[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Advice on system purchase

Am Samstag, 3. November 2012 schrieb Stan Hoeppner:
> On 11/2/2012 7:16 AM, Martin Steigerwald wrote:
> > And thus I say, that I better use a dual core CPU with higher peak
> > performance for typical desktop workloads, than a quad core CPU with
> > lower peak performance. A quad core CPU with as high peak
> > performance might be in order if something compiles software a lot.
> "Average"/"normal" users don't compile source code.  They use a
> browser, maybe an MUA, an office suite, a PDF reader, a media player,
> etc.
> Pick an AMD or Intel system with a dual core CPU @ ~2.5GHz and ~3GHz.
> The extra ~500MHz will yield little noticeable difference in system
> responsiveness, app load times, media playback stutter, etc.

Agreed. I wouldn´t never go for maxed out CPU specs. Not even on
dual cores.

But if I could have a dualcore with 2,5/3.0GHz I would prefer that over
a quad core with 2 GHz for example - on same CPU architecture/design.

> Now, pick one of those frequencies or closest available in a quad core,
> 6 core, and 8 core.  Run the same subjective tests of normal user
> desktop applications.  The perceived performance will be slightly
> higher, but this will be due to the existence of large L3 caches, not
> core count, because most productivity apps are still mostly single
> threaded, simply because there's not much parallelism to be had in such
> work flows.  Productivity apps are not CPU bound, and never have been.

Well thats the point I am making. See my other posts.

Some peak performance by reduce latencies, how noticable that is when
current desktop/server x86 CPUs IMHO already do have lots of peak 
performance on one core.

I do think with current desktop workloads any dual core CPU > 2 GHz from 
Intel or AMD will do.

> Now, if a 2.5GHz dual core CPU launches an app in 1 second, and a 3GHz
> 8 core CPU launches the app in 0.4 seconds, twice as fast, and both
> play media files with the same stutter (caused by packet loss not CPU
> horsepower) then what is the practical performance difference here for
> the average user?  *There is none*

My way to remove stuttering and latencies is still:

apt-get purge pulseaudio

Not even my ThinkPad T23 stutters. Well mostly. BTRFS, quite aged, slow 
fsync, due to "old" 3.3 kernel, apt-get dist-upgrade and then Amarok may 
stutter occassionaly.

But not so on this T520 and also not on a T42 with Ext4.

But then my media files are stored *locally*.

> The only real difference is cost and thermal output, with the big core
> CPUs losing in both categories.
> My entire point in this sub discussion is that AMD/Intel keep pushing
> more speed and more cores, while the vast majority of users need
> neither of these things.  They both still have relatively low TDP dual
> core chips on the market, thankfully, but these aren't the chips we
> tend to see in many/most retail machines.

I do think that improved peak performance on *one* core can matter.

How much it would need to be tested.

I think if I put the Intel SSD 320 from this T520 into my old Pentium 3 
1,13GHz based T23 (with some SATA to IDE converter) I would face higher 
latencies on desktop usage - especially on application 
initialization/startup with typical KDE/GNOME/Iceweasel/Libreoffice kind 
of big complex applications. I also bet that would still be the case with 
my Pentium M based ThinkPad T42.

So I gladly take some improvement there. But I agree that getting highest 
spec CPUs is rather pointless. Cause current CPUs are already quite fast.

And I do agree on the core count.

Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA  B82F 991B EAAC A599 84C7

Reply to: