Performance of new 10900k VS 10940X (Cascade Lake) ?

Hey,

I am planing to build a new Hackintosh and do struggle between these two CPUs.

I have analyzized the charts from scanproaudio.info, which contain the previous models of these CPUs:

DSP Chart: http://www.scanproaudio.info/wp-content/uploads/2019/07/Dawbench-DSP-Chart-2019Q3-2.jpg

Kontakt Chart: http://www.scanproaudio.info/wp-content/uploads/2019/07/Dawbench-VI-Chart-2019Q3-2.jpg

However, I am uncertain about speculating how the newer models would compete in a Mac OS Cubase enviroment (those tests were run on Reaper, Win10).

According to other benchmarks, the 10940X hasn’t improved at all, while the 10900k got a performance boost of 33% compared to their older 9900k version.

I know that figuring out, how exactly DAWs are using CPU power is a never ending topic, but maybe there is a clear answer when its between Intels k- and their X-series.

Btw, I am working mainly with MIDI and running a huge orchestral template, so both the DSP and Kontakt charts seem significant for me.

Edit: The 10940X will probably still perform better than the 10900k, but its also a question of efficiency price wise, as the difference (including mainboard price difference) of these two lies about 500€,

10940X has 14 cores and 28 threads(10900K has 10 cores and 20 threads)
10940X supports 4 memory channels(10900K only 2 memory channels) and up 256gb(10900K only 128gb)
10940X has 48 PCI-E lanes(10900K has 16 PCI-E lanes)
10900K has integrated graphics, 10940X doesn´t and probably runs cooler.

I am currently testing 10980XE against 3970X and 10900X and I am about to setup another 10940X system because of its higher clocking against the 10980XE.
It totally depends on the Plug-ins you are using. Isolated benchmarks like the old DAW Bench show only the one spcific case which is not a real use case. The 3970X for example seams (not sure yet, I am currently trying to find answeres) to be weak for AVX2 Instructions. Too built a real Benchmark it would be better to have alot test sessions with typical Plug-ins.
Its not easy to answere straight which is better, even the much smaller 10900K outperforms the other in some scenarios.

But the 10940X should outperform the 10900K in most cases by far. There are a Plug-ins that are very CPU intense like U-HE Diva etc. these really need the highest possible clock. Other DAWs and lighter Plug-ins are not that high frequency dependant.

Also the 10X CPUs supports 4ch RAM which is very important for some Plug-ins like SIR3. So in any case I would take the 10940X.


If you want you can send in a 16 bar Cubase Song with standard Steinberg Plug-ins, so I can post the results.

Thanks for your reply.

4 memory channels vs 2 memory channels sounds like the game changer when working with Kontakt. That would explain why when DSP processing, both CPUs acted according to their benchmark results, however when processing Kontakt, the X-series reached doubled results.

I think in this case, the memory difference (256GB option) isn’t that important because these CPUs will probably both struggle hard to play Kontakt sounds in quantity of 128GB RAM without any audio crackles. However it’s always nice to heave the option to upgrade.

Also, 14 cores vs 10 faster clocked cores for almost half of the price is the trade I am aware of. But considering that working with Kontakt is one of my main purposes, the 10940X really seems to be worth the 500 bucks more to spend.

But after reading your title, your charts above don’t have any numbers or data
on the performance of the i9 10900k.

That’s right, because the 10900k has been released just recently. I was just comparing the 9900k with the 9940X and added the difference between 9900k and 10900k Benchmarks to my calculation. It’s not reliable, but it gives me an impression as the architecture of the k-Series hasn’t changed with the last model.

I’m running the 10940x and it’s working quite well for Cubase/VIs. Support for 256GB, 4 channel memory, and 48 PCIE lanes were the main factors in choosing that. The extra PCIE lanes are good for getting maximum throughput from a large number of SSDs. They’re decently overclockable as well with adequate cooling.

That’s great to hear! What cooling are you using? I am thinking about getting the Arctic Liquid Freezer II, which is an AIO but got incredible reviews so far.

I’m using a custom liquid cooling loop w/ 3x120 radiator. That AIO system should work, but I’d go for one of the bigger ones (looks like there are a few different variants). As far as overclocking, I wasn’t getting great results with the “automatic” ASUS system. Ended up reducing the voltage a good bit lower than the defaults to 1.190V w/ all cores at 4.5GHz. Feel free to send a PM if you want any more info.

Thanks a lot. I will probably come back to your offer as soon as my build is running and I start messing around with overclocking. :slight_smile:

And yes, I will stick to the 360 variant of that AIO.

Yes but the performance has changed, and it is higher then the older i9 9900K,
and in most tests higher than the 9940x, at a lower price on both CPU and motherboard.
There are benchmark tests all over YouTube.

I know about the 10900k’s performance upgrade which is about 33% better than the 9900k, meanwhile the 10940X hasn’t really changed from the 9940X. But it seems due to the memory channels, a 9940X/10940X still outperforms the 10900k when working with big Kontakt libraries that use a lot of RAM.

However if only DSP would matter, I’d probably go with the 10900k as the price difference is so much bigger than the difference in performance compared to the 10940X.

Could you post the benchmarks where the 10900k beats the 9940X/10940X? This might still be interesting, especially for everyone how doesn’t work with big libraries and just needs raw processing power.

The two benchmarks that I remember looking at were Cinebench and DAW,
because I think they are more relevant to us in pro audio.
I was Googling at the time so I’m sure you can find it.
The one that may interest you the most is DAW benchmark.

I couldn’t find any DAW benchmarks with the 10900k or the 10940x involved.

When comparing Cinebenchmarks, the 10940x has the lead.
https://www.cpu-monkey.com/de/cpu-intel_core_i9_10900k-1139
https://www.cpu-monkey.com/de/cpu-intel_core_i9_10940x-967

However, I know that common benchmarks can be very tricky… there is so many aspects to consider if the aim is preventing audio dropouts. A DAW benchmark of these CPUs would be a great thing.

I got another question. When it comes to DSP processing and works with Kontakt, how much of a difference is there from CL16 to CL14 and is it recommended to use 4 or 8 memory sticks (4x32GB or 8x16GB in total) with the 10940X?

I have not tested, but I am pretty sure that faster ram even quad against dual ch. will not give you that more voice counts.
The amount of voices like 300 for example is not a huge amount of data to stream from M.2 / Ram prealod samples into processing.
But thats all guessing most people worry more about high CPU power demanding synth VSTis or effect Plug-ins.

SIR3 moves large amounts of data (impulse response does not fit into CPU cache) all the time from the ram for calculation, if I interpret it correctly. Thats something completly different.

But better ask someone from VSL or try to “Kontakt” someone from NI, maybe your luky and get a repsonse, they just fired a huge amount of employes.

Low CL wont help if the other RAM parameters are not optimised as well, but in generall its really not worth it for audio stuff, better invest in bigger CPU power.

Yes, thats totally true. But again, if you look at the DAW benchmark chart I posted earlier, there seems to be something more than just raw CPU benchmark power, that affects the polyphony count in Kontakt. The 9940X performs exactly 2 times better in that specific bench than the 9900k, even though their stats and common benchmark results are much closer than this difference would suspect. So maybe the issue with 4 vs 2 memory lanes is making the difference?

I just think that I’ve already covered enough CPU power with the 10940X and would like to avoid having a bad Kontakt workflow just because one other element in my hardware chain is slowing down. According to these guys here, this could be one of the most common reasons for audio dropouts: https://www.da-x.de/de/news/was-ist-wichtig-f-r-gute-daw-echtzeitperformance/
The article is in german but totally worth reading via google translate.

Thanks for the tipp. I might really contact them, as they should know their software. But it’s good to hear that lower CL isn’t a big deal so I can save the money. The 10940X is already expensive enough. :unamused:


I totally agree with you that they advertise their product in this link, however the problem they are talking about does exist. Happened to me so many times that my average cpu load meter within cubase showed less than 50% and I still got audio dropouts all the time.
Of course they won’t show us the solution to this (if they found any), because that’s what they earn their money with. I don’t know if they really managed to fix these issues with an intelligent hardware configuration, but their costumers would probably notice pretty quick if that wasn’t the case. In fact it happened a few times to me that professionals did recommend those guys, but I also live in Berlin so it’s no surprise that they are locally known.

But even if they are just smart advertisers, there must be a solution to this. I mean, this behaviour really doesn’t look like thats what neither Steinberg nor Native Instruments were intending… :frowning: