CPU update, will a newer i7 allow for lower buffer sizes? And cooling

Hey there,

At the moment I am using a i7 8700k . I planned to build a new PC this year with a newer i7 like a 12700k or even a 13700k when it comes out. I expect it to improve performance on my big projects on higher buffer sizes by quite a bit, but what about realtime recording with low buffer sizes?

At the moment I can max, use a buffer size of about 64 (which is 4,6ms latency in total, in+out) with my recording templates.
Will a better CPU be able to handle even lower buffer sizes or did we already reach the limits some years ago and the gain is mainly expected when using higher buffer sizes?

Curious to know if anybody had the chance to test that!

I did this exact upgrade, check out my post with more details (and benchmarks):

All I’ll say is that if you’re already at a sub-5 msec latency, you’re in excellent shape since that’s below the threshold of human perception, so making that even less isn’t really buying you much benefit?

3 Likes

Thanks for letting me know! This was exactly what I was looking for.
Well my monitors have DSP, which introduces its own latency, then I got some nitpicking clients which are used to zero latency direct monitoring and will notice that a little.
If I can set my buffer even lower with a Gen 12/13 CPU, then I am really happy.

Glad that post was helpful to you!

My guess is that at best you’ll be able to shave off 1 - 2 msec of latency off of your already low latency due to the ~50% increase in CPU power, so you might be able to reduce buffer size by a commensurate amount. There’s probably a limit on how low you can go regardless of CPU power, but hopefully this upgrade will give you a nice improvement!

If you have clients that insist on zero latency monitoring, then an audio interface with direct monitoring is the way to go. My Yamaha TF1 (which I use as my audio interface) has that capability (I can even add comfort effects like reverb to that direct monitor signal), but even with my ~10 msec latency, that has been sufficient so far with all clients, so I’m not even using that direct monitoring capability. I prefer to keep everything inside the DAW if possible, it’s just easier.

1 Like

Yeah, thats why I tend to avoid direct monitoring. Its just way easier and foolproof to have everything in the DAW.

If I can cut my latency in half, that would already be enough. No one can tell me they notice 2ms of latency.

1 Like

@Tj99 you could also look at the Quantum 2626. Its inexpensive and faster than most Interfaces.

I have one and it sounds good but its so cheap , like £500, that its worth it just for the 8 IOs. I thought it might sound bad but actually I have an Anubis and had a prism and both gave a little extra polish to the sound. But I probably could not tell the difference between my Apollo and the Quantum.

For latency there is not much that can beat it. I get about 4 ms round trip with the UAD M1 drivers for my Apollo but for the same buffer setting the Quantum is just over 1 ms at a 32 buffer. Input latency 0.4545 ms and output 0.635 ms (I would use 64 for recording which gives me 1.8 ms)

For me, instead of building a new PC for my audio workstation, which I have been doing since 1995. This time I went way left field and bought a Mac book Pro M1. I’ve never had an apple computer before and I’m impressed. Windows is a better operating system for functionality but the Mac for audio is great and silent.

Thanks for the info, but I just got 2 Motu 828es and am mega happy with them. They also can manage about 1,6 ms latency on Buffer 16, but there it begins to stutter on my bigger templates. But on an empty project even this buffer size is possible!

M1 chips are king when it comes to power consumption/efficiency, but performance wise they fall behind comparable intel chips. Also if you don’t need to be mobile, get a Desktop PC, as it will always be faster!

Also I am no fan of the Apple ecosystem in general, their price/performance ratio and how I am stuck to a predefined hardware is something I cannot live with. Also there were way to many plugin compatibility/ general update issues lately and most performance issue threads are about Macs. Never had any problem with Windows since Win10, so I think I am gonna keep using that :wink:

Just out of curiosity, why do you think it is easier to not use direct monitoring? maybe it depends on the interface (I have an RME UCX), but I just have the volume set for the mic channels to the headphones anyway, there is nothing I need to do anymore. Au contraire, i think it is much more annoying to monitor through Cubase, i have to activate the “monitor” button with its different modes that never exactly do what I need at that moment… then you have to be sure that you don’t have any latency inducing plugins already somewhere in the signal path.
Sure for recording e.g. guitar through amp sims, it is unavoidable, but otherwise I never use Cubase monitoring if not necessary. Less trouble, less latency.

1 Like

Because to properly record vocals you need a good set up compressor+eq. The engineer and the artists hear themselves way better, the vocals will sound immediately better, and therefore the takes will be recorded way better. A raw or only half mixed vocal is just not that pleasant to listen to, sounds often weak/muffled and has way too many dynamics, which will make it much more difficult to monitor. Also you can judge way better, if the take fits the mix (or even the voice the mic), because sometimes you don’t hear some details in a raw voice, which are then clearly audbile after compression.
So I put my fav vocal comp on the track (which is also the one which will be used douring mixing and is zero latency btw too) and start recording.

Also if there have to be any changes in the monitor mix, everything can be done through Cubase. When direct monitoring you need to fiddle around with Cubase AND your interface control. Adds unnecessary steps during recording, where everything has to be fast, easy and as untechnical as possible for the artist.

And about the monitor feature:
There is a setting (I think its the tape setting), where monitor will be automatically enabled on recording and disabled on playback. Couldn’t be any easier.

At the moment my minimum latency possible for recording in my templates is about 4.6 ms, which is more than fine for 95% of my clients. I would upgrade my PC anyway, so being able to please these remaining 5% as well, is very welcome :slight_smile:

@Tj99 well thats just not a true statement lol

Thats not how the new M1 works.

I think there is a total misunderstanding out there about M1. It is efficient but it also achieves twice as much processing as the equivalent speed intel.

The M1 used out-of-order-processing which means that you need a 5 GHz or more Intel to get the same processing power. But the intel will running as hot as the sun to do it and probably start to throttle lol.

The M1 native apps have additional attributes within the code which enables the processor to run more things in parallel. That means it can get almost twice as much processing done. So while it might run at 3.2 Ghz it does have an almost 6 ghz intel equivalent rating.

The only intel that can beat the M1 Max is the new 12900k running at full speed and thats going to need water cooling because it gets insanely hot when it does it.

Im not aware of any other intel that can keep up with an M1.

I have a 14 core i9 intel overclocked to 5 ghz with a foot long external radiator and 8 fans in it and the M1 still does better.

I have been building my own audio workstations for 27 years. I always build desktops and always the latest Intel I can afford. Like my i9 14 core which i considered the fastest I could get while controlling the heat.

I don’t think i will ever bother building a new computer again. Apple make something that can run even my largest projects without breaking a sweat. I never even have to freeze tracks anymore.

Well thats the logic behind it lol.

But hey of your not into Apple thats cool. i wasn’t either so i get it.

Sorry, but it looks like you fell into apples marketing hype. They weren’t entirely honest with their clients, when it comes to benchmarks as well.

The 12900K is better than the M1 Max, in fact it is more comparable to the M1 Ultra (single core performance is better on the Intel, which is good for mixing and recording, multicore they are about as good). On top of that you can build your own 12900k PC for less than half the money, the Ultra M1 Mac will cost. Also air cooling it, is really not a big deal. But with that money saved, also water cooling would not be an issue.

Check this: APPLE M1 Ultra vs Intel i9 12900k - Busting Apple’s MYTHs!? - YouTube

@Tj99

Im not getting this from marketing I’m getting it from

That has some detailed programming formation about how it works. Also this is not an Apple thing. They did not come up with the Apple chips out-of-order- processing, it comes from ARM who already had the technology. In fact Intel have the same tech but they do not have as many out of order processors in their chips.

Im not talking about the Ultra, no one needs it. The M1 pro and Max have the same performance at single core and would sit just behind 12900k but you will not be air-cooling the 12900K to get that performance.

Also some real world DAW examples.

If you like youtube:

Here he looks at the ultra and decides the Pro (which I have) is just as good for Audio.

Also apple have never said their M1 Max or Pro will beat any desktop intel chip. They only compared to other laptop chips. Also the 12900K was not released so apple have never said anything about it.

Maybe the Ultra they compared but in some ways it does and in someways it does not.

But for audio no one needs the Ultra so that’s not the comparison to do.

I have had intel desktops for nearly 30 years and I got the M1 and was surprised by what it could do. I took projects from my desktop and load them onto my macbook and was surprised how the laptop just worked. Its my first apple so i know windows well.

I bought the M1 because its silent, i thought i would use to until the projects got so big i would transfer them to my intel desktop. Well i have never had to do that, i just don’t even turn the desktop on anymore. I do not remember apples marketing team popping round whilst i was transferring projects on a USB stick and doing my own testing. They were my real Cubase project tests.

As i said, the single core speed of the Macbook pro 10 core M1 is the same as the Max and Ultra. In fact for audio there is no difference between the Pro and the Max. The ultra does not give much boost for audio either (see youtube above).

The M1 max Studio is actually very cheap. If i had to upgrade to a 12900k i would need a new motherboard, new Ram, new chip. So that would be about £1,400 ish. A New M1 Max studio is £1,900. Sure it’s more but not that much more. But if I needed to add a graphics card and 1T SSD And cooling its probably more expensive than the M1 Max Studio. If you already have some of those parts the upgrade is way way cheaper of course. But the price for the M1 Max Studio 32 gb is not so crazy.

And didn’t i say in my previous post the chip that can beat it is the 12900K ? So your agreeing with me but I’m wrong ?

I spent some time researching the M1 before i bought it. I waited until Nov last year to get it. I did lots of tests with my Intel 5ghz intel to see what the M1 can do.
So i have done my research, i have the laptop, i have done the comparisons for myself. I even tried to find examples where the M1 was not so good and for audio i could not find many. I like to look at the negative reviews to see the worst cases.

So please do not insult my intelligence by accusing me falling for the marketing. Don’t accuse me of being some sort of idiot for buying an M1 pro. Its utterly ridiculous.

Its fine not to like something, its cool to try to link me the video on youtube with extra info about the Ultra (Im already not a fan of the Ultra). That’s all good but don’t suggest I’m a fool for buying the best computer I’ve ever owned.

First of all, I was always talking about Desktop chips, as notebook chips are nothing I am interested in. If you don’t need to be mobile, its nonsense to invest in a notebook. There are way to many downsides to it.

Of course you can. First of all when running Cubase, the CPU will never run at 100%, but more like 50% and secondly, with a proper fan like the Noctua, air cooling in general is doable without big issues. Plenty of people doing that with avg temps at around 70-80 during cinebench tests.

You said the only chip which can beat it is the 12900K, but that is not true. First the 12900k is a Desktop chip, so you were the one comparing Desktop to Notebooks chips first, I was only answering to that.

But when you compare it to mobile chips, the m1 is more comparable to a 11900h or a 12500h.

Well if you think 2300 € for a PC is “cheap” then I don’t know. A DAW windows machine with a 12700k, will cost you a bit more than the half of it. With a 1 tb SSD and a cheaper 4k capable graphics card already included. Then if you have some parts you can reuse for a newer build, it even gets cheaper.

I just said it looks like it, happens to all of us at some time. Also I never called you a fool or “insulted” your intelligence btw, chill down. You claimed my statement was wrong. I was saying that M1 falls behind intel chips as can be seen in different tests and benchmarks, and you said “well thats just not a true statement lol” and added

and

even if thats clearly not the case.

So where is your proof, that there is no Intel that can keep up with an M1? Where are these tests where it achieves twice as much processing as an “equivalent” intel?
Your assumptions are based on your personal subjective eperiences, I am talking about measurements, as everybody can say anything in the net and claim it is true.

@Tj99

It is worth mentioning that I like Intel and I like Windows. I think the 12900K is amazing.

I prefer Windows as an OS. I do not like the way Apple keep upgrading the CPU every 6 months like they are phones, I’m sure I will feel left behind in a years time.

So do not get me wrong, if I had never bought an M1 I would be building a !2900K rig right now.

But once I tried a totally and utterly silent computer that can perform at almost the same top speed as a 12900K single core then I am quite addicted to it lol And I’m not dealing with throttling or adjusting the VRM voltages to get it on Air. I just have a silent M1 24/7…it makes no sound at all ever.

But I do like Intel and Windows, all my other computers and laptops are Intel. So I do not want to sound like I’m saying Apple is better or Intel is worse. I’m saying they are near enough the same and the apple is silent always…that what I like the most, the silence…and if I wanted to I could carry it around with me too.

@Tj99 I hope you enjoy your Intel. I’m sure it will be fantastic.

I would not swap ia 12900k for an M1 pro but thats my personnel thing. While I do not agree with your analysis I see why you come to those conclusions. I have seen more than enough posts from people who have got rid of their desktop workstations in favour of a M1 laptop to think they are good enough for me. I am one of those people that realised that the laptop was in fact a desktop killer, it killed mine for a start. The 12900K looks great but then again I am also a great believer in the 2nd law of thermodynamics.

Both of them will do the job perfectly I’m sure. I’m sure you will be very happy with the result…plus is more interesting to build a computer from scratch.

1 Like

In light of the original concern, I use a Roland studio capture interface/sound card with 18 in and 10 out which you can route whichever way you need. The direct out with DSP has nothing to do with any interference with CB and you can still print audio with your preamp etc with no modifications or adjustments. It simply routes the input signal directly to either its headphone outputs or any of the 10 line outs but sends the virgin signal directly to CB to do with whatever you need for CB processing. Because of this, I experience 0 latency with effects and have no complaints from clients or myself. This way, processing power doesn’t even come into play. I’m still on the Intel i7 6700k and it is barely breathing with 100 tracks with processing on all of them.

1 Like

@Tj99 answered that question very well :slight_smile:

1 Like

Sorry, I did not notice the question was meant for you. I thought he was asking me :sweat_smile: :joy:

1 Like

I would have given the same answer, so you saved me some typing :slight_smile:

Air and all in one water coolers are equivalent.
The only way to make water cooling more efficient is to build your own circuit with externally refrigerated water.
But the same can be done with air cooling if you put a small AC in front of the intake fans.
The difference between a air and and a water cooler is max 3 or 4°C under full load depending on the models.
Air cooling is more effective in dissipating small bursts of heat, while with water coolers, once the water is hot then it stays hot, the water heat won’t dissipate as fast as with big air radiators.

2 Likes