Diva on Ryzen

[EvilDragon]

But why do Xeons which are from the same wafer i7s come from perform as well as CPUs 800mhz faster? Only difference is cache and chipsets.

At this poing you gotta show some numbers. Which exact Xeon and i7 are you comparing here?

[david.beholder]

For sure, and it's also quite a bit faster in pure GHz - single core performance is still king for most stuff.

I'm sorry what data you're based on? That article?

[EvilDragon]

Real-life experience.


Multiple cores are all cool and dandy, but at the end of the day, the higher the CPU clock, the more shit gets done faster, the more stuff you can push onto a core before stuff falls over. It's common logic.

[david.beholder]

Real-life experience.

Very scientific data ))

I think input data vacum fluctuations are could lead to Howking-like information radiation. But still I'd rather see real data with proper tests.

[EvilDragon]

It actually is. Just think about it for a second. Same number of cores, one is running at 3 GHz, the other at 4.5 GHz. Which one is better? Obviously the faster one.

Having 8 cores at 3 GHz still isn't faster than the 4 cores at 4.5 GHz, in MOST cases. If there's some highly-parallel processing (video filters, say, or 3D animation), then more cores might count a bit more, but still not enough to close the 1.5 GHz gap.

[david.beholder]

It actually is. Just think about it for a second. Same number of cores, one is running at 3 GHz, the other at 4.5 GHz. Which one is better? Obviously the faster one.

Well man, you're wrong: 4.5 ghz arm gonna suck to 3ghz x64 even to atom. Pure ghz make no sense.

[EvilDragon]

And where is a 3 GHz Atom? It doesn't exist.

Nor does 4.5 GHz ARM, your argument is moot. Don't make up arguments that don't hold up.

[EvilDragon]

Pure ghz make no sense.

They sure do. They still do. It will ALWAYS be the case.

As I said - more cores is better, but software has to know how to use them. Otherwise they're largely moot. In vast majority of use cases, it is still single core performance that matters most, so the faster the CPU is, the better.

For example, take DAWs for example. There is always one thread that can only occupy one core only, because it's an important one and it cannot be parallelized - and that's the main audio output thread. The faster your CPU is, the more you can tax the audio thread to shell out buffers faster, resulting in lower latency and so on. Not only is this pure logic, it is also the truth.

[david.beholder]

And where is a 3 GHz Atom? It doesn't exist.

Nor does 4.5 GHz ARM, your argument is moot. Don't make up arguments that don't hold up.

The fact you don't like my argument doesn't make it invalid or moot.

It would work in case 1.5 ghz atom and 2.25 ghz arm and doubling frequency wouldn't chane anything except power consumption.
But 1.5 ghz atom blows 2.25 ghz arm.
In similar way way that lower freq intels were beating prev generation amds.

[david.beholder]

Pure ghz make no sense.

They sure do. They still do. It will ALWAYS be the case.

As I said - more cores is better, but software has to know how to use them. Otherwise they're largely moot. In vast majority of use cases, it is still single core performance that matters most, so the faster the CPU is, the better.

For example, take DAWs for example. There is always one thread that can only occupy one core only, because it's an important one and it cannot be parallelized - and that's the main audio output thread. The faster your CPU is, the more you can tax the audio thread to shell out buffers faster, resulting in lower latency and so on. Not only is this pure logic, it is also the truth.

First of all pure core frequency is not the only factor of core performance since Pentium.

Second both DAWs I use can multithread / multiprocess and aware of cores. As well as most cpu greedy plugins I use able to run voices in multithread. Computation/thread segmentation is very different now and if you don't run something that overload single core MP performance become important too.

Performance in mt/mp enviroment is very complicated subject and your approach doesn't look valid why I want to see numbers not "real-world experience".

[EvilDragon]

Yes DAWs are aware of cores, BUT THE MAIN PROCESS THAT PUTS OUT AUDIO is always on one single core. Fact. Run out of steam on that core and you get dropouts. Fact. More GHz results in more stuff processed in the same slice of time. Fact.

Single core performance still matters a lot. Fact.

[diaper@ky]

GHz do Matter, buts it's only one piece of the puzzle.
GHz plus large cache as in my i7 5775C, which is 65watts, performs the same as my 3770, 4790S (OCd @ 4GHz) and 4790k @ 106watts.
I just want the same performance I'm use to with lower heat and less watts. My needs are for live rigs, sometimes in 100% humidity in 100+ degree weather, using 1U ATX Chassis.

I was going to wait for Fall when the rumored CPUs using Vega GFX were coming.
But I'm definitely getting the slow Quad Core Ryzen 5 coming in 3 weeks.
I'll modify the Chassis since DIMMs are perpendicular to the Supermicro Fans.
But be happy to share the results and do any tests with Zebra2 HZ or Diva yuze guys come up with.
If I'm right the 16MBs cache and OC to 3.8ghz will prove it.

If I'm wrong my kid gets a mean HTPC/gaming rig as I'll get a pair of RX 480s for GFX.
But Urs single core tests are very promising.

Ryzen fares better at Cinebench 15 Multicore than Intels so maybe I can get by with Zebra2 HZ on a single core and Diva, Kontakt, Omni, Keyscape, PTeq and PLAY sharing 3 Cores.

Just wish I knew the scientific stuff you guys know.
I'm just a guy with chops, gear and experience at CPUs going back to Athlon MPs, Celeron 200A, Tualatin, Conroe, Prescotts, Wolfdales, and our beloved i7s that really changed the game.
They just haven't really scaled much as 4GHz is where they started shrinking dies and focusing on iGPU/lower watts.

Got a feeling though that 16MBs cache on the Ryzen 5 quad is going to be a keeper.

Zen Master says we'll see.

We're all u-He through and through, it could've been another analog vers. Digital debate which hurts feelings...

Cheerz

[david.beholder]

Yes DAWs are aware of cores, BUT THE MAIN PROCESS THAT PUTS OUT AUDIO is always on one single core. Fact. Run out of steam on that core and you get dropouts. Fact. More GHz results in more stuff processed in the same slice of time. Fact.

Single core performance still matters a lot. Fact.

Writing all caps is kinda lame. If you think I don't read you well enough, you're wrong.

You know there are ways to control and load balance threads between cores that would help with load balancing and load main process core only if load on other cores is high?

Processor frequency is just one parameter and 1Ghz of arm has way less processing than 1Ghz of modern x86/amd64 processors for several times. A if there would be 1ghz 8086 it would lose to 200-300mhz single core i7/ryzen because of superscalar processing, branch prediction, predictive caching, hyperthreading and many other parallelism techincs.

So you're wrong -- more Ghz is more stuff processed only if two similar architectures with proportionally scaled caches being compared otherwise processors with lower frequencies may have better performance including single core performance.

[MFXxx]

Personally Id like to see some translation into the real life.

Cubase loaded with half a dozen instances of DIVA, some nice large string kontakt libraries, several instances of 2cAudio Aether running through half a dozen Nebulas and aquas.

[Melvin J.]

^^^Ditto. I want to see some real audio projects with heavy hitting plug-ins. I appreciate what's Urs did, though. It's a great start.