Review Intel 10th Generation Comet Lake-S Review Thread

[Zucker2k]

Review information on the soon-to-be-released 10th generation desktop lineup, as well as all relevant information will be linked in this thread. OP will be updated as information becomes available in the next few days. Please, post links to reputable sites you want to see in the OP, and I'll add them. Thanks!

Anandtech
Phoronix (Linux Benchmarks)
LTT (YouTube Video)
Gamers Nexus
Euro Gamer
ComputerBase.de
Back2Gaming
HWUB (YouTube Video)
Sweclockers
Nordic Hardware


Reviews Roundup on VideoCardz

 

[ZGR]

I hope to see Intel get more stringent toward motherboard OEM's out of the box power settings.

GamersNexus makes a great point how AMD's power efficiency lead is amplified even more because of motherboard OEM's going overboard on performance and throwing Intel's TDP numbers out the window.

 

[jpiniero]

I hope to see Intel get more stringent toward motherboard OEM's out of the box power settings.

GamersNexus makes a great point how AMD's power efficiency lead is amplified even more because of motherboard OEM's going overboard on performance and throwing Intel's TDP numbers out the window.

The people who do DIY will just overclock anyway. The chips can handle the heat so there's not much downside by having the TDP unrestricted by default.

 

[DrMrLordX]

I hope to see Intel get more stringent toward motherboard OEM's out of the box power settings.

GamersNexus makes a great point how AMD's power efficiency lead is amplified even more because of motherboard OEM's going overboard on performance and throwing Intel's TDP numbers out the window.

AM4 isn't completely free of such shenanigans, though it's less pronounced. Go to OCN and ask The_Stilt about "the thing" on Asus motherboards. He might be able to tell you more. AMD has made it harder since the CPU microcode is designed to aim for a specific package power limit, and it can achieve that no matter what kind of goofy voltage offsets you throw at it. So I guess Asus had to get all complicated or whatever. I never got a good explanation of what is "the thing".

Intel leaves everything up to the motherboard microcode. It can set PL1, PL2, and tau. That design allows mobo OEMs to go nuts with funky out-of-the-box settings.

As with the 9900k (which had some boards feeding it up to 210W PL2 values), I expect Z490 microcode updates to reign in some of the benchmark-oriented power settings in favor of Intel's official spec. Boards like the Z490 Aorus Master will probably not be as crazy in a few months.

 

[coercitiv]

The people who do DIY will just overclock anyway.

Have you ever compared "stock" overclocked settings with those obtained using any decent overclocking guide? Motherboard makers use very poorly optimized settings, often leading to huge power draw spikes and even instability. The point of stock settings on DIY systems is to offer a solid base with predictable performance and thermals for whatever the user feels like doing based on their experience level.

If people who do DIY overclock anyway, why start them off with a very poor default overclock that most likely requires additional steps to correct?

XMP is disabled by default on all systems, why should MCE be enabled by default?

I hope to see Intel get more stringent toward motherboard OEM's out of the box power settings.

You won't see that until they get a product in desktop that can beat the competition at lower power levels.

Winning the performance benchmarks is paramount for Intel marketing, regardless of what Bob Swan is saying.

 

[Zucker2k]

Got mine.

PC Specs:
Intel Core i7 10900K
MSI MEG Z490 ACE
32GB G.SKILL Trident Z RBG DDR4
EVGA GeForce RTX 2080 Ti XC Ultra
512GB Samsung 970 Pro M.2 SSD
Three 4TB Samsung 860 QVO SSDs in 12TB RAID-0
10TB Seagate Enterprise HDD
LG 6X External Slim BD-RW
Corsair Crystal 460X case
NZXT Kraken X72 cooler
Corsair ML fans
Corsair AX1600i power supply
Corsair K70 keyboard
Cooler Master MM520 mouse

That's one badass rig you got there!

 

[jpiniero]

XMP is disabled by default on all systems, why should MCE be enabled by default?

Pretty sure MCE is typically disabled. What's enabled is letting the TDP go to 250+ forever.

 

[coercitiv]

Pretty sure MCE is typically disabled. What's enabled is letting the TDP go to 250+ forever.

Times are changing fast, not even MCE is what it used to be.

This final screenshot is from the ASRock Z490 Taichi, which had the wackiest default behavior of all. The multipliers for active cores are still unchanged, but the multiplier per core is unlimited, meaning that this board will behave similarly to the Ace and the Master, but with any two cores allowed to boost to 5.3GHz rather than only the best cores, which are cores 8 and 9 here.

Now let’s look at the actual CPU frequencies. This is what dictates performance results in games or other applications, and is the most important takeaway from the point of fair performance comparisons between brands.

In this test, the 10900K behaved similarly with the Ace and the Taichi, which we’ll plot first. The Taichi was almost constantly at 5300MHz and boosting more than the expected cores. This behavior would boost the result higher than other boards, which makes ASRock look better on charts and is why they’re doing it; at the same time, it threatens system stability in some use cases, as our viewers have voiced, and it also requires a more serious cooling setup and power source.

Examining a couple boards individually, we can see that the Taichi boosts to 5.3 on any core it pleases, often on three or four cores at once, which breaks even the rules the board has set for itself, since the 3-core active limit on the Taichi is still reported to be 51x.

Source.

 

[lobz]

That's one badass rig you got there!

I appreciate the times when even we can agree on something!

 

[AdamK47]

I appreciate the times when even we can agree on something!

Thanks guys.

 

[piokos]

i5-10500 is really good.

Intel Core i5-10500 Review

The Core i5-10500 is Intel's $200 response to the Ryzen 5 3600 and 3600X. It boosts up to 4.5 GHz more reliably than other Comet Lake CPUs, and gaming performance is excellent, too. Our Core i5-10500 review will also answer whether the Core i5-10500 is worth it over the i5-10400F.

www.techpowerup.com

That would be $430 total for ASRock H470M-ITX/ac + 10500.
3 years ago I paid $360 for ASRock H270M-ITX/ac + 7500 (current USD exchange rate).

But 7500 is fine, so that's more of an emergency scenario.
Rocket Lake (Xe, TB4) and Alder Lake (heterogenous cores) will bring way more new stuff.

 

[Gideon]

i5-10500 is really good.

Why is it significantly better than 10400 or 10400F (if you don't need graphics)?

Honest question. Here the prices are:

• 10400F is 180€
• 10400 is 200€
• 10500 is 230€
• 10600 is 250€
• 10600K is 280€ (though not readily available, promised to be delivered in 1-2 weeks)

What makes the 10500 stand out?

 

[dullard]

Why is it significantly better than 10400 or 10400F (if you don't need graphics)?

Honest question. Here the prices are:

• 10400F is 180€
• 10400 is 200€
• 10500 is 230€
• 10600 is 250€
• 10600K is 280€ (though not readily available, promised to be delivered in 1-2 weeks)

What makes the 10500 stand out?

At the current prices just as the processors hit the market, you are correct, the 10500 does not really stand out. Using your current prices, the 10400F is by far the best value of the i5 processors. It is a similar thing in the US right now, where the 10400 is by far the best value of the i5 processors.

But, these are just typical price fluctuations as new technology gets launched. When the dust settles, the recommended price difference between the 10400 and 10500 is a measly $10 (8.89€). When the price difference becomes that small, the math is pretty simple:

• 6.9% higher base clocks,
• 4.7% higher turbo clocks,
• for what will eventually be only a 5.5% higher price for the CPU and less than 2% for the system as a whole.

 

[piokos]

Why is it significantly better than 10400 or 10400F (if you don't need graphics)?

Honest question. Here the prices are:

• 10400F is 180€
• 10400 is 200€
• 10500 is 230€
• 10600 is 250€
• 10600K is 280€ (though not readily available, promised to be delivered in 1-2 weeks)

What makes the 10500 stand out?

Why 10500 is so much more expensive than 10400? Easy. Because it's vPro-compliant (+SIPP+TXT).
So there's a big difference for OEMs and, as a result, there's a big difference in launch price.
Performance-wise 10500 is a couple % faster .
So probably, much like with earlier generations, actual transaction price gap will be less wide.

If I ordered a CPU today, I'd pay:
970PLN ($246) for the 10400
1050PLN ($267) for the 10500

If you don't need IGP, 10400F is the better option. And it'll remain to be cheaper than 10400.

As @dullard said: you're essentially paying a few % more for a few % more performance.
Actually, it's fairly good value compared to, say, what overclockers pay for pushing their CPUs 200MHz above boost.

 

[fleshconsumed]

At $530 it's $200 more expensive than 3900x (after you include cooling) and gets dangerously close to 3950x pricing of $700. 3950x does not come with a cooler, but it runs cool enough that a quality $40-50 cooler would have no problems keeping up with it. The other day my local Microcenter had open box 3950x for $630, almost snagged it, but decided to wait till the end of the year for cheaper prices, shouldn't be too long now.

 

[Zucker2k]

Ryzen 9 3900X/3950X vs. Core i9 10900K In 380+ Benchmarks...

https://www.phoronix.com/scan.php?page=article&item=3900x-3950x-10900k&num=1

Again, the 10900k stages a strong showing, besting the 3900x in overall tests.

Following our initial Core i5 10600K and Core i9 10900K Linux benchmarks last week, here is a much larger comparison I have been working on since then in looking specifically at the Ryzen 9 3900X and 3950X against the Core i9 10900K. It's the largest to date with nearly 400 benchmarks being tested, most of them real-world test cases.

 

[Markfw]

Ryzen 9 3900X/3950X vs. Core i9 10900K In 380+ Benchmarks...

https://www.phoronix.com/scan.php?page=article&item=3900x-3950x-10900k&num=1

Again, the 10900k stages a strong showing, besting the 3900x in overall tests.
View attachment 22825

I looked at the details. The tests that I had no idea what they were really, is the ones that the 10900k won. There was only one that I knew what it was, that the 10900k won. They did not say what coolers were used, and they did not say anything about power usage, or bios settings (like was PL2 used all the time)

Another flawed set of benchmarks that tell me nothing.

 

[DrMrLordX]

Gee you sure are an avid Linux user now.

 

[Markfw]

Gee you sure are an avid Linux user now.

He is avid with anything that fits his agenda.

 

[RetroZombie]

Ryzen 9 3900X/3950X vs. Core i9 10900K In 380+ Benchmarks...

Very interesting thank you.

Can someone explain this to me, how can this result end up like this:

If it's based on results like this:





How did he made the first graph?
And if the R9 3950X is missing from some of the results how did he made the graph Zucker2k pointed out?

 

[Hitman928]

Very interesting thank you.

Can someone explain this to me, how can this result end up like this:

If it's based on results like this:
View attachment 22828
View attachment 22831
View attachment 22833
View attachment 22835

How did he made the first graph?
And if the R9 3950X is missing from some of the results how did he made the graph Zucker2k pointed out?

There's several peculiarities in these results, for instance:

Polyhedron Fortran Benchmarks
Benchmark: fatigue2
Core i9 10900K: 6 Minutes, 8 Seconds 368s
Ryzen 9 3900X: 46 Minutes, 23 Seconds 2783s
Ryzen 9 3950X: 7 Minutes, 51 Seconds 471s

Those are the results of the Fortran benchmarks used in the geometric mean. Obviously something went very wrong on the 3900x.

Also here:

Facebook RocksDB
Test: Random Fill
Core i9 10900K: 15 Minutes, 17 Seconds
Ryzen 9 3900X: 15 Minutes, 17 Seconds
Ryzen 9 3950X: 15 Minutes, 17 Seconds

All 3 CPUs having the exact same completion time over multiple runs?

And another one:

Selenium
Benchmark: CanvasMark - Browser: Firefox
Core i9 10900K: 7 Minutes, 40 Seconds
Ryzen 9 3900X: 7 Minutes, 31 Seconds
Ryzen 9 3950X: 29 Minutes, 34 Seconds

Here it's the 3950x taking over 4x longer than it should.

And another:

Test: Integer + Elliptic Curve Public Key Algorithms
Core i9 10900K: 12 Minutes, 51 Seconds
Ryzen 9 3900X: 12 Minutes, 51 Seconds
Ryzen 9 3950X: 17 Minutes, 8 Seconds

I really like the doughnut charts Phoronix puts together and he has a very good automated system for running these tests and auto-generating graphs, but it seems that there's some QA (i.e. manual review) that is missing making sure the automated results make sense because clearly this was a very buggy set of benches and I only went through maybe 25% of the results.

Edit: Ok, investigating further in his methodology, this was total run time in my quotes but the results are taken of the average across multiple runs and when there was large deviation in run times the test ran over and over again to get an acceptable deviation. So the crazy run times aren't calculated in the results, just the average of all the runs. There's still some peculiarities though that I think need some review.

 

[Markfw]

Very interesting thank you.

Can someone explain this to me, how can this result end up like this:

If it's based on results like this:
View attachment 22828
View attachment 22831
View attachment 22833
View attachment 22835

How did he made the first graph?
And if the R9 3950X is missing from some of the results how did he made the graph Zucker2k pointed out?

Without going into detail, I kind of observed the same thing. The end result does not seem to equal the details.

You know the saying “There are three types of lies -- lies, damn lies, and statistics.”

Most of all "geometric mean" could mean that on average, Intel loses everything, but the mean makes them winners.

I used to work in statistics, so I know how you can twist facts.

 

[SAAA]

Without going into detail, I kind of observed the same thing. The end result does not seem to equal the details.

You know the saying “There are three types of lies -- lies, damn lies, and statistics.”

Don't worry: Intel themselves said not to look at benches so no lies and no statistics either, right?

 

[lobz]

Very interesting thank you.

Can someone explain this to me, how can this result end up like this:

If it's based on results like this:
View attachment 22828
View attachment 22831
View attachment 22833
View attachment 22835

How did he made the first graph?
And if the R9 3950X is missing from some of the results how did he made the graph Zucker2k pointed out?

He might have weighted the bzip2 compression with a factor of 5.

 

[lobz]

Ryzen 9 3900X/3950X vs. Core i9 10900K In 380+ Benchmarks...

https://www.phoronix.com/scan.php?page=article&item=3900x-3950x-10900k&num=1

Again, the 10900k stages a strong showing, besting the 3900x in overall tests.
View attachment 22825

Hate to burst your bubble, but this is one of the laziest comparison reviews I have ever witnessed in my life. Phoronix or not, this article is outright horrible.

 

[Zucker2k]

Edit: Ok, investigating further in his methodology, this was total run time in my quotes but the results are taken of the average across multiple runs and when there was large deviation in run times the test ran over and over again to get an acceptable deviation. So the crazy run times aren't calculated in the results, just the average of all the runs. There's still some peculiarities though that I think need some review.

Great comeback.

I looked at the details. The tests that I had no idea what they were really, is the ones that the 10900k won. There was only one that I knew what it was, that the 10900k won.

Okay, I'll spare you "OneAPi," "Machine Learning," BioInformatics, and even "Imaging." But you don't know "Java," "AV1," "Cryptography," "Algebra, "Python," "Video-Encoding," and the several other virtual ties?

They did not say what coolers were used, and they did not say anything about power usage, or bios settings (like was PL2 used all the time)

Or, motherboards under-reporting power usage. Be careful what you wish for. Anyway, these are extended tests from the original review so same everything. It says so in my previous post, along with the link to the original review.

Gee you sure are an avid Linux user now.

Nice concession speech.

He is avid with anything that fits his agenda.

If by my "agenda" you mean posting links that show the true capabilities of the 10900k in pure linux environment against the competition in the Comet Lake-S review thread, then my actual sin is only being the messenger. Phoronix are the real culprits here.

Can someone explain this to me, how can this result end up like this:

Nice try! Lol