climateprediction.net home page
What's your lowest s/TS?

What's your lowest s/TS?

Message boards : Number crunching : What's your lowest s/TS?
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · Next

AuthorMessage
Profile Honza
Volunteer moderator

Send message
Joined: 5 Aug 04
Posts: 390
Credit: 2,475,242
RAC: 0
Message 8825 - Posted: 7 Feb 2005, 14:34:30 UTC - in response to Message 8823.  

> > Experiment 1 is SM3 (slab model). Both exp1 and SM3 apply to both current work and the next step which is adding a sulphur cycle.

That's correct.
Plus a coupled model, perhaps some special atmospheric model...
An overlapping or even parallelism of different models/experiments has a high propability (as in case of THC experiment).

> Anyone know of any risks of changing the memory command rate from 2T to 1T?
I would run a memtest and prime95 (or similar) for several hours to make sure that memory modules and mainboard is fine with aggresive timing.
A stable machine is crucial in order to running CPDN without unexpected crashes and premature model upload.
ID: 8825 · Report as offensive     Reply Quote
crandles
Volunteer moderator

Send message
Joined: 16 Oct 04
Posts: 692
Credit: 277,679
RAC: 0
Message 8831 - Posted: 7 Feb 2005, 14:59:40 UTC - in response to Message 8825.  

> Plus a coupled model, perhaps some special atmospheric model...
> An overlapping or even parallelism of different models/experiments has a high
> propability (as in case of THC experiment).
>

Not quite sure what you are referring to when you say 'perhaps some special atmospheric model...'

AFAIK after the sulphur cycle, experiment 2 with coupled model is likely to be next. These are the only things I have heard talk about in the sense of preparations being made (as opposed to things like applications for funding which may have been turned down but may or may not be resurrected).


It seems likely that CP classic will continue a long time because of the OU short course that has been set up. So this at least will run at same time as sulphur cycle and probably exp2 as well. I am expecting that they will want more sulphur cycle runs than can be done before experiment 2 starts so I would also expect those to run at the same time.

The funding applications talk was about doing the same as we are doing but with a different model than the Hadley Model which we are using.


Visit BOINC WIKI for help

And join BOINC Synergy for all the news in one place.
ID: 8831 · Report as offensive     Reply Quote
old_user52381

Send message
Joined: 5 Feb 05
Posts: 30
Credit: 77,091
RAC: 0
Message 8832 - Posted: 7 Feb 2005, 14:59:51 UTC - in response to Message 8825.  

> > Anyone know of any risks of changing the memory command rate from 2T to
> 1T?
> I would run a memtest and prime95 (or similar) for several hours to make sure
> that memory modules and mainboard is fine with aggresive timing.
> A stable machine is crucial in order to running CPDN without unexpected
> crashes and premature model upload.

Sounds like a good idea. I'll do so.
<br><br>_______________________<br><a href="http://doodle.naylor.se/">Doodle Theme for Firefox</a><br><a href="http://david.naylor.se">My Web Gallery</a><br><a href="//naylog.blogspot.com/">My Blog</a>
ID: 8832 · Report as offensive     Reply Quote
crandles
Volunteer moderator

Send message
Joined: 16 Oct 04
Posts: 692
Credit: 277,679
RAC: 0
Message 8833 - Posted: 7 Feb 2005, 15:03:12 UTC

If you are talking about 'HADAM3 Model' which is an (unused) option in the sulphur cycle alpha testing, that is for the current work, I think.
Visit BOINC WIKI for help

And join BOINC Synergy for all the news in one place.
ID: 8833 · Report as offensive     Reply Quote
Kenneth Larsen

Send message
Joined: 26 Aug 04
Posts: 59
Credit: 438,133
RAC: 0
Message 8845 - Posted: 7 Feb 2005, 18:02:03 UTC - in response to Message 8810.  

&gt; Well, couldn't it be because of the specs you just mentioned? His RAM and CPU
&gt; are faster. (Not proportionally to the s/TS difference though, admittedly.)
&gt;
He also mentioned an "AthlonXP 3000+ (166/333 FSB)" doing 2.45 s/TS. This is what I can't understand.
ID: 8845 · Report as offensive     Reply Quote
old_user52381

Send message
Joined: 5 Feb 05
Posts: 30
Credit: 77,091
RAC: 0
Message 8861 - Posted: 7 Feb 2005, 20:40:20 UTC - in response to Message 8845.  

&gt; He also mentioned an "AthlonXP 3000+ (166/333 FSB)" doing 2.45 s/TS. This is
&gt; what I can't understand.

Oh, sorry, I missed that bit. Yes, that seems like a to big a difference just to be due to the RAM.
<br><br>_______________________<br><a href="http://doodle.naylor.se/">Doodle Theme for Firefox</a><br><a href="http://david.naylor.se">My Web Gallery</a><br><a href="//naylog.blogspot.com/">My Blog</a>
ID: 8861 · Report as offensive     Reply Quote
Profile geophi
Volunteer moderator

Send message
Joined: 7 Aug 04
Posts: 2167
Credit: 64,403,322
RAC: 5,085
Message 8872 - Posted: 8 Feb 2005, 1:22:55 UTC - in response to Message 8845.  

&gt; &gt; Well, couldn't it be because of the specs you just mentioned? His RAM and
&gt; CPU
&gt; &gt; are faster. (Not proportionally to the s/TS difference though,
&gt; admittedly.)
&gt; &gt;
&gt; He also mentioned an "AthlonXP 3000+ (166/333 FSB)" doing 2.45 s/TS. This is
&gt; what I can't understand.
&gt;
While it may not all be memory performance, no doubt some of it is. How much RAM do you have? Do you have integrated graphics, i.e. is the video using system memory instead of it's own discrete memory? This can be a significant performance hit. Download <a href="http://www.lavalys.com/products/download.php?pid=1&amp;lang=en&amp;pageid=3">Everest</a> and checkout the Benchmark section for Memory Read, Memory Write, and Memory Latency. The XP 3000+ is in an A7N8X-Deluxe (NForce2 chipset) motherboard and had 2500, 984, and 96.8 for those three benchmarks.
ID: 8872 · Report as offensive     Reply Quote
Profile dubbi

Send message
Joined: 29 Aug 04
Posts: 31
Credit: 155,185
RAC: 0
Message 8905 - Posted: 8 Feb 2005, 10:49:42 UTC
Last modified: 8 Feb 2005, 10:56:59 UTC

Iam running it on my P4 prescott 3000mhz@3800Mhz
1024Mb Ram 2-2-2-4
my timestep is arround 2.4929 (one Cpu of HT)

ID: 8905 · Report as offensive     Reply Quote
Kenneth Larsen

Send message
Joined: 26 Aug 04
Posts: 59
Credit: 438,133
RAC: 0
Message 8908 - Posted: 8 Feb 2005, 11:14:07 UTC

Yes, my memory have quite low benchmarks: read 2139 MB/s, write 649 MB/s and latency 134.1ns as Everest reported. The components are starting to get old, I think...
ID: 8908 · Report as offensive     Reply Quote
old_user52381

Send message
Joined: 5 Feb 05
Posts: 30
Credit: 77,091
RAC: 0
Message 8917 - Posted: 8 Feb 2005, 13:08:22 UTC
Last modified: 8 Feb 2005, 13:11:29 UTC

Done a memory benchmark using Everest now: 3193 MB/s, 1080 MB/s and 62.0 ns.

I've found that using the 'Dynamic Overclocking' function of my motherboard (MSI K8T Neo-FSR), having it set to 5 % (which at full usage makes 2200 MHz -&gt; 2310 MHz) i can reduce my s/TS somewhat, from about 2.27 to 2.20 :-)

The memory data above were with the 5 % dynamic overclocking switched on.

One thing I'm seriously wondering about: How can some people using computers substantially slower than mine get higher RAC than me - I've had mine running basically 100 % for 36 h now (since I joined).
<br><br>_______________________<br><a href="http://doodle.naylor.se/">Doodle Theme for Firefox</a><br><a href="http://david.naylor.se">My Web Gallery</a><br><a href="//naylog.blogspot.com/">My Blog</a>
ID: 8917 · Report as offensive     Reply Quote
old_user1132

Send message
Joined: 25 Aug 04
Posts: 28
Credit: 6,522,252
RAC: 0
Message 8918 - Posted: 8 Feb 2005, 13:25:19 UTC - in response to Message 8917.  

&gt; One thing I'm seriously wondering about: How can some people using computers
&gt; substantially slower than mine get higher RAC than me - I've had mine running
&gt; basically 100 % for 36 h now (since I joined).
&gt;

The BOINC RAC calculations are effectively meaningless in CPDN due to the very long smoothing times and the "bug" which sometimes allots ca 5000 credits to a user - which is then corrected by a database correction program which runs every 4 hours or so. The effect of the cancelled overcredit is not removed from the RAC.

If you don't win the credit lottery and wait 3 or 4 weeks running 24/7 then your RAC will converge on a sensible value....


Andrew
Andrew

<a href="http://cpdnforum.info">CPDNforum<a>
ID: 8918 · Report as offensive     Reply Quote
old_user52381

Send message
Joined: 5 Feb 05
Posts: 30
Credit: 77,091
RAC: 0
Message 8920 - Posted: 8 Feb 2005, 14:13:55 UTC - in response to Message 8918.  

&gt; If you don't win the credit lottery and wait 3 or 4 weeks running 24/7 then
&gt; your RAC will converge on a sensible value....

OK, thanks! I sortof guessed something like that might be the case.
<br><br>_______________________<br><a href="http://doodle.naylor.se/">Doodle Theme for Firefox</a><br><a href="http://david.naylor.se">My Web Gallery</a><br><a href="//naylog.blogspot.com/">My Blog</a>
ID: 8920 · Report as offensive     Reply Quote
old_user52381

Send message
Joined: 5 Feb 05
Posts: 30
Credit: 77,091
RAC: 0
Message 9001 - Posted: 9 Feb 2005, 14:07:27 UTC

So, to sum up, the lowest value so far in this thread is an equivalent of 1.20 s/TS when running two models simultaneously on a 3.4 GHz HT Intel.

I just realized there isn't such a thing as a 4.0 Intel (apart from OC'd ones of course).

Anyone have any exact numbers for the 3.6 and the 3.8?
<br><br>_______________________<br><a href="http://doodle.naylor.se/">Doodle Theme for Firefox</a><br><a href="http://david.naylor.se">My Web Gallery</a><br><a href="//naylog.blogspot.com/">My Blog</a>
ID: 9001 · Report as offensive     Reply Quote
Profile old_user248

Send message
Joined: 6 Aug 04
Posts: 65
Credit: 1,605,224
RAC: 0
Message 9120 - Posted: 11 Feb 2005, 10:19:05 UTC

I don't have numbers for a P4 3.6+, but my FX-53 running an 11.5% overclock gets 1.55 +/- 0.2 sec/TS depending on the particular model run on average. You have to be careful comparing these results as depending on the computational requirments of the particular parameter set the model can run faster or slower.

Before anyone makes any changes in their BIOS to the memory timings I highly recomend they read up on the Internet for several days to a week or 2 on overclocking the CPU and memory as well as your particular motherboard. Changes to the BIOS without knowing what the possible effects can be as well as how to recover from mistakes can lock your computer up solid. Also, overclocking your PC increases the heat output and can definitly be detrimental to the stability of your PC.

Naylor83: check out MSI's forums at http://forum.msi.com.tw/index.php for good information on your particular motherboard.

Here is one post that is quite good at describing many of the possible settings in the BIOS as well as the effects and possible inter-relations.

http://www.overclock.net/showthread.php?t=91


ID: 9120 · Report as offensive     Reply Quote
Profile Bonsai911

Send message
Joined: 9 Sep 04
Posts: 228
Credit: 30,229,255
RAC: 3,258
Message 9145 - Posted: 11 Feb 2005, 20:38:22 UTC - in response to Message 8773.  
Last modified: 11 Feb 2005, 20:39:48 UTC

&gt; Without overclocking...
&gt; AMD64 3400+ (2.4 GHz Winchester) In fastest run currently at 1.86 sec/TS
&gt; AMD64 3200+ (2.0 GHz ClawHammer) About 2.15
&gt; P4 3.4 GHz (Northwood) Fastest model 1.36, Fastest two concurrent
&gt; models using HT 2.30 and 2.49
&gt; AthlonXP 3000+ (166/333 FSB) 2.45
&gt;

Hello,

I have an amd AMD64 3500+ on ASUS A8V Board. In fastest run currently at 4.14 sec/TS - is the "cool &amp; quiet"-feature a 'slow downer'? CPU-Frequency is only at 1000 MHz and CPU Voltage at 1.0750 V.
Thanks for help and understanding.

Greetings

Bonsai911
ID: 9145 · Report as offensive     Reply Quote
Profile geophi
Volunteer moderator

Send message
Joined: 7 Aug 04
Posts: 2167
Credit: 64,403,322
RAC: 5,085
Message 9149 - Posted: 11 Feb 2005, 21:09:50 UTC - in response to Message 9145.  

&gt; Hello,
&gt;
&gt; I have an amd AMD64 3500+ on ASUS A8V Board. In fastest run currently at 4.14
&gt; sec/TS - is the "cool &amp; quiet"-feature a 'slow downer'? CPU-Frequency is
&gt; only at 1000 MHz and CPU Voltage at 1.0750 V.
&gt; Thanks for help and understanding.
&gt;
If it's only running at 1000 MHz, your sec/TS is about what I would expect.
ID: 9149 · Report as offensive     Reply Quote
old_user52381

Send message
Joined: 5 Feb 05
Posts: 30
Credit: 77,091
RAC: 0
Message 9155 - Posted: 11 Feb 2005, 23:08:51 UTC - in response to Message 9145.  

&gt; I have an amd AMD64 3500+ on ASUS A8V Board. In fastest run currently at 4.14
&gt; sec/TS - is the "cool &amp; quiet"-feature a 'slow downer'? CPU-Frequency is
&gt; only at 1000 MHz and CPU Voltage at 1.0750 V.

Yes, that's a known problem with Cool n Quiet which I've read about somewhere. (The fact that it stays at 1000 MHz even when under full load.) Your processor should probably be doing 2,2 GHz or something like that if it were running at full steam.
<br><br>_______________________<br><a href="http://doodle.naylor.se/">Doodle Theme for Firefox</a><br><a href="http://david.naylor.se">My Web Gallery</a><br><a href="//naylog.blogspot.com/">My Blog</a>
ID: 9155 · Report as offensive     Reply Quote
Profile Bonsai911

Send message
Joined: 9 Sep 04
Posts: 228
Credit: 30,229,255
RAC: 3,258
Message 9191 - Posted: 12 Feb 2005, 14:22:09 UTC - in response to Message 9155.  

&gt; &gt; I have an amd AMD64 3500+ on ASUS A8V Board. In fastest run currently at
&gt; 4.14
&gt; &gt; sec/TS - is the "cool &amp; quiet"-feature a 'slow downer'? CPU-Frequency
&gt; is
&gt; &gt; only at 1000 MHz and CPU Voltage at 1.0750 V.
&gt;
&gt; Yes, that's a known problem with Cool n Quiet which I've read about somewhere.
&gt; (The fact that it stays at 1000 MHz even when under full load.) Your processor
&gt; should probably be doing 2,2 GHz or something like that if it were running at
&gt; full steam.
&gt;

Thanks to naylor83 and geophi for your interest!

I disabled Cool &amp; Qiet-feature, and now cpdn rocks - 2.16 sec/TS
CPU-Frequency is at 2200 MHz.

Have a nice day and happy crunching - Greetings from hamburg

Bonsai911
ID: 9191 · Report as offensive     Reply Quote
old_user147

Send message
Joined: 5 Aug 04
Posts: 30
Credit: 422,225
RAC: 0
Message 9211 - Posted: 12 Feb 2005, 17:52:24 UTC
Last modified: 12 Feb 2005, 17:54:29 UTC

Hello,

I have some (rough) numbers from a Pentium M (with 1 MB Cache) under linux. The DDR memory is only running at 266MHz CL 2.5, that slows things a bit down.

1600 MHz - 2.83 sec/ts
1400 MHz - 3.00 sec/ts
1200 MHz - 3.25 sec/ts
1000 MHz - 3.55 sec/ts
800 MHz - 4.00 sec/ts
600 MHz - 4.95 sec/ts

josti
ID: 9211 · Report as offensive     Reply Quote
Profile old_user504
Avatar

Send message
Joined: 7 Aug 04
Posts: 33
Credit: 215,841
RAC: 0
Message 9224 - Posted: 12 Feb 2005, 22:19:05 UTC

AMD Duron 1.8GHZ - 4.25 s/TS
mobile AMD Athlon XP 1900+ (1.6GHZ) - 5.25 s/TS (due to slow SD-RAM memory)
ID: 9224 · Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · Next

Message boards : Number crunching : What's your lowest s/TS?

©2024 climateprediction.net