KVR Audio

schnapsglas · Post by **schnapsglas** » Tue Mar 26, 2013 12:51 am

AMD has bad floating point handling, and I am afraid intel, in terms of power consumption hence heat hence acoustic noise, eats any CPU on AMD line. Maybe one day as AMD wants, we will get GPU to do all the floating point... but not yet.

I really want to vouch for AMD, but really, there is no point anymore unless you are looking for CPU below $100.

And if you can afford a Sandy-E, LGA2011 may surprise you in the future with Ivy-E that are not out yet. I think it might last a little longer than AM3+.

digitalboytn · Post by **digitalboytn** » Tue Mar 26, 2013 1:02 am

The Xeons should run a little cooler than the regular chips because they don't have a GPU whatsit on the CPU

I just use these tools to make music and the Xeon's work fine...

Ashe37 · Post by **Ashe37** » Tue Mar 26, 2013 1:07 am

digitalboytn wrote:The Xeons should run a little cooler than the regular chips because they don't have a GPU whatsit on the CPU

I just use these tools to make music and the Xeon's work fine...

Actually, some Xeons do have the GPU, and not all of the Core CPUs do.

Kaine · Post by **Kaine** » Tue Mar 26, 2013 9:44 am

digitalboytn wrote: The Coolermaster Hyper 212+ had to work overtime to keep the i7 960's cool and it sounded like a bloody jet engine

The noise drove me nuts...

You got unlucky with an extremely poor choice of cooler. Its a cheap gaming/overclock solution for those on a budget where noise doesn't matter... and you can tell.

If you'd spent more on that on a Noctura/Thermalright you'd have had a far quieter experience.

q_h · Post by **q_h** » Tue Mar 26, 2013 11:24 am

Kaine,

that shared cache on the AMD design doesn't seem to be doing it any favours, at least when it comes to audio testing.

Two symmetrical 128-bit FMAC (fused multiply-add capability) floating-point pipelines per module that can be unified into one large 256-bit-wide unit if one of integer cores dispatch AVX instruction and two symmetrical x87/MMX/SSE capable FPPs for backward compatibility with SSE2 non-optimized software

http://www.kvraudio.com/forum/viewtopic ... highlight=

Changing a single byte in the reaxcomp.dll gives AMD 20% more performance

.
or
switching to 64bit (floating point operations on x64 is always SSE2) provides more than 50% power for AMD( for Intel - 12-20%)
Microsoft 64-bit compiler - 30% less performance for AMD (for Intel -50%)

jcschild · Post by **jcschild** » Tue Mar 26, 2013 1:28 pm

digitalboytn wrote: I use Xeon 1230 v2 and 1240 CPUs in my main 3 DAWs and they work great...

Powerful,quiet and relatively cheap...

Xeons are no different than regular desktop, the 1230/vs is slightly slower than the 2600 and ofers absolutely no perofmrance different or benefit over the 2600 other than being slightly slower. the 1240v2 is like the 3770 only again slower

Scott
ADK

Kaine · Post by **Kaine** » Wed Mar 27, 2013 12:17 pm

q_h wrote:Kaine,
that shared cache on the AMD design doesn't seem to be doing it any favours, at least when it comes to audio testing.

Two symmetrical 128-bit FMAC (fused multiply-add capability) floating-point pipelines per module that can be unified into one large 256-bit-wide unit if one of integer cores dispatch AVX instruction and two symmetrical x87/MMX/SSE capable FPPs for backward compatibility with SSE2 non-optimized software
http://www.kvraudio.com/forum/viewtopic ... highlight=

Changing a single byte in the reaxcomp.dll gives AMD 20% more performance .
or
switching to 64bit (floating point operations on x64 is always SSE2) provides more than 50% power for AMD( for Intel - 12-20%)
Microsoft 64-bit compiler - 30% less performance for AMD (for Intel -50%)

Yeah, regarding that original thread. I spent 3 weeks with Vin and Justin who put together that plugin where we re-compiled it into the following builds : VC6 ICC, VC6 no ICC, VC2005 ICC, VC 2005 SSE, VC2005 SSE2 with the original 2008 build giving AMD it's best result so it remained in the kit.

Vin did ask you : "In the meantime, q_h can maybe offer up some examples of DAW hosts and plugins where the developers have not used the suspect compiler and "crippled code""

It's still relevant. Everyone we could think of was building using the Intel compilers, if you have some examples where they are not then we'll build another test built around those examples and re-test, the's no point however creating a test that doesn't represent real world code builds.

You also have to ask yourself if changing compiler is such a game changer, why have AMD not developed the uber solution and gone out hammering on every developers door until it's adopted as standard?

q_h · Post by **q_h** » Thu Mar 28, 2013 1:15 pm

the following builds : VC6 ICC, VC6 no ICC, VC2005 ICC, VC 2005 SSE, VC2005 SSE2 with the original 2008 build giving AMD it's best result so it remained in the kit.

but

Changing a single byte in the reaxcomp.dll gives AMD 20% more performance

2008 - Intel compiler 10.1 (reaper) - x87 for AMD
2009 - http://www.intel.com/pressroom/archive/ ... corp_a.htm
2010 - Intel compiler 11.1 (Cubase)- SSE2 for AMD
Sonar,Studio One,Tracktion etc. - Microsoft compiler. FL Studio - Borland.

Vin did ask you : "In the meantime, q_h can maybe offer up some examples of DAW hosts and plugins where the developers have not used the suspect compiler and "crippled code""

You also have to ask yourself if changing compiler is such a game changer, why have AMD not developed the uber solution and gone out hammering on every developers door until it's adopted as standard?

Scan the plugin/program folders with Intel Compiler Patcher.For example, I have 97 files in my VST folder and only 4 VSTs compiled with a Intel ICC.

INTEL Core i7-3820 3.60GHz vs. AMD FX-8350 4.00GHz