S@H and RAM settings question

[SETIdude]

I am wondering which of the following configurations will run S@H faster. I'm hoping someone has already done this and can give me some feedback 😀

System is a Dual P3-550 Katmai 100MHz FSB setup with 256MB RAM.

- Configuration #1 -
Ram speed is Hostclock+PCI = 133MHz
CAS timings set to 3-3-3

- Configuration #2 -
Ram speed is Hostclock = 100MHz
CAS timings set to 2-2-2

I've done a few tests and there is a large difference in Sandra memory benchmarks between the two configurations, with the 133/3-3-3 config easily winning the memory bandwidth MB/Sec benchmark.

I'm using PC133 ram that is CAS 2 up to 100MHz and CAS 3 up to 133MHz.

I need to know which config would be faster for SETI@home. I hope some of you have already figured out the answer from past experiences. 😀

thanks

 

[mechBgon]

Interesting, what motherboard is that? I'm afraid I don't have a definite answer for you but I am guessing PC133 will be faster, even at CAS3. BTW which version of the SETI client are you running?

 

[SETIdude]

The board is a Tyan Tiger 133. I cannot decide which would be faster because the 133 setting would provide a higher bandwidth, but the processor would be waiting longer for the information at CAS 3. Maybe someone will be able to answer this 😀

I'm running the 3.0 winnt-cmdline client X 2

 

[Fardringle]

With PC100 CAS2, you are running at 3/4 the bus speed of PC133, but with only 2/3 of the latency. Just by the numbers, this should make the PC100 CAS2 faster overall, but I'm really not certain.

Also, you may be able to get that memory to run at PC133 CAS2 if you change your memory speed settings (I forget the exact name of the setting) to NORMAL instead of TURBO..

 

[office boy]

Running it at 133Mhz, will be faster.

 

[BurntKooshie]

Okay, I missed it.....mech....when'd the Cheeta get added? I knew you made it...but....🙂

Congrats mech 🙂

I agree that 133mhz will be better, as S@H isn't nearly as latency sensitive as it used to be (still is rather sensitive however).

One thing about asynchronous timing of any chipset - the simple fact that it is asyncrhonous adds an additional cycle per access, which is actually rather significant. But, Its better to run it at 133mhz CAS 3 asyncrhonously than at 100mhz CAS 2 in many situations....and I think OB is right in this case, where 133mhz would be better....why not try it, and find out?

 

[BadThad]

IMO, 133MHz RAM is faster. While CAS2 is indeeded faster, I believe it is far offset by the benefits of the additional 33MHz. Based on all the benchmarks I've ever read, increasing system memory speed has HUGE positive effects on performance, whereas CAS2 v CAS3 effects are relatively minor.

 

[mechBgon]

Thanks, BurntKooshie, it was added earlier today. BTW it's spelled "cheetah." 😉 I think I'll try it myself for a while, a smile is more true to my personality than a scowl. 🙂

SETIdude, if you want to do a direct comparison between your two options, how about getting two instances of CLI 3.0 running with two instances of SETISpy, letting the computer get into the middle of a couple of WU's, and then take note of the typical Instantaneous Performance; then shut down, switch the RAM settings in the BIOS, reboot and compare the Instantaneous Performance to what it was with the other settings. If there's a major difference, that should reveal it.

 

[cory]

well i have found that with the older 2.x client
the cas 2 vs 3 was worth about 20mhz bus speed.

i am running a lot of boards in my cluster at 150 mhz bus
and seti would run faster at 124 cas 2 then 143 cas 3.
i ended up getting better ram so i could run 143 to 150 cas 2
to really be fast.
that was with the old client and it does not seem to
mean as much with the newer client.
but a 33mhz bus difference i think would mean more than the
cas 2 vs 3 so go with 133 and try and force the motherboard to
use cas 2 and see what happens.