Intel Core i9-9900K Re-Reviewed, 95-Watt TDP Results

IdiotInCharge

NVIDIA SHILL
Joined
Jun 13, 2003
Messages
14,679
Tell Dell that and a large amount of businesses with power users, e.g. smaller editing rigs, image shoop'n etc. There is no 9900 currently that I can see. So if they want the same performance for twice the price of AMD, becoz muh_intel or some shitty software optimisation, or some single thread application while needing mulithreading for others, their only option is a 9900k.

Why would you share an image of an XPS tower? Where's the 1:1 comparison with an AMD-based tower?

And remember that the Intel systems regularly make up for fewer cores with higher clocks. Both more transistors and more clocks increase power draw, so you should really be comparing task efficiency.
 

N4CR

Supreme [H]ardness
Joined
Oct 17, 2011
Messages
4,888
AMD tends to be more egregious with their 'TDP-busing'. Regardless, as you can see, both are being called out.

[of course your memory may be a little short, if you don't remember Netburst- and the real concern with AMD's TDP on Ryzen was that it was limiting performance, and further, much of this was traced to the inexplicably poorly designed motherboards available on release...]
Yes both get called out but one is protected by a legion of rabid fans and paid forum manager types, Intel is above criticism and any attempt to hold them to that will result in goalpost shifting. Hence this thread is a perfect example of the typical arguments used.
Ryzen was process limited, as Zen+ is and as the 9900k is. They are all too close to their upper maximums. Just like the RMA2080Ti and Vega, Fiji, etc.

Yup, just like with the 2990WX, right?
Sure, just like Intels HEDT too and Xeon, but we are discussing a mainstream desktop part here.
But if you want to go to the big end of things, there are a few reviews showing Xeons going up against EPyc, with Xeons getting asses handed to them even with 100W extra use (breaching TDP by 1/3rd!)

And they run OEM systems, and should evaluate the performance of the system in addition to what is expected of the CPU.
Which is pretty hard when most reviews show it at much higher clocks than what is allowed by the actual TDP ;)
Aka, they are not getting the performance they think they are getting and that is the point of this article.

At stock, the Intel CPU does more with less watts ;)
As we just saw with the 95-watt limit, it’s barely any faster than the Ryzen 7 2700X. In fact, in some tests it’s slower, and that’s an awful result for a CPU that costs ~70% more.
Did we read a different article? It does about the same, with ~800MHz more clock and over double the price. As you said netburst, yeah, Intel sure looks like Netburst 2.0 right now.

AMD is entirely dependent on third-parties for production. They don't have a process, they have a contract.

You mean AMD might finally catch up?!?
AMD is already at parity with their desktop process especially from a performance per Watt metric as you can clearly see and ahead in server where perf/W is everything. Zen is even more efficient at Epyc clock levels. 7nm will turn that into domination on both sides, especially at like-like TDPs. I can only hope Kyle does a IPC+TDP comparo before clocking the shit out of it.

Why would you share an image of an XPS tower? Where's the 1:1 comparison with an AMD-based tower?

And remember that the Intel systems regularly make up for fewer cores with higher clocks. Both more transistors and more clocks increase power draw, so you should really be comparing task efficiency.

I shared the XPS tower to show that it has a 9900k as an option. And that 9900k will be running at... gasp.. 95W TDP! Or are you going to try and convince [H] that Dell overclocks their business towers? I don't think they even OC Alienware.
1:1 comparison is shown in this article between a 95W TDP Intel and AMD flagship.
As above, 95W TDP vs 95W TDP, 8 core vs 8 core, Intel doesn't do any more work for the same power. Maybe a small single core advantage when everyone does their superPi benchmarks at work...
 

IdiotInCharge

NVIDIA SHILL
Joined
Jun 13, 2003
Messages
14,679
Yes both get called out but one is protected by a legion of rabid fans and paid forum manager types, Intel is above criticism and any attempt to hold them to that will result in goalpost shifting. Hence this thread is a perfect example of the typical arguments used.

Do you not see the hypocrisy in your post? Have you not heard of Reddit?

Sure, just like Intels HEDT too and Xeon, but we are discussing a mainstream desktop part here.

We're discussing the top-of-the-line enthusiast part.

Aka, they are not getting the performance they think they are getting and that is the point of this article.

Not if they review the system.

Did we read a different article? It does about the same, with ~800MHz more clock and over double the price. As you said netburst, yeah, Intel sure looks like Netburst 2.0 right now.

Doesn't look anything like Netburst; Intel tried something new with Netburst, and it was at least competitive. The Dozers weren't, but at least Ryzen is.

And no, I'm not talking about clockspeed- I'm talking about task energy, where the AMD part uses more. You even quoted me!

AMD is already at parity with their desktop process especially from a performance per Watt metric as you can clearly see

Except that they're not?

I shared the XPS tower to show that it has a 9900k as an option.

Yet you failed to show the AMD system that you base your comparison and rant on.
 

N4CR

Supreme [H]ardness
Joined
Oct 17, 2011
Messages
4,888
Do you not see the hypocrisy in your post? Have you not heard of Reddit?
I was referring to here. Not reddigg where there are fanboys for all camps segregated to their own areas usually.

We're discussing the top-of-the-line enthusiast part.
Not these days it isn't since x99 came about. HEDT is enthusiast part and a whole different league. This is a a top of the line desktop part, which is employed by many OEMs without overclocking.
This TDP limitation (without OC) is what a vast majority of 9900k users (which are not enthusaists!) will experience.
Unless you want to tell me that Dell, HP, etc will sell less desktop CPUs than enthusiasts? Businesses buy custom PCs from nobodies now?

Not if they review the system.
Yes sure, but the whole point is here that these are often compared on a '95W tdp' basis. Of course total system watts is measured, but typically when shown in reviews the '95W' figure (which both AMD and Intel will breach under OC) is used as a specification.
But now, when an article shows AMD and INtel desktop flagships competing neck for neck, it's thrown out as 'irrelevant' because it's a 95W TDP and a few people on [H] won't run that?

[/quote]
Doesn't look anything like Netburst; Intel tried something new with Netburst, and it was at least competitive. The Dozers weren't, but at least Ryzen is.

And no, I'm not talking about clockspeed- I'm talking about task energy, where the AMD part uses more. You even quoted me!
Task energy aka performance per watt, where AMD has a lead in server performance per watt, or where AMD is tied in deskop perf/watt as per this article? Keeping in mind skylake is more efficient uarch than the larger core bretheren used in Xeon, thus for some operations on desktop Zen (slightly less efficient that Epyc as it's over the 2nd critical) AMD has a slightly higher W/transaction/ops figure.. yes at a microscale for some operations intel is more efficient. That's what you are seeing in this benchmark, each CPUs strength in perf/W. Overall, it's a wash.

And overall, one is twice the price and has the same performance, when used in an OEM environment,.


Except that they're not?
>this article doesn't exist
>article shows two competing 95W CPUs performance is about the same
>they're not the same
>go to line 5
>argue in circles


Yet you failed to show the AMD system that you base your comparison and rant on.

I didn't show an AMD system because I don't need to, it's not the selection we are discussion - not talking systems, we are talking CPUs. And that CPU in question (the 9900k) is found in OEM systems (as I proved), which sell in far higher volume than the few of us on [H] and enthusiast circles buy. That is what I am trying to make clear and you don't seem to understand this or wilfully are not after the third time explaining it: in the real world, aka OEM market, without OC, the 9900k is the same speed as the 2700k as per this article and costs over twice as much. /end
 

Meeho

Supreme [H]ardness
Joined
Aug 16, 2010
Messages
5,393
Upset about what? That a company wants to sell product? Misleading and lies and BS is something all these companies do. Go to your local store and if you're intelligent enough every product out there will be misleading in one way or another. It's to be expected.

It's all posturing and marketing BS. It's fun and I love it. I'm smart enough not to get into too deep that I get my feelings hurt unlike some of you guys. Intel does it, AMD does it, nVidia does it.

You guys need to relax and care less.
"I don't care." and "Everyone does it so it's not wrong."

Market is as consumer does.
 

IdiotInCharge

NVIDIA SHILL
Joined
Jun 13, 2003
Messages
14,679
I was referring to here. Not reddigg where there are fanboys for all camps segregated to their own areas usually.

Reddit users support the point, but I'm talking about here too.

Not these days it isn't since x99 came about.

HEDT is HEDT, and yet it's not a 1:1 matchup for the desktop parts. And further, if it's a 'K' part, it's an enthusiast part. Period.

This TDP limitation (without OC) is what a vast majority of 9900k users (which are not enthusaists!) will experience.

They'll experience what the system the CPU is installed in allows- the same as with a 2700X and many other modern CPUs.

Yes sure, but the whole point is here that these are often compared on a '95W tdp' basis.

There is no 'basis'. TDP is typically a 'class' that a CPU is placed in. CPUs with different core counts and clockspeeds in the same CPU family have the same wattage rating and have been for at least a decade. Arguing for a hard and fast TDP limitation is arguing for something that does not exist and has not ever existed.

Task energy aka performance per watt

Uh, no. Total energy that it takes to perform a task. Which could be performance/watt*time, but may not be.

In any case, the 2700X is less efficient here than the 9900K.
 

Crystoff

Weaksauce
Joined
Jun 3, 2017
Messages
111
Misleading and lies and BS is something all these companies do.

So you equate what all companies do, which is puffery, with BS. Well, OK. Then you equate BS to being misleading. Could take it that way. Then you equate being misleading with lying. Gee, I guess so.

Thus, we shouldn't complain about lying.

I never thought of it that way - thankfully.
 

Advil

[H]ard|Gawd
Joined
Jul 16, 2004
Messages
2,036
It has never been this. It was never spec'd to be this. It was never designed to be this. EVER by ANYONE.

I understand I oversimplified it. I know it's a thermal design limit and power consumption regularly defies it especially on desktop CPUs which essentially are unlocked on power consumption.

But that's stupid. In earlier times CPUs regularly followed the TDP to within a few watts when tested under high loads.

The problem here is that we need the CPUs to work like graphics cards are. They need to be 100% power target default, and WE have to tell it to break it and by how much.

The default mode for a CPU marked 95W should be right about 95W. And there can be a big honking button on the front of the BIOS that says "PERFORMANCE MODE 130W TDP."

It can even be supported by the factory but the number needs to be understood and obvious how the motherboard is set and how the CPU will behave.

I realize it's not a "big issue" but it should still be more straight up than this.

How long did I take reviewers and enthusiasts to even recognize all the motherboards weren't actually running the CPU at spec? That right there tells us this situation is way too complicated.
 

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
52,939
I understand I oversimplified it. I know it's a thermal design limit and power consumption regularly defies it especially on desktop CPUs which essentially are unlocked on power consumption.

But that's stupid. In earlier times CPUs regularly followed the TDP to within a few watts when tested under high loads.

The problem here is that we need the CPUs to work like graphics cards are. They need to be 100% power target default, and WE have to tell it to break it and by how much.

The default mode for a CPU marked 95W should be right about 95W. And there can be a big honking button on the front of the BIOS that says "PERFORMANCE MODE 130W TDP."

It can even be supported by the factory but the number needs to be understood and obvious how the motherboard is set and how the CPU will behave.

I realize it's not a "big issue" but it should still be more straight up than this.

How long did I take reviewers and enthusiasts to even recognize all the motherboards weren't actually running the CPU at spec? That right there tells us this situation is way too complicated.
About a decade + ago? Actually, AMD Threadripper CPUs actually following TDP under Precision Boost 2 was the biggest surprise we have had in awhile.
 

Brian_B

2[H]4U
Joined
Mar 23, 2012
Messages
3,356
My 2c

You can:
Say it’s faster with an unlocked clock
-or-
Say it’s 95w and therefore more efficient

But you can’t say both at the same time.

It feels like some fuckery going on. 9900k isn’t a bad CPU, but compared to 2700x and 8800k, It sure feels like someone was trying a little too hard to differentiate it
 

N4CR

Supreme [H]ardness
Joined
Oct 17, 2011
Messages
4,888
You didn't read the article here about TR overclocking?
I said
Or 'TDP doesn't matter, this is [H]'.

You said:
Yup, just like with the 2990WX, right? - in relation TDP, probably as some far flung excuse for the 9900k

From the fingers of Kyle himself;
'The 2990WX has a TDP of 250 watts'
'Running Prime95 at stock PB2, the 2990WX CPU Package Power stayed right at ~250W (which is its TDP rating) while pulling 425W at the wall, with our CPU clock settling out at 3.125GHz.'

So AMD HEDT with four damn dies doesn't breach TDP, but the 9900k does and that's fine becoz 'muh_enthusiast' desktop CPU (found in a dell XPS as I posted, at stock TDP) but it also means that it can't actually soundly beat the 2700x without an OC.
So therefore, in OEM scenarios across an average of tasks, the 9900k and 2700k are for all intents and purposes at performance and power parity give or a take a few percent for the traded bench wins. Remember, the article clearly says it's barely any faster and in some tests slower than the 2700x. You seem convinced it's not.

As we just saw with the 95-watt limit, it’s barely any faster than the Ryzen 7 2700X. In fact, in some tests it’s slower, and that’s an awful result for a CPU that costs ~70% more.
.
 

DrBorg

Gawd
Joined
Jan 22, 2005
Messages
555
Isn't overclocking Literally "What We Do Here"?

Fanbois aside, I could give a fuck about the logo on the package; I want to make it crash, repeatedly, until I get it Just Barely Stable across days of operation.

I've bought $100+coolers to hit that; I will not buy 2x 1000W power supplies, and a Vapochill rig to get a 1G oc. It's just not going to happen.

If the computer won't run on 1x 20A breaker; I guess Overclocking is dead for me.

None of the 4 running systems I have on right now are at stock clocks, except for the one with ISA slots; the cards in it crash at abnormal bus speeds, for some reason... :)

The other 3 systems are running at least a GHz overclock.

It's such a part of this site, You'd think it was in the name or something, lol.
 
  • Like
Reactions: N4CR
like this

Meeho

Supreme [H]ardness
Joined
Aug 16, 2010
Messages
5,393
There is no 'basis'. TDP is typically a 'class' that a CPU is placed in. CPUs with different core counts and clockspeeds in the same CPU family have the same wattage rating and have been for at least a decade. Arguing for a hard and fast TDP limitation is arguing for something that does not exist and has not ever existed.
b_intel-power-limit-tau-vs-ewma.png

c_intel-power-limit-100-seconds-10ms.png

d_intel-pl2-vs-pl1-spec.png

2_intel-tdp-investigation-frequency-chart_asus-max.png
 

Johan Steyn

Weaksauce
Joined
Dec 1, 2016
Messages
85
You're missing some context: If you want to see how a 9900K performs in a particular Dell or HP, you go read reviews of those systems. People here are putting them in enthusiast boards.

Further, that's assuming that you'd get one in your garden variety Inspiron or Pavilian- which is unlikely. Nearly all of these OEMs have gaming lines, and can set up the CPUs as they see fit for their systems.

Nope. The issue is not about what the CPU can do for us, but the impression Intel is officially giving. If they say that this is what it can do as an overclocked cpu, fine. This is not what they did

They gave performance numbers as if this is what it is at stock. So yes, I understand your point for enthusiasts, but this is not the issue. It is not about what happens on this board.
 

Advil

[H]ard|Gawd
Joined
Jul 16, 2004
Messages
2,036
Marketing Exec: "This CPU is super efficient at 95W TDP!"

Engineer: "Uh Mr. Executive sir, I'm pretty sure there isn't even one motherboard maker that isn't setting this CPU to 140W power target as default out-of-the-box."

Marketing Exec: "Don't mess with my narrative! It IS super efficient at 95W TDP whether or not anyone ever runs one at that!"

Engineer: "But the benchmarks will all be run around 50W above that..."

Marketing Exec: "So? This just proves that the CPU is both efficient AND fast! No need to confuse buyers trying to explain this."

Engineer: "Technically, don't you mean efficient OR fast sir?"

Marketing Exec: "Shut up and get back to work on the next processor."
 

Olle P

Limp Gawd
Joined
Mar 29, 2010
Messages
331
It is more people upset at intel misleading people...
Also reviews of the CPU get screwed up when the performance at "stock" settings differ significantly based on what motherboard is used.
That's where this common denominator in the form of the official TDP has its place.

From there it's possible to compare and see which motherboard step away the furthest from the norm when it comes to "overclock" out of the box.

You're missing some context: If you want to see how a 9900K performs in a particular Dell or HP, you go read reviews of those systems. People here are putting them in enthusiast boards.
Good luck finding any reliable reviews of such systems...

There is no 'basis'. TDP is typically a 'class' that a CPU is placed in. CPUs with different core counts and clockspeeds in the same CPU family have the same wattage rating and have been for at least a decade. ...
The same official TDP is often (for simplicity) used for a couple of CPUs in the same line, yes. That TDP value is determined by the most demanding of those CPUs, which in this specific case is the i9-9900K.

For me there's a very good reasoning behind the 95W TDP:
Go a bit over a year back, when Coffee Lake was released...
Then it was a great outcry amongst enthusiasts about the fact that their Z270 motherboards weren't allowed to support the new CPUs. The reason given by Intel was that the 200-series of boards couldn't be relied upon to handle the 95W TDP, while 300-series boards were designed for it.
Now the "refresh" is launched, and they're supported by all 300-series motherboards, built for a 95W TDP.

There should be no difference in performance of an i9-9900K if you run it on an MSI Z390 Godlike or some more than one year old Z370 board. Not with stock settings nor at overclocking.
 
  • Like
Reactions: N4CR
like this

DrBorg

Gawd
Joined
Jan 22, 2005
Messages
555
...

For me there's a very good reasoning behind the 95W TDP:
Go a bit over a year back, when Coffee Lake was released...
Then it was a great outcry amongst enthusiasts about the fact that their Z270 motherboards weren't allowed to support the new CPUs. The reason given by Intel was that the 200-series of boards couldn't be relied upon to handle the 95W TDP, while 300-series boards were designed for it.
Now the "refresh" is launched, and they're supported by all 300-series motherboards, built for a 95W TDP.

There should be no difference in performance of an i9-9900K if you run it on an MSI Z390 Godlike or some more than one year old Z370 board. Not with stock settings nor at overclocking.



I didn't realize that the published reason for the non-supported processors was TDP; usually it's the lack of bios updates to handle the new processors, or a new socket type.

Socket 2011 v1 has processors up to 150W at stock clocks; newer processors went to DDR4, and that limited future compatibility.

This x79 Sabertooth mobo doesn't even run warm at 175W, so there's more headroom on the board power.

The lack of unlocked processors except for the E5-16XX processors made an "upgrade" unfeasible; an E5-1680 has two more cores, and 10MB of cache, but unless I can buy one for $100, why upgrade. (last I saw, they're over $1k, used)

If anyone ever figures out how to unlock all intel processors, they'll never sell another Extreme Edition again; The ones I bought for a work computer wouldn't reach the overclock I'm running now, lol.
 

Olle P

Limp Gawd
Joined
Mar 29, 2010
Messages
331
I didn't realize that the published reason for the non-supported processors was TDP; ...
I think the exact wording used by Intel was closer to: "... unable to handle the power demanded by six-core CPUs..." (as opposed to the previous up to four cores).

This next leap from six to eight cores should logically produce the same added demand for more power and consequential outcry about foul play from Intel for not providing backwards motherboard compatibility, resulting in additional bad reputation.
With a 95W TDP any motherboard officially supporting "Coffee Lake" CPUs will (formally) be compatible to the new CPUs.
 

juanrga

2[H]4U
Joined
Feb 22, 2017
Messages
2,804
For when a re-review of the 1800X with TDP limited to 95W and the 2700X with TDP limited to 105W?
 
Top