Need low latency project for "other" cores on quad-core

[jasong]

At the moment, I'm running two Prime95 instances and two LLR instances for the Prime Cullen Prime search. Because of the high bandwidth of all four programs, my timings are terrible.

Are there any low cpu bandwidth projects that we are pushing in, like OGR or ABC?

[vaughan]

Try BOINC Depspid (www.depspid.net)

[em99010pepe]

At the moment, I'm running two Prime95 instances and two LLR instances for the Prime Cullen Prime search. Because of the high bandwidth of all four programs, my timings are terrible.



1 instance LLR 5.473ms
2 instances (cpus 0 and 2) 5.449ms
3 instances (cpus 0, 1 and 2) cpu1=5.804ms and cpu2=5.466ms
4 instances (all 4 cpus) 6.674ms

(Jasong benches running LLR on Cullen numbers)

So AMD was right....hehe..thanks for the benches.


‘‘Barcelona’’ processors feature dedicated L2 cache and new 128-bit floating-point math units in each of its cores. And a new shared L3 cache can reduce data latency to main memory.


Carlos

[Digital Parasite]

It will be interesting to see how the new AMDs fair with quad core and if the new Intel Penryn CPUs do any better.

But 4 clients running at 6.674ms still gets a lot more work done overall than just 1 client running at a faster rate.

[em99010pepe]

It will be interesting to see how the new AMDs fair with quad core and if the new Intel Penryn CPUs do any better.

But 4 clients running at 6.674ms still gets a lot more work done overall than just 1 client running at a faster rate.

Agree but you are paying for two dual core and not a true quad core.