SETI Performance and Scaling factors?

[Ionizer86]

On the Anandtech Distributed Computing Seti site, it says that factors for SETI include CPU clock, memory bandwidth, CAS latency, etc. I also read somewhere that for Folding@Home, having SSE matters quite a bit (about 15%). What have people found for Seti? How are the results and what are your configs?

I tested out the 3.08 gui with my system: 2600+ (12.5*166), CAS = 2.5, clock = 2.08, and WU time is about 4 hrs. I jumped onto 3.03 (controlled with the SETI driver) and it looks like I should be getting about 3 hrs on my current WU.

 

[Assimilator1]

Yeah the CLI version is much faster than the GUI version (25-75%!) & v3.03 is about 12% faster again🙂

You'll probably get less than 3hrs/WU😀
Btw there is no SSE or 3DNow support in SETI

I use the v3.03 CLI connecting straight to SETIQ on my rig
XP1700 @2.04GHz,177MHz FSB,Nfrc2 Dual channel,CAS 2.5-2-2-6 ,average WUs are done in 2hrs 28mins🙂

You can look at some stats of mine here🙂

 

[Ionizer86]

With the 3.03 CLI, it looks like I cranked out a WU in around 2:50, and another one is going to be only around 2:20. Not bad 🙂

XP1700+ @ 2600+ (166 * 12.5) CL2.5, dual channel nForce 2 Ultra 400.

 

[NTB]

Originally posted by: Ionizer86
With the 3.03 CLI, it looks like I cranked out a WU in around 2:50, and another one is going to be only around 2:20. Not bad 🙂

XP1700+ @ 2600+ (166 * 12.5) CL2.5, dual channel nForce 2 Ultra 400.

CPU speed makes that much of a difference? I'm sitting here with a non-OC'd Athlon XP 1600 (1400 MHz) and 1 WU takes me ~4 hrs. Of course, I'm working with slower memory and a slightly older chipset too: Epox 8K5A-2+ (KT333 chipset) & 512MB PC2100 RAM.

I've been thinking about dropping a 2400+ (2GHz, and the last of the 266 FSB chips) in here for a while, to pick up some speed on video encoding and other things. A DC project would just be another reason. And heck, for $76 it's a fairly cheap upgrade.

Nate

 

[Ionizer86]

NTB, my bro has a XP1600+, and I've found that since it has about 2/3 of my chip's clock, it takes about 50% longer to do WUs.
XP1600+ (10.5 * 133) CAS 2.5, KT266A.

It's a cheap upgrade, so why not spring for it? Do spend the extra to get the 2400+ unless your board takes the 2500+ (Bartons are good for other daily activities). I got a 1700+ to oc and found that it won't boot with multipliers greater than 12.5. Good thing my 2100 ram does 2700!

 

[Assimilator1]

Thats because any XP from 1500-2100 is stuck in the multiplier range from x6(?) to x12.5 (even unlocked),without breaking some L3 & L5 bridges that is.
If you do mod these bridges then you can jump to the next multiplier range from x13 upwards.

NTB
Its not just about CPU speed ,memory bandwidth & low latency are also big factors in speed for SETI.
Also the Nfrc 2 chipsets have slightly lower latency than the KT ones.
I used to have an XP1500 @1.52GHz ,KT266a mbrd, PC2100 Crucial RAM at 152MHz CAS 2-2-2-6 ,this did WU's in about 3.5hrs🙂
I briefly had my XP 1700 Tbred B in this same rig at 1.9GHz & it knocked about 1/2hr off WU times

If you wanted a significant boost in speed for your rig then you'll want to buy a 333FSB XP & get some good quality PC2700 RAM so you can run it at CAS 2-2-2-5/6.Unless your PC2100 overclocks to 166MHz😉

 

[Ionizer86]

Assimilator1, I looked at your main rig, and I noted that you have PC2700 Crucial. What's the highest speed your ram can do while maintaining CAS 2.0? How high can it go if you set it to CAS 2.5?

My 2100 gets to about 151 CAS 2.5, and after setting it to 2.5, I've gotten it to 166 stable (I haven't tried pushing it more yet). It's possible the CAS affects it quite a bit.

I may go about lowering my memory bus to run async with my FSB so I can up the CPU clock more via FSB. This may possibly give more performance. I may also decide to run CAS 2.5 vs 2 vs even 3.0. Single vs. dual channel may be another interesting test, hehe.

BTW, what's Win98SE Lite?

 

[Assimilator1]

What's the highest speed your ram can do while maintaining CAS 2.0? How high can it go if you set it to CAS 2.5?

I did test the RAM at CAS2 to see how high it would go ,I can't remember exactly as it was a while ago😱 ,but it was somewhere around the 168-170MHz area ,quite dissapointing 🙁 (considering my old Crucial PC2100 CL2.5 went to 152MHz @ CAS 2-2-2-6 !).This was at 2.6v.

At CAS 2.5 my PC2700 RAM goes upto 178MHz @2.6v & upto 181MHz @2.7v(I only just tested this a few mins ago!😉).
My RAM timings are 2.5-2-2-6 ,I just tried 177MHz,2.5-2-2-5 @ 2.7v but it failed the Prime95 test after a few mins


My 2100 gets to about 151 CAS 2.5, and after setting it to 2.5, I've gotten it to 166 stable (I haven't tried pushing it more yet). It's possible the CAS affects it quite a bit.
CAS does affect your RAMs top speed significantly ,though remember the higher the CAS setting the higher the latency which reduces performance.When I had my KT266a rig I found that CAS2 @152MHz was faster than CAS2.5 @160MHz!
Because my PC2700 RAM goes much higher at CAS 2.5 than 2 it's worth it.


I may go about lowering my memory bus to run async with my FSB so I can up the CPU clock more via FSB.
Running asynch introduces more latency into the memory bus ,again unless you gain a significant boost to CPU speed you may loose performance!.Only 1 way to find out for sure(on SETI) is to run the same WU over & over again ,or at least compare the same Angle Rangle WU (with 0.417 be an 'average speed' WU).Easiest way is to use SETI spy to see what the AR is of your current WU

You'll probably find that CAS3 will hurt performance too much ,re Dual Channel ,I've done that comparision on my rig & I can tell you on an AMD rig it doesn't make much difference🙁 ,with the FSB at 178MHz & cpu at 2.05GHz going from single to dual channel knocked off about 5-6mins/WU .That was from about 2hrs 33 mins to 2hrs 27mins

Win98SE Lite is a special patch that de-intergrates Internet Explorer from Windows 9x (when used at the 'sleek' setting) ,this makes Windows faster,more stable & uses less RAM.
Go here to find out more