News Intel Core i9-10850K CPU Benchmarks: Cheaper, but Nearly Identical to 10900K

[Admin]

The Intel Core i9-10850K offers nearly identical performance to the expensive Core i9-10900K, but at a big discount.

Intel Core i9-10850K CPU Benchmarks: Cheaper, but Nearly Identical to 10900K : Read more

 

[NightHawkRMX]

I'm glad to see this pop-up. I won't argue about better value.

 

[everettfsargent]

While the MSRP for the i9-10900 (without or with the F) is lower then the i9-10850K/i9-10900K/KF, I have NEVER seen the i9-10900 selling, at retail, for less then i9-10900K. It has almost always been $50US higher (and a very similar story can be told for the i9-9900 vs the i9-9900K).

I would be very interested in a direct test of the i9-10900 against both the i9-10900K/i9-10850K. Basically power consumption stress test 100% on all 20 threads. As the non-K parts have a purported 65W TDP while we already know that the K parts have a 125W TDP but in reality act as 250W parts (at maximum loads but without OC) ...
https://www.tomshardware.com/reviews/intel-core-i9-10900k-cpu-review/2
"Intel set its official new PL2 watermark at a 250W TDP, which is double the 125W PL1 rating, and recommends that motherboard makers keep boost activity limited to 56 second bursts (Tau)." ...
During those tests, we recorded up to 332W of power consumption when paired with either the Corsair H115i 280mm AIO watercooler or a Noctua NH-D15S air cooler. Yes, that's with the processor configured at stock settings. "
In other words. do the non-K parts approach 250W or something closer to 125W (i. e. both max out at ~2X their listed TDP with no OC on the K parts)?

Oh and if such a test were ever conducted, please do so on a non-OC board (e. g. any LGA1200 like B460, H470 or H410). Thermals, power and speeds.

 

[everettfsargent]

Oh and teh Egg is currently selling the I7-10700 at $90US below the i7-10700K (or $70US below the i9-9900), I can only wish for a similar price delta for the i9-10900 vs the i9-10090K (provided that the i9-10900K is plentiful, which it is not at the moment and will never be during the fall holiday cycle).

 

[TerryLaze]

During those tests, we recorded up to 332W of power consumption when paired with either the Corsair H115i 280mm AIO watercooler or a Noctua NH-D15S air cooler. Yes, that's with the processor configured at stock settings. "

Not to be pedantic but just like ryzen intel's clocking strategies are opportunistic so the cooler is part of what dictates the maximum power draw when using stock settings and using a way better than 125W cooler will raise that limit.
All of pl1 pl2 and tau only work if there is headroom for them to work,if you want to see stock settings you would have to use the recommended cooler and not more.

 

[everettfsargent]

Not to be pedantic but just like ryzen intel's clocking strategies are opportunistic so the cooler is part of what dictates the maximum power draw when using stock settings and using a way better than 125W cooler will raise that limit.
All of pl1 pl2 and tau only work if there is headroom for them to work,if you want to see stock settings you would have to use the recommended cooler and not more.

Actually, you are 100% wrong. Power = Power. A million dollar custom water cooler doesn't change the power requirements one iota. Currently the i9-10900 consumes ~340W regardless of cooling solution.

Now thermals are a different matter, better thermals leads to higher OC's AND higher power requirements.

Take this nonsense article, for example ...
https://www.tomshardware.com/features/overclocking-intel-core-i9-10900k

Notice how total CPU power draw is NEVER mentioned. Funny that. In fact, the power draw curve is highly nonlinear, meaning concave up with increasing slope the entire way. This is really very basic EE stuff.

See also ...

First Core i9-10900K Review Pits Intel’s Flagship Processor Against AMD’s Ryzen 9 3900X and 3950X

The embargo for Comet Lake-S hasn't lifted yet, but China's TecLab (via VideoCardz) has gone ahead and published the world's first review of Intel's latest flagship processor, the Core i9-10900K. As some of you might...

www.thefpsreview.com

Everyone, and I do mean everyone, is hitting 330W-340W regardless of cooling solution with the i9-10900K.

So obviously, the i9-10900K isn't a power efficient chip and was never meant to be a power efficient chip.

 

[everettfsargent]

Oh and you want to compare energy consumed (e. g. killowatt-hours) for the same task at a set of CPU frequencies for the total system. This is, in fact, quite similar to finding the most energy efficient speed in going from point A to point B. In that case, you are fighting wind speed resistance which goes up ~quadratically (and if you are not moving you are still consuming energy, your body or an IC car engine, keeping in mind that going from point A to point B is a minimal requirement).

I'm not trying to win any race, I am trying to use a given amount of kilowatt-hours most efficiently, that way I get more work done in a given amount of time by using more threads to more than offset the time taken for a single thread running at that same slower speed.

 

[TerryLaze]

Actually, you are 100% wrong. Power = Power. A million dollar custom water cooler doesn't change the power requirements one iota. Currently the i9-10900 consumes ~340W regardless of cooling solution.

Now thermals are a different matter, better thermals leads to higher OC's AND higher power requirements.

It does not consume 340W, it reaches 340W for a whopping 10ms...

for whatever reason intel made it so although there is no conceivable reason for a CPU to hit higher power draws for just 10 milliseconds,that's not enough time for anything other than for benchmarks to pick it up as a max.
My interpretation of "controllable turbo" is that there is enough cooling for the CPU to do this,otherwise it will stick with Pl2 of 250W but that also is only a upper limit that is only sustained if there is enough cooling.

(whenever i say cooling I mean all the things that would influence turbo)
https://www.intel.com/content/www/u...e/10th-gen-core-families-datasheet-vol-1.html

 

[everettfsargent]

It does not consume 340W, it reaches 340W for a whopping 10ms...

for whatever reason intel made it so although there is no conceivable reason for a CPU to hit higher power draws for just 10 milliseconds,that's not enough time for anything other than for benchmarks to pick it up as a max.
My interpretation of "controllable turbo" is that there is enough cooling for the CPU to do this,otherwise it will stick with Pl2 of 250W but that also is only a upper limit that is only sustained if there is enough cooling.

(whenever i say cooling I mean all the things that would influence turbo)
https://www.intel.com/content/www/u...e/10th-gen-core-families-datasheet-vol-1.html

You will have to point me to a direct link to the i9-10900K datasheet in *.pdf format. The two volume set with any updates that would explicitly cover the i9-10900K and others like the i7-10700, i7-10700K and i9-10900. The product datasheet that you are pointing appears meant mainly for U-Series and Y-Series. TIA

PS - I don't know what is up with the most recent Intel documents and/or document numbers. The one you appear to point to is REV5 VOL1 (July 2020) but does not mention the entire list of their most recent 10th generation i9/i7/i6/i3 AFAIK.

 

[everettfsargent]

Oh and you are showing page 53 of this document ...
https://www.intel.com/content/www/u...e/10th-gen-core-families-datasheet-vol-1.html

Subtitled ...
Volume 1 of the datasheet covers processor information for the 10th Gen Intel® Core™ processor families, Pentium® processors, and Celeron® processors for U- and Y-processor lines including technologies, power and thermal management, memory, and more.

I need omething showing PL1=125W and PL2=250W or whatever for the i9-10900K.

 

[everettfsargent]

OK found it ...
https://www.extremetech.com/computi...l1-pl2-and-tau-values-on-10th-generation-cpus

https://cdrdv2.intel.com/v1/dl/getContent/615211?wapkw=10th%20gen%20datasheet

Same 10ms for Tau for the i9-10900K AFAIK.

But PL2 defaults to 28 seconds (?) or 56 seconds (?), no downtime metric AFAIK, so technically PL2 could consume a large total fraction of runtime for stock settings (so ~250W continuous at stock settings with 10ms bursts to ~340W at a TBD frequency in Hz). Would have to sample at a pretty high frequency (e. g. over 100 Hz) to see what all is really going on.

My main problem though, is with this and other review websites, no specifics on total power (sic that should be energy, like watt-seconds=joules) used in a benchmark, misleading reports without explanation that ~340W is ~instantaneous and NOT continuous. That is, if you care at all about energy efficiency (which probably means only myself given the overriding need for speed most gear heads demand).

Perhaps that is the main reason that I don't post on these websites.

The so-called experts just don't know what they are doing in terms of power management and measurements to derive total power consumption for targeted benchmarks.

 

[NightHawkRMX]

I do think this power consumption is too high, however:

The CPU will only draw that much power if the motherboard has enough vrm to power it and the CPU has enough cooling to allow it.

If you have a beefy mobo and aio, why not use them to the fullest and get the most amount of performance from your CPU?

If you don't want it to, it's not like it can't be disabled in 5 min.

 

[everettfsargent]

OK better link to current Intel docs ...
10th Gen Intel® Core™ processor families, Pentium® processors, and Celeron® processors for U, H and S processor lines
https://www.intel.com/content/www/us/en/products/docs/processors/core/core-technical-resources.html

 

[everettfsargent]

I do think this power consumption is too high, however:

The CPU will only draw that much power if the motherboard has enough vrm to power it and the CPU has enough cooling to allow it.

If you have a beefy mobo and aio, why not use them to the fullest and get the most amount of performance from your CPU?

If you don't want it to, it's not like it can't be disabled in 5 min.

You could at least try to read AND understand what all I am trying to say. I will always take more threads than MHz because more threads is a nearly linear process (e. g. going from 16 to 20 is an improvement of ~25% with consummate power increase) while going from 4GHz to 5GHz (also an ~25% improvement) consumes much more than 25% additional power.

Same goes for OC'ing memory and OC'ing in general today (it is so last century for what now is much less than ~10% improvement).

You want to play games and I get that. I do not want to play games and perhaps you should also get that?

 

[NightHawkRMX]

My comment was not aimed at you or anyone in particular so no I didn't bother reading your comments in completion. Those were just my thoughts in general about 10900k power consumption.

Yes, its more linear performance to power consumption, but ONLY if the application can scale perfectly across those extra cores. Some applications do, many do not.

I overclock my hardware and honestly don't care THAT much about efficiency. The cost of energy is under 6 cents per kilowatt-hour here.

 

[everettfsargent]

My comment was not aimed at you or anyone in particular so no I didn't bother reading your comments in completion. Those were just my thoughts in general about 10900k power consumption.

Yes, its more linear performance to power consumption, but ONLY if the application can scale perfectly across those extra cores. Some applications do, many do not.

I overclock my hardware and honestly don't care THAT much about efficiency. The cost of energy is under 6 cents per kilowatt-hour here.

I write my own code, have done so for nearly fifty years now. In one particular case, heck just about all situations, the code is single threaded, so that however many threads there are that is how many copies of the code I run. It all goes into ensemble averaging.

 

[NightHawkRMX]

Keep in mind, I am not saying this from an Intel fanboy. I have owned many AMD and Intel systems. My PC is all AMD.

 

[TerryLaze]

My main problem though, is with this and other review websites, no specifics on total power (sic that should be energy, like watt-seconds=joules) used in a benchmark, misleading reports without explanation that ~340W is ~instantaneous and NOT continuous.

Techpowerup is the only site I know of that does that but it's extremely limited,only one single and one multi threaded bench.
https://www.techpowerup.com/review/intel-core-i9-10850k/18.html

 

[NightHawkRMX]

I would assume the reason the r9 draw so much power 1 thread is due to the boosting algorithms.

 

[PCWarrior]

See also ...

First Core i9-10900K Review Pits Intel’s Flagship Processor Against AMD’s Ryzen 9 3900X and 3950X

The embargo for Comet Lake-S hasn't lifted yet, but China's TecLab (via VideoCardz) has gone ahead and published the world's first review of Intel's latest flagship processor, the Core i9-10900K. As some of you might...

www.thefpsreview.com

Everyone, and I do mean everyone, is hitting 330W-340W regardless of cooling solution with the i9-10900K.

So obviously, the i9-10900K isn't a power efficient chip and was never meant to be a power efficient chip.

Why do you cite a questionable Chinese review released before the review embargo of the 10900K? Why don’t you cite any of the legitimate reviews from reputable sites? Anyway, if you watch the video of that Chinese site (here) you would know that this 338W power figure was the TOTAL SYSTEM power draw and that FROM THE WALL. By the way for the 3900X the same test reads 262W and for the 3950X it reads 306W. Also if you noticed the screen for the 10900K, the software power reading for the package was only 211.8W with a peak recorded power 212.62W. Here is the screenshot.

I would assume the reason the r9 draw so much power 1 thread is due to the boosting algorithms.

The reason is that Intel’s 14nm++++ process is more power efficient than TSMC’s 7nm process for frequencies higher than say 4.35GHz . It is fairly comparable to TSMC's 7nm efficiency at lower frequencies. The thing is that AMD’s equivalent priced cpus pack more cores, cache, etc so they can outscore Intel cpus at purely multithreaded workloads while running at 4GHz. At 4GHz, TSMC 7nm, runs at an optimal point in the power frequency curve - it is around the upper limit where power and performance scale fairly linearly. Intel, on the other hand, in order to get as much MT performance as possible they push their fewer cores (10cores in the case of the 10900K) at much higher frequencies (4.9GHz in the case of the 10900K). This means they operate in a less efficient part of the power-frequency curve - definitely Intel 14nm++++@4.9GHz is a much less efficient operating point compared to TSMC's 7nm @4-4.2GHz .