|
PureBytes Links
Trading Reference Links
|
Given the recent news reported by Tom's Hardware, it
appears that Intel will be dropping HyperThreading for most
of its new processors. Some CPUs destined for the
specialized server market will continue to have
hyperthreading.
So what follows is more of academic interest than practical
significance.
I am not surprised that some applications work fine (or
even faster) with HyperThreading. My comment was about
speed loss was for programs that have not been re-written
for optimal performance under hyperthreading.
I recall Tomasz indicating that at some point Amibroker
will be optimized for multiple core CPUs. This should also
work for HyperThreading which makes a single core CPU
behave like a dual core, if my understanding of
HyperThreading is correct.
As for the question about page filing settings, I used the
computers as they came. My assumption (perhaps wrong) is
that the manufacturer would have set the page filing to the
optimal setting for their CPU and memory type.
b
--- Mark Keitel <mkeitel@xxxxxxxxxxxxxxx> wrote:
> Not so sure since that was not my understanding on Hyper
> threading, and my
> lap top that has it is much faster on mathematical
> equations than my other
> computer that does not.
>
>
>
> I take it that the page filing settings on all computers
> were the same too
>
>
>
> Mark
>
>
>
>
>
> _____
>
> From: amibroker@xxxxxxxxxxxxxxx
> [mailto:amibroker@xxxxxxxxxxxxxxx] On Behalf
> Of b
> Sent: Friday, August 26, 2005 1:11 PM
> To: amibroker@xxxxxxxxxxxxxxx
> Subject: RE: [amibroker] Off-Topic: AMD vs Intel CPU
> Speed Comparison
>
>
>
> My test P4 was a non- Hyperthread model (516J or 519J, I
> can remember which).
>
> I would expect a P4 with HyperThreading turned ON to be
> about 1/2 the speed.
>
> My understanding is that unless an application is
> optimized
> for HyperTreading, the application will run at half speed
> since Windows will assign that application to 1 of the 2
> virtual CPU cores that exist when HyperThreading is ON.
>
> The advantage of the 2 virtual CPU cores is that a single
> application can not "hog" all the CPU cycles. Thus a
> HyperThread CPU is appears to be a lot more responsive to
> the user when running multiple applications (at the cost
> of
> doing individual jobs at half speed).
>
> So my guess would be that a P4 (2.93 MHz) with
> HyperTreading would take twice as long to run my AB test
> code. That would mean it would be 30% slower than my four
> year old Athlon 1.4 GHz computer.
>
> b
>
> --- Mark Keitel <mkeitel@xxxxxxxxxxxxxxx> wrote:
>
> > Curious if you used a Hyperthread P4 on how it would
> > compare in that test
> >
> >
> >
> >
> >
> >
> >
> > Mark
> >
> >
> >
>
<file:///C:\Documents%20and%20Settings\Owner\Application%20Data\Microsoft\Si
> > gnatures\www.aajonahfish.com\astronomyhaven.htm>
> >
> >
> >
> >
> >
> > _____
> >
> > From: amibroker@xxxxxxxxxxxxxxx
> > [mailto:amibroker@xxxxxxxxxxxxxxx] On Behalf
> > Of b
> > Sent: Friday, August 26, 2005 12:37 PM
> > To: amibroker@xxxxxxxxxxxxxxx
> > Subject: [amibroker] Off-Topic: AMD vs Intel CPU Speed
> > Comparison
> >
> >
> >
> >
> > Due to recent purchases at work (where Intel reigns)
> and
> > at
> > home (where AMD has a fighting chance), I had the
> > opportunity to do a side by side comparisons of how
> fast
> > each is for doing long optimization runs in Amibroker.
> >
> > For a baseline, I used my four year old "workhorse"
> > computer with an Athlon CPU.
> >
> > As best as I can tell all three computers are about the
> > same except for the CPU and RAM. All are running
> Windows
> > XP
> > with SP2. All three use 7,200 hard drives.
> >
> > They differ in ram size: my workhouse has 1.5 GB of DDR
> > running at 266 MHz; the new Athlon 64 has 1 GB of 400
> MHz
> > DDR in a dual channel set up; and the P4 at work has
> just
> > 512 MB of DDR. To make sure the size of RAM is not an
> > issue, I made a special test database with just 30
> stocks
> > in it. That way all the stock data would fit into RAM
> > cache.
> >
> > My test AFL code contained items usually found in the
> > type
> > of code I use. In particular:
> >
> > 3 - IIF statements
> > 1 - Foreign call (I use an index for timing)
> > 20 - EMA formulas (a bit more than my typical code)
> > 10 - AND calls
> > 3 - comparisons "<" or ">".
> > 15 - Portfolio size
> > 1 - Positionscore
> > plus a few "SETOPTION" statements
> > 0 - ApplyStops
> >
> > For those still reading, here are the times in minutes
> > for
> > a 4,000 cycle optimization run.
> >
> > Athlon.......(1.4 GHz)....65 minutes..... 1.00x
> > Pentium 4....(2.93 GHz)...35 minutes..... 1.86x
> > Athlon 64....(2.0 GHz)....28 minutes..... 2.32x
> >
> > Notes:
> >
> > - The final column gives a score to each with the
> Athlon
> > 1.4 GHz being the baseline of 1.0x
> >
> > - To make sure hard drive speed was not a factor, the
> > optimization was started and stopped after a couple of
> > cycles. That gets all the stock data into RAM cache so
> > hard
> > drive speed no longer matters. Then the optimization
> was
> > restarted and the time remaining was recorded after 100
> > cycles were completed.
> >
> > Observations:
> >
> > For my type of AFL code, the Athlon 64 is about 25%
> > faster
> > than the P4.
> >
> > AMD continues to give the most bang for buck. Both the
> > new
> > P4 and new AMD computers were within a few dollars of
> > each
> > other with virtually identical features except for the
> > CPU.
> >
> > Moore's law (processing power doubles very 18 to 24
> > months)
> > appears to no longer be working. If it were, the Athlon
> > 64
> > would be 4 to 5 times faster than my four year old
> > Althon.
> > However, maybe Moore's law still is working: the new
> > Athlon
> > 64 computer cost about half what my four year old one
> > did:
> > Twice the speed for half the cost is the equivalent of
> > two
> > doubles in four years.
> >
> > AMD's model numbering appears to understate its power.
> > The
> > Athlon 64 (2.0 GHz) has a model number of 3200+ which
> is
> > supposed to indicate it is approximately equal to a
> > Pentium
> > 4 running at 3.2 GHz. However, since the Athlon 64 is
> 25%
> > faster than the 2.93 GHz Pentium, the AMD model number
> > could have been 3660 (at least when running my type of
> > AFL
> > code).
> >
> > Conclusions:
> >
> > - AMD is still the best deal for for number crunching
> > work
>
=== message truncated ===
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
------------------------ Yahoo! Groups Sponsor --------------------~-->
Get fast access to your favorite Yahoo! Groups. Make Yahoo! your home page
http://us.click.yahoo.com/dpRU5A/wUILAA/yQLSAA/GHeqlB/TM
--------------------------------------------------------------------~->
Please note that this group is for discussion between users only.
To get support from AmiBroker please send an e-mail directly to
SUPPORT {at} amibroker.com
For other support material please check also:
http://www.amibroker.com/support.html
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/amibroker/
<*> To unsubscribe from this group, send an email to:
amibroker-unsubscribe@xxxxxxxxxxxxxxx
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
|