intel-core-i9-9900k-9th-gen-cpu,5847-11.html

Tom's does power measurement over multiple task types. Only Prime95 is putting out enough load to exceed TDP by that large of an amount. So at least within their own gaming and workstation benchmarks the numbers are reflective with a lower power consumption.

Gamer's Nexus -
https://www.gamersnexus.net/hwrevie...paste-delid-gaming-benchmarks-vs-2700x/page-3

Here they test power consumption while streaming. This would be higher CPU load test from a gamers perspective. At here it suggests that any cooler that is specced for the 2700x's 95w would performance functionally well enough for the 9900k's 95w TDP in this type of work load with respect to their benchmarks.

Techreport -
https://techreport.com/review/34192/intel-core-i9-9900k-cpu-reviewed/13

Here they test a high CPU load using Blender. The power draw itself is high but move down to the total task energy section. You'll now see it isn't an efficiency issue per say which is something that's being brought up with respect to the power numbers.

 

[french toast]

Is this going to be the case though?

I think a lot of this is due to a combination of how Intel's CPU uarch, their behind the times power management/clocking as well as the test suite used and how reviewers present that data.

GPUs now heavily throttle down in something like Furmark to stay within power targets, but we don't say that they are inadequately specced and that gaming performance benchmarks are then misleading.

Using something like Prime95 leveraging AVX2 to that extent itself is misleading. If you are going to use that data then at least I feel you should provide a corresponding benchmark.

Tomshardware -
https://www.tomshardware.com/reviews/intel-core-i9-9900k-9th-gen-cpu,5847-11.html

Tom's does power measurement over multiple task types. Only Prime95 is putting out enough load to exceed TDP by that large of an amount. So at least within their own gaming and workstation benchmarks the numbers are reflective with a lower power consumption.

Gamer's Nexus -
https://www.gamersnexus.net/hwrevie...paste-delid-gaming-benchmarks-vs-2700x/page-3

Here they test power consumption while streaming. This would be higher CPU load test from a gamers perspective. At here it suggests that any cooler that is specced for the 2700x's 95w would performance functionally well enough for the 9900k's 95w TDP in this type of work load with respect to their benchmarks.

Techreport -
https://techreport.com/review/34192/intel-core-i9-9900k-cpu-reviewed/13

Here they test a high CPU load using Blender. The power draw itself is high but move down to the total task energy section. You'll now see it isn't an efficiency issue per say which is something that's being brought up with respect to the power numbers.

You raise good points but I am going to stick to my guns on this one, I think Anandtech had it spot on when they said we need tdp low and tdp peak numbers (something to that effect).

 

[coercitiv]

So whats it gonna be? I remember you specifically calling the 9900K as 'poor value' for productivity when the platform cost of a 1920X + X399 motherboard is about the same as a 9900K + Z390 motherboard, and they are essentially the same tier of performance

Did you read the post you quoted? The 1950X is $680. Heck, even the 9800X will be coming in at $589.

The advantage of the 9900K is that it is also the best gaming CPU you can buy on top of having 1920X/7900X levels of application performance.

Yeah, people who buy these chips for productivity will be so relieved, you never know when your career takes a turn towards professional gaming.

But people will naturally moan about the negatives such as power consumption and price, conveniently forgetting that its in a whole other performance tier to the $300 - $400 CPUs it is being compared to such as the 2700X and 9700K.

Naturally moan?! I'll tell you what's going to be moaning, all the 1kg air coolers and 280mm AIOs will be moaning very hard to keep this productivity beast happy:

Good thing the 360mm custom cooling price gets offset by that costly 2080Ti, otherwise pricing might have been an issue.

 

[french toast]

Excuse me, where's the overclock?

You compared the 2700x @ 4.2ghz...

 

[Zucker2k]

Edit; And another thing, Intel advertises the 9900k as being great for overclocking due to it's new re invention..soldered Tim, in reality the 9900k is a poor overclocker, it draws massive amounts of power and emits soldering iron type temperatures to hit all core @ single core turbo frequency...some people have squeezed 100mhz higher...right in line with the oft derised 2700x, which does similar but AMD never bigger up its OC.

9900k is a FX 9590 for 2018...just with better performance yet decieving marketing and ratings.

Are you seriously saying this about the 9900k when the 2700x is also rated 105watts and consumes up to 254watts for less performance? This chip does over 5GHz on the 14nm process, what more did you expect? It's beating HEDT chips with up to 50% more cores, what more do you expect? The 9590 is did nothing even close. Sorry, but you need to try harder, if you're going to paint this chip in a bad light.

 

[IntelUser2000]

I think you are understating the issue, some consumers are going to be spending $500+ on this CPU, buying a respectable 120w air cooler and getting a big shock when some of the tasks are not getting full performance they expected.

The tdp should be set from the worst case scenario, and was usually from prime 95 or AIDA64.. something like that, any overreach was usually minor and didn't matter. (~10-15w)

Intel conveniently changed its tdp policy at this time to have an excuse for what was to come, this was complete BS and led to where we are now...a mess.

Exactly. TDP used to be the average worst case scenario power use. It might peak higher than TDP, but it'll have to come down after few mins, otherwise the cooler won't be able to dissipate the heat. That was the whole definition of Turbo 2.0 that came with Sandy Bridge. PL2 was the short Turbo, and it was made for responsiveness, not sustained performance. It took advantage of the fact that power usage goes up instantly while a heatsink takes sometime to heat up.

Datasheets also said moving average power has to be in line with PL1 value, meaning because of power limits it has to come down.

They won't be able to pull such shenanigans in mobile without serious consequences. TDP = power for such constrained form factors. That's kind of true in desktop too, but being able to use oversized HSF combos change that somewhat.

https://www.anandtech.com/show/13400/intel-9th-gen-core-i9-9900k-i7-9700k-i5-9600k-review/21

Look at the page above. Old PL1 = New PL2. What about when they are selling those things? New PL1 = TDP, which is less than necessary to keep clock speeds as advertised.

1.25V for 5GHz across 8c/16t is a little hard to believe as a typical result.

We made little progress since Kabylake for maximum overclocks. I doubt anything has changed. All they are doing is being able to make better golden samples. Anything above 5GHz is still the "press clock". What the new process allows is to reach those clocks that were previously meant for overclockers.

 

[Zucker2k]

I think you are understating the issue, some consumers are going to be spending $500+ on this CPU, buying a respectable 120w air cooler and getting a big shock when some of the tasks are not getting full performance they expected.
Worse still any hope of some mild overclocking is a non starter and they might be a bit closer to their psu limit than they would have liked or planned, even the mobo they are rocking might not provide the power requirements for max performance?
Even some reviewers with moderate water coolers are experiencing some throttling as linked earlier.

This TDP issue is not solely Intel, 7700k and 1800x maybe started this recent stupidity, but it was not that bad.
The tdp should be set from the worst case scenario, and was usually from prime 95 or AIDA64.. something like that, any overreach was usually minor and didn't matter. (~10-15w)

6900k was fine.
https://www.tomshardware.co.uk/intel-core-i7-broadwell-e-6950x-6900k-6850k-6800k,review-33569-9.html

7700k @ 1800x = where it starts, 8700k; Things take a turn for the worst...2700x..not too bad here, must another test I saw that it exceeded tdp...
https://www.tomshardware.co.uk/amd-ryzen-7-2700x-review,review-34307-12.html
You can see there the 8700k is where this starts to get out of hand...in prime95 8700k consumes 160w...way past it's tdp and a complete difference to past processors which mostly stayed within it's rating or near it so not to matter.

Intel conveniently changed its tdp policy at this time to have an excuse for what was to come, this was complete BS and led to where we are now...a mess.

1800x, 2700x 7700k could be coaxed past it's tdp a bit, no danger really, from the 8700k onwards it has gotten silly and a consumer can no longer look at the box for a guide on cooling/power requirements...sad.

If Intel would stop this dishonest crap with benchmarks, tdp rating and overclock boasts and was just upfront then I would have only one complaint..price, but that is subjective and could be argued it is worth it for it's leading performance..even if I wouldn't agree with that..this is a $400 CPU in my opinion.

All this sudden concern for tdp rating is funny to watch. Moreover, your issue with both AMD and Intel is PBO and Turbo respectively. It's like you're mad your chip is going faster than advertized; that's right, that's where the power is going. The idea is simple, if conditions allow, run faster and finish a task earlier. Why's that a bad thing?

 

[The Stilt]

Please can your share the Voltage/Frequency chart?

thanks

Once I've mapped it for non-256b workloads as well.

 

[IntelUser2000]

The idea is simple, if conditions allow, run faster and finish a task earlier. Why's that a bad thing?

If you are playing games, does a faster CPU allow you to finish playing the game in an hour instead of two? What about when you need to run a server farm with constant load? Will it speed that up?

There are some things that by nature aren't bursty.

We made little progress since Kabylake for maximum overclocks.

Continuation of the above.

Remember Devil's Canyon? With the 4790K they said 5GHz overclocks. Actual samples couldn't do it. They claimed 5GHz for 6700K. Then with 7700K they were finally able to get it with few real world samples.

5GHz is a practical barrier with air overclocks just like Mach 3 is the practical barrier for airplanes. Some fundamental limit that makes leakage and power consumption go parabolic.

 

[Zucker2k]

You compared the 2700x @ 4.2ghz...

And where does PBO clock it to?

You know what's funny, if I were to say the 2700x is a poor overclocker, I'll have an army descend on this thread quoting this same 4.2 and even 4.35Ghz as max clocks for this chip, yet when it's about power consumption, all of a sudden some have a problem. You can't eat your cake and still have it. Imagine a "fair" comparison at 4.7GHz!

 

[french toast]

Are you seriously saying this about the 9900k when the 2700x is also rated 105watts and consumes up to 254watts for way less performance? This chip does over GHz on the 14nm process, what more did you expect? It's beating HEDT chips with up to 50% more cores, what more do you expect? The 9590 is did nothing even close. Sorry, but you need to try harder, if you're going to paint this chip in a bad light.

Wait..what? 254watts? I think you are conflating things...

All this sudden concern for tdp rating is funny to watch. Moreover, your issue with both AMD and Intel is PBO and Turbo respectively. It's like you're mad your chip is going faster than advertized; that's right, that's where the power is going. The idea is simple, if conditions allow, run faster and finish a task earlier. Why's that a bad thing?

I understand how turbo's work, but tdp used to take this into consideration, also to get that turbo frequency you need a much better cooler than most would realise..including experienced testers apparently....
All I am advocating for is some upfront honesty, the 9900k overclocks similarly to the 2700x..which is a poor overclocker and wasn't advertised as such..but was derided for it on here.
The 9900k needs exotic cooling and lots of power to reach its 'grestest gaming cpu' tag...which the 8700k does not and it is nearly as fast.

If the tdp was 140-160w I wouldn't be complaining and I would ignore prime 95...but 95w is a joke.
9900k is not a good buy at current prices.

 

[french toast]

And where does PBO clock it to?

You know what's funny, if I were to say the 2700x is a poor overclocker, I'll have an army descend on this thread quoting this same 4.2 and even 4.35Ghz as max clocks for this chip, yet when it's about power consumption, all of a sudden some have a problem. You can't eat your cake and still have it. Imagine a "fair" comparison at 4.7GHz!

We were talking about STOCK power consumption...ie turbo's but not OC...related to tdp..what is difficult about this?

 

[IntelUser2000]

Intel got caught with its pants down when AMD started doubling cores with Ryzen, while catching up to them in per core performance.

So this is a result of them rushing in panic. Similar issue regarding rated power and actual used power happened in the Netburst days. Because Netburst chips used way more power than they'd like they had to obfuscate things.

 

[beginner99]

Did any reviewer check SSD performance of 9-series vs 8-series and if the mentioned hardware fixes bring it back up again?

 

[The Stilt]

And where does PBO clock it to?

Based on my testing on several samples, the typical maximum all-core frequency on a 2700X is around 4.05GHz.
The variation is 3.975GHz - 4.075GHz, depending on the specimen.

 

[french toast]

Exactly. TDP used to be the average worst case scenario power use. It might peak higher than TDP, but it'll have to come down after few mins, otherwise the cooler won't be able to dissipate the heat. That was the whole definition of Turbo 2.0 that came with Sandy Bridge. PL2 was the short Turbo, and it was made for responsiveness, not sustained performance. It took advantage of the fact that power usage goes up instantly while a heatsink takes sometime to heat up.

Datasheets also said moving average power has to be in line with PL1 value, meaning because of power limits it has to come down.

They won't be able to pull such shenanigans in mobile without serious consequences. TDP = power for such constrained form factors. That's kind of true in desktop too, but being able to use oversized HSF combos change that somewhat.

https://www.anandtech.com/show/13400/intel-9th-gen-core-i9-9900k-i7-9700k-i5-9600k-review/21

Look at the page above. Old PL1 = New PL2. What about when they are selling those things? New PL1 = TDP, which is less than necessary to keep clock speeds as advertised.



We made little progress since Kabylake for maximum overclocks. I doubt anything has changed. All they are doing is being able to make better golden samples. Anything above 5GHz is still the "press clock". What the new process allows is to reach those clocks that were previously meant for overclockers.

I agree entirely with your statement, intel has changed things probably in response to AMD xfr and their process issues, this is the wrong way to go about it though, we are seeing the old intel recently..winning at all costs.

 

[french toast]

Intel got caught with its pants down when AMD started doubling cores with Ryzen, while catching up to them in per core performance.

So this is a result of them rushing in panic. Similar issue regarding rated power and actual used power happened in the Netburst days. Because Netburst chips used way more power than they'd like they had to obfuscate things.

Arise! sir IntelUser2000....

 

[Zucker2k]

I agree entirely with your statement, intel has changed things probably in response to AMD xfr and their process issues, this is the wrong way to go about it though, we are seeing the old intel recently..winning at all costs.

So, as you can see, Intel didn't start this. The 9900k is a good 400-500Mhz above all core turbo clocks of 4.7GHz. What about the 2700x?

 

[PotatoWithEarsOnSide]

If a CPU never runs at base clockspeed, what use is it to label its TDP at that base?
A 95W cooler may well be enough at 3.6GHz, but once you go above that then your cooling solution isn't enough. The result is that the CPU temperature goes through the roof. Now, if Intel are saying that their CPUs can run hot for years then there is no issue. I'm not seeing 5-year warranties on these things. On the contrary; it's a K chip, so overclocking is what it's designed for, and overclocking invariably invalidates any warranty.
If you're selling a CPU designed to be overclocked, the least you should expect from a rated TDP is what is full load at the spec ACT clockspeed.
You're gonna say that ACT isn't a 24/7 load, so how about a new metric?
If ACT can only be sustained for 10 seconds at the rated TDP, then this piece of information should be clearly labelled.

 

[Zucker2k]

If a CPU never runs at base clockspeed, what use is it to label its TDP at that base?
A 95W cooler may well be enough at 3.6GHz, but once you go above that then your cooling solution isn't enough. The result is that the CPU temperature goes through the roof. Now, if Intel are saying that their CPUs can run hot for years then there is no issue. I'm not seeing 5-year warranties on these things. On the contrary; it's a K chip, so overclocking is what it's designed for, and overclocking invariably invalidates any warranty.
If you're selling a CPU designed to be overclocked, the least you should expect from a rated TDP is what is full load at the spec ACT clockspeed.
You're gonna say that ACT isn't a 24/7 load, so how about a new metric?
If ACT can only be sustained for 10 seconds at the rated TDP, then this piece of information should be clearly labelled.

It will run at base clock if any number of mitigating conditions are met, including running even lower clocks at idle. Once the situation calls for more power and conditions are met, it ramps up power. This is good for the consumer. All of a sudden, XFR and Turbo Boost are bad.

 

[french toast]

So, as you can see, Intel didn't start this. The 9900k is a good 400-500Mhz above all core turbo clocks of 4.7GHz. What about the 2700x?

400-500mhz?? Really?.
5ghz seems the average..and on past history you can bet intel seeded golden samples to reviewers.
But max turbo is 5ghz for two cores...which is about the OC, it is a bit better than 2700x..but not much and you don't need crazy cooling to hit it's weak OC, indeed you could probably hit it on its bundled cooler.

AMD never advertised the 2700x as a 95w gaming and overclocking monster by the way...intel did.
Better off with a 8700k or an OC 2700.

XFR is not the same scenario.

 

[PotatoWithEarsOnSide]

Ryzen tends to ship with coolers that meet the CPUs advertised specs, and then says that a better cooling solution should provide better results, even if not much better.
This 9900k ships without a cooler but states 95W TDP. If you put a 95W cooler with it, it is never going to hit that ACT for even a second. In fact, I saw some links to reviews claiming that it wouldn't even post.
The lesson is simple; ship a cooler that allows the CPU to hit the rated specs. Right now, none of us know what cooler is the minimum required. Nor do we know how a 9900k performs with it.
Sure, no-one will ever use that cooler (or so you'll claim), but what relevance does that have when 720p low are used as a basis for CPU bound gaming tests?
The reviews so far suggest that the regular Joe is going to get punked by this 9900k. That's not good for longterm mindshare.

 

[french toast]

Ryzen tends to ship with coolers that meet the CPUs advertised specs, and then says that a better cooling solution should provide better results, even if not much better.
This 9900k ships without a cooler but states 95W TDP. If you put a 95W cooler with it, it is never going to hit that ACT for even a second. In fact, I saw some links to reviews claiming that it wouldn't even post.
The lesson is simple; ship a cooler that allows the CPU to hit the rated specs. Right now, none of us know what cooler is the minimum required. Nor do we know how a 9900k performs with it.
Sure, no-one will ever use that cooler (or so you'll claim), but what relevance does that have when 720p low are used as a basis for CPU bound gaming tests?
The reviews so far suggest that the regular Joe is going to get punked by this 9900k. That's not good for longterm mindshare.

Exactly! Someone is going to get punked by this nonsense.

 

[inf64]

I will have to say it out loud. Anyone who is justifying a purchase of this hot piece of turd for nearly 600USD needs to have their heads checked. Sorry but I had to.

 

[William Gaatjes]

Did any reviewer check SSD performance of 9-series vs 8-series and if the mentioned hardware fixes bring it back up again?

That is a good question. :thumbsup:
It may not have any serious impact while gaming, but it is good to see how the fix affects theoretical performance versus 8 series , patched and unpatched.