Is multiple cores in a CPU more important than speed for recording? Advice Appreciated

[21tones]

As mentioned in another thread I am looking for a reasonably cheap, refurbed computer. The tech guy who does computer stuff for me, has offered the following machine for £500.
It has a E5-2698 v4 20-Core processor. Apparently, this is an Intel Xeon which runs at 3.6 ghz.
He says "The rest of the machine is based on a high end ASUS X99 motherboard, 64GB RAM, 1TB storage (and I can slap a 2TB drive in there also for more storage), and a baby graphics card for video output. There’s 10 USBs on back including one USB C, and one USB 3.0 on the front along with two USB 2.0."
I will be using Waveform for my DAW, and an SSL 2 audio interface.
I mainly do recording of guitars and bass and numerous vocals, plus Jamstix drum software, and some VST instruments for other sounds e.g. keyboards, saxophone and (rarely) strings.
I rarely go beyond 24 tracks in a song. but each track could have a number of VST effects e.g. reverb, delay etc.

Would this 20 core processor be beneficial or is fewer cores more important and a faster CPU?

Thanks for any advice you can offer.

[BertKoor]

I'd say that's a pretty good deal. Quite hard to predict how much of all that power will remain idle.
Can you put a project on a USB stick, try it out on the spot and watch the CPU activity monitor?

[Proteinshake]

It has a E5-2698 v4 20-Core processor. Apparently, this is an Intel Xeon which runs at 3.6 ghz.

Keep in mind this is not the base clock of that CPU, which is "only" 2.2 GHz. Even the base Model Mac Mini will run circles around that thing for music production and the power consumption ..

[21tones]

Thanks for taking the time to respond. I didn't realise the chip had been overclocked. (Is that right?)
What do you both think about CPU cores versus CPU speed. If running lots of VSTs and tracks, would different cores be allocated to them? I am concerned about latency which I guess is more to do with CPU speed?
Forgive my ignorance about this technical stuff. All help and advice appreciated. Thanks

[VariKusBrainZ]

Thats a beast and not s stupid price

[BertKoor]

Some daws utilize multiple cores better than others. I think its better to have 4 cores at 4GHz than 40 cores at 0.4 GHz, but that's not at hand here. To be fair, your demans will be relatively modest. You're not running 20x Diva through convolution reverbs with a symphonic sampled orchestra, are you?

Latency is purely due to efficiently coded drivers.

[sjm]

A lot of audio processing is inherently serial in nature. You need to process an entire FX chain on a mixer channel in order, for example. You can process multiple mixer channels in parallel, but depending on how your buses are set up, sooner or later you are going to have more cores than you can process stuff on. Every bus needs to wait for every channel feeding into that bus to finish processing before you can start processing the bus. The master bus needs to wait for everything else, before processing can begin. This can lead to a lot of waiting for threads to finish processing and a lot of idle cores.

This means that single core speed is far more important for audio than other applications that lend themselves to parallelisation. Obviously you will benefit from additional cores, but you are better off going for a processor that is twice as fast with half as many cores, all other factors being equal.

This very discussion came up semi-recently on a Discord server I'm a member of, which includes a guy who works for Intel. He was trying to disuade the guy who wanted to use a Xeon for audio who was also blinded by the number of cores and assumed that would mean amazing performancer. Those processors are designed for servers, where you benefit from all the additional cores. There are far better options out there for audio.

You can think of this problem with waiting for threads to finish like a multi-lane road, with each core being its own lane. It doesn't matter how many lanes of traffic you have, traffic throughput is 0 when you're waiting because that light is red. All that matters in that situation is how quickly the light changes to green; i.e. how quickly you can process the audio you are still waiting for. And that is where the clock speed comes in.

[21tones]

Thanks for the further responses. sjm thanks for the eloquent explanation, reinforced by BertKoors view. To be fair to my tech guy he asked me to ask on a suitable forum what the demands on a computer are for music applications. So would you say that the 64gb RAM is overkill and I should focus on getting the fastest processor I can? (Which means not this machine presumably)
Thanks again for helping me understand more than I did before and for your advice.

[BertKoor]

You don't use huge sampled instruments, but mainly record real audio?

Suppose you record not in 16 or 32 bit but 64 bits, not mono or stereo but quadrophonic, not 44.1, 48 or 96 but 192kHz. That needs (64/8bit= ) 8 * 4 * 192.000 = 6.144 mb per second = 368 mb per minute. 64 gb then holds 174 minutes. Even with 16 tracks, that is 10 minutes. I'd have to look up the transfer speed of DDR4 RAM to verify whether that's realistic.

Yeah, maybe it is somewhat overkill. It does make sense for a web application server, 3.2gb per core is plenty.

[lobanov]

You don't need plenty of cores for recording and moderate mixing (without Acustica stuff for example but this is an exception, most of plugins aren't greedy to such extent). The top speed isn't needed too. I'm sure for your tasks any high mid processor will be sufficient. More or less new Intel Core i7 is absolutely enough. Even Intel Core i5. But more is always more.

SSL 2 is 2-ins/2-outs, this level of load for a processor is negligible (I mean recording).

For your tasks 64 Gb of RAM is an overkill but if you can afford it buy and use it. Mostly you need memory if you use big sampler libraries (Kontakt etc.). If you moderately record and mix you need RAM for wavs of your project + some memory for OS + some free memory (to avoid freezing and swaping). But again, more is always more.

[lfm]

This means that single core speed is far more important for audio than other applications that lend themselves to parallelisation. Obviously you will benefit from additional cores, but you are better off going for a processor that is twice as fast with half as many cores, all other factors being equal.

All that matters in that situation is how quickly the light changes to green; i.e. how quickly you can process the audio you are still waiting for. And that is where the clock speed comes in.

No matter how many cores you still share the same memory bottle neck. CPU cannot do much with just internal registers. So boosted memory like motherboards support at 3200 MHz and up is a good thing.

Building my current daw I did a mistake and did not look at all details for memory, DDR4 in this case, with that MB and stopped at 2400 MHz. You sensed this right away when 3200 Mhz was supported getting the right memory.
- this was in 2023, so gather now higher speed with DDR5 is the norm and worth the pricetag if MB support it

So I share that single core speed is more important than cores. Then memory speed , then cores IMO.
- and don't be blinded by Efficiency Cores, those are basically half speed than Performance cores
- talking Intel then
- 64 GB memory seems over the top for described purposes

As I understand it multiple cores allow for less information need to be swapped to/from stack when to perform a threads job. This saves time/clock cycles, but still there are many, many threads on the same core and those have info of cpu registers to be swapped to/from stack too. Each time slice a thread gets have to swap registers back and forth on stack, which is regular memory too.
- so memory/ram is crucial
- many are fooled to think cores are more cpu's, but memory/ram and ssd are shared resources
- then drivers for audio interface

You can get a feel for what counts at www.cpubenchmark.net and compare same cpu's but with different model number and see that increase in rating is all to do with clock speed.

[uOpt]

That's a good machine. My server is also a (dual) Xeon v4.

It will have high idle power, though. Maybe not good if you want to leave it on 24/7.

[21tones]

Thanks for the further responses. So, CPU speed rather than masses of cores, 32gb RAM ( as fast as possible) is what I need to look for in my budget range, plus the SSD specs mentioned at the start of this thread.
Thank you so much to everyone who has responded, helped me understand this better and given advice.
Be interesting to see what I can get for my bucks!
Best Wishes All

[jancivil]

get the best audio interface you can. The difference here can be huge. 20 cores is overkill for your use case in a big way. I don't think there is a one-size-fits-all answer to speed vs cores.

[eLawnMust]

Too funny... Modern DAWs & Plugs are so inefficient to need so much RAM & speed... One of my workhorses is an old Toshiba P-25 from @ 2003 it's a single-core at 2.85 GHZ (very fast for the time) & I have the full 2 GB RAM... Runs all the 32bit stuff like a stallion, Can make just as gooda music as any other in it...

Folks, modern plugs & DAWs are nothing special, they are just taking advantage of modern power machines & in a bad way... There is no more thought or effort into making resource efficient anything anymore... If these plugs & DAWs were presented back in 2000 they would be considered junk as nobody could use them...

By memory Back in 1999 Cubase VST 24 3.7 had minimum req of 128 MB RAM... Present Cubase is what? 8 GB RAM minimum? For WHAT?... I have in the past ran 32bit versions of Serum, SynthMaster & Harmor in old Cubase VST24... I don't use it as if you have too many rewires registered it takes up tracks, which is an old Cubase flaw as they just had Rebirth as a rewire then...