@ Who?
Maybe you can help me out here? (it might interest others who read this thread)

I'm quite new (7 weeks) to the s.775, DDRII and PCIe.
I jumped strait from s.478 (NW & hot Prescotts) to C2D E6600 and a very cheap ASRock Dual775-VSTA mainboard.
This is only a temporary 'soft-migration' solution for me - as the MB takes old DDR (and DDRII) and also AGP (and PCIe x4).

The MB is not OC-friendly at all. I managed to run stable at a lousy 280Mhz FSB @2520 with stock cooler and goob.
Load temp is amazing 42-44C (room 25C) measured with CoreTemp.
I use old Geil DDR PC3500 (217Mhz) wich are locked (by BIOS) at 200Mhz - but i run these at very low timings 2-3-3-6 on default voltage - which gives me a 1600+ RAC over 7 weeks.
Not bad I think?

When looking at "msattler" E6600 it has a 1900 RAC - but is running @3500 - which are some 40% more Mhz than my E6600@2520.
The difference here is 1Ghz - but the RAC doesn't indicate that much of a gain. I can't see if he has lowered the multiplier (default is 9).

This leads me to think, that tight and low RAM-timings and the divider at 1:1, is still very important when cruncing - rather than a high or highest acchived CPU-frequency (by lowering the multiplier).
Am I wrong here?

Kiva

it is all about balancing it.
The prefetchers of Core 2 can be efficent only if you memory modules can deliver in time the cache lines ordered.
DDRII is the minimum i ll recommand for Core 2, and 667 is ok, but there is no secret: a PC is sensitive to all the component.

You do not want cheap tires on a Ferrarie :)

Who?

Snip
When looking at "msattler" E6600 it has a 1900 RAC - but is running @3500 - which are some 40% more Mhz than my E6600@2520.
The difference here is 1Ghz - but the RAC doesn't indicate that much of a gain. I can't see if he has lowered the multiplier (default is 9).

This leads me to think, that tight and low RAM-timings and the divider at 1:1, is still very important when cruncing - rather than a high or highest acchived CPU-frequency (by lowering the multiplier).
Am I wrong here?

Kiva[/quote]
According to prior exchange of info with msattler, he is indeed running his e6600 at 3500, but with a fairly low memory clock rate. It appears that this unit is processing 60RAC in about 4100 seconds or so.
I have my e6600 running at 3500 also, but with a memory speed of 970 and am processing 60 RAC in around 3600 seconds.
I recommend the use of a good temp monitor program and use it to help you find the maximum performance at the lowest possible CPU and MB tempature while maintaining 100% stability.

---

When we finally figure it all out, all the rules will change and we can start all over again.

Snip
When looking at "msattler" E6600 it has a 1900 RAC - but is running @3500 - which are some 40% more Mhz than my E6600@2520.
The difference here is 1Ghz - but the RAC doesn't indicate that much of a gain. I can't see if he has lowered the multiplier (default is 9).

This leads me to think, that tight and low RAM-timings and the divider at 1:1, is still very important when cruncing - rather than a high or highest acchived CPU-frequency (by lowering the multiplier).
Am I wrong here?

Kiva

According to prior exchange of info with msattler, he is indeed running his e6600 at 3500, but with a fairly low memory clock rate. It appears that this unit is processing 60RAC in about 4100 seconds or so.
I have my e6600 running at 3500 also, but with a memory speed of 970 and am processing 60 RAC in around 3600 seconds.
I recommend the use of a good temp monitor program and use it to help you find the maximum performance at the lowest possible CPU and MB tempature while maintaining 100% stability.[/quote]
Items of importance to S@H crunching speed appear to be CPU speed, L2 cache size, memory bandwidth, and HD speed. In other words increasing any of these too much (other than L2) at the expense of the other items may not help much.

---

BOINC WIKI

Thanks for replies ;)

@ Pilot

I recommend the use of a good temp monitor program and use it to help you find the maximum performance at the lowest possible CPU and MB tempature while maintaining 100% stability.

I use "Core Temp" which can be found here: http://www.thecoolest.zerobrains.com/CoreTemp/
(worldwide xtreme OC'er recommend this for measuring temps on C2D/Q and other types of CPU's)
I have no trouble with OC, cooling and stability!

@ John McLeod VII
In other words: "Nothing new under the sun" - when it comes to building/tuning a s.775 C2D performance 'cruncher-rig' compared to a P4 socket478.

You are mentioning HD speed:
I used to run SETI Classic on RamDisks on my P3's with Win98 and Win2K - which gave me some 8-10% performance gain.

I was wondering if there's anything to gain by running BOINC from a RamDisk nowadays?
In 12 hrs BOINC has written (I/O) 280MBytes to the harddisk (according to TaskManager) - but whereto is BOINC writting this?? Is it in the BOINC installed folder or is it in Windows temp-folder (or other places) ?

Kiva

---

Greetings from Norway

Crunch3er & AK-V8 Inside

My current hdd rotation speed is 7,200 rpm, Soon It will be 10,000rpm and hence I'm putting in sound absorbing pads[Akasa](I need another 2 kits and a 4th new case).

Yeah some of those 10,000rpm hard drives make a lot of noise.

Akasa is one of more common soundproofing mats used, Acoustipack is another.

---

My current hdd rotation speed is 7,200 rpm, Soon It will be 10,000rpm and hence I'm putting in sound absorbing pads[Akasa](I need another 2 kits and a 4th new case).

Yeah some of those 10,000rpm hard drives make a lot of noise.

Akasa is one of more common soundproofing mats used, Acoustipack is another.

Yeah I found this over at silentpcreview on the subject. :D
http://www.silentpcreview.com/article71-page1.html

Interesting article. The results with the Sonata were nice... but pretty expensive for the foam.

---

I don't think putting in a 10,000 RPM hard disc drive is going to help speed up Seti appreciably. I've never gamed, but it's possible it could help there. I believe Seti only takes about a tenth of a second to write to disc, every minute. I heard the writes every minute on my Pentium III 450 but now with the PD950s I don't notice any writes except an occasional flash of the LED indicator in front of the cases. How can they make those arm-wiggles so quiet? I don't think that the insides of disc drives are completely evacuated; otherwise there would be no air lubricant on which the read/write heads could glide, over the disc surface. The proximity of the head to the disc surface was only about 0.5 micron some ten years ago. It's gotta be closer now.

---

I don't think putting in a 10,000 RPM hard disc drive is going to help speed up Seti appreciably. I've never gamed, but it's possible it could help there. I believe Seti only takes about a tenth of a second to write to disc, every minute.

I agree. Depending upon your preference settings, hard drive performance should not impact crunch times much at all. I have my settings set to write to hard drive at most every 300 seconds, so I don't see how one hard drive access every 5 minutes, even with a very slow hardrive subsystem, would have much of an effect.

---

"Time is simply the mechanism that keeps everything from happening all at once."

Well someone poo poohed an Windows XP x64 64bit Boinc Client and app saying It wouldn't be compatible with Boinc or Seti

The statement was that it would not gain much in performance over a similarly optimized 32 bit app as most of the work is done in floating point - which is already 80 bits.

---

BOINC WIKI

Even if you use four PIII Xeons they would be extremely slow compared to the more recent processors. You'd also have to worry about support, not only for the motherboard, but the processors. Maybe a more recent dual-socket system would be better. A Dempsey perhaps.

He is correct. If you used 4 of the fastest PIII Xeon's they would still not equal the speed of the fastest Core 2 Duo chip.

---