BadSweetums
Member
Tom's HW Guide published an article comparing how various hardware platforms performed at various memory timings. They concluded that memory timings no longer contributed a lot to system perfomance.
They compared a P4 3.2 EE system to a P4 3.2 system to an Athlon 3.2 FX-51 system to an Athlon 64 3.2 system to an Athlon XP 3200+ system. Their results compared memory operating at 3 timings; 2-2-2-5, 2.5-3-3-6, and 3-4-4-8.
The best results (IMO) showed that using lower latency memory enabled a P4 3.2 to out-perform a P4 3.2 EE by one second.
The P4 EE spread in tenths of seconds between 2-2-2-5 and 3-4-4-8 for encoding 1.2 GBs of DV to MPEG-2 (DVD) at 720x576 at 25 fps was 3.6 seconds. (Fewer seconds are better. )
Best score; P4 3.2 EE, 172.5 seconds at fastest memory timings to 176.1 seconds at the slowest memory timings with 1 GB of installed memory.
Worst score; Athlon XP 3.2, 224.9 seconds at fastest memory timings to 227.6 seconds at the slowest memory timings with 1 GB of installed memory.
Nicest discovery: P4 3.2, 175.1 seconds at fastest memory timings to 178.0 seconds at the slowest memory timings with 1 GB of installed memory. (a 3 sec faster EE ain't worth the price of admission IMO)
So, reliable memory which does not fault and cause fatal errors is still critical; BUT, low latency memory can save you only about 4 seconds for every three minutes of time spent performing memory intensive tasks.
OR, citing their 32-bit color UT2003 at 1024x768 results
Low latency memory is worth an extra 4 (four) fps in systems that generate 214 (Athlon XP) to 284 fps (Athlon 64 FX-51).
My conclusion, ultra fast low latency memory is not worth as much as a faster (or more capable) processor, or platform.
Was there much discussion of this topic closer to when it was originally published, January 19, 2004? Did THG screw up? What is low latency memory worth beyond bragging rights?
They compared a P4 3.2 EE system to a P4 3.2 system to an Athlon 3.2 FX-51 system to an Athlon 64 3.2 system to an Athlon XP 3200+ system. Their results compared memory operating at 3 timings; 2-2-2-5, 2.5-3-3-6, and 3-4-4-8.
The best results (IMO) showed that using lower latency memory enabled a P4 3.2 to out-perform a P4 3.2 EE by one second.
The P4 EE spread in tenths of seconds between 2-2-2-5 and 3-4-4-8 for encoding 1.2 GBs of DV to MPEG-2 (DVD) at 720x576 at 25 fps was 3.6 seconds. (Fewer seconds are better. )
Best score; P4 3.2 EE, 172.5 seconds at fastest memory timings to 176.1 seconds at the slowest memory timings with 1 GB of installed memory.
Worst score; Athlon XP 3.2, 224.9 seconds at fastest memory timings to 227.6 seconds at the slowest memory timings with 1 GB of installed memory.
Nicest discovery: P4 3.2, 175.1 seconds at fastest memory timings to 178.0 seconds at the slowest memory timings with 1 GB of installed memory. (a 3 sec faster EE ain't worth the price of admission IMO)
So, reliable memory which does not fault and cause fatal errors is still critical; BUT, low latency memory can save you only about 4 seconds for every three minutes of time spent performing memory intensive tasks.
OR, citing their 32-bit color UT2003 at 1024x768 results
Low latency memory is worth an extra 4 (four) fps in systems that generate 214 (Athlon XP) to 284 fps (Athlon 64 FX-51).
My conclusion, ultra fast low latency memory is not worth as much as a faster (or more capable) processor, or platform.
Was there much discussion of this topic closer to when it was originally published, January 19, 2004? Did THG screw up? What is low latency memory worth beyond bragging rights?
