New PC: Cores > RAM/Speed?

I know this has been asked and answered multiple times, but I’d like to get the latest thinking on a new PC build.

My current i7 4 Core processor is struggling once I get to 30 or so tracks and start adding in reverb, delay and guitar sims. I’m planning to havea new PC built so I don’t have to constantly freeze tracks and end up having to drop to 2048 samples just to get a complex project to play.

Am I right that the latest thinking is to prioritise processor cores over speed? Will I better off heading towards a 12 or 16 core jobby, perhaps at the expense of sticking with 32Gb RAM instead of 64 or even 128 if I need to cut a few corners financially?

Is the processor speed a consideration, or is there a balance I should strike - is 12 core enough with a faster processor?

Also, i7, i9 or AMD … suggestions on that route?

Any thoughts from the community would be most welcome.

High CPU speed is still important, tons of cores less so. My current Intel 9900k has 8 real cores, core load is normally fairly uneven. Some cores(often just 1 , probably involved with real-time/low latency stuff) get hammered, others less so.
If latency is not a problem and you can use a higher buffer(say 512) core load is more even.
A Z790 board with an Intel 13900k will be plenty for even the most demanding tasks. It will need a beefy PSU and some serious cooling, though.
32 GB RAM is fine unless you use giant sample libraries. NVME is faster than SSD’s, nice for quickly loading big sampled instruments. DDR5 can be a bit faster than DDR4, but it doesn’t make that much of a difference.
Unfortunately powerful PC parts have become expensive.
It is a good idea to buy from a company that caters for the DAW markets, you pay extra but they will have tested the system for use with Cubase and help you if something doesn’t pan out. Also, if you’re sitting next to your computer you better make sure it is silent.

Ryzen needs 3600 MHZ RAM (DDR4 platforms) to maximize performance and latency (Infinity Fabrix, etc.). Diminishing returns beyond that. I am not sure about the new CPUs. Im not up for a new desktop MOBO/system, so I haven’t bothered to look.

Beyond that, you can easily upgrade RAM so always invest in CPU if your budget is limited and you have to make a choice. I wouldnt even try to compromise. CPU is always the better investment as it will likely remain for the lifecycle of the PC in many cases. RAM often becones quite cheap.

Exceptions for non-servicable platforms like Macs, PC AiOs and Laptops with soldered components, of course.

DDR5 makes the biggest diff for applications thag conatantly move large amounta of data into and out of RAM.

M-Series Macs (paired with PCIe 4 SSD for Swap, DaVinci Resolve/Fusion, Photoshop (large images), After Effects and similar applications).

Its not that big a deal for music production as a lot of data is loaded (more storage-dependent) but isn’t often changed and written back/moved about the way these other application do.

It’s also why the average Joe or Jane doesn’t really notice PCIe 4 SSD and GPU speed/bandwidth uplifts. The workloads are still not taxing PCIe 3, so you have to benchmark to see the differences.

With some of the modern CPUs out there you’re actually getting the fastest single core boosts on the best CPUs, which may be the ones with the most cores. It seems odd compared to how things were in the past, but it actually makes sense. Fortunately I really don’t think it’ll make a big difference. If you’re thinking about getting a new AMD CPU for example and you’re choosing between 8, 12 and 16 core parts the lineup looks like this:

As you can see the one with the most cores has the fastest clock speed. But even if this is the case we’re not talking about huge differences. Between the 7950x and the 7700x the difference is roughly 5% in peak clock speed.

Higher speed for longer/heavier (serial) processing chains.
More cores for more tracks / more processing in parallel.

Both are great. Just know that Intel’s current platform is end-of-line. So no more new architectures for that socket. AMD’s platform is new and you should expect to see at least 3-4 more CPUs for it.

Single-core performance is important. If you’re like me, and you just keep on stacking stuff on your mix bus, including mastering plugins like Ozone, you will hit your single-core limit more quickly… It doesn’t matter how many cores you have for this. <— this is confusing for a lot of folks. Some processing simply can’t be spread over multiple cores, like mix bus processing.

My suggestion is don’t worry about cutting costs too much. Put it on a credit card. Just know what you need, and then buy MORE than you need… because at some point, you’re gonna wish you had it. :slight_smile: That’s just a sad fact of life… The hard part, however, is KNOWING what you actually need, and that depends on the size of your largest projects.

Some people have a hard time managing debt, so I really don’t think recommending that is a good idea.

Hmm… well, people have to know and work within their own limits. My recommendation in this instance stands. Hey, I don’t recommend buying too much more than you need here… just more. Skip the fancy car, and fancy vacations. You don’t need those things. Save there, spend here.

Great response! Kudos to everyone that has taken the time out for me :smiley:

I have been putting quotes together using - I could probably put a machine together myself by buying in parts, but I reckon paying the extra to have it done properly is the best option!

Two options I have looked at come in at approximately the same price:

  1. AMD 5950X 16 Core 3.4 GHz - 4.9 GHz / 72 MB Cache/AMA
    coupled with 32 GB Corsair Vengeance DDR4 3600 MHz

2)AMD 7700X 8 Core CPU (4.5GHz-5.4GHz/40MB CACHE/AM5)
Coupled with 32GB Corsair Vengeance DDR5 4800MHz

Having taken into account all of the comments, I’m leaning towards option 2 - a faster machine, but compromising on the number of cores.

Any thoughts?

I am not so sure, the cpu speed only differs 10% vs 100% for the number of cores. Also how many cores can run at a high speed and at what speed?
But both will be great I guess. Maybe scan pro audio has some results that give some insight?

Are those broken down into performance vs. efficiency cores? And do the efficiency cores kick in to performance mode (if that’s even a thing).

My guess is that option 2 will be fine, in fact great. But I personally would wonder if 8 is enough because, on my MacBook Air M1, I have 8 cores (4 performance, and 4 efficiency), and I couple that with a Waves server that handles Waves plugins separately… and I do freeze tracks (but I’d rather not), and that mostly works. But I do run out of single core performance, and I do seem to max out my efficiency cores too (especially if I try to run my external monitor at its native HiDPI resolution). So, again, how hard you are actually pushing your machine matters, and only you can know that.

So, for me, I’d probably go for an option #3: a 4+GHz CPU with 16 cores, since I currently have a 3.2Ghz, 8-core Apple M1 machine. But you’ve got boost speeds on those AMD chips that I don’t have. I don’t know anything about boost speeds and whether or not they can be sustained for very long. Perhaps those boost speeds will be all the extra you need, but I would personally not want to rely on that.

I don’t use a ton of sample libraries, but I do end up adding more and more tracks to these pop tunes I write, and one thing leads to another and eventually I’m out of processing power, usually because of the mixbus. BUT, freezing tracks actually is a perfecly good option, and not too time-consuming when you have a powerful CPU, so… regardless, I will continue to do that no matter what machine I use.

I’m pleased I asked the question here because the core performance has been a subject that seems to have cropped up recently on the forum since Cubase started to work with multi cores more efficiently. It seems to be a consideration for everyone, but there doesn’t appear to be a great consensus on whether this is more important than MOBO/processor speed.

My existing rig is around 5 or 6 years old - I did buy an off-the-shelf i9 PC a couple of years back, but saw little or no improvement in performance, so it went back.

It is definitely adding more and more tracks that is causing my PC to grind to a halt - I have started leaning on the Izotope plugins more in recent times, and end up with Neutron plugs on most tracks once I get down to serious mixing, and it’s at this point things get problematic.

I can have multiple tracks and instances of plugins working fine, but it when I start separating channels (such as HALion, Sampletank or Maschine) and begin adding processing on each that the meter hits the red. Cubase occasionally crashes (or vanishes!) or at least stops playing until I freeze some more tracks, or print reverbs and disable the FX channel temporarily.

Not wanting to divert into another subject, but performance is definitely worse in 12 than any of the previous iterations …

Copying text from Gearspace where preliminary results were given for the DAWbench test on the 5950 vs 7950:


Scan AMD 5950X DDR4 / 7950X DDR5

064 - 381 / 471
128 - 405 / 529
256 - 426 / 575
512 - 432 / 601

The above is the “DSP” test, which is basically just processing. 30% uplift at 128 sample buffer, which is great. And for Virtual Instrument:

Scan AMD 5950X DDR4 / 7950X DDR5

064 - 1860 / 2260
128 - 2920 / 3800
256 - 4020 / 5380
512 - 5620 / 8060*

Also +30% at 128 samples.

Yes, you’re cutting cores in half by picking the 8 core instead of the 16 core. If you have fewer tracks but heavier processing that should be fine. If you have tons of tracks and less heavy processing you would want more cores.

Keep in mind too though that the 5950X is on the AM4 platform which won’t get new CPUs. If you get the 7700X not only can you upgrade later to a 7950X, but a year from now you could probably also upgrade to their fourth processor on that socket, in the fall. Or, if you want to look a bit further into the future you could upgrade to the end-of-the-line CPU of theirs once their next socket/platform comes out and that eol platform drops in price.

I would go for the new platform.

There’s always the 12-core parts too. I see AMD is opting to release new CPUs soon that are non-X, lower power. See if you can find a chart on how much they boost on individual cores compared to the X counterparts. Because it’s possible that for what you do the non-X part may be just fine, and if it is you could be better off with a non-X 12-core as opposed to a 7700X 8-core, at almost the same price (?).

I think that’s worth looking into.

@MattiasNYC - More useful and pertinent info. Thanks for your efforts.

I suppose everyone has a different idea of what tons of tracks might entail … in my case 30 - 40 tracks including busses and FX channels would be typical with some plugin processing on each.

I see other users on the forum mentioning hundreds of instances of symphonic VSTs and I think wow … how can that even be possible without a NASA sized computer! But again, I guess it depends on the amount of processing they are putting on each track.

Definitely close to settling on the 7700X, but I will look into the 12-core options as well.

Just wanted to post a thank you to everyone that contributed to this thread. I am now the pround owner of an AMD 7700x with 32Gb DDR5 RAM , I couldn’t be a happier bunny :smiley:

All of the tricky projects of 40 or so tracks with heavy processing on each are now running smooth as can be, without any compromise in the way of frozen tracks or needing printed reverbs and delays.

I have noticed (and I brought it up on the Gearspace thread) that VSX has a huge CPU hit, nothing I could have noticed when EVERYTHING was running hot, but it is nevertheless a bigger CPU hog than I had thought.

So again, kudos to the helpful thoughts of @MattiasNYC @shagazulu @vinark @Trensharo @zip


Often the CPUs with the most cores will have the highest Boost Clocks, but may have lower Multi-Core Boost Clocks. You’re basically getting like 100 MHz base clock speed upgrade that you can depend on. Anything heavily multi-threaded is going to drive clock speeds down as the CPU cannot run at those high clock rates without ramping up thermals and forcing a throttling scenario.

The marketing material often only cites the former, while the latter is generally in the product specs or the fine print. However, this has been fairly common with most Intel and AMD Multi-Core CPUs.

A good example of this is the AMD website, where you must mouse over the ‘?’ to actually see this fact:

AMD Ryzen™ 9 7950X Desktop Processors | AMD

Max boost for AMD processors is the maximum frequency achievable by a single core on the processor running a bursty single-threaded workload. Max boost will vary based on several factors, including, but not limited to: thermal paste; system cooling; motherboard design and BIOS; the latest AMD chipset driver; and the latest OS updates. GD-150. Learn More.

Yes, I know. I think that’s what I said above.