So what does SETI depend on, hardware wise

[Illah]

Is it just CPU and mem or is there any other part of the system it uses?

--Illah

 

[TC]

CPU processing power and fast memory/system bandwidth. It has to access the work unit on the hard drive, so that has some role, but not nearly as much as the aforementioned.

 

[Illah]

Hmmm, cuz mine seems kinda slow. I'm at 164x11.5=1886 MHz w/ an XP-1600 and I just hit 'compare' under the performance tab and it ranks me under a PIII! Huh. Are some WU 'tougher' than others, making it bog down or something?

--Illah

 

[lonewolf1983]

yep, cpu and ram speed, and high FSB's tend to determine overall speed, with ondie cache also a factor...

and yes seti units vary in the time to cruch depending on the AR of the unit

 

[Ned Slider]

CPU processing power and fast memory/system bandwidth. It has to access the work unit on the hard drive, so that has some role, but not nearly as much as the aforementioned.

... but no affect on performance. On a t/bird 1.4 @1.6GHz I found absolutely no difference when using an old slow 1Gig HD and then swapping it out for a Matrox UDMA133 7200rpm 40Gig drive. I expect you could almost run seti from a floppy and not see any performance loss

Ned

 

[rogerdugans]

Another post in agreement with what has already been stated

Factors, in order of Importance that I have seen:

FSB speed (memory bandwidth)
CPU speed
How much you are using your computer for other, less worthy things
Memory timings

Memory bandwidth is king, so high FSB will rock the house when it comes to fast worrk unit times

Fast cpus will be done quicker than slow ones- but 1.7ghz at 133 FSB will be slower than 1.6 at 166

I think most of us shut off EVERYTHING possible when doing the benchmark work unit: I went from 3.07 hrs to 2.9 by shutting off all services and software that were not required for the pc to operate (web services, and a bunch of other things for work)

All other aspects being equal, faster ram timings will be faster (cas 2 over cas 2.5, etc)

With the variation in work units, the only way to really accurately compare time is by running a benchmark WU- OCSetiTeam's is right HERE

 

[Mictlan]

As said before:

1.-CPU MHz
2.-Memory bandwidth
3.-Memory timings

The other items in your rig won't be of importance if you:

1.-Video card:this item will be important if you are using the screensaver, as this will unload the CPU from processing the pretty graph and charts.

2.-HD: if you have enough memory available and your HD defragged, you should access your HD only a few times to store data, so this is not a bottlenbeck in your system.

3.-Network or modem: if you are caching units you won't see no downtime when uploading/downloading WU as you will be having at least two instaces of the client running, one to handle the data transfer to/from berkeley and the other one to crunch.

If you find a comparable entry in the benchmark database and see that your have have higher times, try the following:

a)In AMD systems, try upping your FSB to the max, even if that mean lowering your multiplier.

b)Defrag your HD and move the swap file (in Windows) to the start of the HD. This can be done with a more professional solution than the defrag utility bundled with Windows (try Norton Speed Disk).

c)Remove unnnecessary programs from the system memory. Minimize the icons on the tray bar. Each and every one of them grab some cycles, so drop WinZip tray icon, WinAmp, Real Player, etc.....just leave the ones you really need.

d)If you are using the latest installment of SETISpy (3.1.1) you don't need to calibrate anymore the Spy. Grab it from the site and use it.

Happy crunching.

 

[LandShark]

Hmmm, cuz mine seems kinda slow. I'm at 164x11.5=1886 MHz w/ an XP-1600 and I just hit 'compare' under the performance tab and it ranks me under a PIII! Huh. Are some WU 'tougher' than others, making it bog down or something?

--Illah

well, first of all, Spy does NOT do the "comparison" right! not before at least 3.1.1 version. so, don't bother to "compare" that.

also, from start, all monitoring software, Driver/Spy, all tend to over-estimate the time, and in the middle, under-estimate it. it only get close to the real complete time at the end.

with ur system, u should be doing right around 3hr/wu ave. (or for our benchmark work unit) if u r using the CLC client.

also, 0.417AR is consider ave. WU. anything under is low-AR which is harder to crunch, and anything above is much easier to crunch.

good luck.

 

[jokrswild]

I think most of us shut off EVERYTHING possible when doing the benchmark work unit: I went from 3.07 hrs to 2.9 by shutting off all services and software that were not required for the pc to operate (web services, and a bunch of other things for work)

Hmm..I know what alot of the processes in task manager are for, but the following ones i'm not so sure on:

SMSS.EXE
WINLOGON.EXE (does that have to be running for windows to run? I know explorer has to (at least i think so))
CSRSS.EXE
SERVICES.EXE
LSASS.EXE
svchost.exe (a few instances of this)
spoolsv.exe
hidserv.exe
nvsvc32.exe
regsvc.exe
mstask.exe
WinMgmt.exe
mspmspsv.exe
EM_EXEC.EXE


Which ones of these can i shut down to improve on my seti bench scores? As an alternative, would booting into safe mode be easier?