$100 CPU for budget DAW?

Looking to build a second rig in a separate office for light duty post-prod work.

Just curious what ~$100 CPU’s would get me the most bang-for-buck. I have decent overclocking skills.

I see a lot of positive comments on the Intel G3268 and AMD FX’s. Would those beat out i3’s?

THANKS!

You can find performance comparisons here:
http://cpubenchmark.net/high_end_cpus.html

G3258
Performance: 4004
Cores: 2
Performance/core: 2002
FX-4200
Performance: 4454
Cores: 4
Performance/core: 1113.5

While FX-4200 is 10% faster overall, it’s single core performance is poor. And that’s what counts when it comes to latency figures etc. When you decide your CPU you many times have to balance between total performance and single core performance. Sometimes you go for total performance (if single core performance is close enough or irrelevant), but in this case I recommend to concentrate on sigle core performance, since difference in those figures is so great (but total performance figures are close).

As you can see from the chart both G3258 and FX4200 are slower than best i3s, but faster than worst i3s.

After quick look I wouldn’t recommend any AMD FX series processors for DAW use when latency matters, because they seem to achieve better performance by therowing in more slower cores than competeng Intel processors. If latency doesn’t matter (no VSTi:s, monitoring done with external equipment) things change and FX series becomes good option.

Thanks Jarno!

I’m super curious about the G3258, especially when it comes to single thread performance. It’s super-overclockable, reaching speeds of 4.5 with relative ease even on standard motherboards…even up to 4.8.

It’s only dual-core though, and doesn’t have hyper-threading. Is that an issue?

http://techreport.com/review/26735/overclocking-intel-pentium-g3258-anniversary-edition-processor/3

http://www.anandtech.com/show/8232/overclockable-pentium-anniversary-edition-review-the-intel-pentium-g3258-ae

THANKS!!! :slight_smile:

It not just the cost of the CPU - the overall platform to run the CPU on tends to be slightly cheaper for AMD CPUs. Modern Intel CPUS tend to outperform AMD CPUs and generate less heat which makes building quiet systems easier. However, AMD floating point performance is very good and the difference isn’t as much between Intel and AMD for apps like cubase - for your buck you should be able to build a water-cooled machine over clocked with more cores (perhaps FX-6300) for the cost of an Intel with a standard cooler. That said the G3258 is a nice CPU for the price - I’d take the extra cores as Cubase and VSTs are only going to become better multithreaders as time goes on - but either should do a solid job.

I think I’ve just recently ruled out the G3258 due to it being only dual-core. Gamers highly tout it as I think games run great on super fast and single or dual cores, and aren’t utilizing quad+ cores.

That said, how well does Cubase handle quad+ cores? Is 6 or 8 cores excessive, and will they go unused?

I’m looking strongly at the AMD FX-6300/6500, and even the FX-4350.

Thanks!

WARNING! WARNING! Will Robinson.
One thing that will severely skew your thinking re DAW requirements is to place any reliability upon what gamers want.

Gamers tend be biased to video performance and system noise is generally not a consideration, both being fairly oppositely viewed from a DAW perspective.

Also, once using all SSDs (even SATAII) and a modest video card, DAW/sampler performance becomes almost totally dependent upon CPU performance. Cubase handles multiple cores well, but I have noticed a favouring of the first available core (turn off core 0, and the next core gets used more), which means that using a consumer CPU over a XEON will enable getting that little extra performance when required for heavy projects by overclocking.

I run an 8 core cpu and regularly have resource monitor running while I’m mixing - I just don’t see the core 0 loading issue - ever. Perhaps it’s setup related?

(PS three SSD system) no traditional harddisks installed on my studio PC.

Just wondering…do you see that all 8 cores are being utilized by Cubase?

I’m torn between faster quad cores and slower 6-8 cores.

THANKS! :mrgreen:

Yes, Cubase is a pretty good multithreader, just loaded up a fairly CPU intensive project (with some 32 bit plugs so the 32bit bridge is active as well) and hit play - see for yourself…

Admittedly, it was under C7.? that I was still noticing some slight bias. I have not tested under C8P.

I had better watch that I am not contributing to a continuing of obsolete performance rumours.

I do think CB8 balances workload better than CB7 - I did see some thread bunching on one core in CB7 - but not necessarily on core 0 - the core which got overloaded was pretty random.

Why do you rule out a processor due it being “only dual-core”? Excuse my ignorance and narrow point of view (after writing my bachelor thesis on computer performance measurements) but I would always prefer systems with less processing units (cores) if the systems have similar overall performance.

Sorry, but this only shows how your operating system (Windows) is able to use available resources (processor cores) evenly. User software (Cubase) has no ability to decide in which processor cores it runs in. That’s the job of the operating system. Only thing it can do is to spread out processing into different processes or threads.

And without exact details of both operating system (Windows) scheduling algorithms and application software (Cubase) multiprogramming algorithms you can’t tell a sh*t about what’s going on … unless you do some highly intelligent benchmarking (or comparisons with two different systems with only one variable).

What are your thoughts on the Intel G3258 OC’d up to ~4.8hz then? The CPU has created quite the buzz among budget minded enthusiasts that like to OC. Again, it’s dual-core with no hyperthreading though.

http://techreport.com/review/26735/overclocking-intel-pentium-g3258-anniversary-edition-processor/3

Thanks Jarno… :smiley:

Software can be structured to allow fairly even processing, or it can confine too much to one thread, so that the core running that thread is more used than others. Hopefully, the thread load splitting is improved with each version.

When I tested Cubase, I ran it solo, so that it would tend to bias to the OS allocation preference, which at the time (possibly pre W8.1) seemed to be to use the lowest numbered available core first. That was fairly consistent. However, if I were running other services or apps, the low cores could well be busy at the time I start Cubase, and so give inconsistent results.

As you hint at, proper evaluative testing is a PITA to design, set up, and get consistent and reliable results from. Too many unknown inputs, with all elements evolving as their designers ‘move the cheese’.

Precisely - Cubase multithreads nicely.

Sigh, what it shows is Cubase is producing many threads none of which are “big” core hugging threads and can therefore be distributed evenly across my 8 cores, compared to a two core CPU and with the same number of threads I will be performing 4 times fewer PSW loads and unloads and will be thrashing the CPU cache less as different threads want their turn on a processor core causing different register sets and instructions to be brought back in for execution. The nice even loadings are just that - it shows Cubase multithreads nicely and will therefore benefit from more cores being available. I have been in IT for well over 30 years (at a very technical level often involved in coding operating system software) I know very well how computers work at a bits and bytes level - and probably did before you were born so cut the sanctimonious “know it all cr@p”. :unamused:

Indeed you can theorise to you’re blue in the face but sometimes the results are not as expected. I was tasked recently with building 15 standalone systems to a tight budget (geographical and network issues meant they didn’t want blades or virtualisation) to playback a production Linux workload with testers logging on performing their testing simultaneously, this would mean an average of 30,000 threads running per system - I built two test systems - all the specs said the Intel machine (much better core performance) would walk it - but given this workload the AMD system pasted it - simply by having more cores and dealing with more threads without pipeline stalling which was causing massive slowdowns on the “faster” Intel based workstations. If I was a gamer I wouldn’t look at AMD twice. Horses for courses as they say.

I’m a bit confused…did AMD or Intel win here?

“the AMD system pasted it” the workloads which were running perfectly well on the 8 core AMD system were stalling and glitching on the 4 Core Intel system - tried builds with and without hyperthreading.

This happens a lot with DAWs, mainly because they are such an atypical use – multiple near-realtime streams with complex processing on each – compared to common computer use, that they are not catered for in the test results published for most of the hardware required.

A case in point is for SSDs, where even the SATAII spec is plenty fast enough for almost all DAW users, but yet there is a lot of discussion about new expensive super fast products, like PCIe ones, which are often tested under non-DAW-friendly conditions to get their ‘fantastic’ results.