Leaked benchmark shows Core i9-10900K beating 9900K by 30%

midian182

Posts: 9,632   +120
Staff member
Something to look forward to: It seems we’re edging ever closer to the launch of Intel’s Comet Lake desktop chips, and that means leaks are arriving thick and fast. The latest of these comes in the form of a Geekbench database entry, which shows the i9-10900K performing 30 percent better than its Core i9-9900K predecessor.

The alleged Geekbench 5 result was spotted by regular leaker Tum_Apisak. The benchmark, performed using an ASRock Z490M Pro4 with 32GB of DDR4 memory, shows the upcoming flagship reaching a single-core score of 1,437 and a multi-core of 11,390. The Core i9-9900K, for comparison, has a single-core score of 1,340 and a multi-core result of 8,787.

With its ten cores, it’s not a surprise to see the i9-10900K outperforming the eight-core i9-9900K. Intel is also expected to have increased the base frequency from 3.6GHz to 3.7GHz, while the boost clock goes from 5.0GHz to 5.1GHz. While all that will have contributed to the higher benchmark, the extra performance likely comes from further refinement of the 14-nanometer process, which Intel has been using since Skylake.

Back in January, an alleged leaked document (below) from Intel also showed the i9-10900K outperforming the 9900K by up to 30 percent. It was noted that the chip had a massive power draw, but it appears Intel has managed to get this under control.

We still don’t know exactly when the Comet Lake desktop CPUs will arrive. It looks as if they’ll be announced sometime in April, though a recent report claimed Covid-19 disruption to the supply chain could see the launch delayed until June, but that now seems unlikely.

Intel’s 10th-generation desktop CPUs will be competing with AMD’s 7nm Ryzen 3000 chips at launch, so Chipzilla will have to be smart with its pricing. And with the fourth-gen Ryzen series expected sometime in Q3, Comet Lake could be facing some tough competition.

Permalink to story.

 
1437 vs 1340, not even 7% better if I calculated good.. 30% is indeed extra cores.

It's actually a bit funky with the math in those figures:
-- Increasing the core frequencies should give about a 2-2.7% improvement (based on the predicted Turbo/base frequencies), assuming no other changes are made
-- As you pointed out, the single-core performance improvement is around 7% total, so there was probably some tinkering to get the extra 4-5% extra performance out of the chips
-- However, the multi-core performance is weird. The total performance improvement was about 29.6%. Once you account for the 25% increase in cores, that means you had a combined effect from frequency increase & basecode improvements of only ~3.7%. Take out the effect of the frequency increases, & you're left with maybe an improvement of only ~1-1.6%.

I'm wondering if there were possibly some other factors involved in those results -- maybe some extra cooling, maybe just a fluke because only 1 test was run, reliability of the source, etc.
 
GeekBench5 seems to be rather variable, when it comes to scoring - for example, my i7-9700K system got 1350 (single) | 7950 (multi):


A 5.1 GHz CPU vs a 4.9 GHz CPU is 4.8% faster on a single thread: that same increase for a score of 1350 would give 1405 - this is less than the 1437 claimed for the i9-10900K but GeekBench's own database is stating 1312 for a 9700K.

There's a 3% difference between my result and the database; if that's down to just normal test variance, then we're not seeing evidence of architecture tweaking in the GeekBench 5 results. I'm suggesting that there hasn't been any kind of improvement, just that one shouldn't read too much into the test's output and/or values from their database.
 
What are you guys even talking about? What architecture tweaking? The improvements are because of bigger caches found on 10900K vs 9900K.
 
This is margin of error stuff on geekbench, it would suggest there hasn't been a great deal of advancement on the single core. It could just as easily come down to minor cache and memory tweaks.

Still it's smoking fast and no doubt Intel will make sure they charge plenty for it again. Even assuming Intel can squeeze the full extra 7 percent from single core performance it's hard to imagine how AMD won't manage more from Zen 3, and close the gap further.
 
I'll believe it when independent testers get a hand on it.

This is just hype news.

If you're asking yourself why?
1. They have been caught working with(paying sums of money) Benchmark companies to make sure their processor would be treated more favourable vs. the competition and they were caught.


Wait and see, if it even comes out. They had huge delays lately. If they do we will do some independent testing..

 
So Intel gained 7% single threaded, zen 3 should be a 30% ipc boost. Not sure if Intel is aware if both estimated are true they're in trouble.

I have my doubts about these data that they put in the article. I myself have a 9900ks in a hero xi configured in bios to 1.34v in adaptive and in real use it never goes from Vcore 1.27v with avx0 and passing a geek5 gives me other data ...




https://ibb.co/PMxx76M
https://ibb.co/x5TvzrY


I hang some links where you can see a little more info about my 9900ks oc! and a post on page 50 where all the photos of the bios are

 
Yeah, OK whatever, Intel

As a 9900K owner, I've already decided that I'll be making the switch to Zen 3 and not going with Intel for my next build in Q4 2020 or Q1 2021

12c 24t 4900X for me, with the clockspeed, IPC bumps & a Gen 4 SSD,it should be quite a jump over my current 9900K rig.
 
Yeah, OK whatever, Intel

As a 9900K owner, I've already decided that I'll be making the switch to Zen 3 and not going with Intel for my next build in Q4 2020 or Q1 2021

12c 24t 4900X for me, with the clockspeed, IPC bumps & a Gen 4 SSD,it should be quite a jump over my current 9900K rig.
Why replace your 9900k?
 
Why replace your 9900k?

I like to build a new PC (CPU & Mobo) every 2yrs, selling the old parts whilst they still retain a bit of value, to help fund the new one.

...all I'll need to buy is the Zen 3 CPU, Mobo & a Gen 4 SSD, I'll be keeping my RTX 2080,32Gb RAM, Case & PSU, so it won't be a huge outlay of cash.
 
I have my doubts about these data that they put in the article. I myself have a 9900ks in a hero xi configured in bios to 1.34v in adaptive and in real use it never goes from Vcore 1.27v with avx0 and passing a geek5 gives me other data ...




https://ibb.co/PMxx76M
https://ibb.co/x5TvzrY


I hang some links where you can see a little more info about my 9900ks oc! and a post on page 50 where all the photos of the bios are


This is more accurate core i9-9900K (or KS) data. Benchmarkers need to make sure no background processes are interfering with the benchmark run and almost nobody actually pays attention to that.

I run GB on a number of different systems (no 9900Ks yet though) and my runs are usually within 1% of the max you'll see on any runs in the database because I pay attention and want consistent results. No tricks though like disabling background processes, just waiting until they're all calmed down.
 
What I dont want to see is 1920x1080 benchmarks with this cpu because no one with this cpu will ever game at that res because they just wont.
soon as you do that its gaming advantage will be down to only 5% and unnoticeable but any shill will do those benchmarks as instructed by intel.
 
Yeah, OK whatever, Intel

As a 9900K owner, I've already decided that I'll be making the switch to Zen 3 and not going with Intel for my next build in Q4 2020 or Q1 2021

12c 24t 4900X for me, with the clockspeed, IPC bumps & a Gen 4 SSD,it should be quite a jump over my current 9900K rig.
That 9900k is going to be a force for a long time. Why replace it so soon?

Anyway, Geekbench is pretty cool. Wondering how did I miss it so long ?
I ran it on my laptop. The cooling is excellent but I'm sticking to 5 GHz (I game at 4.6).

https://browser.geekbench.com/v5/cpu/1635195
 
That 9900k is going to be a force for a long time. Why replace it so soon?

I would have gone with a Zen 2 build, instead of a 9900K, but they still weren't quite there in terms of clock speed & IPC.

I'm heavily into using emulators, where single core IPC & clockspeeds are king, hardly any emulators (apart from RPCS3, which is brutal on my 9900K) make use of more than 4-6 cores.

So the 9900K was still the best option at the time, and I don't regret it, fantastic CPU and yes it will still be a top end CPU, in performance terms, long after I switch to a Zen 3.

With what I'm hearing about the Zen 3 CPU's, they should now go above 4.5Ghz and have an IPC of about 10-15% greater than my 9900K at 5Ghz, whilst running cooler and using less power, added to the fact I'll have an extra 4c/8t on top of that, switching to a 4900X is a no brainer, TBH.
 
" alleged leaked document (below) from Intel "
" It was noted that the chip had a massive power draw, "

Should be taken with a grain of salt though !
 
Ya, beating 30% sure, becouse it has more cores.
The most expected are 6c/12threads i5 processors, and reasonably priced against Ryzen 3600.
But becouse of the virus, release date delayed. Not a lots of peoples gonna have enough money for throwing away 400- 500$ for Mobo and CPU.
Well, intel is going to sell that to OEM pc builders, and they again gonna mount 65W Coolers on 150W boosted 8core i7, on a poor vrm boards.
Intel saying they are 65W :)))
 
Last edited:
10900K, I bet its gonna be a space heater, 300 or even 400 Watt chip, mandatory delidding and expensive cooling solution and a very well ventilated case is a must.
I own 6 core 65w 8700 non K, its a 110W on 4.3GHz.
Delidded with coollaboratory liqud pro, and very close to 70 degrees under full load with dual tower 6 heatpipes heatsink.
 
Last edited:
So Intel gained 7% single threaded, zen 3 should be a 30% ipc boost. Not sure if Intel is aware if both estimated are true they're in trouble.
I think a 30% IPC might be overly optimistic but I wouldn't be surprised (or unhappy) with a 15-20% increase.
 
I would have gone with a Zen 2 build, instead of a 9900K, but they still weren't quite there in terms of clock speed & IPC.

I'm heavily into using emulators, where single-core IPC & clock speeds are king, hardly any emulators (apart from RPCS3, which is brutal on my 9900K) make use of more than 4-6 cores.

So the 9900K was still the best option at the time, and I don't regret it, fantastic CPU and yes it will still be a top-end CPU, in performance terms, long after I switch to a Zen 3.

With what I'm hearing about the Zen 3 CPU's, they should now go above 4.5Ghz and have an IPC of about 10-15% greater than my 9900K at 5Ghz, whilst running cooler and using less power, added to the fact I'll have an extra 4c/8t on top of that, switching to a 4900X is a no brainer, TBH.

You'll see far more benefit and value for money if you instead upgrade your GPU to Nvidias next top-end card (3080 Ti?) and wait for Zen 4 which is only due a year later. AMD has offered excellent longevity with its platforms, you're talking about selling up and buying into a platform at the end of its life - AM5 is coming next year and will likely be supported for several generations afterward offering you more flexibility in terms of upgrades. The move to a much more powerful GPU will give you better performance than the CPU upgrade you plan, do yourself a favour and skip a CPU gen.
 
Back