Author: Robert Hyatt
Date: 12:22:10 12/17/02
Go up one level in this thread
On December 17, 2002 at 12:38:01, Vincent Diepeveen wrote: >On December 17, 2002 at 12:08:04, Matt Taylor wrote: > >>On December 17, 2002 at 11:50:59, Enrique Irazoqui wrote: >> >>>On December 17, 2002 at 11:28:27, Matt Taylor wrote: >>> >>>>On December 17, 2002 at 11:20:38, Enrique Irazoqui wrote: >>>> >>>>>Deep Fritz 7 runs about 8% faster on a dual Xeon 2400 (detected as 4 cpus)than >>>>>on a dual AMD MP 2200+ (2x1800 GHz). Both with DDR 333. >>>>> >>>>>Enrique >>>> >>>>Which board/chipset? >>> >>>This friend of mine tells me that the 2x2400 Xeon machine has RDR, not DDR, >>>sorry. Here are some specs (in Spanish): >>> >>>Placa Base SuperMicro P4DC6+ >>> >>>- Chip-Set Intel© 860, Bus 400. Soporta RIMM 800Mhz. 4 bancos de memoria. >>>Ampliación máxima 2Gb. 2,5v. >>>- Controladora Adaptec 7899W Ultra 160 Dual. >>>- Slots de expansión: 2 PCI 64bits, 4 PCI 32bits, 1 AGP 4x Pro. >>>- Salidas externas: 2 serie, 1 paralelo, 4 USB 1.1., 2 PS/2. >>>- Soporta Adaptec SCSI RAID 2005S. >>> >>>Memoria RAM 1Ghz Rambus 800 Kingston >>> >>>- Velocidad de 8ns. RIMM de 800Mhz. >>>- Garantía de por vida. Índice de fallo: 1%. >>> >>>Enrique >>> >>>> I haven't seen any SMP boards that support better than a >>>>133 MHz FSB, and it's been a bitter disappointment. >>>> >>>>-Matt >> >>That's only 100 MHz. I was interested in the Athlon, though. My Athlon SMP >>system has a 133 MHz bus, and I've been sorely disappointed that no 166 MHz bus >>has popped up. I'm not suprised since JEDEC didn't standardize DDR-2 until this >>past summer, and it won't likely be out to market until next summer. >>-Matt > >It's no coincidence that dual machines have always slower latencies than >single cpu machines can have. > >It isn't easy to let them cooperate with each other! The latency is not significantly different on a dual vs a single cpu machine. And on a good dual, the bandwidth is 2x higher overall so the bandwidth is not significantly worse either. Simplest test is to run one program, using one thread, on a dual. Then run two _separate_ copies of that program at the same time and see if the times reported are any slower. If not, there is no loss at all. Any loss is directly attributable to the parallel hardware interactions...
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.