advertisement
Forums

The Forum is sponsored by 
 

AAPL stock: Click Here

You are currently viewing the Tips and Deals forum
Dirty Little Secret About the Touted Speed Gains in Intel-Macs
Posted by: MacMagus
Date: January 10, 2006 05:38PM
In case you were wondering where Steve could claim 3-5x speed increases in the Intel-based Macs when Intel dual-cores consistently test slower than single-proc G5's in real-world tests...

[weblog.infoworld.com]
Options:  Reply • Quote
Re: Dirty Little Secret About the Touted Speed Gains in Intel-Macs
Posted by: M A V I C
Date: January 10, 2006 05:46PM
I noticed that too. Apple will probably start using non-altivec optimized filters for PS benchmarks.




Help MacInTouch: Buy from Amazon? use this link [amazon.com]
Mac News & Info: [macintouch.com] [macnn.com] [tuaw.com]
Mac Benchmarks: [barefeats.com]
Used Mac Stuff [FS/T]: LowEndMac Swap List
Mac Software Updates: [macupdate.com]
Fonts: [dafont.com] [fontspace.com]
Online Computer Store With Mac Support: [macsales.com]
Options:  Reply • Quote
Re: Dirty Little Secret About the Touted Speed Gains in Intel-Macs
Posted by: TheTominator
Date: January 10, 2006 05:50PM
From TFA,

Quote

Single CPU, or single-core machines do positively lousy on SPEC*_rate2000 tests

That doesn't make sense to me. I would think the test would spawn one process when running on a single processor machine and two when running on a dual.

It sounds like the blogger wants single processor machines to be compared to dual processor machines with the dual only doing work on one processor.

I guess it depends on your definition of fast. If the type of work you do is not multithreaded, then a higher clock speed single CPU is better. If you do work in parallel, then dual or quad (or 8, 16, whatever) is better.




Edited 1 time(s). Last edit at 01/10/2006 05:54PM by TheTominator.
Options:  Reply • Quote
Re: Dirty Little Secret About the Touted Speed Gains in Intel-Macs
Posted by: MacMagus
Date: January 10, 2006 06:05PM
> That doesn't make sense to me. I would think the test would
> spawn one process when running on a single processor machine
> and two when running on a dual.

Apple used a test that was optimized for dual-cores. The tests were designed to stress that particular kind of chip. Even a dual-cpu computer would do worse on that test than a dual-core.
Options:  Reply • Quote
Re: Dirty Little Secret About the Touted Speed Gains in Intel-Macs
Posted by: OWC Larry
Date: January 10, 2006 06:06PM
I think the test is designed specifically for a dual processor or dual core machine...

Essentially - it'd be like trying to stuff two balls through the same slot at the same time running on single processor... Versus a test that has the balls all lined up.

It's real world tests that will reveal the truth. Real world tests that pit these machines running the same apps with the full benefits of whatever the app is able to do on the processor running on.

Marketing Myth.... megahertz myth redux.

but the jury still is out till those realword tests ON THE RELEASED machines are done. That said... I was very surprised by the claims made seeing what we've seen ourselves of the OS X on Intel - and explains that potentially.



OWC Larry
Other World Computing
[www.macsales.com]
Options:  Reply • Quote
Re: Dirty Little Secret About the Touted Speed Gains in Intel-Macs
Posted by: Silencio
Date: January 10, 2006 07:20PM
Oh, the minefield of benchmarks. Like any other form of statistic, they can be spun and manipulated any which way by both supporters and detractors. Now the AMD fanboys (and the few remaining IBM fanboys) will get all worked up in a lather about this, and for no good reason IMO.

I only really pay attention to application-specific benchmarks from independent sources. Barefeats is the best of the bunch, though I wish he could get ahold of more hardware on a regular basis to paint a truly complete picture.
Options:  Reply • Quote
Re: Dirty Little Secret About the Touted Speed Gains in Intel-Macs
Posted by: Marcos Malo
Date: January 10, 2006 07:30PM
OWC Larry Wrote:
-------------------------------------------------------
> I think the test is designed specifically for a
> dual processor or dual core machine...
>
> Essentially - it'd be like trying to stuff two
> balls through the same slot at the same time
> running on single processor... Versus a test that
> has the balls all lined up.
>

I wonder if they're going to put dual core's in the iHooker. Then she can suck two golf balls through two garden houses simultaneously.



_____________________________________________
I think….
there….
4a.m.
Options:  Reply • Quote
Re: Dirty Little Secret About the Touted Speed Gains in Intel-Macs
Posted by: bfd
Date: January 10, 2006 08:20PM
… the average user will claim they notice a difference (especially after they fork over a few thousand dollars for one)

Options:  Reply • Quote
Re: Dirty Little Secret About the Touted Speed Gains in Intel-Macs
Posted by: shadow
Date: January 11, 2006 09:26AM
Let me sum up this "dirty little secret":

If you have to complete 32 calculations and they can be done in parallel, with all other things being equal, a dual core/proc machine will finish in about half the time over a single core/proc.

Captain Obvious strikes again!

Usually, SPECint is locked down to a single core or proc on a multiple core/proc machine so the benchmark will accurately compare raw unit speed. The rate tests do not have this limitation and, ultimately, can be used to compare complete systems.

Imagine you have two processors, one a single core and the other a quad core. Everything about them is the same (per-core speed, cache, etc). SPECint will benchmark both procs at about the same speed, while the rate tests will show the quad core about 4 times faster.

Given we tend to multi-task with our machines, the rate tests are more representative of actual speed (although they are _highly_ optimistic).

- Shadow
Options:  Reply • Quote
Sorry, only registered users may post in this forum.

Click here to login

Online Users

Guests: 287
Record Number of Users: 52 on November 20, 2014
Record Number of Guests: 2330 on October 25, 2018