Question on memory speed and throughput

th3
th3
Joined: 24 Aug 06
Posts: 208
Credit: 2208434
RAC: 0

You make a good point, and i

You make a good point, and i agree that a P4 530 is more than fast enough. I sold my last P4 (630 3GHz 2MB HT) just a few weeks ago, i kept it this long because it was still performing quite well. Otoh, i didnt use that CPU for Boinc, too unefficient considering the heat it produced. In an aircon room you must multiply the power consumption by 1.1 to get an idea of the real expense in electicity, that was too much for my taste.

Anywayyy.. $45 to upgrade memory is also wasted money IMO, the difference will be only a few percent.

th3
th3
Joined: 24 Aug 06
Posts: 208
Credit: 2208434
RAC: 0

Typo in last post, multiply

Typo in last post, multiply by 2.1 to get the total power consumption in an airconditioned room. Thats Intels figure for datacenters, 1.1W to the aircon for each watt the computers consume.

Alinator
Alinator
Joined: 8 May 05
Posts: 927
Credit: 9352143
RAC: 0

RE: Typo in last post,

Message 82642 in response to message 82641

Quote:
Typo in last post, multiply by 2.1 to get the total power consumption in an airconditioned room. Thats Intels figure for datacenters, 1.1W to the aircon for each watt the computers consume.

Hmmm...

Well, I'd say the figure given by Intel for datacenters is accurate in that context since the primary reason for the airconditioning is for the benefit of the machines themselves and thus that cost is a direct operating expense.

However, for the case of the home crunching enthusiast I think that breaks down a bit. I doubt the 500 or so watts all my hosts are dissipating is doubling the cooling load for the A/C units in my house. In addition, they get turned on for my benefit, not the computers'. IOW's if the house is comfortable for me with the A/C off and a couple of window fans turned on, then the computers just have to make due with the ambient temperature. ;-)

Also, if the datacenter utilizes heat recovery, then they can offset fuel/energy costs in the heating season and that would tend to reduce the 'bite' from straight DC cooling during the summer months. For the home user that happens by default. ;-)

Alinator

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1364
Credit: 3563704011
RAC: 123529

RE: RE: ...As DanNeely

Message 82643 in response to message 82637

Quote:
Quote:
...As DanNeely said, Core2 CPU + MB + RAM, that will give you far more performance per buck.
New 2GB 533MHz ram costs $45 U.S. currency, can't buy a new computer for that. Let me state precisely that I'm no computer geek when it comes to hardware; I can open the case and insert a mem stick and reboot, that's about the extent of my prowess. Otherwise I will simply buy a newly built computer off the shelf, quad core maybe... technical innovation occurs so fast one cannot keep up with it. Then there is my wife who thinks I should NOT spend money on computer upgrades ever. As long as she can do word processing and email she is content to save her money.

IMO the ram upgrade will provide little if any performance gain for E@H, if you other application uses are causing it to hit the swap file adding more ram'll speed things up but faster ram won'tr help much in and of itself.

My bang for the buck comment was RE a CPU upgrade, which would cost significantly more.

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1364
Credit: 3563704011
RAC: 123529

RE: However, for the case

Message 82644 in response to message 82642

Quote:
However, for the case of the home crunching enthusiast I think that breaks down a bit. I doubt the 500 or so watts all my hosts are dissipating is doubling the cooling load for the A/C units in my house. In addition, they get turned on for my benefit, not the computers'. IOW's if the house is comfortable for me with the A/C off and a couple of window fans turned on, then the computers just have to make due with the ambient temperature. ;-)

When the AC us running however you are adding roughly 500W to it's thermal load from your boxes. If you have a large house it's probably only a small fraction of your total cooling bill, but it's still roughly doubling the cost of running your machines when it's on.

vetrivel
vetrivel
Joined: 12 Jul 08
Posts: 1
Credit: 3803
RAC: 0

Hey guys , i m a newbie and

Hey guys , i m a newbie and got a question. It takes about 82000 secs per work unit on my Pentium 4 running @ 3.6 GHz with 2 X 1 GB 533MHz RAM . Is this normal and can i do anything to improve the timing ?
And How long does it take to finish a work unit with a Core @ duo ?
Thank you

th3
th3
Joined: 24 Aug 06
Posts: 208
Credit: 2208434
RAC: 0

RE: Hey guys , i m a newbie

Message 82646 in response to message 82645

Quote:
Hey guys , i m a newbie and got a question. It takes about 82000 secs per work unit on my Pentium 4 running @ 3.6 GHz with 2 X 1 GB 533MHz RAM . Is this normal and can i do anything to improve the timing ?
And How long does it take to finish a work unit with a Core @ duo ?
Thank you


You should use the Windows Beta app 4.46, will give you a very significant boost,
HERE

With the fastest Linux Beta app 4.49 (a bit faster than windows beta app) i get between 11 and 13 thousand seconds per wu on a Core2duo E8400 at stock speed (3GHz). Overclocked to 3.6 it takes between 9000 and 10500 seconds, and 3.6 isnt a high overclock on an E8x00.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5877
Credit: 118574194239
RAC: 17523213

RE: Hey guys , i m a newbie

Message 82647 in response to message 82645

Quote:
Hey guys , i m a newbie and got a question. It takes about 82000 secs per work unit on my Pentium 4 running @ 3.6 GHz with 2 X 1 GB 533MHz RAM . Is this normal and can i do anything to improve the timing ?

The best way to get an idea of what is possible for any particular architecture is to browse the Top Computers link on the front page of the website. When you see a host of interest, "drill down" into the results and see what sort of times it is producing. As a comparison for your P4, I have a 2.8 Northwood with HT enabled which is being shared 50/50 with Seti. One virtual cpu in that host is able to finish tasks in around 65-70 Ksecs - rather faster than yours. It is running the 4.36 version of the science app (same as what is in the 4.46 app already mentioned).

P4 is an architecture I tend to avoid like the plague. I have quite a few Celeron 1300s (PIII) admittedly running around 1600MHz and admittedly running Linux, that are able to complete results in around 50-60 Ksecs. In fact, for the tasks that have the fastest sequence numbers, they can finish a result in not much more than 12 hours - well less than 50 Ksecs. They each have a massive 256MB of very ordinary PC133 SDRAM - none of this 2GB nonsense :-). I bought a fairly big batch of these boxes and they cost me around 10 bucks each :-). They even have Windows XP COAs on them :-). The Linux app is significantly faster than the Windows app at the moment but that could change in the future.

Quote:
And How long does it take to finish a work unit with a Core @ duo ?

Check the top computers list and see for yourself :-).

Cheers,
Gary.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.