I’ve been reading that Clock Speed can be more beneficial than Cores, so I have TWO QUESTIONS:
Base Clock Speed is the max non-Turbo speed (when it is disabled in the BIOS).
With Turbo your i9 it will most likely go up to 5 GHz. Actually this clock speed information is stated on any shop website.
And an i5 will never outperform an i7 or i9, even if its clock speed is double or triple, this is a matter of number of transistors and core count. You cannot compare with a CPU from 3 or 4 gen ago whatsoever.
This isn’t entirely true. A high-end i5 with good cooling and good memory can outperform a low-end i7 with bad cooling and bad memory on many benchmarks. You really have to try the particular workload you’re wanting to use to make sure, because there are too many variables to count.
That being said – there exist virtual instruments that don’t run well on my 3.6 GHz Ryzen 1950X. I have tons of cores, but the Straylight (or is it Farlight?) plugin from Native Instruments crackles and it only gets reasonable (but not perfect) if I turn the buffer size up to 2048. Anything else on this system runs great at 128. So, a higher clock frequency can absolutely help in cases like this. (And, yes, I’m just waiting for any kind of next-gen Threadripper to ACTUALLY become available before I upgrade. It’s been… a troubled last few years for that.)
When people talk about the clock speed in reference to how a DAW performs it’s typically the upper end performance of the CPU that matters, not the base speed.
You can take a look at what Intel itself says about what “clock speed is” (as opposed to trusting some ‘rando’ on the internet): CPU Speed: What Is CPU Clock Speed? | Intel
Excerpt: “Intel CPU specifications list both Max Turbo Frequency and Processor Base Frequency. The processor base frequency refers to the CPU’s regular operating point, while the Max Turbo Frequency refers to the maximum speed the processor can achieve using Intel® Turbo Boost Technology.”
AMD and Intel update more than just speed generation to generation, so there’s more to it than just those numbers.
Also, with a more ‘wide’ or ‘parallel’ workload you’ll benefit from having more cores (and threads) rather than speed, whereas with heavy processing in a single sequential path you typically want a processor that can do a lot of work on one core in one second.
So, if I buy a CPU from the past 3 or 4 years, what would be considered an acceptable ‘Clock Speed’ so that it has enough Single-Core processing power?
I’m a singer-songwriter… home enthusiast. probably no more than 30 tracks per song, with some orchestration sections.
3.4Ghz?
3.6Ghz?
3.8Ghz?
more?
Not really your question, but the motherboard, memory, and storage device(s) matter too.
Both Intel and AMD processors tend to add various features/abilities with each generation, and all motherboards are not equal in terms of implementing/supporting those technologies. Not all of them get steady firmware updates as new CPUs come out with little features that can sometimes be significant improvements.
Examples are things like enhancements for thunderbolt devices on some Intel chipsets, or enhanced NVMe performance for some AMD models.
As others have suggested, benchmarks are weird things. One model of CPU might do better at a given highly specific task than another. All systems will eventually run into some kind of bottle neck at some point (if you fully load every aspect of the system and observe it for a while).
With all that in mind, I’d suggest charting out the things you’ll be using the PC for over the long haul. When it comes to a DAW, describe your typical session. Does it use a ton of virtual instruments? How many active interface audio inputs and outputs do you tend to work with? How long/big are your usual projects (song/piece length/numbers of audio tracks, and clock frequency you wish to use for sampling)? How many, and what size screens? Do you deliver drives and such directly to clients alot (what storage options perform well, but might save you some money if you need to buy them in bulk and deliver lots of them to clients. I.E. NVMe, SATA, SAS, Thunderbolt, etc.)?
If projects are fairly modest, almost any PC meeting minimum specs should be fine for most casual users.
If you’re planning to build a real work-horse of a professional DAW, I’d recommend communicating with some of the PC builders who specialize in building DAWs. It’s about much more than the CPU itself…
Pretty much anything i7 or i9, or equivalent AMD will handle that load regardless of clock speed and single core performance will be more than adequate.
i5, quite likely as well in that age range - you can pretty much think i5 is previous years i7, i7 is previous years i9.
Really though, for that load my old i7 3700k (which is a decade old) will handle it.
So don’t sweat, buy a bargain and write some music - that’s what its all about.
Sorry for repeating again, but unlike some would say, an i7 will always outperform the higher end i5 from the same generation. (Starting from the principle that thermal paste is applied properly, of course, and the RAM speed is irrelevant for a DAW, 2600 MHz is enough and is already faster than how fast a 4 GB/s M.2 can feed it, so the performance probably won’t be any better with 4000 MHz RAM). So there is just no need to compare a current i7 / i9 to an i5 from several generations ago…
You better pay more and have a CPU ready for years to come.
It’s not about the number of tracks, depending on many factors, you can have less than 10 tracks and smash your CPU. It’s about the number of transistors and the core count.
I dont think it mattes that much. I would get a modern gen i7 w/16gb ram. its possible an i5 will work – but hard to know.
Though I suppose we can quibble over speed vs bandwidth etc., as well as whether or not the difference is relevant to any specific user.
Oh come one, you are comparing DDR4 vs DDR5 …
And you are showing the polyphony graph, generaly once you bounce the MIDI to Audio this becomes irrelevant, but whatever.
No user will ever reach 1080 polyphony at a buffer of 64, this is a benchmark software, this does not show real-life usage at all 
The real processing graph is that one, and it shows almost no difference between RAM speed :
Yes, and the CPUs in question benefit either from lower latency or higher bandwidth. The point being that apparently to some platforms RAM makes a difference.
It’s also true on Ryzen that increasing RAM speed to a point will yield better DAW performance. Last figure I saw was 10% improved performance comparing (I think) 3200Mhz to 3600Mhz. The return-on-investment so to speak ends at 3733Mhz if I recall correctly since that’s the ‘last’ point at which the infinity fabric is 1:1 with memory speed. Beyond that fabric speed drops to half and gains are lower. There’s a graph on this somewhere.
Of course. But not all users will continuously bounce VI to audio because they either don’t like doing it for various reasons or because it’s just not practical (for professionals). It’s really beside the point though because I was merely pointing out that there is a difference.
The type of CPU to choose depends on your typical workload. If your priority is low latency with a relatively low amount of plugins you should select a CPU with a high frequency and TDP.
If you don’t care about latency but are running a high amount of plugins you should favor CPUs with a high amount of cores.
These are just the extreme cases. The as real world often is something in between and what fits into the budget.
Regarding the question about the 4 year old CPU: It very much depends on the CPU model, but generally the more modern CPU architecture is usually much more efficient and can perform better at lower frequencies than an old CPU with a higher frequency.
Some more information can be found here: https://helpcenter.steinberg.de/hc/en-us/articles/206929270-DAW-Components (in the Processor section)
@easybullet3 as a bit of an overclocker, I look at it like this.
If you have a CPU rated at 4 ghz but it can turbo up to 5 ghz. Then i would say base = 4 ghz and I’m never going to mess with that.
5 ghz = the total i can get out of my CPU, and if its a K series Intel, then thats the values I’m aiming at with my multiplier.
The base clock is set in the bios so whatever it says in what ever website it doesn’t matter how or who describes it. Its your bios that sets what the base will be and that will be set to 4 ghz. No one is trying to overclock that anymore. No one has a base set to the same as the turbo max because it would not boot or it would corrupt the whole system. Never change your base clock.
Your turbo to 5 ghz can be overclocked using the multiplier by setting a multiplier of 1.25 of the base clock which is 4*1.25 = 5
So the base is never the max you can get out of a CPU, its the value if you had no turbo applied and its set in the bios.
But overclocking involves a lot more than that, there are voltages to be set and power management settings to stop the voltage deviations under load (or lack of load) and of course cooling.
Now the only thing that throws this would be Out of Order processing like the Apple M1 which tries to solve calculations in parallel. If there is enough information available and the result is not waiting on a another result then the CPU will search for other calcs to solve further up the incoming list. This is why a M1 cpu running at 3 ghz but Apple rate it at an equivalent of 6 ghz. Intel also have this in their CPUs, its nothing new, but the M1 is loaded with them. That might be a difference in i5 or i7 too, its Out of Order Processing through put. I do not know how far Intel had got with it.
But you cannot compare on the base clocks because it’s the turbo speed and number of cores that matters more. If your talking about a full turbo i5 vs an full turbo i9, well i would be asking what i9 is that running full turbo at 2.4 ghz? What i9 is that lol
@RTT1 Do you confuse “base clock” and BCLK ?
The “base clock” is indicated in the CPU specifications, and like you say, it is the maximum non-turbo clock, but, this cannot be changed in the BIOS ! Such information is directly written in the CPU microcode, and I don’t know of any BIOS that can change that.
And this is the same with the Turbo value, we cannot change it, as such data is specified by the CPU itself.
What you can change is the Ratio and the BCLK :
If the Ratio is set to Auto (which is the default setting), the CPU will operate as intended and the Ratio will be able to move, from 8 (which is Intel minimum) to the max Turbo Ratio (for example 50 for a 5 GHz Turbo CPU.
If you set the Ratio to a manual value, it will no longer be able to move and it will stay fixed to that value, meaning that the CPU will always operate at the same clock. This is what is actually called a manual overclock.
The Ratio is linked to the BCLK (which means Base Clock, but not the “base clock” you are talking about).
The BCLK is the main operating frequency of the CPU, which is 100 MHz by default. This value is then multiplied by the Ratio and it gives the final Clock.
The BCLK can be modified even when the Ratio is set to Auto, which can effectively increase the maximum Turbo frequency, but this does not work like you explained ,you simply cannot set it to 125 to increase the final frequency by 25%, it is intended for small changes only.
For example, if Turbo Ratio is 50, you set the BCLK to 101.5 and the total clock will be 5.075 GHz. This is the only way to change the “Turbo frequency” without a manual fixed frequency overclock, but don’t expect drastic changes. This setting is only intended as a fine-tune.
(And please MattiasNYC do not start arguing once again, I see you coming already.)
Why? I don’t disagree with that…
Ok, thats kinda hot discussion here haha. IMO I think that the best choice would be to get current gen Intel Alder Lake (or wait till Raptor Lake 13th gen arrive this fall) or AMD current Zen 3 CPUs (or just wait the new Zen) as I would say that all previous Intel 4,5,6,7,8,9,10 and 11 generations will be much slower then Alder Lake even on same frequencies. On AMD side its a little bit different as even previous gen 3900x or 3950x will be very capable today. Which CPU is better? I guess it will always depend on how much you can spend on CPU… Clock speeds will vary and 5GHZ 4 year old Intel will be up to 35-40% slower then current 5GHZ 12th gen… So I think I would suggest to stick to your budget and make a research which CPU/System you can buy/build with the best performance for that price… About RAM I would advice to choose more RAM capacity over higher speed. As its not guaranteed that every IMC (Integrated memory controller) on CPU will handle super expensive fast RAM sticks on rated XMP profiles and situation can go even worse with for example 4 stick double ranked RAM kits on ‘‘Daisy Chain’’ topology mobos… For DDR4 sweet spot is 3600MHz, and for DDR5 I would advice something like 5600MHz kits (but DDR5 is kinda new and it will need couple of years to become ‘‘mature’’ and cheaper). I think we need some sort of topic for overclockers and PC enthusiasts here so we can share our experience on how to squeeze those extra bits of performance from our systems haha))
@Louis_R it’s always complicated lol.
I am not looking in my bios so this from my memory, which is not perfect, I was typing using my old brain.
but yes I am miss remember it I guess lol a lot actually.
I was thinking of the strap that no one is touching anymore but back in the late 1990’s that was what we did lol.
These days I had xmp profile 2 with all core ratio set to 46 on my i9 7 7940x which has a base frequency of 3.10 Ghz and turbo frequency of 4.3 ghz.
But that was on my Asus ROG Board but now I have a X299X DESIGNARE 10G and it is a whole different ball game. On the Asus I can get to 5 Ghz, it’s basically stable but not perfect so setting it down to All Cores at 46 worked perfectly. The Gigabyte is very difficult.
On the Gigabyte I have to open it up to let more voltage through (and I mean changing the safe limits set in the bios) and set all the voltages across a host of items and turn off a whole host of things to make it work. It’s not an easy board to overclock, or my CPU is getting old. I have never had to change so many things to keep it at 4.6.
But yes, the base clock speed of a CPU is the one, like on mine, the none turbo speed of 3.1 ghz. and the max turbo is 4.3 ghz (but I run it at 4.6 with a foot long external radiator to keep it cool or it would go through the roof with heat lol)
thats why I got the M1 because I was sick of the water pump and fan noise on the radiator. Though I could have rebuilt it quieter but the M1 makes no noise so goodbye to overclocking for now.
Its strange because after many years of building and tweaking my computers I have never done a thing to the M1, just press start and it works. But then I like messing with things so I still fire up the i9 and see what I can get out of it every now and again but never bettered 4.6 on the gigabyte.
As you can tell, I hate the Gigabyte board but I needed Thunderbolt at the time.
Generally, it only does this on one or two cores, though, and when you use all the cores, it throttles back towards the base speed. Exactly how much it throttles, with how many cores, depends on how good the cooling is. This is why the Threadrippers are so effective for heavily multi-threaded workloads, because their base clock frequencies (which are “guaranteed” across all cores, assuming you have the specified amount of cooling) are higher than the equivalent (or even more expensive) Xeon chips.
The RAM speed also determines how fast you fill.a cache line after you miss the cache, and large sample patches and convolution reverbs are likely to keep missing the cache. So it turns out, memory speed does matter for DAWs. Not just the MHz, but also the T time, because lower T time will lead to lower latency and thus faster cache line fill, as well as higher MHz at the same T time leads to faster cache line fill.
