KVR Audio

Dschungelfieber · Post by **Dschungelfieber** » Wed Jul 21, 2004 12:31 pm

Hi!

I´ve searched the web for real world audio benchmarks and I mean VSTi and VSTfx performance. I`ve found nothing. There seems to be no conclusion how much faster the 64 is. Of course in gaming benchmarks and the like it shines, but what about VST? I'm currently using an 2400+ with DDR266 and would like to know if it does make sense to switch to a Athlon 64 3500+ with a 939 board. Any hard facts?

DF

DevonB · Post by **DevonB** » Wed Jul 21, 2004 12:35 pm

http://www6.tomshardware.com/cpu/200406 ... 39-23.html

Probably as close as you're going to get. Doens't look like it's worth the extra cash to me.

Devon

daverich · Post by **daverich** » Wed Jul 21, 2004 12:50 pm

It depends on if you do a lot of low-latency work.

Athlon64 shines at this - but for higher latencies (20ms and up) then the benefit is a lot smaller.

Over on the sonar forums there is a thread with an audio benchmark which is pretty good - do a search for it.

Kind regards

Dave Rich.

Dschungelfieber · Post by **Dschungelfieber** » Thu Jul 22, 2004 5:58 pm

Thanx Guys!

The Sonar benchmarks are the first usable ones I've seen. Very interesting! But i would also like to see systematic benchmarks using Cubase SX.
So a Athlon 64 3500+ should be almost 50% faster than my 2400+. The rating seems to be correct. But if I consider how old my PC is, one can see how slow CPU technology is advancing nowadays. I remember the rapid race til 1 Ghz and above. That were cool times!

DF

seldom · Post by **seldom** » Fri Jul 23, 2004 8:56 pm

here's a highly non-systematic benchmark for you.

i previously had an athlon 1800XP, and have just upgraded to an athlon64 3200. i just tried one of my tunes that had my CPU meter at over 90% with a fair bit of crackling on the 1800XP, and on the athlon64 the CPU meter sits at around 25-30% with my ASIO latency half what it was before.

seems ok to me.

Dschungelfieber · Post by **Dschungelfieber** » Sat Jul 24, 2004 10:23 am

Thats interesting and much more one would expect! What sequencer do you use and what VSTis and VSTfx did you use in the tune?

HHaynes · Post by **HHaynes** » Mon Jul 26, 2004 12:56 am

http://cubase.freezope.org/perftest/shoresults

Someone lists results for an AMD64 built by eMachines.

There was another Cubase SX performance test site that seemed to start as a Mac vs. PC thing - and has grown to include information on dual proc G5 and dual Opteron systems. I can't for the life of me find the site, though. Look around - it's out there somewhere.

DevonB · Post by **DevonB** » Mon Jul 26, 2004 3:43 am

You mean this one?

http://www.fivetowers.moar.net/cubasete ... erformance

I take it with very much a grain of salt. A AMD 2200+ with the same performace as several AMD 3000+ procs? Ummm sure.

I hate these websites because honestly, you don't know if it's BS or not.

Devon

HHaynes · Post by **HHaynes** » Mon Jul 26, 2004 3:55 am

It's true that you can never completely trust data on sites like that - they could be hedging by a percentage or two - or they could be completely skewing the results on purpose.

It's kind of like what Mac did with the early audio performance "benchmarks" of the G5...

The load does not always map 1-to-1 with the processor. So many other factors weigh in, such as the quality of the drivers and the audio interface itself. Then you also have the issue of RAM performance and latency settings.

It's all on a slide rule until you get ONE machine and rotate components in and out and find out which part affects the end result. On the agregate, you can begin to see a trend - but you'd need a LOT more data with higher reliability before you can call the results definitive.

That's a long way of saying "you're right".

ObsoleteAcc99 · Post by **ObsoleteAcc99** » Mon Jul 26, 2004 4:48 am

[DELETED]

DevonB · Post by **DevonB** » Mon Jul 26, 2004 12:18 pm

TTOZ wrote:i wish there was a method that coule be foolproof!

If you remove the fool, you have the proof!

What they should require is a screenshot.

Devon

seldom · Post by **seldom** » Thu Jul 29, 2004 10:54 pm

i just thought i'd drag this topic up again after having a bit of a muck about with the chaps in #k-v-r and my new athlon64 system. not a particularly scientific study but usefully reproducable for comparison.

here's the system -
athlon64 3200 socket 754
1BG DDR400 ram single channel
audiophile2496
hard drives and other computery bits.

ASIO latency set to 64 samples (1ms)

using FLStudio, i made a pattern using polyiblit, selecting the "flutterfly" preset, turning the polyphony of polyiblit up to 32, and making a big 32 note chord in the pattern.

the first play though the pattern used 18% on the cpu meter, which rose to around 45% and then 69% on the third loop through, which remained stable.

if i turned down the decay on all envelopes of polyiblit to zero, the cpu usage remained at 18% with 32 notes playing.

1ms latency, decent (96 notes with the envs turned up) polyphony, and no crackling or any kind of audio problems.

this is a test you can try at home!

[edit] thanks to griels for putting me on to the cpu-killer that is flutterfly.

Athlon XP vs. Athlon 64 Audio Benchmarks