Eight Core Mac Pro - Performance test results.

telekom · Post by **telekom** » Sat May 26, 2007 9:47 am

spiral wrote:
knotkranky wrote:
nebulae wrote: How many of these piggies could a quad-core handle?

wow, that made me have a pull.
where are the other teats?

Is that Princess Superstar?

Johnisfaster · Post by **Johnisfaster** » Sun May 27, 2007 1:29 am

nebulae wrote:
Johnisfaster wrote:
nebulae wrote:Anyone have other suggestions to give it a good baseline test that would be a good indicator of modern processors?
if I were making a test I'd definetly put at least 1 of each of lives plugs including all fx and all instruments (operator, simpler, sampler, impulse) as well as a bunch of audio tracks. alot of reverbs and alot of delays and even some clever track routing. basically, use all of the features available including midi effects and track routing and the works.
The issue with this is that not everyone has all of the plugs, so a large portion would miss out on a true benchmark. I'd say Simpler and Impulse for sure.

The other issue is that true trackcount has always been hard because I can't put out a 1gb project that streams 50 tracks...

is there any reason they couldn't run operator and sampler in demo mode for the test? or would that give different performance results?

longjohns · Post by **longjohns** » Sun May 27, 2007 1:48 am

RobertX -

If you have the EIC, I'd be curious if you are successful in doing a Simpler -> Sampler conversion and loading the entire Grand Piano patch into RAM

nebulae · Post by **nebulae** » Sun May 27, 2007 3:04 am

Johnisfaster wrote:
nebulae wrote:
Johnisfaster wrote: if I were making a test I'd definetly put at least 1 of each of lives plugs including all fx and all instruments (operator, simpler, sampler, impulse) as well as a bunch of audio tracks. alot of reverbs and alot of delays and even some clever track routing. basically, use all of the features available including midi effects and track routing and the works.
The issue with this is that not everyone has all of the plugs, so a large portion would miss out on a true benchmark. I'd say Simpler and Impulse for sure.

The other issue is that true trackcount has always been hard because I can't put out a 1gb project that streams 50 tracks...
is there any reason they couldn't run operator and sampler in demo mode for the test? or would that give different performance results?

Yeah, I suppose you're right.

Johnisfaster · Post by **Johnisfaster** » Sun May 27, 2007 3:05 am

nebulae wrote:
Johnisfaster wrote:
nebulae wrote: The issue with this is that not everyone has all of the plugs, so a large portion would miss out on a true benchmark. I'd say Simpler and Impulse for sure.

The other issue is that true trackcount has always been hard because I can't put out a 1gb project that streams 50 tracks...
is there any reason they couldn't run operator and sampler in demo mode for the test? or would that give different performance results?
Yeah, I suppose you're right.

I don't know if I am or not, maybe demo mode performs differently.. I don't know either way.

nebulae · Post by **nebulae** » Sun May 27, 2007 3:07 am

Johnisfaster wrote:
nebulae wrote:
Johnisfaster wrote: is there any reason they couldn't run operator and sampler in demo mode for the test? or would that give different performance results?
Yeah, I suppose you're right.
I don't know if I am or not, maybe demo mode performs differently.. I don't know either way.

The only disadvantage with the demo is that it disables saving the project, so I think there's some benefit to using Operator and Sampler. I'll put it in. Will get to work tomorrow - just got back from a deathmatch.

mercyplease · Post by **mercyplease** » Sun May 27, 2007 11:45 am

RobertX wrote:I got the RAM [8 GIG] and drives [150GB 10k] in this morning and repeated the test...

Same results as before.

The fanboys are desperate to see this test perform better on such a mega machine. It doesnt matter HOW MANY FUCKING TIMES they are told live is the most hungy CPU mo fucker in the known universe they still claim this softwares engine as sleek and fast. Like fuck it is, add a few audio tracks and ditch some of those abletons plugins replacing with vsts and you will see that CPU go up to 40% in no time at all.

nebulae · Post by **nebulae** » Sun May 27, 2007 3:15 pm

oh how we've missed you...

knotkranky · Post by **knotkranky** » Sun May 27, 2007 6:48 pm

Forget little tests. Do real tests.

What can Live do? What is it's ceiling? A real benchmark begins with it's true peak performance.

Also, the cpu bar has been calibrated for what? if the audio gets weird before it reaches it's end (100%) what's the point? If below 50% is what's needed to have a stable session, what is 100% for? Why does this meter go over 100%?

My point?

The meter is crap and means nothing.

32 tracks of continuous audio, a slew of quality plugs and instruments, tons of edits.

Put this very reasonable session load into the box, do a simple mix, render and print a resample version too. Then post the audio somewhere, at 16bit 44.1 and we will "all" have a listen.

a phat intel mac w/ a bunch of ram and 10k drives will give live all that it needs to do this simple sessions load.

OvertoneZero · Post by **OvertoneZero** » Sun May 27, 2007 8:27 pm

This is not an accurate test, because as everyone knows, the CPU Meter on 8-Core Mac goes to 111%

sweetjesus · Post by **sweetjesus** » Wed Jun 06, 2007 6:29 am

mercyplease wrote:
RobertX wrote:I got the RAM [8 GIG] and drives [150GB 10k] in this morning and repeated the test...

Same results as before.
The fanboys are desperate to see this test perform better on such a mega machine. It doesnt matter HOW MANY FUCKING TIMES they are told live is the most hungy CPU mo fucker in the known universe they still claim this softwares engine as sleek and fast. Like fuck it is, add a few audio tracks and ditch some of those abletons plugins replacing with vsts and you will see that CPU go up to 40% in no time at all.

i find myself still agreeing with you more and more these days...

no offence to folks at ableton ( i love them dearly) but the way plugins behavine inside of LIVE is so not cool!

JACKAL & HYDE · Post by **JACKAL & HYDE** » Wed Jun 06, 2007 6:57 am

OvertoneZero wrote:This is not an accurate test, because as everyone knows, the CPU Meter on 8-Core Mac goes to 111%

Wow, I've seen both the Mac & PC "meters" go up to 200% during the Live stress test.. Whats your point here? Your telling me that within Ableton, on an 8 core Mac, that the cpu meter defaults to 111%? C'mon

The test was done using Lives Benchmark/Stress test. PC/XP won -> 5% usage over time, end of story.

And with that, lets keep in mind whats on the box when you buy Ableton...

MADE for M.A.C.!!!!!! and..................p.......c

OvertoneZero · Post by **OvertoneZero** » Wed Jun 06, 2007 8:10 am

JACKAL & HYDE wrote:
OvertoneZero wrote:This is not an accurate test, because as everyone knows, the CPU Meter on 8-Core Mac goes to 111%

Wow, I've seen both the Mac & PC "meters" go up to 200% during the Live stress test.. Whats your point here? Your telling me that within Ableton, on an 8 core Mac, that the cpu meter defaults to 111%? C'mon

The test was done using Lives Benchmark/Stress test. PC/XP won -> 5% usage over time, end of story.

And with that, lets keep in mind whats on the box when you buy Ableton...

MADE for M.A.C.!!!!!! and..................p.......c

Haha, naw man it's a Spinal Tap joke

baffleson · Post by **baffleson** » Tue Jul 17, 2007 12:26 pm

RobertX wrote:
tacvbo83 wrote:although 35C is too hot, gonna wait for a cooler Mac Pro
So far the test has been going for almost 5 hours and the temp is still at 35°. The fans are barely moving and the computer is making no noise at all.

I have a Core 2 Duo rig at home and that damn thing sounds like a leaf-blower.

When the new 45nm Intel chips [Penryn, Wolfdale & Yorkfield] come out they should run quite a bit cooler than present models.

Is your c2d rig a macbook pro?

forge · Post by **forge** » Tue Jul 17, 2007 12:43 pm

longjohns wrote:RobertX -

If you have the EIC, I'd be curious if you are successful in doing a Simpler -> Sampler conversion and loading the entire Grand Piano patch into RAM

yeah I was going to say

EIC ould be a pretty damn good test

just being able to run more than 2 at once! lol

the EIC is all about the HD and Ram