P4 floating point error...

T

Tusk

New member
. . this is what i've read in another forum:
"if you are going to use this computer for making music do not buy a pentuim 4..it goes into a denormal mode when it tries to process very small numbers..making it quite useless..you can forget about using most plug ins..athlon is now the chip of choise for music.." maybe an Amd Engineer write this?.. or there's some truth?
Thanx in adv. Tusk.
 
Sounds like absolute bullshit. I'm an AMD user and fan (give my money to the underdog, in the hopes of avoiding the $1000 CPU days again), but that is ridiculous. Both processors use the same precision on internal registers, unless you're talking about SSE2, and even though it does some tricks with the integer units the final floating point output is identical.

What the guy might have meant is that the P4 could be slower than an Athlon, since the Athlon has very powerful native floating point execution units, while the P4 has releatively weak FP units and relies on software optimization to keep in on par (SSE2). If the audio software isn't coded properly for SSE2 support, then the Athlon in FP intensive applications will likely be quite a bit faster.

However, the output of both will be identical. They are both 32 bit x86 architectures, and both use the same internal precision on registers (80 bit I believe).
 
The P4 is slower at floating point math than the Athlon...however, its considerably faster clock speeds make up for it (hence AMD's "P4 equivalent" naming sheme, and hence the fastest P4 is still the fastest processor on the market).

For price/performance, the author is right, AMD is king right now. Now, if you can just find the right motherboard.... :)

Slackmaster 2000
 
Slackmaster2K said:
The P4 is slower at floating point math than the Athlon...however, its considerably faster clock speeds make up for it (hence AMD's "P4 equivalent" naming sheme, and hence the fastest P4 is still the fastest processor on the market).

For price/performance, the author is right, AMD is king right now. Now, if you can just find the right motherboard.... :)

Slackmaster 2000
Click to expand...

The P4 is fastest in most office and home applications, games, and such. Some software that has been optimized for SSE2 is also quite a bit faster on the P4. However, any software still using legacy x87 FPU code will benefit from the Athlon's beefy floating point execution units (three of them).

I use FEA (finite element analysis) software at work, and the basic code structure hasn't changed since the early 80's. It runs much faster (in some cases 50%) on an Athlon than a P4 of equivalent rating (e.g., XP 2000+ vs. P4 2.0 GHz). In the case of this type of software, even the fastest P4 on the market will fall short of the Athlon chips.
 
Bigus Dickus said:
The P4 is fastest in most office and home applications, games, and such. Some software that has been optimized for SSE2 is also quite a bit faster on the P4. However, any software still using legacy x87 FPU code will benefit from the Athlon's beefy floating point execution units (three of them).

I use FEA (finite element analysis) software at work, and the basic code structure hasn't changed since the early 80's. It runs much faster (in some cases 50%) on an Athlon than a P4 of equivalent rating (e.g., XP 2000+ vs. P4 2.0 GHz). In the case of this type of software, even the fastest P4 on the market will fall short of the Athlon chips.
Click to expand...

Thats fine and dandy, but as soon as you get into other niches like DAW's, the AMD is less appealing because of chipset incompatabilities with audio cards.
 
Absolutely! DO NOT get any AMD mobo's for the Motu. In fact, don't get anything that uses DDR ram. I went through several mobos until I found one that works really well: The Asus P4TE (Intel 850 chipset -uses Rambus). The newest DDR is rumored to keep up with Rambus, but its still experimental so I wouldn't risk it.
 
so what does a denormal bug look like to the average p4 daw user? i'm curious. will it slow down my machine or something?
 
Here's another good page about the denormal problem:

http://phonophunk.phreakin.com/p4denormal.html

Bottom line, it's a software-resolvable problem, not P4 specific, more problematic with older plugins, interesting, but does not render the P4 useless for audio, as the original quote mentioned. It's definately good to know this, and on that page above there is a reasonable workaround to the issue.

Be assured that all reputable plugin manufacturers are aware of this issue. Be assured also that there are plenty of people doing audio work on the P4 :)

Slackmaster 2000
 
You must log in or register to reply here.
Back
Top