Do we need to re-review the Core i9-9900K?

[Steve]

Benchmarking the i9-9900K with unlimited and 95-watt limited results might show two very different sides of the same CPU. Since the Core i9-9900K reviews went live, a few strange anomalies have been spotted, mostly relating to its power and thermal performance...

Read the full article here.

https://www.techspot.com/news/77313-do-need-re-review-core-i9-9900k.html

 

[Puiu]

It is really hard to check the VRM config of a motherboard when making purchases if you don't find some reviews that specifically mention it. You can get burned even if you buy something from a big brand (BTW thanks Techspot for the B450 "best of" article for helping me pick a good cheap mobo)

TL;DR if cheaper boards or boards that have bad VRMs affect performance by more than 5% then you should do the additional testing that has the 95W TDP enforced (or how you think it's best).

 

[ET3D]

I think it would be good to do a full comparative review of the two modes, to highlight the real life difference and let people make their minds about whether it's important or not. ASUS users will have this limit by default, and other users might care about power use and will want to run the CPU at its specced TDP, or at least know how it affects performance.

Personally I'd be interested in seeing such a comparison, and also throwing a 9700K into the mix. Not going to buy either, but that's precisely the kind of in depth investigation I want from a good tech site.

 

[illusio12]

Excellent preliminary analysis @Steven Walton, and to your final questions:
1: " is it misleading........"; YES, to not show default ( Spec Tdp ) configuration ........, by default! Right?

2: Short and Long form tests are great ( loads of extra work for you dudes though! ); in reality, most tasks will heat-load the cpu ( All of gaming, besides some benchmarks methinks! ) and default performace is crucial, before any vendor-applicable turbo-boost / PBO is activated and then comparative ( boosted ) benchmarks are charted.

Personally I prefer the default-spec testing for baseline purposes, then see with boosted TDP etc as part of the overclocking section of the analysis. But really.....; possible 20% reduction in performance ( which must put the 9900K closeer to 5% faster/slower over-all than the 2700X at the same clocks? ), methinks the 2700X is sitting really pretty right now as THE buy for performance AND efficiency!

Also, its outstanding to be using those two words with respect to AMD. Long time !

Edit 1: for clarity.

 

[godrilla]

What ever happened to the extreme line up thats right it turned to core i9 and even some core i7 nomenclature and this came out with similar pricing to the lowest extreme cpu but without all the bells and whistles. The chip runs too hot for me to run long term ( I mean im still running a i7 980 xe at 4.3 ghz for past 8.5 years lol that runs cooler )
All the links here for availability show anywhere from $900 to $999 pricing currently.
I'll definitely wait for next years Ryzen 3000 series on paper it hits all the bells and whistles especially the 15% ipc gain ( isent the i9 9990k around 15% better than a 2700k) the gap closes and the there is support for next pcie standard 4.0 talk about innovation.
If you must have the fastest today at the almost the worst value in history compared to the competition then by all means go with the i7 9990k lol oops I mean i9990k.

 

[Forebode]

Both should be shown with price differences.. It should also be shown how's the best way to differentiate from a board that supports unlimited and not. Those like me, will read all data before picking out parts (remember when overclocking was settings on the physical board and IRQ management was required if you wanted sound and video, among other things?).

To many, building a pc is like buying legos and plugging them together and it "should" work. I don't know if PCpartpicker differentiates between the 2 types.

 

[MetXallica]

If you're buying a "K" series CPU, you bought it to overclock. If you're overclocking, you don't care about TPD. DUH!

 

[meric]

This is intel version of FX 9590

 

[Vulcanproject]

This is intel version of FX 9590

Except this is very fast and good

Quoted TDP is a bunch of nonsense and has been for years with most high end consumer desktop parts.

If you're buying a platform and CPU for servers you better take careful note of true performance per watt but if you're buying high end for consumers then really, who cares.

 

[SpaceSecks]

If you're buying a "K" series CPU, you bought it to overclock. If you're overclocking, you don't care about TPD. DUH!

Meh, I kinda care, mainly because I make an assumption when buying an Intel CPU about the amount of overclocking headroom I will have. If I buy the chip with stated frequencies around 5ghz, I'm going to assume I can push the chip well past that without running in to temp issues. What it sounds like here is if you want to get that 5ghz you will already be running hot with not as much room to push it on your own. I'm a long term Intel fan but it does sort of sound like they are stat padding on this one, which is kind of disappointing imo.

 

[Lew Zealand]

K series CPU: 2 tests— unrestricted TDP as it's an overclocker's chip, and restricted to stated TDP for reference and level playing field with AMD.

Non- K series CPU: at stated TDP only. This may be obvious but even my B360 board allows me to set the TDP on my Core i5-8400 above the rated 65W, which could be useful as it will consume 72W at all-core 3.8GHz in Handbrake with a 1080p h.265 conversion. In this case, I instead choose to undervolt 0.08 W and it runs at 63W with the same test so no TDP increase needed.

 

[ACE76]

Intel hasn't done anything to improve their architecture in years...they are basically factory overclocking their CPUs to keep pace with AMD...This dumpster fire CPU is literally two 7700k CPUs slapped together...and the power consumption shows it...if you test it with the intended TDP, AMD 2700x is hands down the better CPU.

 

[pawel04]

In an ideal world I would like to see how the CPU performs in both instances just as you mentioned in the last paragraph. I guess as a reviewer your job is to inform the consumer on how a product performs and to get the clearest idea id love for a reviewer to show me the whole picture.

You've already done a fantastic job of bringing this to light as I wasnt aware of the issue before (partly because I have no interest in buying this or any other Intel CPU), but for others to get a complete picture of the product I believe its necessary to see how the product performs under all situations.

PS. im also on the 'let it burn' boat...

 

[ACE76]

This is intel version of FX 9590

Except this is very fast and good

Quoted TDP is a bunch of nonsense and has been for years with most high end consumer desktop parts.

If you're buying a platform and CPU for servers you better take careful note of true performance per watt but if you're buying high end for consumers then really, who cares.

The 9900k when overclocked reaches 80-90c even with a custom loop water cooling setup...it's a garbage slap job of a CPU that Intel put out to screw their customers.

 

[Ser01]

The big question is what would you prove by providing different testing scores ?
This abomination of a CPU is a clear attempt to downgrade the opposition (read AMD) and milk the cash cow for a little while longer . Leave tests as they are and wait for a proper CPU to show up,meaning a new generation from Intel , whenever that is ...

 

[ACE76]

If you're buying a "K" series CPU, you bought it to overclock. If you're overclocking, you don't care about TPD. DUH!

Meh, I kinda care, mainly because I make an assumption when buying an Intel CPU about the amount of overclocking headroom I will have. If I buy the chip with stated frequencies around 5ghz, I'm going to assume I can push the chip well past that without running in to temp issues. What it sounds like here is if you want to get that 5ghz you will already be running hot with not as much room to push it on your own. I'm a long term Intel fan but it does sort of sound like they are stat padding on this one, which is kind of disappointing imo.

Intel took any overclocking headroom away just to show up fake numbers to stay relevant...if you push the 9900k to 5.1-5.2ghz, the temps skyrocket past 90c

 

[Sausagemeat]

I would say review it as the consumer would use it. If a consumer has to actively go in and change settings to get lower speeds to keep to the manufacturers TDP then benchmarking that setup makes about as much sense as benchmarking Ryzen in gaming with game mode switched on.

Of course call Intel’s bs out with the publishing the 95watt figure. But publishing results from gimped motherboards to match the manufacturers bs wouldn’t really help consumers much. AMD fans would love it but if people who actually buy the products are going to get much better performance than that (along with much higher temps) then the review would be misleading.

I have always used reviews to determine heat output and power consumption and take TDP figures with a mountain of salt. I imagine most other users do the same.

 

[mrjgriffin]

I think this is the first cpu I've ever not liked lol.

 

[Maxiking]

Intel hasn't done anything to improve their architecture in years...they are basically factory overclocking their CPUs to keep pace with AMD...This dumpster fire CPU is literally two 7700k CPUs slapped together...and the power consumption shows it...if you test it with the intended TDP, AMD 2700x is hands down the better CPU.

If you have no clue what you talk about, just be silent?

You would need liquid nitrogen cooling to reach the 9900k default frequency on 2700x so what's the point of this comment.

Also, according to you, Intel hasn't done anything in years, yet still faster clock to clock and consuming less power than AMD, woah. Silence is golden.

 

[WorthySK1LLZ]

I say that we Test both TDP and unlimited! atleast for a few benchmarks. After that is done I would just go to Unlimited while showing all core frequency.

 

[Maxiking]

Also, I don't remember you testing AMD cpus with TDP locked, AMD TDP is also above 105W in plenty of tasks.

 

[Lew Zealand]

Also, I don't remember you testing AMD cpus with TDP locked, AMD TDP is also above 105W in plenty of tasks.

Yup. Have a look at the Ryzen 7 2700X review here at TS. https://www.techspot.com/review/1613-amd-ryzen-2700x-2600x/page4.html

The 8th gen Intel i7 and Ryzen 2xxx 6 and 8 core systems are consuming between 179 and 214W in Handbrake before overclocking, while the Core i5-8400 is consuming 116W. My 8400 pegs at it's 65W TDP in HB if not modified so that argues that at least the "95W" Intel CPUs should only be using around 150-160W system power if consuming 95W at the CPU level. I can't speak to the 6 and 8 core Ryzen system draw but even the 2600X using 179W seems to be a high total system power draw for a "95W" chip. The rest of the board uses 84W? Seems high.

 

[Brahman05]

I think cpu's should be reviewed both stock and maxed. Most sites do this anyway already. Both intel and amd have 'reccomended' settings so this is nothing new. It just means if you buy a machine from hp or dell its gonna run at stock. Us enthusiasts like to push that OC button and any reviewer that dosent do the same is lame. I feel this highlights a problem with motherboard manufacturers not clearly marketing their product. I almost dont see the point to this article. Excessive heat and wattage is why we buy aio liquid loops and aftermarket waterblocks, so to me this is nothing new. It does, though, highlight the fact that available power delivery is NEVER listed for any motherboard sold ever, a metric that to me, someone with a degree in electronics, is a no brainer.

 

[wukayinij]

Would also love to see the 9900k (or a future i7/i9 CPU with 2MB L3$) benchmarked with hyper-threading disabled. This would be helpful for folks who are disabling (or considering disabling) SMT due to recent published exploits and also for comparing performance of a HT-disabled 2MB L3$ CPU to a 1.5MB L3$ CPU like the 9700k.

 

[BadThad]

Great article! IMO, the TDP limit should be enforced for testing. Anything else invalids an apples to apples comparison.